Jun 27, 1985 - We thank Mr. Greg Christiansen, Mr. Dave Evans, and Ms. J. Kramer for technical assistance; Ms. Betty Perrin and Ms. Jan Liebsch for typing, ...
RAPID COMMUNICATION
Fragments of Fibronectin Are Potent Inhibitors of Endothelial Cell Growth Heparin-Binding
GENE A. HOMANDBERG, PhD, JAMES E. WILLIAMS, MD, DEBRA GRANT, MS, BARBARA SCHUMACHER, BS, and REUBEN EISENSTEIN, MD
From the Departments of Medicine and Pathology and Laboratory Medicine, Mount Sinai Medical Center, University of Wisconsin, Milwaukee Clinical Campus, Milwaukee, Wisconsin
Two heparin-binding proteolytic fragments of fibronectin -an amino-terminal 29-kd segment and a carboxylterminal 40-kd segment- are apparently specific, potent inhibitors of the growth of cultured bovine aortic endothelial cells and inhibit growth in a reversible, dosedependent manner. In contrast, native fibronectin at
higher dosages neither inhibits nor interferes with the effects of the 29-kd fragment. The data, therefore, suggest that fibronectin fragments may participate in the regulation of vascular growth. (Am J Pathol 1985, 120:327-332)
THE DEVELOPMENT of new blood vessels is an important event in embryology, reparative processes, and a number of pathologic conditions. This complex process involves several activities of endothelial cells (ECs), including budding and, subsequently, migration, proliferation, and invasion.' As might be expected from the ubiquity of this phenomenon, a growing list of naturally occurring agents have been identified which either stimulate EC growth in culture or induce capillary proliferation in experimental models. Of these, relatively few, including fibroblast growth factor2 and molecules from a murine chondrosarcoma3 and bovine brain,4 have been purified as yet. There also have been efforts to find natural inhibitors of capillary or EC proliferation. Factors with this property have been identified in several normally avas-
EC migration," and supports human EC growth.'3 Paradoxically, however, heparin inhibits bovine aortic EC growth in culture, although large doses are required.6 On the basis of these observations, Taylor and Folkman'4 used protamine, a fish sperm nucleoprotein, which strongly binds heparin, to retard tumor-induced angiogenesis in mice. Administration of at least one commercial heparin, together with cortisone, has also been reported to result in dramatic regression of transplantable murine tumors.'5 These somewhat confusing data led us to test the effects of other heparin-binding molecules on EC activity. The heparin-binding glycoprotein fibronectin can readily be purified in large amounts from both blood plasma and tissues. This molecule is fairly well characterized and, upon fragmentation, yields several peptides with different properties.'6-20 The amino-teminal 29-kd fragment and less well characterized carboxyl-terminal fragments, have strong avidity for heparin. Other fragments have biologic activities, and the intact molecule is chemotactic for EC21 (a current conceptual schema of fibronectin domains is presented in Figure 1).
cular tissues, including cartilage,' aorta,6 optic lens,7 and vitreous humor.8 9 Adrenal corticosteroids retard neovascularization; and medroxyprogesterone, a synthetic steroid which inhibits the expression of collagenase, can retard tumor-induced neovascularization, perhaps by suppressing the collagenolytic component of invasion.10 Several lines of evidence suggest that heparin may be involved in regulating capillary growth. "I Mast cells, rich in this glycosaminoglycan, are found in large numbers near proliferating capillaries." Heparin enhances the expression of collagenase,'2 an enzyme thought to be involved in tumor cell invasion, stimulates capillary
Supported by NIH Grants HL27330, EY04154, and HL28444. Accepted for publication June 27, 1985. Address reprint requests to Gene A. Homandberg, PhD, Department of Medicine, Mount Sinai Medical Center, University of Wisconsin, Milwaukee Clinical Campus, Milwaukee, WI 53201.
327
328
AJP * September 1985
HOMANDBERG ET AL Heparin Ill DNA
Collagen Heparin I Polyamine Fibrin I Factor XIlIla Actin Bacteria
H2N-j-
50 kd
-[
Cellbinding
140 -160 kd
Heparin 11 Fibrin 11
II-COOH ss
S -C
H2N-
OOIHCO
Figure 1-Scheme of fibronectin domains. Binding domain assignments and size in kilodaltons are based on information cited in reviews! '20 The 29-kd
fragments used in these studies are isolated from thrombin digests of plasma fibronectin.25.26 The 72-kd fragment (containing the 29-kd and 50-kd domains) and a 40-kd fragment (containing Heparin 11 domain) were formed by cathepsin D digestion of plasma fibronectin.2' Digestion of the 72-kd fragment with thrombin25 produced the 50-kd fragment and the 29-kd fragment.
