Prothrombotic Effects of Fibronectin Isoforms Containing the EDA Domain
Objective— Fibronectin (FN) plays an important role in the formation of stable arterial thrombi at the site of vascular injury. FN containing Extra Domain A (EDA+FN) is absent from normal plasma, but elevated plasma levels of EDA+FN are found in several pathological conditions. We hypothesized that EDA+FN plays a special role in thrombosis.
Methods and Results— We used mouse strains constitutively including (EDA+/+) or excluding (EDA−/−) the EDA domain in all tissues and plasma. Using a flow chamber and the ferric-chloride injury model we found that EDA+FN accelerates thrombosis both in vitro and in vivo at arterial shear rates. In EDA+/+ mice thrombi (>30 μm) grew faster when compared with EDAWT/WT (6.6±0.2 minutes versus 8.3±0.6 minutes, P<0.05) and the mean vessel occlusion time was shorter (9.9±0.4 minutes versus 14.6±1.7 minutes, P<0.05). However, the presence of EDA+FN affected neither single platelet adhesion to subendothelium nor thrombosis in veins. In addition, the mortality rate of EDA+/+ mice after collagen/epinephrine infusion was twice that of EDAWT/WT or EDA−/− mice.
Conclusions— Our findings reveal that EDA+FN has prothrombotic activity, and its presence in plasma may worsen pathological conditions in which this form is elevated.
- plasma fibronectin
- fibronectin splice variants isoforms
- arterial thrombosis
- intravital microscopy
- arterial and venous injury
The formation of a thrombus in an injured vessel wall is a complex process that involves multiple adhesion molecules and their respective receptors on the platelet surface. von Willebrand factor (vWF) and fibrinogen are considered the major ligands mediating platelet adhesion and aggregation. However, in an experimental model 73% of the injured vessels of mice lacking both fibrinogen and vWF still formed occlusive thrombi either at the site of injury or downstream,1 suggesting that other major adhesive proteins, such as fibronectin (FN), might contribute to the process. Incorporation of FN into a growing thrombus was shown both in vitro2 and in vivo.3 Additionally, it was documented in vivo that the depletion or lower levels (50%) of plasma FN result in serious defects in arterial thrombosis.3,4 FN is a dimeric multidomain glycoprotein playing an important role in adhesion, migration, growth and differentiation of cells.5,6 FN generates protein diversity as a consequence of alternative processing of a single primary transcript at 3 sites: the Extra Domain B (EDB, EDII, or EIII-B), Extra Domain A (EDA, EDI, or EIII-A), and the Type III Homologies Connecting Segment (IIICS) (Figure 1A).7–9 Two major forms of FN exist: soluble plasma FN (pFN), which lacks both the EDA and EDB domains (Figure 1A); and cellular FN (cFN), which is deposited as insoluble fibrils in the extracellular matrix (ECM) and contains these domains at variable proportions.
FN is a ligand for many members of the integrin receptor family and binds to thrombosis-related proteins including heparin, collagen, and fibrin.10 Platelets contain both EDA+FN and EDA−FN,11 suggesting that EDA+FN may have some special role in thrombosis. Disease states such as atherosclerosis, pulmonary and acute vascular injury, diabetes, thrombocytopenic purpura, and ischemic stroke are accompanied by elevated plasma levels of EDA+FN.12–17 In spite of this, the role of both the EDA+FN present in plasma and that derived from platelets in platelet aggregation, thrombosis, and hemostasis remains unknown. To address the role of EDA+FN in these processes, genetically engineered mice that constitutively include (EDA+/+) or exclude (EDA−/−) the EDA domain18 were used. We used arterial and venous models of ferric chloride–induced endothelial injury, a perfusion chamber with escalating shear stress and a collagen-epinephrine induced model of pulmonary thromboembolism. We found that mice exclusively having the EDA+FN isoform in plasma and platelets had increased thrombosis, revealing that this FN isoform has prothrombotic activity.
Materials and Methods
The generation of the mice devoid of regulated splicing at the EDA exon has been previously described18 and is briefly detailed in the Supplemental Data I Section (available online at http://atvb.ahajournals.org). The mice used for intravital microscopy were male and female young mice (approx 3 to 4 weeks old), weighing 14 to 18 g. Infused platelets were isolated from 4- to 6-month-old mice of the same genotype. Animals were bred and housed at the ICGEB and the CBR Institute for Biomedical Research. Experimental procedures were approved by the Animal Care and Use Committees of each institution.
