Heme triggers TLR4 signaling leading to endothelial cell activation and vaso-occlusion in murine sickle cell disease

During hemolysis, heme may accumulate in plasma derived from hemoglobin (Hb) released by lysed erythrocytes. Heme exerts multiple adverse effects, including leukocyte activation and migration, adhesion molecule and cytokine upregulation, and oxidant production.^1-4 

Sickle cell disease (SCD) is characterized by chronic unrelenting hemolysis, inflammation and coagulation, adhesion of blood cells, vaso-occlusion, and ischemia-reperfusion injury leading to strokes and organ infarctions. Endothelial cells (ECs), leukocytes, and platelets are chronically exposed to Hb and heme that may promote activated proinflammatory and prothrombotic phenotypes.^5-8  To explain the basis for such phenotypes, substantial attention has been directed to hemolysis in SCD; indeed, in some patients, hemolysis may be especially brisk (hyperhemolysis) during acute painful episodes.⁹  Endothelial dysfunction with diminished nitric oxide (NO) bioavailability and increased adhesion molecule expression is a known consequence of hemolysis and systemic inflammatory states.¹⁰  Interest also resides in the oxidative cytotoxic effects of heme in SCD.¹¹ 

The present study examined a potential mechanism that links these 3 salient phenotypic features of SCD: hemolysis, vaso-occlusion, and inflammation. Specifically, this study examined the hypothesis that heme, released from plasma Hb during hemolysis, is an extracellular cell-signaling molecule and a causal factor in the development of vascular inflammation and vaso-occlusion in SCD.

Stroma-free Hb was a gift from Sangart. Methemoglobin (metHb) and cyanomethemoglobin (cyanometHb) were prepared as described previously.¹²  Initial studies employed clinical grade Panhematin (Ovation Pharmaceuticals). Subsequent studies employed hematin prepared immediately before use by mixing 10 mg hemin chloride (Frontier Scientific), 10 mg D-sorbitol (Sigma-Aldrich), and 6.9 mg sodium carbonate (Sigma-Aldrich) in 5.7 mL sterile saline (Baxter) for 30 minutes in the dark. All heme preparations, appropriately diluted in saline, were filtered at 0.22 µm. Protoporphyrin IX (PPIX; Frontier Scientific) preparations were prepared identically to heme. The structures of heme (ferrous PPIX), hemin (ferric PPIX), and PPIX (no iron) are shown in supplemental Figure 1 (available on the Blood Web site). The term “heme” is used generically to refer to both heme and “hemin.” Endotoxin levels were monitored using a Limulus amebocyte lysate test (GenScript). All Hb/heme preparations contained <0.1 endotoxin units (EU)/mL. Human haptoglobin was a gift from Bio Products Laboratory. Size-exclusion high-performance liquid chromatography profiles of haptoglobin showed the following molecular weight distribution: 60% with 2 αβ (dimer, haptoglobin 1-1), 21% with 3 αβ (trimer, mostly haptoglobin 1-2), and 19% larger forms (polymer, mostly haptoglobin 2-2). Hemopexin was purified from rabbit plasma or human recombinant hemopexin was purchased (Athens Research & Technology).¹³ 

All animal experiments were approved by the University of Minnesota’s Institutional Animal Care and Use Committee. We used male and female NY1DD¹⁴  and HbSS-Townes¹⁵  transgenic sickle mice. The NY1DD and HbSS-Townes mice are on C57BL/6 and mixed genetic backgrounds, respectively. The NY1DD mice are homozygous for deletion of the mouse β^major globin and express a human α and β^S globin transgene. NY1DD mice have no anemia and a red blood cell (RBC) half-life of 7 days (J.D.B. and G.M.V., unpublished data); C57BL/6 mice (RBC half-life, 24 days)¹⁶  served as controls. The HbSS-Townes mice were created by knocking in human α and β^S globins into the sites where murine α and β globins were knocked out. HbSS-Townes mice have severe anemia and an RBC half-life of 2.5 days.¹⁶  HbAA-Townes control mice (RBC half-life, 16 days)¹⁶  were created by replacing the β^S with β^A. HbAS-Townes mice (RBC half-life, 11 days)¹⁶  were created by breeding HbSS with HbAA mice. TLR4^−/− mice (The Jackson Laboratory; C57BL/6 background) were transplant recipients and donors of murine pulmonary vein ECs (mPVECs).

