Publications

Thrombospondin is an adhesive glycoprotein that supports cell attachment, spreading, and migration. The murine thrombospondin gene is approximately 18 kb in length and includes 22 exons. Interspecific backcross analysis using progeny derived from matings of (C57BL/6J x Mus spretus) F1 x C57BL/6J mice indicates that the thrombospondin gene is tightly linked to the Fshb, Actcl, Ltk, and B2M loci on murine chromosome 2. The sequence of the murine gene is very similar to that of the human gene in (1) regions of the promoter, (2) the coding region, and (3) the 3'-untranslated region. The predicted amino acid sequence of the mature murine thrombospondin subunit is 95.1% identical to that of the human. The sequences of these two species are most similar at the regions containing the type 1, 2, and 3 repeats as well as the COOH-terminal globular domain. The thrombospondin promoter is similar to the 5' flanking region of some housekeeping and growth control genes in that it contains multiple GC-rich regions and lacks a CAAT box. The presence of various consensus sequences suggests that thrombospondin gene expression is regulated by cAMP, cytokines, and steroid hormones.

Thrombospondin is a multifunction adhesive protein with the ability to bind proteoglycans, cell surface receptors, other proteins, and calcium ions. Several sequence motifs for some of these interactions have been identified in human thrombospondin. To evaluate the potential functional significance of these sequences and to begin a study of the evolution of thrombospondin, we have isolated and sequenced thrombospondin cDNA clones from a chicken embryo library. Comparison of the chicken and human sequences reveals that the NH2-terminal heparin-binding domains are only 34% identical. By contrast, the type 3 repeats and the COOH-terminal domains are 80 and 82% identical, respectively, when comparing human and chicken sequences. Potential cell recognition sequences of RGD and VTCG are conserved, with the chicken sequence containing an additional copy of the VTCG sequence. Whereas substitutions occur in the two potential heparin-binding motifs that have human counterparts, the chicken sequence contains a third potential heparin-binding motif. The results indicate that the evolutionary constraints on the various types of cell-binding motifs may be quite different.

The complete amino acid sequence for human erythrocyte band 4.2 has been derived from the nucleotide sequence of a full-length 2.35-kilobase (kb) cDNA. The 2.35-kb cDNA was isolated from a human reticulocyte cDNA library made in the expression vector lambda gt11. Of the 2348 base pairs (bp), 2073 bp encode 691 amino acids representing 76.9 kDa (the SDS/PAGE molecular mass is 72 kDa). RNA blot analysis of human reticulocyte total RNA gives a message size for band 4.2 of 2.4 kb. The amino acid sequence of band 4.2 has homology with two closely related Ca2(+)-dependent cross-linking proteins, guinea pig liver transglutaminase (protein-glutamine gamma-glutamyltransferase; protein-glutamine: amine gamma-glutamyltransferase, EC 2.3.2.13) (32% identity in a 446-amino acid overlap) and the a subunit of human coagulation factor XIII (27% identity in a 639-amino acid overlap), a transglutaminase that forms intermolecular gamma-glutamyl-epsilon-lysine bonds between fibrin molecules. The region of greatest identity includes a 49-amino acid stretch of band 4.2, which is 69% and 51% identical with guinea pig liver transglutaminase and the a subunit of factor XIII, respectively, within the regions that contain the active sites of these enzymes. Significantly, within the five contiguous consensus residues of the transglutaminase active site, Gly-Gln-Cys-Trp-Val, band 4.2 has an alanine substituted for cysteine (which is apparently essential for activity). Consistent with this active site substitution, erythrocyte membranes or inside-out vesicles, which contain band 4.2, show no evidence of transglutaminase activity by two types of in vitro assay.

In addition to the increased synthesis of the classical heat-shock proteins (28,000, 71,000, 73,000, 90,000 and 100,000 Mr polypeptides) there is also an increase of thrombospondin in the growth medium of endothelial cells exposed to hyperthermia. The effect of a monoclonal antibody to thrombospondin on the recovery of endothelial cells from hyperthermia as it relates to cytoskeletal organization and cell spreading was assessed. The antibody interacts with the heparin-binding domain of thrombospondin in the extracellular matrix of cells. We report that during recovery from thermal insult at 37 degrees C, intermediate filaments, stress fibres and microtubules show distinct time-recovery characteristics in bovine aortic endothelial cells; that in the presence of this antibody the cytoskeleton is notably altered; that this antibody causes retraction of endothelial cell processes; and that the recovery of the cytoskeleton in endothelial cells exposed to hyperthermia is prevented by the thrombospondin antibody in the time frame examined. Our data suggest that the recovery of cells from heat shock requires the integrity of thrombospondin and its interactions.