Blood Sampling and Platelet Preparation
Blood sampling and platelet preparation were done as previously described19 and is briefly detailed in the Supplemental Data I Section.
Platelet Aggregation Test
Platelet count was adjusted to the same concentration (3×108 platelets/mL) with modified Tyrode’s buffer containing 1 mmol/L CaCl2 and maintained at 37°C. Aggregation was initiated by adding the following agonists: ADP, thrombin, collagen (Nycomed), and protease-activated receptor-4 (PAR4) to platelet rich plasma (PRP) or to washed platelets and was monitored by light transmission that was recorded over 20 minutes on a Chrono-Log 4-channel optical aggregation system.
Pulmonary thromboembolism was induced as described previously20,21 with slight modifications. Briefly, thromboembolism was induced by intravenous injection of a mixture of soluble-collagen and epinephrine (40 μg [Nycomed] and 4.8 μg [Sigma-Aldrich], respectively, per 30 g body weight) into the tail veins of mice. Mice were randomized and genotypes kept unknown until the end of the experiment. The animals were observed for 1 hour from the time of injection. The time to death was normally less than 3 minutes.
Platelet interaction with immobilized collagen under flow conditions was studied as previously described22 and is briefly detailed in the Supplemental Data I Section.
Thrombus in Arterioles and Venules
A previously described model for arterial and venous thrombosis was used.1,23 Mice were anesthetized with 2.5% tribromoethanol (0.15 mL/10 g) and fluorescent platelets (2.5×109 platelets/kg) were infused through the retro-orbital plexus of the eye. An incision was made through the abdominal wall and mesentery arterioles of approx 100 μm in diameter, and mesentery veins of 200 to 300 μm were studied. Whatman paper saturated with ferric chloride (10%) solution was applied topically for 5 minutes, which induced denudation of the endothelium, and the vessel was monitored and video-recorded for 40 minutes after injury or until occlusion.
Results are reported as the mean±SEM, unless otherwise noted. The statistical significance of the difference between means was assessed by the ANOVA followed by the Bonferroni test. A value of P<0.05 was considered statistically significant. Thromboembolism data were analyzed by the Fisher test.
Constitutive Inclusion of EDA Domain in FN Results in Decreased Levels of FN in Platelets
Previously we have documented that constitutive inclusion of the EDA exon results in ≈70% to 80% decrease of FN in plasma of EDA+/+ mice compared with EDAWT/WT (Figure 1B).18,24 Because most of the FN present in platelets is endocytosed from plasma, we hypothesized that the amount of total FN in platelets was also decreased. As expected, platelet-FN in the EDA+/+ mice was ≈30% to 35% of the EDAWT/WT levels (Figure 1C). The FN levels in the platelets of the EDA−/− mice were similar to those of EDAWT/WT (Figure 1C). Western blot analysis using a monoclonal antibody specific for the EDA domain showed that EDAWT/WT platelets had roughly 20% of the amount found in the EDA+/+ platelets (not shown). No significant differences among the genotypes were observed for vitronectin and fibrinogen in washed platelets (not shown).
Normal Platelet Aggregation, Whole Blood Clotting Time, and Bleeding Time
Because various pathological conditions are accompanied by elevated levels of EDA+FN in plasma, we evaluated whether EDA+FN will affect platelet aggregation. In vitro aggregation was initiated by adding agonists (ADP, thrombin, collagen, and PAR4) to platelet-rich plasma (PRP) or to washed platelets (no plasma), and it was monitored by light transmission. There were no significant differences in platelet aggregation of the EDA+/+ or EDA−/− PRP compared with EDAWT/WT PRP, when triggered by different agonists at 2 different concentrations of the agonist (Supplemental Data II Section). Similarly to PRP, washed platelets from either genotype did not show any difference in platelet aggregation (data not shown). These results show that the presence of the EDA+FN isoform in plasma or in platelet α-granules does not contribute to platelet aggregation at low shear in vitro.
pFN is an important component of fibrin clot.5 However, reduction of FN (50% of WT levels) or deficiency of plasma FN does not affect clotting time and bleeding time in the mice.3,25 We evaluated whether the presence of EDA+FN in plasma may affect hemostatic parameters such as whole blood clotting time, clot rate, and bleeding time. The onset and clotting rate of whole blood was similar in the EDA+/+ mice compared with EDAWT/WT (Supplemental Data III Section). Additionally, EDA+FN neither affected tail bleeding time (Supplemental Data III Section) nor other hemostatic values such as PTT, Antithrombin III, dimer and fibrinogen concentration in plasma (not shown).