Dorsal skinfold chambers (DSFCs) were implanted on NY1DD, HbSS-, HbAS-, and HbAA-Townes and C57BL/6 mice at age 8 to 10 weeks and NY1DD/TLR4^−/− and NY1DD/TLR4^+/+ chimeric bone marrow (BM) transplant mice at age 20 to 22 weeks as described elsewhere.¹⁷  At 3 days postimplantation, flowing subcutaneous venules (n = 20-30) in the DSFCs were randomly selected at baseline in each mouse. Mice were then given a bolus infusion (0.012 mL/g body weight) via tail vein with the indicated doses of saline, water, Hb, metHb, cyanometHb, heme, or lipopolysaccharide (LPS; Escherichia coli, serotype O111:B4; Sigma-Aldrich) or exposed to 1-hour hypoxia (7%O₂/93%N₂) followed by 1-hour H/R. Venules selected at baseline were reexamined 1 hour after infusion or H/R. Venules without visible blood flow were counted as static and the percent static venules calculated. In some studies, HbSS and HbAA mice were not given a stimulus and venules were reexamined on days 1 and 2 after baseline to determine steady-state stasis. For inhibitor studies, we infused haptoglobin, hemopexin, methylene blue (Sigma-Aldrich), PPIX, apocynin (Santa Cruz Biotechnology), diphenyleneiodonium (DPI) (Sigma-Aldrich), calphostin C (Tochris Bioscience), starch-conjugated deferoxamine (S-DFO; Bo Hedlund), N-acetylcysteine (NAC; Sigma-Aldrich), TAK-242 (InvivoGen), or blocking immunoglobulin G (IgG) to P-selectin (clone RB40.34), E-selectin (clone 10E9.6), vascular cell adhesion molecule 1 (VCAM-1) (clone 429), intercellular adhesion molecule 1 (ICAM-1) (clone 3E2), platelet EC adhesion molecule 1 (PECAM-1) (clone 35-95; BD Biosciences), αVβ3 and α4β1 integrins (clones LM609 and PS/2; Millipore), von Willebrand factor (VWF) (Cedarlane; CL20176A-R), thrombomodulin (clone MAB3894; R&D Systems) or LPS (clone WN1 222-5; Hycult Biotech). For negative controls, we infused the appropriate control IgG (BD Biosciences, BioLegend, Jackson ImmunoResearch), vehicle, or 6% hetastarch in saline (B. Braun Medical). Total infusion volumes (inhibitors + Hb/heme/LPS) were limited to 0.012 mL/g body weight (∼20% blood volume). Hb concentrations were normalized to heme (4 hemes per Hb).

Plasma heme was measured by pseudoperoxidase activity.¹⁸ 

Primary human umbilical vein ECs (HUVECs) were cultured as described previously.¹²  Primary mPVECs from TLR4^−/− and C57BL/6 (TLR4^+/+) mice were isolated and cultured as described elsewhere.¹⁹  ECs were positively identified by costaining for VWF, VE-cadherin, and CD31.

HUVECs in 0.1% fetal bovine serum (FBS) were incubated with heme (20 µM) for 15 minutes, hemoglobin A (HbA) or hemoglobin S (HbS) (20 µM heme) for 24 hours, or Hb for 24 hours followed by heme for 15 minutes or H/R (3 hours in 95% N₂/5% CO₂ followed by 1 hour in 95% air/5% CO₂). Haptoglobin, hemopexin, apocynin, calphostin C, or TAK-242 was used as an inhibitor. Cells were fixed in 4% paraformaldehyde, stained with anti–P-selectin (Santa Cruz Biotechnology) and anti-VWF IgG (Cedarlane), and visualized with appropriate secondary IgG labeled with fluorescein isothiocyanate (FITC) (P-selectin) and tetramethylrhodamine-5(and 6)-isothiocyanate (TRITC) (VWF). Nuclei were counterstained with 4′,6-Diamidino-2-phenylindole (DAPI; Santa Cruz Biotechnology). Cells were mounted using Fluoromount-G (Southern Biotech) and visualized, and images were acquired using a FluoView FV1000 BX2 upright confocal microscope (Olympus, Center Valley, PA) with a ×60 objective at room temperature and processed with FluoView (Olympus) and Adobe Photoshop software (San Jose, CA).

For tissue staining, C57BL/6 and NY1DD mice were administered saline, heme, or histamine, a known Weibel-Palade body (WPB) agonist. After 15 minutes, mice were sacrificed; organs were removed and placed into optimal cutting temperature compound (Sakura Finetek), snap-frozen, and stored at −85°C. Six-micrometer tissue sections were mounted on glass slides and fixed in 4% paraformaldehyde. Slides were blocked with 3% donkey serum, stained, mounted, and visualized as described for HUVECs. Nonspecific staining was assessed using nonimmune IgG. Dorsal skin samples from C57BL/6 and NY1DD mice were fixed in Zamboni’s and processed and analyzed as described elsewhere²⁰  using anti-VWF IgG.