We investigated Plasmodium falciparum parasitized erythrocyte binding to proteolytic fragments of thrombospondin and the effects of anti-thrombospondin monoclonal antibodies on this binding. Purified human platelet thrombospondin was cleaved by trypsin, chymotrypsin or thrombin. Fragments were separated by heparin-agarose affinity chromatography, removing the amino-terminal heparin-binding region. Trypsin at 5.0 micrograms ml-1 of thrombospondin cleaved thrombospondin to reduced 140 and 120 kDa fragments plus a reduced 25-kDa heparin-binding fragment. Infected erythrocytes bound to intact thrombospondin (3420 +/- 460 infected erythrocytes mm-2) and the carboxy-terminal fragment, yielding 120-140-kDa fragments on sulfhydryl reduction, but not to the 25-kDa fragment (144 +/- 104 infected erythrocytes mm-2 (mean +/- s.d., N = 4). Similar results were obtained with chymotrypsin and thrombin cleavage. When the anti-thrombospondin monoclonal antibody MA-I was added to immobilized thrombospondin prior to infected erythrocytes, adherence was inhibited by 99%. At the same concentration, MA-I inhibited adherence to C32 melanoma cells by only 35%. MA-I binds to a calcium-dependent structure at the C-terminal globular region of thrombospondin. Monoclonal antibody MA-II inhibited adherence to thrombospondin by 46%, while MA-III had no effect. These antibodies bind to the N-terminal globular region which includes the heparin-binding site and the segment connecting the two globular regions, respectively. The site(s) for infected erythrocyte binding on thrombospondin reside in the large, 140- or 120-kDa, proteolytic cleavage fragments, and not in the N-terminal heparin-binding region.

The cDNA for human erythrocyte ankyrin has been isolated from a series of overlapping clones obtained from a reticulocyte cDNA library. The composite cDNA sequence has a large open reading frame of 5636 base pairs (bp) with the complete coding sequence for a polypeptide of 1879 amino acids with a predicted molecular mass of 206 kDa. The derived amino acid sequence contained 194 residues that were identical to those obtained by direct amino acid sequencing of 11 ankyrin proteolytic peptides. The primary sequence contained 23 highly homologous repeat units of 33 amino acids within the 90-kDa band 3 binding domain. Two cDNA clones showed evidence of apparent mRNA processing, resulting in the deletions of 486 bp and 135 bp, respectively. The 486-bp deletion resulted in the removal of a 16-kDa highly acidic peptide, and the smaller deletion had the effect of altering the COOH terminus of the molecule. Radiolabeled ankyrin cDNAs recognized two erythroid message sizes by RNA blot analysis, one of which was predominantly associated with early erythroid cell types. An ankyrin message was also observed in RNA from the human cerebellum by the same method. The ankyrin gene is assigned to chromosome 8 using genomic DNA from a panel of sorted human chromosomes.

Defects involving alpha spectrin (Sp) are found in patients with hereditary elliptocytosis and a related disorder, hereditary pyropoikilocytosis (HPP). We have previously found that the severity of hemolysis was related to the total spectrin content of the cells and the percentage of unassembled dimeric Sp (SpD) in the membranes, which, in turn, reflected the amount of mutant Sp in the cell. However, no data are available comparing differences in the function of various alpha Sp mutations to clinical severity. We now report studies of nine homozygotes or double heterozygotes for four alpha Sp mutations: alpha 1/74, alpha 1/46, alpha 1/65, and alpha 1/61, whose red blood cells (RBCs) contained only the mutant Sp and no normal Sp. Sp alpha 1/74, Sp alpha 1/46, and alpha 1/65 homozygotes differed strikingly in the severity of hemolysis that correlated with the severity of mutant Sp dysfunction, as reflected by the fraction of unassembled SpD in the membranes and the self-association of mutant Sp on inside-out vesicles. Homozygotes for Sp alpha 1/74 had a very severe hemolytic anemia and their SpD were virtually incapable of self-association, whereas SpD alpha 1/46 were not as severely affected. The Sp alpha 1/65 homozygotes had a relatively mild hemolytic anemia and their SpD showed the least impairment of function. Ultrastructural examination of membrane skeletons from subjects whose SpD self-association was severely impaired showed gross skeletal disruption and loss of hexagonal structure. In striking contrast, the homozygote for the mildly dysfunctional Sp alpha 1/65 had only a moderate disruption of the skeleton. Some of the homozygous or doubly heterozygous subjects also exhibited a partial deficiency of Sp that correlated with a RBC morphology characteristic of HPP, namely, marked microspherocytosis with virtual absence of elliptocytes. These data demonstrate striking differences in the function and structure of various alpha Sp mutants that underlie differences in clinical expression.