To explore the in vivo role of EDA+FN in platelet aggregation, we used a well-established pulmonary thromboembolism model.20,21 A preliminary dose-response experiment was done using EDAWT/WT mice to select the minimal dose which gave reproducible 40% to 50% mortality (not shown). We found that the mortality rate in EDA+/+ mice after collagen/epinephrine infusion was almost twice that of EDAWT/WT (66% as compared with 38% in EDAWT/WT, Fisher Test P<0.007, Figure 2). The mortality rate in EDA−/− mice was similar to EDAWT/WT (43% in EDA−/− mice) and significantly lower than in EDA+/+ mice (EDA+/+ versus EDA−/−, P<0.025). This observation indicated to us that the presence of EDA+FN in the plasma might have a prothrombotic activity.
EDA+FN Augments Thrombus Growth at Arterial Shear Rate in a Flow Chamber
Because the EDA+/+ mice were more sensitive in the pulmonary thromboembolism model, we hypothesized that EDA+FN might be prothrombotic, playing a special role in promoting thrombus growth. We performed in vitro flow chamber studies with whole blood from EDAWT/WT, EDA+/+, and EDA−/− mice at venous and arterial shear rates in a parallel plate chamber at 37°C over glass coverslips coated with fibrillar collagen type I (Figure 3). To determine the size of thrombi, the surface area covered by the mepacrine-labeled fluorescent platelets was quantified. The mean ±95% confidence intervals of 6 different optical fields in 3 separate experiments is plotted in Figure 3B. We found that EDA+/+ blood formed significantly bigger thrombi covering a larger area than EDA−/− or EDAWT/WT samples when perfused over collagen for 1 minute at shear rate of 1500 s−1 (mean±SEM, 15.5±2.4%, 30.5±5.2%, and 18.7±2.7% for EDAWT/WT, EDA+/+, and EDA−/−, respectively; ANOVA Test, P≤0.0001, Bonferroni correction EDAWT/WT versus EDA+/+, P≤0.001, EDA+/+ versus EDA−/−, P≤0.01) suggesting a role of EDA+FN in augmenting thrombus growth. However, at venous shear rate of 250 s−1 the percentage surface area covered were similar among EDAWT/WT, EDA+/+, and EDA−/− mice (mean±SEM, 7.0±0.9%, 9.2±2.6%, and 10.1±2.2% for EDAWT/WT, EDA+/+, and EDA−/−, respectively; Figure 3). These studies suggest that the EDA+FN plays a prominent role only at arterial shear rates.
Inclusion of EDA Domain in FN Accelerates Thrombus Formation
After finding that the EDA+FN promotes thrombus growth in vitro we asked whether the presence of EDA+FN in platelets and plasma would also accelerate thrombosis in vivo. We first evaluated whether EDA containing FN could enhance platelet adhesion at an early time point after ferric chloride-induced endothelial injury. The number of single platelets adhering within 2 to 3 minutes after injury was not significantly different in EDA+/+ mice when compared with EDAWT/WT or EDA−/− mice (EDAWT/WT=328±54, EDA+/+=299±33, EDA−/−=317±31) suggesting that the EDA+FN does not promote early platelet interaction with the subendothelium.
Using the ferric chloride injury model, it was previously documented that deficiency or lower levels (50%) of pFN result in serious arterial thrombosis defect.3,4 Strikingly, in spite of the decrease in pFN, using the same injury model we found that thrombi grew faster in the EDA+/+ mice (Figure 4A). The thrombi ≥30 μm in diameter were seen in the EDA+/+ arteries at 6.6±0.2 minutes compared with 8.3±0.6 minutes in the EDAWT/WT mice (P<0.05, Figure 4B). The mean vessel occlusion time was also shorter in EDA+/+ mice compared with EDAWT/WT mice (9.9±0.4 minutes and 14.6±1.7 minutes for EDA+/+ and EDAWT/WT mice, respectively; P<0.05, Figure 4C). In the EDA−/− mice the mean time to form the first thrombus and occlusion of the artery was similar to EDAWT/WT mice (Figure 4A through 4C). Mice from both sexes were used and there were no significant differences in the occlusion time between male and female mice (not shown).