Intracellular reactive oxygen species (ROS) levels were measured using a cell-permeable fluorescent probe, dichlorodihydrofluorescein diacetate (DCFH-DA), as described previously.²¹ 

mPVECs from TLR4^+/+ and TLR4^−/− mice were incubated with heme, Hb, or LPS for 15 minutes (n = 6 wells/treatment). Cells were fixed and cell surface VWF was measured by enzyme immunoassay using sheep anti-VWF IgG (Abcam) and donkey anti-sheep IgG-alkaline phosphatase conjugate (Jackson ImmunoResearch).

Chimeric mice were generated by harvesting BM from NY1DD mice (β^S+/+) and transplanting the BM into lethally irradiated TLR4^−/− or TLR4^+/+ (C57BL/6) mice. Recipients (8-10 weeks of age) were irradiated with 2 doses of 5 Gy (X-RAD 320 Biological Irradiator) 3 hours apart. During the 3-hour interval, BM donors were killed and BM was collected from both femurs. Ten million BM cells were injected via tail vein into each irradiated recipient. Drinking water containing 0.2% neomycin sulfate (Sigma-Aldrich) was given to transplanted mice for 3 weeks immediately after transplantation. Twelve weeks posttransplant, globin phenotype was determined by Hb isoelectric focusing. Chimeric mice were employed 12 to 14 weeks after transplant.

Leukocyte rolling and adhesion were measured in mice with DSFCs as described previously.^17,22 

HUVEC or mPVEC isolated from TLR^+/+ and TLR^−/− mice, grown to confluence, were incubated with serum-free media containing LPS, TNF-α, or heme or pretreated with TAK-242 followed by heme. Cells were washed and nuclear extracts were prepared using a nuclear extraction kit (Affymetrix). Western blots were performed using a rabbit IgG to nuclear factor κB (NF-κB) phospho-p65 (Ser536)²³  (Cell Signaling Technology).

HbSS and HbAA mice (n = 6 mice per treatment) were given a bolus infusion of heme at time zero. One treatment group received TAK-242 30 minutes before heme infusion. Time of death after heme infusion was recorded.

Analyses were performed with SigmaStat 3.5 for Windows (Systat Software). Comparisons of multiple treatment groups were made using 1-way analysis of variance (Holm-Sidak method). An unpaired t test was used for 2 groups with equal variances.

Because hemolysis is fundamental to the pathobiology of SCD, we tested whether hemolysis and plasma Hb induce vaso-occlusion (percent stasis) in 2 SCD models (NY1DD and HbSS-Townes) and 2 nonsickling control groups (C57BL/6 and HbAA-Townes). NY1DD sickle mice administered saline had 6.6% stasis after 1 hour (Figure 1A). NY1DD mice administered water to induce hemolysis²⁴  had 39.8% stasis (P < .05, water vs saline). Likewise, HbA at 32, 3.2, or 0.32 µmol/kg induced 38.9%, 31.1%, and 29.9% stasis 1 hour after infusion (P < .05, HbA vs saline).

Methemoglobin A (metHbA), which can release heme,²⁵  induced 36.2% stasis, whereas equimolar heme-stabilized cyanomethemoglobin A (cyanometHbA) induced 8.2% stasis (Figure 1A; P < .05, metHbA vs cyanometHbA), suggesting that metHb plays a role in stasis. Infusion of HbA and methylene blue, a metHb reducing agent,²⁶  inhibited stasis (7.0%) compared with HbA alone (31.1%) (P < .05, methylene blue + HbA vs HbA). Binding HbA with equimolar haptoglobin also inhibited stasis (6.9%) (P < .05, haptoglobin + HbA vs HbA). Similarly, the high affinity heme-binding protein hemopexin inhibited HbA-induced stasis (5.8%) (P < .05, hemopexin + HbA vs HbA). These data indicate that Hb induces stasis in NY1DD mice through the release of heme from metHb.

HbS infused into NY1DD mice induced 38.3% stasis, which was 7.2% more than the stasis induced by equimolar amounts of HbA (Figure 1A; P < .05, HbS vs HbA). This suggests that HbS may lose heme more readily than HbA in NY1DD mice.

We also infused HbA into HbSS-, HbAS-, and HbAA-Townes mice (Figure 1A). HbA (0.32 µmol/kg) induced 30.0%, 19.8%, and 10.0% stasis, respectively (P < .05, all pairwise comparisons). At this physiological dose, percent stasis in HbSS mice was not significantly different than NY1DD mice. Finally, HbA (32 µmol/kg) infused into C57BL/6 mice induced 5.9% stasis that was significantly lower than stasis in NY1DD mice (38.9%, P < .05) but similar to stasis in C57BL/6 mice infused with saline (5.5%), indicating that Hb induces stasis in sickle, but not normal, mice.