Materials and Methods General procedures for fibronectin fragment isolation have been described by many workers. 16-20 Fibronectin was isolated from human plasma by adsorption to gelatin-Sepharose (prepared by CNBr activation of Sepharose 4B and coupling to gelatin as described23) and desorption22 with 3 M urea-phosphate (0.1 M)buffered saline (0.15 M NaCl) (PBS). Partially reduced fibronectin was prepared as described23 by a method that selectively cleaves the interchain disulfides connecting the subunits. Fibronectin fragments were isolated from cathepsin D (Sigma Chemical Co.) and human a-thrombin (provided by Dr. J. Fenton of the New York State Department of Health) digests of fibronectin. Aminoterminal 72-kd and central 150-kd fragments and carboxyl-terminal fragments of 35-40-kd were generated by exposure of 0.5-1 mg/ml of fibronectin in 0.1 M formate buffer, pH 3.7, to 0.8 l.g/ml of cathepsin D.24 After 3-4 hours at 300 the reaction was quenched by addition of pepstatin (0.2 pg/ml final) and adjustment of pH to 7.5. The mixture was then diluted with one volume of water and applied to gelatin-Sepharose. The 72-kd fragment bound and was eluted with 3 M urea-PBS. The nonadherent material was concentrated by adjustment to 40% (wt/wt) ammonium sulfate and redissolved. The solution was then applied to a Sephacryl S-300 column, equilibrated in PBS, which resolved a 150-kd fragment from smaller 20-40-kd fragments. The smaller fragments were then applied to heparin-Sepharose 4B (prepared as described23) in 0.05 M NaCl, 20 mM Tris, pH 7.4. A 40-kd fragment bound and was eluted with 0.3 M NaCl, 20 mM Tris buffer, pH 7.4. The 72-kd fragment was digested further with human a-thrombin as described25 for generation of the amino-terminal 29-kd fragment and a 50-kd gelatinbinding fragment. The 72-kd fragment, in 50 mM NaCl
20 mM Tris buffer, pH 7.4, was adjusted to 1 U/ml human a-thrombin. The proteolysis was complete after 5 days at 22C as monitored on 15% sodium dodecyl sulfate (SDS) polyacrylamide gels. The thrombin digest was then applied to gelatin-Sepharose in order to adsorb the 50-kd fragment. The nonadherent material, the 29-kd fragment, was concentrated by adjustment to 60% (wt/wt) ammonium sulfate. The bound 50-kd fragment was eluted with 4 M urea-PBS. Some of the 29-kd fragment material studied here was prepared by digestion of fibronectin with human a-thrombin26 by
the following method. Fibronectin (0.5-1 mg/ml) in 10 mM phosphate, pH 7.0, was adjusted to 1 U/ml human a-thrombin. Digestion was monitored by SDS polyacrylamide gels.27 The digestion was complete after 2-4 days, and the digest was then applied to DEAE-cellulose in 50 mM NaCl, 20 mM Tris buffer, pH 8.5. The 29-kd amino-terminal fragment was recovered in the wash and a 190-kd fragment was eluted with 300 mM NaCl, 20 mM Tris buffer, pH 8.5. Dialyzed and lyophilized fragments were tested on second to sixth passage ECs, bovine aortic smooth muscle cells, 3T3 fibroblasts, mouse spleen lymphocytes, and B16 murine melanoma cells cultured as previously described.6 To test the effects on growth, we seeded 0.25 ml of cell suspension into Corning 24-well trays (5000 cells/well). After 24 hours, cultures were refed media containing test material. Cells were harvested at various time intervals and counted in a Model ZF Coulter Counter. In some experiments, 3H-thymidine (1 iC/ml) was added to the medium 24 hours before harvest, and cell incorporation was measured. The results paralleled those obtained by cell counts. In some experiments, the bovine calf serum used in the culture medium was passed through heparin-Sepharose columns. To test for reversibility of growth inhibition, we exposed ECs to the 29-kd fragment for 4 or 24 hours. Then they were refed with control medium. Cell counts were done 24,
FIBRONECTIN FRAGMENTS INHIBIT EC GROWTH
Vol. 120 * No. 3
Table 1-Effect of 29-kd Fragment on Growing Endothelial Cells % Inhibition 29-kd (nM) (mean ± SEM) 34.5 42.5 51.1 68.7 87.4 95.5
8.6 17.0 34.0 68.0 136.0 272.0
329
t ;
;
'';2
± 2.57 ± 1.88
± 1.29 ± 7.55 ± 8.03 ± 13.1
48, and 72 hours later and compared with cultures either continuously or never exposed to the fragment. Polyacrylamide gel electrophoresis in the presence of SDS was performed by Laemmli's method."