These results show that EDA+FN promotes platelet aggregation/cohesion in the growing thrombus under arterial shear rate.
Thrombogenesis in Veins
Previously it was documented that venous thrombosis is not affected by the deficiency or decrease in FN levels.3 Because the EDA+/+ mice had such a striking increase in arterial thrombogenesis, we were interested to determine whether EDA+FN could also affect venous thrombosis. Similar to the artery injury model used above, endothelial damage was induced by ferric chloride. The mean time to form the first thrombus ≥30 μm in diameter (EDA+/+=9.5±0.5 minutes, EDA−/−=9.2±0.6 minutes, EDAWT/WT=9.7±0.6 minutes, Figure 5A) and the mean occlusion time (EDA+/+=17.6±1.6 minutes, EDA−/−=17.3±1.3 minutes, EDAWT/WT=18.2±1.4 minutes, Figure 5B) were similar in the EDA+/+ when compared with EDAWT/WT or EDA−/− mice. Thus it appears that the presence of EDA+FN in plasma does not stimulate venous thrombosis.
Plasma fibronectin plays an important role in thrombus growth and stability under arterial shear conditions, as was demonstrated using pFN-deficient mice and heterozygous mice having one null FN allele (FN+/−).3,4 The absence of the EDA domain was previously shown not to affect thrombus formation in the ferric chloride injury model,3 but nothing was known about the role of the EDA+FN isoform in thrombosis and hemostasis.
To study the role of EDA, we used mouse strains unable to undergo regulated splicing of the EDA exon. In this study we show physiological features of EDA+FN in thrombosis, revealing a prothrombotic role. Despite decreased plasma and platelet FN levels in EDA+/+ mice, the presence of EDA+FN in whole blood produced larger thrombi in vitro using a flow chamber, accelerated thrombosis and stabilized thrombi in injured arterioles. Additionally, the EDA+/+ mice were more sensitive to collagen-epinephrine induced pulmonary thromboembolism. The plasma contains particularly low levels of the EDA domain (cFN, 1.3 to 1.4 μg/mL12,13) but several pathological conditions are accompanied by 3- to 6-fold increase in the plasma levels of EDA+FN, including diabetes (4.3 μg/mL)13 and acute stroke (7.3 μg/mL).12 Consequently, elevated EDA+FN levels in plasma of patients should be considered a risk factor for thrombosis. EDA+/+ mice used in the present work have higher plasma EDA+FN concentration than that found in human pathological conditions (roughly 10 to 15 times more elevated). It will be interesting to determine the correlation between EDA+FN levels in plasma and arterial thrombosis in patients affected by the disease states mentioned above.
During blood coagulation, Factor XIIIa mediates the cross-linking of pFN to fibrin, enhancing the stability of the clot.26 Additionally, differential incorporation of pFN and cFN into clots has been reported.27 Previous studies done in different mouse models documented that polymerization and gelation of fibrinogen was not affected by either a partial reduction in FN levels or depletion of pFN.3,25 However, nothing is known about the role played by EDA+FN. In this study, we document that EDA+FN in plasma and platelets does not affect hemostatic parameters such as clotting rate and clot retraction. Thus, the effects of EDA are likely linked to its interaction with platelet receptors.