Because liberated heme appeared to be the active, stasis-inducing molecule, we examined stasis in mice after infusion of heme. Heme at 32, 3.2, or 0.4 µmol/kg induced 38.9%, 33.0%, and 19.7% stasis 1 hour after infusion (P < .05, heme vs saline; Figure 1B), which is comparable to stasis induced by HbA. Equimolar amounts of hemopexin inhibited heme-induced stasis (2.4%, P < .05, hemopexin + heme vs heme). Similar to Hb, heme (32 µmol/kg) infused into C57BL/6 mice induced only 6.1% stasis (P < .05, C57BL/6 vs NY1DD; Figure 1B), indicating that, like Hb, heme induces stasis in sickle, but not wild-type, mice.

To evaluate the specificity of heme-induced stasis, we examined whether PPIX, a tetrapyrrole ring identical to heme but lacking iron, antagonizes heme-mediated stasis. PPIX (40 µmol/kg) was administered by intraperitoneal (IP) injection to NY1DD mice followed 1 hour later by an IV infusion of heme (3.2 µmol/kg). PPIX inhibited heme-mediated stasis at 1 hour after heme infusion (9.1% stasis with PPIX + heme vs 33.0% stasis with heme alone, P < .05; Figure 1B). Thus, redox-active iron in heme is required for heme-mediated stasis.

Mean percent stasis was plotted against mean total plasma heme (Figure 1C) in NY1DD mice 1 hour after infusion of heme, Hb, metHb, or cyanometHb or exposure to H/R. Stasis was correlated to total plasma heme (r² = 0.81). Since the half-life of RBCs in HbSS mice (2.5 days) is short, we hypothesize that sickle mice are chronically hemolyzing and releasing Hb and heme into plasma. We examined steady-state stasis in HbSS- and HbAA-Townes mice, which have total plasma heme levels of 52.6 ± 18.3 µM and 28.8 ± 3.1µM, respectively (P < .05). DSFCs were implanted and 3 days later flowing venules were selected at baseline (day 0). The same venules were reexamined for stasis on days 1 and 2 in steady state without any additional stimulus. Untreated HbSS mice had 8% to 10% stasis in steady state on days 1 and 2, compared with 0% to 1% stasis in HbAA mice (P < .05, Figure 1D). Steady-state stasis in HbSS mice was abrogated after infusion of haptoglobin or hemopexin (P < .05), indicating that chronic hemolysis is continually driving stasis in HbSS mice. On day 3, infusion of Hb (0.32 µmol heme/kg) into HbSS mice increased stasis to 30% (data not shown), suggesting that vaso-occlusion occurs during steady state in HbSS mice and small increases in hemolysis can induce additional vaso-occlusion.

Multiple adhesion molecules have been previously implicated in the adhesion of blood cells to the endothelium and vaso-occlusion.^5,22,27-31  We examined the role of adhesion molecules in heme-induced stasis (Figure 2A). Infusion of IgG against P-selectin, E-selectin, VWF, VCAM-1, α4β1 integrin, ICAM-1, PECAM-1, or αVβ3 integrin inhibited heme-induced stasis compared with control IgG (P < .001). In contrast, an antibody to thrombomodulin on endothelium had less stasis than control IgG but more stasis than anti–adhesion molecule IgG (P < .002).

Two of the adhesion molecules required for stasis, P-selectin and VWF, are stored in EC WPBs.³²  Agonists of WPB mobilization, such as histamine and thrombin, induce transport of WPBs to EC surfaces within minutes. Since stasis occurs less than 30 minutes after heme infusion (data not shown), we examined P-selectin and VWF expression on the cell surface of HUVECs treated with Hb, heme, and/or H/R. Heme stimulated the expression of P-selectin and VWF “strings” on the cell surface within 15 minutes of heme exposure (Figure 2B) and as early as 5 minutes after heme addition (data not shown). Heme-mediated expression was inhibited by hemopexin. Treatment of HUVECs with HbA or HbS for 24 hours stimulated P-selectin and VWF expression that was inhibited by haptoglobin or hemopexin, supporting a role for free-heme in Hb-mediated WPB degranulation. Treatment of HUVECs with Hb for 24 hours followed by heme for 15 minutes enhanced P-selectin and VWF expression on the cell surface to levels higher than with heme or Hb alone. H/R stimulated VWF expression that was enhanced in ECs pretreated with HbA or HbS for 24 hours. P-selectin and VWF surface expression was greater in ECs pretreated with HbS + H/R vs HbA + H/R, suggesting greater release of heme from HbS after H/R.

Heme was infused into C57BL/6 and NY1DD mice to examine P-selectin and VWF expression in vivo. Infusion with histamine (WPB agonist) or saline provided positive and negative controls, respectively. After 15 minutes, mice were sacrificed and organs examined for expression of P-selectin and VWF on vessel surfaces. Varying degrees of P-selectin and VWF expression occurred on venules in the skin (Figure 2C, only VWF shown), lungs (supplemental Figure 2A), brain (supplemental Figure 2B), liver (supplemental Figure 2C), and kidneys (supplemental Figure 2D) of heme- and histamine-treated mice. Expression of P-selectin and VWF in saline-treated mice was visibly lower but higher in saline-treated NY1DD than in C57BL/6 mice. Thus, heme markedly induces expression of P-selectin and VWF on vessel walls in both normal and sickle mice in vivo.