Results The 29-kd and 40-kd fragments with heparin-binding capacities inhibited EC growth significantly at very low dose. In some experiments (Table 1), a concentration lower than 34 nM resulted in 50% inhibition. The 50kd gelatin-binding domain, carboxyl to the 29-kd segment, also inhibited EC growth in some experiments but was much less effective. Our studies focused on the 29-kd fragment, because it is the most completely characterized fragment of those we studied, but other larger fragments with heparin-binding properties were also inhibitory (Figure 2). Intact dimeric fibronectin was not inhibitory at concentrations as high as 700 nM, the level in normal human plasma. Partially reduced fibronectin inhibited approximately half as much as either of the two heparin-binding fragments. Figure 3 is a 5-20% linear polyacrylamide gradient slab gel of the 72-, 50-, 40-, and 29-kd fragments. 10090z 80706050z 403020-
T
10-
FRAGMENT Figure 2-Inhibition of EC growth by fibronectin fragments. ECs were seeded at 20,000 cells/ml; and after 24 hours, the medium was removed and replaced with medium containing 70 nM of fragments indicated. The cells were harvested 72 hours later and counted in a Coulter Counter. PR-PFn is partially reduced plasma fibronectin (PFn) prepared by subjecting PFn to 17 mM dithiothreitol. Functionally active monomeric fibronectin is the product.23
:-
..
..
1,r
Figure 3-SDS polyacrylamide gel (5-20% linear gradient), of 72-kd (1), 50-kd (2), 40-kd (3), and 29-kd (4) fragments. Each band represents about 60 Ag of protein.
When an excess of intact fibronectin (125-700 nM) was added to cultures in addition to the 29-kd fragment (8-35 nM), there was no change in the effect of the fragment on EC growth. When an excess (15 of hepa-
j*M)
rin (Sigma: Grade I, porcine intestinal mucosa, 158 U/mg), itself a mild inhibitor of bovine aortic EC growth,6 was added to cultures together with the 29-kd fragment, the effect on EC growth was additive. If the serum used in the culture medium was made deficient of heparin-binding material by passage through a heparin-Sepharose column, EC growth inhibition by the 29-kd fragment was also not diminished. At higher doses of 29-kd fragment, many floating cells were noted, an observation generally considered to indicate cytotoxicity. Alternatively, the fragment might interfere with cell attachment rather than growth. This was evaluated by studies of plating efficiency. When plating efficiency was compared in 35-mm plastic dishes coated with water or water containing 29-kd fragment (200 nM) and subsequently dried, there was no significant difference in plating efficiency except after 24 hours, when the growth inhibitory effect became evident (Figure 4). Similarly, addition of the fragment to cell suspensions prior to plating did not affect plating efficiency. To test whether these effects of the 29-kd fragment were primarily related to cell detachment, we cultured
330
AJP * September 1985
HOMANDBERG ET AL
44
.1 (5
HOUM Figure 4-Plating efficiency in the presence of the 29-kd fragment. ECs were added to 35-mm plastic dishes either precoated with 1 ml of water or 1 ml of the 29-kd fragment (200 nM) and then dried. Cells that attached to the dish were determined at various times for control (open bars) or for the 29-kd fragment (shaded bars).