At arterial shear, vWF and GPIb mediate the initial interaction of platelets with the subendothelium. FN available for interaction with platelets in vivo is pooled from different sources: FN from plasma, tissue FN released at the site of injury, and FN released from α-granules of platelets. FN present in EDA+/+ mice is similar to cFN, as the EDB domain is almost absent in adult cFN.28–30 We observed no differences in platelet adhesion to the subendothelium in EDA+/+ animals when compared with EDAWT/WT or EDA−/− mice suggesting that the EDA+FN present in the subendothelium has a minor role in initial platelet adhesion. Similar results have been also obtained with FN+/− mice and pFN null mice, which have decreased pFN levels, using the same ferric chloride injury model.3,4
There may be several possible explanations for the enhanced thrombus formation observed in the presence of EDA+FN. The main integrin involved in thrombus formation is αIIbβ3, a receptor for adhesive proteins such as vWF, fibrinogen and pFN. Other integrins such as αvβ3 and α5β1 are also capable of platelet adhesion. FN can bind to αIIbβ3, αvβ3, and α5β1 through the Arg-Gly-Asp (RGD) sequence present in the FN type III-10 domain. The EDA domain is present between type III domains 11 and 12 and is a ligand for integrins α9β1 and α4β1,31 but neither of those integrins is present on platelets. Mechanical stretching of FN or inclusion of the EDA domain are known to augment FN-FN adhesion and binding to integrins, its cell binding and spreading activity.32,33 We hypothesize that the enhanced function of the EDA+FN isoform could be the consequence of a conformational change in the FN molecule caused by the inclusion of the EDA domain, leading to a more extended form of fibronectin displaying increased exposure or local unfolding of the type III-10 module, as suggested previously.34,35 Cryptic sites present in FN-type III domains 10 to 12 could be more exposed in the extended conformation when the EDA exon is included in the FN molecule, thus allowing a more efficient interaction with the platelet integrins. Another possibility could be the increased exposure of the 70 kDa N-terminal FN domain and subsequent binding to fibrin or to activated platelets facilitating the interaction with the αIIbβ3 integrin on the platelet surface as suggested previously.2,36
We showed that the presence of EDA+FN in whole blood did not increase thrombus growth at venous shear rate using an in vitro flow chamber. This result is consistent with those obtained in experimental venous thrombosis, where we showed the presence of EDA+FN in plasma and platelets influenced neither first thrombus formation nor mean occlusion time in veins, strongly suggesting that the presence of EDA+FN in plasma plays a prominent role only at arterial shear. This result is similar to that observed in the FN heterozygous (FN+/−) mice where a concentration dependent role of FN was documented only at arterial shear.3
We conclude that the presence of the EDA domain gives FN an important prothrombotic potential that is revealed at arterial shear rates. Thus the presence of the EDA+FN may aggravate peripheral arterial disease, coronary artery disease, stroke, and other conditions in which thrombi may develop at arterial shear rates.
We thank Lesley Cowan for help in preparing the manuscript.
Sources of Funding
This work was supported by National Heart, Lung, and Blood Institute of the National Institute of Health grant R37 HL41002 to D.D.W. and by Telethon Grant GGP06147 to F.E.B. L.D.M. was supported by a grant of the ASI-DCMC.
Original received June 7, 2007; final version accepted October 29, 2007.
Cho J, Mosher DF. Enhancement of thrombogenesis by plasma fibronectin cross-linked to fibrin and assembled in platelet thrombi. Blood. 2006; 107: 3555–3563.
Matuskova J, Chauhan AK, Cambien B, Astrof S, Dole VS, Piffath CL, Hynes RO, Wagner DD. Decreased plasma fibronectin leads to delayed thrombus growth in injured arterioles. Arterioscler Thromb Vasc Biol. 2006; 26: 1391–1396.
Ni H, Yuen PS, Papalia JM, Trevithick JE, Sakai T, Fassler R, Hynes RO, Wagner DD. Plasma fibronectin promotes thrombus growth and stability in injured arterioles. Proc Natl Acad Sci U S A. 2003; 100: 2415–2419.
Hynes RO. Fibronectins. New York: Springer-Verlag; 1990.
Mosher DF. Fibronectin. New York: Academic Press; 1989.
Tamkun JW, Schwarzbauer JE, Hynes RO. A single rat fibronectin gene generates three different mRNAs by alternative splicing of a complex exon. Proc Natl Acad Sci U S A. 1984; 81: 5140–5144.
Pankov R, Yamada KM. Fibronectin at a glance. J Cell Sci. 2002; 115: 3861–3863.
Paul JI, Schwarzbauer JE, Tamkun JW, Hynes RO. Cell-type-specific fibronectin subunits generated by alternative splicing. J Biol Chem. 1986; 261: 12258–12265.
Castellanos M, Leira R, Serena J, Blanco M, Pedraza S, Castillo J, Davalos A. Plasma cellular-fibronectin concentration predicts hemorrhagic transformation after thrombolytic therapy in acute ischemic stroke. Stroke. 2004; 35: 1671–1676.