NADPH oxidase (NOX) is a major source of oxidants in ECs and has been implicated in vessel wall adhesion in SCD mice.³³  NOX inhibitors apocynin or DPI inhibited heme-induced stasis in NY1DD mice (P < .05, Figure 3A). Inhibition of NOX activator protein kinase C (PKC) with calphostin C or treatment with the iron chelator S-DFO or the antioxidant NAC inhibited stasis (P < .05; Figure 3A). Also, heme-mediated VWF expression on HUVECs was blocked by apocynin or calphostin C treatment (Figure 3B). Moreover, apocynin, DPI, calphostin C, and the antioxidants NAC, quercetin, or DFO prevented heme-mediated ROS production in HUVECs (P < .025; Figure 3C). These data indicate that PKC, NOX, and oxidants are required for heme-induced WPB degranulation and stasis.

Heme activates TLR4 signaling and tumor necrosis factor α (TNF-α) production in macrophages.⁴  TLR4 signaling was examined in heme-treated HUVECs. We inhibited TLR4 signaling with TAK-242 (resatorvid), a specific small-molecule inhibitor of TLR4 signaling that selectively binds to Cys747 on the intracellular domain of TLR4 and interrupts interactions between TLR4 and its adaptor molecules.³⁴  TAK-242 inhibited heme-induced P-selectin and VWF expression on the surface of HUVECs (Figure 4A), indicating that heme induces the release of P-selectin and VWF via TLR4 signaling.

We quantified VWF expression on mPVECs in primary culture from TLR4^+/+ and TLR4^−/− mice by enzyme immunoassay. LPS, the classic TLR4 ligand, or heme induced VWF expression within 15 minutes on TLR4^+/+, but not TLR4^−/−, mPVEC cell surfaces (Figure 4B). This finding confirms that TLR4 is required for heme- and LPS-mediated release of VWF to the EC surface. Treatment of cells with Hb for 15 minutes did not induce VWF expression on either TLR4^+/+ or TLR4^−/− mPVECs, indicating that Hb does not release significant amounts of heme in the first 15 minutes of cell culture.

We examined whether TLR4 signaling is required for stasis. TAK-242 inhibited heme-induced stasis in NY1DD mice at 1 hour (Figure 4C; P < .01, 26.4% stasis with vehicle + heme vs 9.3% stasis with TAK-242 + heme).

LPS also induced stasis in NY1DD mice through TLR4. Mice were treated with vehicle + LPS or TAK-242 + LPS (Figure 4C). LPS induced stasis (45.0%), but pretreatment with TAK-242 inhibited LPS-induced stasis (11.0%, P < .001). Thus, both heme and LPS initiate TLR4 signaling, P-selectin and VWF expression, and vaso-occlusion in sickle mice. LPS did not induce stasis in C57BL/6 mice (4.2%; Figure 4C), indicating the β^S globin is required for stasis.

All Hb/heme preparations contained <0.1 EU/mL. Anti-LPS IgG directed against the LPS core did not inhibit heme-induced stasis (27.7% stasis with isotype control IgG + heme vs 24.4% stasis with anti-LPS IgG + heme) but significantly inhibited LPS-induced stasis (40.8% stasis with isotype control IgG + LPS vs 8.6% stasis with anti-LPS IgG + LPS, P < .001) (Figure 4C), confirming that heme-specific TLR4 activation is not caused by LPS contamination.

TAK-242 also inhibited H/R-induced stasis in NY1DD mice at 1 hour (23.3% stasis with vehicle + H/R vs 7.4% stasis with TAK-242 + H/R, P < .01) (Figure 4C). Thus, H/R-induced stasis requires TLR4 signaling that may be initiated by increased plasma heme levels. In corroboration, H/R-induced stasis in sickle mice was accompanied by an increase in total plasma heme (Figure 1C).

To confirm that heme-induced stasis requires endothelial TLR4 signaling, we generated chimeric NY1DD sickle mice lacking endothelial TLR4 expression by transplanting BM from NY1DD sickle mice into lethally irradiated TLR4^−/− or TLR4^+/+ mice. Three months posttransplant, the presence of the β^S globin in blood was confirmed by isoelectric focusing (Figure 4D, bottom), DSFCs were implanted, and stasis was measured 1 hour after heme infusion. Nontransplanted NY1DD mice and TLR^−/− mice transplanted with BM from C57BL/6 mice served as positive and negative controls, respectively. Heme-induced stasis was prevented in NY1DD β^S/TLR4^−/− chimeric mice, but not in NY1DD β^S/TLR4^+/+ chimeric mice or nontransplanted NY1DD mice (Figure 4D, top). These results confirm that heme induces stasis in sickle mice through endothelial TLR4.