a dose at which
growing ECs in 500 nM 29-kd fragment, many cells became detached from the dish. Fifteen and 40 hours later, the viability of the floating cells was assessed by trypan blue staining. At both times, 28% of these cells appeared viable, while 9007e of attached cells excluded the stain. When aliquots of the floating cells were replated, plating efficiency assessed 24 hours later was only 3.5%7o. Also, the 29-kd fragment neither affected incorporation of tritiated thymidine nor total cell count in contact-inhibited confluent monolayers of ECs. The distinction between cytotoxicity and growth inhibition is difficult. This was approached by experiments in which ECs were either continuously exposed to the 29-kd fragment (approximately 100 nM) for 72 hours or exposed 4 or 24 hours, at which time the fragment was removed and the cells were refed with fragment5040A
30* .1
V 0
"
ol
.10 --e
.e-ll
20
x
..,
-01
(a)
Ji LU
10'
I
,
I ,
I
4r
I
-I
I
oll,.1
11
-I
.01
-I
e 1.
I
24
48
HOURS Figure 5-Reversibility of the 29-kd fragment affect (100 nM). Inhibition of no exposure to the EC growth after exposure for 4 or 24 hours. * --, exposure to the 29-kd fragment for 4 hours fol29-kd fragment; lowed by refeeding with control medium; _- ---_, exposure to the 29kd fragment for 24 hours followed by refeeding; _--_, continuous exposure to the 29-kd fragment. -
-
free medium. The cells recovered their rate of growth even after a 24-hour exposure to the fragment, suggesting reversibility and thus cell growth inhibition rather than cytotoxicity (Figure 5). Additional support for this thesis comes from the experiments with detached cells described above. Tests of the 29-kd fragment on growth of 3T3 fibroblasts, primary bovine aorta smooth muscle cells, B16-F10 melanoma cells, and mouse spleen lymphocytes, showed no significant effect. Sprouting is a phenotypic expression of ECs30 when they are maintained past confluence. Addition of the 29-kd fragment to such cultures was followed by the disappearance of the sprouts, while the confluent monolayers remained unaffected (Figure 6).
Discussion Thus, the two heparin-binding regions of fibronectin, but not the parent molecule or some other fragments, appear to be relatively cell-specific, potent inhibitors of bovine aortic EC growth, and this inhibition is reversible and dose-dependent. Several points in relation to these observations merit discussion. Intact fibronectin in doses several times that of the 29- or 40kd fragment did not inhibit EC growth. This suggests either that there is a lack of accessibility of certain domains in the intact molecule or that the liberated fragments have properties altered from those expressed when these segments are in the native molecule. There is precedent for masked activities in fibronectin. A 30kd gelatin-binding fragment has tumor-enhancing activity not expressed by native fibronectin,29 and a 29kd fragment has affinity for glycosaminoglycans also not expressed in the native molecule.30 These fragments may occur in vivo. Circulating proteases such as thrombin and plasmin, as well as others expected in inflammatory reactions, such as cathepsin G and leukocyte elastase, can generate amino-terminal fibronectin fragments.16-24 Although neither the 40-kd segments nor the 29-kd segments have yet been demonstrated in tissue or blood, large 170-205-kd fragments have been found in plasma.31'32 These experiments do not settle the question of how the heparin-binding capacity of these fibronectin segments is related to their effects on ECs. The failure of heparin to counteract growth inhibition suggests that the heparin binding is not completely adequate to explain growth inhibition. Additional support for the thesis that heparin-binding may not be the sole mechanism of EC growth inhibition comes from the observation that antithrombin III, a strong heparin binder, did not significantly affect EC growth. Recent structural homology studies33 revealed a striking similarity in the primary structures of the 29-kd frag-
331
FIBRONECTIN FRAGMENTS INHIBIT EC GROWTH
Vol. 120 * No. 3