Erythrocytes infected with P. falciparum express knob-like adhesion structures that allow the infected cells to cling to the postcapilliary endothelium of characteristic host organs. At present, the mechanism of cytoadherence is not fully understood. While parasitized erythrocytes have been shown to specifically bind to the platelet/matrix molecule thrombospondin, adherence to suitable target cells can also be blocked by monoclonal antibody OKM5, which recognizes a surface molecule expressed by hematopoietic cells and endothelium. In apparent reconciliation of these findings, it has been reported that the OKM5 antigen (CD36) is a receptor for thrombospondin. Here we report that expression of a CD36 cDNA clone in COS cells supports cytoadherence of parasitized erythrocytes but does not support increased binding of purified human thrombospondin.

The members of the integrin family of membrane glycoprotein heterodimer complexes function as cell surface receptors for adhesive proteins. We report here on the identification of two integrins on the surface of human platelets that bind to thrombospondin. When platelet membrane proteins are radiolabeled with 125I-lactoperoxidase, solubilized in n-octylglucoside, (Boehringer Mannheim Biochemicals, Indianapolis, IN), and applied to a column of thrombospondin-Sepharose, both complexes are bound to the column and specifically eluted with the peptide GRGDSP. One of these integrins, glycoprotein (GP) IIb-IIIa, appears to bind relatively weakly. The second integrin shares the same beta subunit (beta 3 or GPIIIa), but has a distinct alpha subunit that comigrates with the alpha subunit (alpha v) of the vitronectin receptor (VnR) on endothelial cells and reacts with a monoclonal antibody, LM142, which was raised against an integrin from M21 melanoma cells. The alpha v beta 3 integrin is present on a variety of cell types and appears to act as a receptor for thrombospondin on endothelial and smooth muscle cells. On endothelial and M21 melanoma cells this receptor is also involved in adhesion to fibrinogen, vitronectin, and von Willebrand factor (vWF). The alpha v beta 3 integrin is present at approximately equal levels on normal and thrombasthenic platelets, whereas levels of GPIIb-IIIa are greatly reduced on thrombasthenic platelets. The alpha v beta 3 integrin on thrombasthenic platelets also binds to thrombospondin-Sepharose and can be eluted with the peptide GRGDSP. These data indicate that the alpha v beta 3 integrin on platelets, endothelial cells, and smooth muscle cells functions as an Arg-Gly-Asp (RGD)-dependent receptor for thrombospondin.

Two patients are described who presented with severe hemolysis and erythrocyte fragmentation. One patient had renal allograft rejection and disseminated intravascular coagulation, and the other had thrombotic thrombocytopenia purpura. The severity of hemolysis and the red cell abnormalities were considerably more profound than usually seen in patients with thrombotic microangiopathies. After evaluation of blood smears prepared before the onset of the disease and biochemical characterization of proteins of the red blood cell skeleton, a mutation of the skeletal protein spectrin, designated Sp alpha l/65, was identified. In the heterozygous form, this mutation manifests as mild, often asymptomatic, hereditary elliptocytosis. We conclude that in these two patients with thrombotic microangiopathy, the intrinsic red cell membrane instability resulting from the underlying skeletal defect aggravated the mechanical red cell fragmentation, producing morphological features similar to the severe hemolytic form of hereditary elliptocytosis or hereditary pyropoikilocytosis.