Leukocyte rolling and adhesion were measured in subcutaneous venules in NY1DD with DSFCs at baseline and after infusion of heme ± TAK-242. Leukocyte rolling flux (Figure 4E) and adhesion (Figure 4F) increased significantly after heme infusion in mice not pretreated with TAK-242. Pretreatment with TAK-242 blocked the heme-mediated increase in rolling and adhesion.

We previously demonstrated that NF-κB is activated and adhesion molecules are overexpressed in transgenic sickle mice.⁷  To determine whether heme activates NF-κB via TLR4 signaling, we used TAK-242 to inhibit TLR4 signaling in HUVECs (Figure 5A). LPS, TNF-α, and heme increased nuclear NF-κB phospho-p65, a specific marker of NF-κB activation.²³  TAK-242 blocked nuclear NF-κB phospho-p65 induction by heme. Similar findings were observed in mPVECs isolated from TLR4^+/+ and TLR4^−/− mice (Figure 5B). Heme, LPS, and TNF-α activated nuclear NF-κB phospho-p65 in TLR4^+/+ mPVECs and TAK-242 inhibited heme-mediated activation. In TLR4^−/− cells, NF-κB was activated by TNF-α, but not by heme or LPS. These findings demonstrate that heme and LPS activate nuclear NF-κB phospho-p65 in ECs via TLR4 signaling.

In studies in HbSS-Townes mice, we found that heme infusion was lethal within 24 hours (Figure 6). Autopsies could not determine a definitive cause of death. Heme was not lethal in HbAA-Townes mice, suggesting the β^S mutation is needed for heme lethality. Pretreatment with TAK-242 completely prevented heme-induced lethality, indicating a role for TLR4 signaling in heme-induced lethality.

Heme, released during hemolysis, activates TLR4 signaling on ECs leading to delivery of WPB constituents P-selectin and VWF to the vessel wall and NF-κB activation. These events trigger vaso-occlusion in sickle, but not normal, mice. The effects of heme are mimicked by TLR4 ligand LPS, thereby supporting a model that links hemolysis- and infection-induced vaso-occlusion in SCD through TLR4 signaling leading to NF-κB, PKC, and NOX activation; ROS production; and WPB degranulation (Figure 7).

Heme induces TNF-α secretion in monocyte/macrophages through TLR4 signaling.⁴  Our studies expand this finding by demonstrating that heme activates EC WPB degranulation, NF-κB, and vaso-occlusion in SCD mice. Heme, not ferrous Hb, was the active molecule in TLR4 induction of vaso-occlusion. MetHb, which gives up heme, induced stasis, whereas cyanometHb, which tightly binds heme,²⁵  did not. Neither heme nor Hb was able to induce stasis or WPB degranulation in the presence of the high-affinity heme-binding protein hemopexin. Our data suggest that generation of metHb is a necessary step in the release of heme. The metHb reducing agent methylene blue at clinically relevant concentrations also inhibited Hb-induced stasis. MetHb can be generated in vivo when oxyhemoglobin reacts with NO. Plasma metHb levels are elevated in SCD¹⁰  and are further increased by inhaled NO.³⁵  Thus, plasma oxyhemoglobin S in the pro-oxidative environment of SCD¹¹  may react with NO or other oxidants to generate methemoglobin S, which can release heme.

Plasma heme in SCD patients is approximately 4 μM compared with 0.2 μM in normal volunteers.¹⁰  However, plasma heme levels in SCD patients can exceed 20 µM.^10,36,37  In our studies, plasma heme is elevated in sickle mice and positively correlated with stasis. Endothelial TLR4 signaling and vascular stasis were induced at plasma heme levels that are in the pathophysiologic range (20-60 µM). Steady-state plasma heme levels reflect increased heme release during hemolysis and degradation by heme oxygenase but do not completely reveal the excessive heme throughput in SCD. Sustained extravascular hemolysis in SCD and shortened RBC lifespans, especially during crisis,⁹  continually deliver heme from sickle RBC to the vessel wall.

Although heme and Hb can activate endothelium in wild-type C57BL/6 mice and human and mouse ECs in culture, the β^S mutation was necessary for stasis. In C57BL/6 mice, supraphysiologic infusions (32 µmol heme/kg) of Hb or heme that overwhelm plasma haptoglobin or hemopexin induced similar amounts of stasis (5.9% to 6.1%) as saline. Stasis was dose responsive to β^S expression as heterozygous HbAS mice had stasis intermediate between HbAA and HbSS mice.