FIgure 6-Effect of the 29-kd fragment on sprouting of contact-inhibited confluent monolayer cells.~ When sprouts appeared in monolayer (A), the 29-kd fragment was added (700
t
._
nM final concentration). Af-
ter 72-96 hours, the sprouts
disappeared completely
(B).6
ment and platelet factor 4, which also binds heparin with high affinity. Addition of low concentrations (4-8 Mg/ml) of platelet factor 4 to cultures of growing ECs results in marked growth inhibition (data not shown). Nonetheless, a likely candidate molecule with which these fibronectin fragments may interact is heparan sulfate, a molecule related to heparin which has been suggested to play a role in cell growth, attachment, and shape.34 The 29-kd fragment binds to a glomerular heparan sulfate proteoglycan.35 ECs, as well as other cells, synthesize heparan sulfate and display it at their surface.36 Molecules which are mitogens for ECs have recently been purified from brain with the use of heparin affinity columns.37 Klagsbrun and Shing37 have suggested that cell-surface heparan sulfate serves the function of binding such molecules to the cell. This hypothesis would seem even more attractive now that heparin-binding molecules have been shown to inhibit EC growth. Our experiments used bovine aortic endothelium. Neither studies using cells from other species or the microcirculation nor in vivo studies have yet been done. These experiments appear obligatory, because several less highly purified inhibitors of bovine aortic EC growth have been shown to inhibit some type of ex-
perimental neovascularization.5-9 Note Added in Proof A recent report (Hayman EG, Pierschbacher MD, Ruoslahti E. Detachment of cells from culture substrate by soluble fibronectin peptides. J Cell Biol 1975, 100: 1948-1954) shows that synthetic cell-attachment-promoting peptides from fibronectin detach cultured cells,
including ECs, from a fibronectin-coated substratum and that this effect is best demonstrated in serum-free medium. We therefore tested the 29-kd fragment (800 nM) against ECs according to their experimental protocol. No detectable detachment was demonstrated after 4 hours of exposure whether cells were maintained on plastic, fibronectin-coated plastic, or plastic coated with extracellular matrix synthesized by ECs.
References 1. Eisenstein R, Sorgente N, Soble LW, Miller A, Kuettner KE: The resistance of certain tissues to invasion. Am J Pathol 1973, 81:765-774 2. Gospodorowicz D, Moran J, Braun D, Birdwell C: Clonal growth of vascular endothelial cells: Fibroblast growth factor as a survival agent. Proc Natl Acad Sci USA 1976, 73:4120-4124 3. Klagsbrun M, Smith S: Purification of a cartilage-derived growth factor. J Biol Chem 1980, 255:108959-108966 4. Maciag T, Mehlman T, Friesel R, Schreiber AB: Heparin binds endothelial cell growth factor, the principle endothelial cell mitogen in bovine brain. Science 1984, 225:932-935 5. Eisenstein R, Kuettner KE, Neopolitan C, Soble LW, Sorgente N: The resistance of certain tissues to invasion: III. Cartilage extracts inhibit the growth of fibroblasts and endothelial cells in culture. Am J Pathol 1975, 81:337-348 6. Eisenstein R, Harper E, Kuettner KE, Schumacher B, Macijevitch B: Growth regulators in connective tissues: II. Evidence for the presence of several growth inhibitors in aortic extracts. Arerial Wall 1979, 5:163-170 7. Williams GA, Eisenstein R, Schumacher B, Hsiao K-C, Grant D: Inhibitors of vascular endothelial cell growth in the lens. Am J Ophthalmol 1984, 97:366-371 8. Jacobson B, Sullivan D, Raymond L, Basu PK, Hasony SM: Further studies on a vitreous inhibitor of endothelial cell proliferation. Exp Eye Res 1983, 36:447-450 9. Lutty GA, Thompson DC, Gallup JY, Mello RJ, Fenselau A: Vitreous: An inhibitor of retinal extract-induced
332
AJP * September 1985
HOMANDBERG ET AL