Kanters SD, Banga JD, Algra A, Frijns RC, Beutler JJ, Fijnheer R. Plasma levels of cellular fibronectin in diabetes. Diabetes Care. 2001; 24: 323–327.
Tan MH, Sun Z, Opitz SL, Schmidt TE, Peters JH, George EL. Deletion of the alternatively spliced fibronectin EIIIA domain in mice reduces atherosclerosis. Blood. 2004; 104: 11–18.
van der Plas RM, Schiphorst ME, Huizinga EG, Hene RJ, Verdonck LF, Sixma JJ, Fijnheer R. von Willebrand factor proteolysis is deficient in classic, but not in bone marrow transplantation-associated, thrombotic thrombocytopenic purpura. Blood. 1999; 93: 3798–3802.
Muro AF, Chauhan AK, Gajovic S, Iaconcig A, Porro F, Stanta G, Baralle FE. Regulated splicing of the fibronectin EDA exon is essential for proper skin wound healing and normal lifespan. J Cell Biol. 2003; 162: 149–160.
Chauhan AK, Motto DG, Lamb CB, Bergmeier W, Dockal M, Plaimauer B, Scheiflinger F, Ginsburg D, Wagner DD. Systemic antithrombotic effects of ADAMTS13. J Exp Med. 2006; 203: 767–776.
Yang J, Wu J, Kowalska MA, Dalvi A, Prevost N, O’Brien PJ, Manning D, Poncz M, Lucki I, Blendy JA, Brass LF. Loss of signaling through the G protein, Gz, results in abnormal platelet activation and altered responses to psychoactive drugs. Proc Natl Acad Sci U S A. 2000; 97: 9984–9989.
Odawara A, Kikkawa K, Katoh M, Toryu H, Shimazaki T, Sasaki Y. Inhibitory effects of TA-993, a new 1,5-benzothiazepine derivative, on platelet aggregation. Circ Res. 1996; 78: 643–649.
Mazzucato M, Spessotto P, Masotti A, De Appollonia L, Cozzi MR, Yoshioka A, Perris R, Colombatti A, De Marco L. Identification of domains responsible for von Willebrand factor type VI collagen interaction mediating platelet adhesion under high flow. J Biol Chem. 1999; 274: 3033–3041.
Chauhan AK, Kisucka J, Lamb CB, Bergmeier W, Wagner DD. von Willebrand factor and factor VIII are independently required to form stable occlusive thrombi in injured veins. Blood. 2007; 109: 2424–2429.
Moretti FA, Chauhan AK, Iaconcig A, Porro F, Baralle FE, Muro AF. A major fraction of fibronectin present in the extracellular matrix of tissues is plasma-derived. J Biol Chem. 2007; 282: 28057–28062.
Sakai T, Johnson KJ, Murozono M, Sakai K, Magnuson MA, Wieloch T, Cronberg T, Isshiki A, Erickson HP, Fassler R. Plasma fibronectin supports neuronal survival and reduces brain injury following transient focal cerebral ischemia but is not essential for skin-wound healing and hemostasis. Nat Med. 2001; 7: 324–330.
Mosher DF. Cross-linking of cold-insoluble globulin by fibrin-stabilizing factor. J Biol Chem. 1975; 250: 6614–6621.
Wilson CL, Schwarzbauer JE. The alternatively spliced V region contributes to the differential incorporation of plasma and cellular fibronectins into fibrin clots. J Cell Biol. 1992; 119: 923–933.
Kornblihtt AR, Pesce CG, Alonso CR, Cramer P, Srebrow A, Werbajh S, Muro AF. The fibronectin gene as a model for splicing and transcription studies. Faseb J. 1996; 10: 248–257.
Manabe R, Ohe N, Maeda T, Fukuda T, Sekiguchi K. Modulation of cell-adhesive activity of fibronectin by the alternatively spliced EDA segment. J Cell Biol. 1997; 139: 295–307.
Zhong C, Chrzanowska-Wodnicka M, Brown J, Shaub A, Belkin AM, Burridge K. Rho-mediated contractility exposes a cryptic site in fibronectin and induces fibronectin matrix assembly. J Cell Biol. 1998; 141: 539–551.
Johnson KJ, Sage H, Briscoe G, Erickson HP. The compact conformation of fibronectin is determined by intramolecular ionic interactions. J Biol Chem. 1999; 274: 15473–15479.