Stasis (∼10%) in untreated HbSS mice in steady state, but not HbAA mice or in HbSS mice treated with haptoglobin or hemopexin, supports a model in which hemolysis-driven vaso-occlusion is occurring continuously in SCD. HUVECs treated with Hb for 24 hours to mimic chronic hemolysis had increased P-selectin and VWF on the cell surface relative to untreated HUVECs. The addition of heme or H/R to the Hb-treated cells, simulating enhanced local hemolysis or hypoxic venules, elicited an exaggerated P-selectin/VWF response. Unstable HbS³⁸  elicited even greater inflammatory responses and stasis than HbA. The WPB agonists heme and histamine stimulated expression of P-selectin and VWF on vessel walls of sickle and normal mice in all tissues examined (skin, lung, liver, brain, and kidney). However, in tissues, there were variable responses (eg, EC activation and vaso-occlusion) that may depend on local factors in different vascular beds including EC heterogeneity, hemolysis, cytokines, haptoglobin/hemopexin levels, heme oxygenase activity, infections, and O₂ gradients. We hypothesize that local or global changes in hemolytic rates, infection, or hypoxia may precipitate vaso-occlusive crises.

Hemolytic rates in SCD correlate with the quantity of “activated” VWF in plasma.³⁹  Large VWF multimers promote thrombosis in SCD and increase adhesion of SS-RBC to ECs.³¹  In SCD plasma, the ADAMTS13 proteolysis site in VWF can be blocked by Hb bound to the A2-domain of these VWF multimers,⁴⁰  fostering an adhesive milieu. Our data demonstrate that heme induces VWF expression on endothelium and provide an explanation for the correlation among hemolytic rates, plasma VWF levels, and vaso-occlusive crises.

Haptoglobin minimizes heme loss from Hb by transporting Hb to CD163 on macrophages for catabolism.⁴¹  SCD patients have low plasma haptoglobin levels.¹⁰  Plasma haptoglobin levels were lower in sickle mice compared with normal control mice (supplemental Figure 3). Infusion of human haptoglobin blocked Hb-induced stasis in sickle mice. These observations provide a rationale for the use of haptoglobin as a new therapeutic approach to SCD.⁴² 

When haptoglobin is overwhelmed by excess plasma Hb, Hb can be oxidized to metHb with subsequent liberation of heme. The released heme can bind to plasma hemopexin (K_d < 1 pM) or albumin (K_d ∼10 nM) from which it is normally transferred to hemopexin.⁴³  However, when haptoglobin and hemopexin are depleted, albumin becomes a reservoir for heme.⁴⁴  Normally, hemopexin safely transports heme away from the vessel wall to CD91 on hepatocytes where the heme is degraded and hemopexin recycled back to the plasma.^45-47  Thus, hemopexin protects the endothelium from heme-induced oxidative stress, inflammation, and cellular injury.^47-49  In SCD patients³⁷  and sickle mice (supplemental Figure 3), hemopexin levels are lower than in normal subjects. Despite the shorter half-life of RBCs in HbSS-Townes mice compared with NY1DD mice, plasma haptoglobin and hemopexin levels appear similar, perhaps explaining the similar amounts of stasis (30%) seen when infused with low doses of Hb (0.32 µmol/kg). However, it is noteworthy that stasis plateaued at supraphysiological heme doses (32 µmol/kg) and was approximately 10% higher in HbSS-Townes (50%) vs NY1DD (40%) mice (data not shown). These data suggest that hemopexin supplementation might blunt EC dysfunction in haptoglobin/hemopexin deficiency states such as SCD.⁴² 

Infusion of antibodies to WPB P-selectin or VWF inhibited stasis. Adhesion molecules that are not found in WPBs but are found on activated endothelium are also required for vaso-occlusion. Antibody blockade of adhesion molecules VCAM-1, α4β1 integrin, ICAM-1, PECAM-1, E-selectin, or αVβ3 integrin inhibited heme-induced stasis. Remarkably, blockade of only one adhesion molecule inhibited stasis. Isotype control antibodies or antibodies to thrombomodulin on the EC membrane had limited effects on stasis. PECAM-1 is required for leukocyte extravasation and for the first time has been implicated in vaso-occlusion. Endothelial αVβ3 integrin mediates adhesion of sickle RBCs via activated intercellular adhesion molecule 4 ligands on SS-RBC.²⁹  In addition, α4β1 integrin on leukocytes and SS-reticulocytes is a counter-receptor for VCAM-1 and can form multicellular aggregates in plasma linked by fibronectin.²⁷  Thus, multiple cell adhesion molecules participate in vaso-occlusion. Further studies are warranted to examine the effects of TLR4 signaling on the expression and activation state of these adhesion molecules.