neovascularization. Inv Ophthalmol Vis Sci 1983, 24: 52-56 10. Gross G, Azizkahn RG, Besivas C, Bruns RR, Folkman J: Inhibition of tumor growth vascularization and collagenolysis in the rabbit cornea by medroxyprogesterone. Proc Natl Acad Sci USA 1981, 78:1176-1180 11. Folkman J, Taylor S, Spillberg C: In: Development of the Vascular System. Edited by J Nugen, M O'Connor. London, Pitman Press, 1983, pp 132-142 12. Sakamoto S, Sakamoto M, Goldhaber P, Glimcher MJ: Studies on the interaction between heparin and mouse bone collagenase. Biochem Biophys Acta 1975, 385:41-50 13. Thornton SC, Mueller SN, Levine EM: Human endothelial cells: Use of heparin in cloning and long-term serial cultivation. Science 1983, 222:623-625 14. Taylor S, Folkman J: Protamine is an inhibitor of angiogenesis. Nature 1982, 297:307-312 15. Folkman J, Langer R, Linhardt RJ, Haudenschild C, Taylor S: Angiogenesis inhibition and tumor regression caused by heparin or a heparin fragment in the presence of cortisone. Science 1983, 221:719-725 16. Mosesson MW, Amrani DL: The structure and biologic activities of plasma fibronectin. Blood 1980, 56:145-158 17. Yamada KM: Cell surface interactions with extracellular material. Ann Rev Biochem 1983, 52:761-799 18. Hynes RO, Yamada KM: Fibronectins: Multifunctional modular glycoproteins. J Cell Biol 1982, 95:369-377 19. Ruoslahti E, Engvall E, Hayman EC: Fibronectin: Current concepts of its structure and functions. Collagen Res 1981, 1:95-128 20. Mosher DF: Fibronectin. Prog Hemost Thromb 1980, 5:111-151 21. Bowersox JC, Sorgente N: Chemotaxis of aortic endothelial cells in response to fibronectin. Cancer Res 1982, 42:2547-2551 22. Engvall E, Ruoslahti E: Binding of soluble form of fibroblast surface protein, fibronectin, to collagen. Int J Cancer 1977, 20:1-5 23. Homandberg GA, Amrani DL, Evans DB, Kane CM, Ankel E, Mosesson MW: Preparation of functionally intact monomers by limited disulfide reduction of human plasma fibronectin dimers. Arch Biochem Biophys 1985, 238:652-663 24. Balian G, Click EM, Crouch E, Davidson JM, Bornstein P: Isolation of a collagen-binding fragment from fibronectin and cold insoluble globulin. J Biol Chem 1979, 254:1429-1432 25. Furie MB, Rifkin DB: Proteolytically derived fragments of human plasma fibronectin and their localization within the intact molecule. J Biol Chem 1980, 255:3134-3140
26. Furie MB, Frey AB, Rifkin DB: Location of a gelatinbinding region of human plasma fibronectin. J Biol Chem 1980, 255:4391-4394 27. Laemmli UK: Cleavage of structural proteins during the assembly of the beard of bacteriophage T4. Nature 1970, 227:680-683 28. Schwartz SM: Selection and characterization of bovine aortic endothelial cells. In Vitro 1978, 14:966-980 29. DePetro G, Barlati S, Vartio T, Vaheri A: Transformationenhancing activity of proteolytic fragments in fibronectin. Proc Natl Acad Sci USA 1981, 78:4965-4969 30. Sekiguchi K, Hakomori S-I, Funahashi M, Matsumoto I, Seno N: Binding of fibronectin and its proteolytic fragments to glycosaminoglycans. Exposure to cryptic glycosaminoglycan-binding domains upon limited proteolysis. J Biol Chem 1983, 258:14359-14365 31. Ruoslahti E, Hayman E, Engvall E, Cothran WC, Butler WT: Alignment of biologically active domains in the fibronectin molecule. J Biol Chem 1981, 256:7277-7281 32. Amrani DL, Homandberg GA, Tooney NM, WolfensteinTodel C, Mosesson MW: Separation and analysis of the major forms of plasma fibronectin. Biochim Biophys Acta 1983, 748:308-320 33. Erickson J: Manuscript in preparation 34. Hook M, Robinson J, Kjellin L, Johansson S: Heparan sulfate: On the structure and function of the cell associated proteoglycans, Extracellular Matrix. New York, Academic Press, 1982, pp 15-23 35. Atherton M, Kanwar Y: Heparan sulfate proteoglycan binds specifically to heparin binding sites in fibronectin. (Abstr) J Cell Biol 1984, 99 (Part 2): 77a 36. Oohira AT, Wight N, Boorstein P: Sulfated proteoglycans synthesized by vascular endothelial cells in culture. J Biol Chem 1983, 258:2014-2021 37. Klagsbrun M, Shing Y: Heparin affinity of anionic and cationic endothelial cell growth factors: Analysis of hypothalamus derived growth factors and fibroblast growth factor. Proc Natl Acad Sci USA 1985, 82:805-809
Acknowledgments We thank Mr. Greg Christiansen, Mr. Dave Evans, and Ms. J. Kramer for technical assistance; Ms. Betty Perrin and Ms. Jan Liebsch for typing, and Drs. J. Erickson, R. Marlar, R. Montgomery, and M. W. Mosesson for critical comments. We also thank Dr. Marlar for generously providing AT-3.