NOX may be involved in TLR4 signaling. LPS can activate WPB degranulation and NF-κB activation in ECs that requires oxidant production by NOX.^50,51  In our studies, inhibition of NOX, its upstream activator PKC, or antioxidants inhibited heme-mediated stasis, WPB degranulation, and oxidant production. NOX in sickle RBCs and leukocytes and xanthine oxidase are also possible sources of oxidants at the vessel wall.⁵²  Further studies in chimeric HbSS/NOX^−/− mice are warranted.³³ 

Autopsies were performed on the HbSS mice after they succumbed to intravenous heme. RBC congestion was present in the lungs, but the cause of death could not be definitively determined. Intravenous heme was recently shown to induce acute lung injury and death in sickle mice that can be averted by TLR4 inhibition, TLR4 knockout in nonhematopoietic tissues, or hemopexin replacement therapy.⁵³  Suppression of TLR4-induced vaso-occlusion in the pulmonary vasculature may explain the survival of HbSS mice treated with TAK-242.

TNF-α, which is produced in response to heme-induced TLR4 signaling in monocytes,⁴  can induce P-selectin–dependent vaso-occlusion in SCD mice.³⁰  TNF-α signaling, which is independent of TLR4, induces WPB degranulation at the vessel wall.⁵⁴  Thus, multiple agonists (eg, TNF-α, LPS, interleukin-1β, thrombin, and histamine) in ECs can activate WPB exocytosis and vaso-occlusion in SCD. However, the defining and chronic nature of hemolysis in SCD suggests that heme-induced TLR4 signaling on monocytes, platelets, and ECs may be a primary driver of cell activation, inflammation, cytokine production, and vaso-occlusion.

EC TLR4 plays a critical role during sepsis and hemolytic states.⁵⁵  Chronic hemolysis in SCD may promote a proinflammatory tone mediated by TLR4 signaling and oxidant production leading to WPB degranulation, activation of NF-κB, and increases in adhesion molecules and proinflammatory cytokines and chemokines.⁶  Thus, heme may be central to the evolution of EC dysfunction and vasculopathy in SCD. These studies uncover new pharmacologic modalities that interrupt heme-mediated inflammation and vaso-occlusion in SCD, including haptoglobin and hemopexin and inhibitors of TLR4 and adhesion molecules. Improved understanding of the regulatory elements of TLR4 signaling pathways in ECs may facilitate the development of novel therapies for SCD and sepsis. These data add support for the therapeutic impact of therapies that decrease hemolysis in SCD, scavenge Hb or heme (haptoglobin or hemopexin), detoxify heme, or modulate TLR4 activation. Furthermore, these therapies would likely impact the pathobiology of disseminated intravascular coagulation, transfusion reactions, crush injuries, and sepsis.

This work was supported by grants from the National Institutes of Health, National Heart, Lung, and Blood Institute This work was supported by grants from the National Institutes of Health, National Heart, Lung, and Blood Institute (P01HL55552; R.P.H. and G.M.V.), (HL115467-01; G.M.V. and J.D.B.), (HL1109000-01A1; A.I.A.), and by a research grant award from Sangart (G.M.V. and J.D.B.). Dr Ann Smith thanks the University of Missouri Research Board and School of Biological Sciences at University of Missouri–Kansas City Research Incentive Funds for support. K.A.N. received support from the National Institutes of Health, National Institute of Diabetes and Digestive and Kidney Diseases (DK47060). We would like to thank Graham Brown for preparation of metHb and cyanometHb, Dr Heather Bechtel for assistance with the graphics in Figure 7, Dr Mark Young and Ashok Malavalli at Sangart for the preparation and generous gift of stroma-free Hb, Bio Products Laboratory for the preparation and generous donation of human haptoglobin, and Bo Hedlund for the generous gift of S-DFO. Michael J. Franklin read the manuscript and was extremely helpful in making editorial suggestions.

Contribution: J.D.B. designed the experiments, analyzed data, and wrote the paper; C.C. collected stasis, leukocyte rolling/adhesion, NF-κB, lethality, and plasma haptoglobin/hemopexin data; J.N. collected and analyzed cell culture, plasma heme, and immunohistochemistry data; L.M. and F.A. bred and phenotyped mice and performed transplants; A.I.A. designed the haptoglobin experiments and wrote the paper; A.S. designed the hemopexin experiments and wrote the paper; K.A.N. discussed the data and wrote the paper; R.P.H. analyzed data, designed experiments, and wrote the paper; and G.M.V. designed the experiments, analyzed data, and wrote the paper.

Conflict-of-interest disclosure: J.B., G.V., and R.H. receive research funding from and are consultants for Sangart.

Correspondence: John Belcher, University of Minnesota, 420 Delaware St SE, MMC 480, Minneapolis, MN 55455; e-mail: belcher@umn.edu.