The distribution of antigenic determinants in rat skin collagen.

Immunological studies of rat skin collagen were carried out with a sensitive and quantitative radioimmunoassay. Hyperimmune rabbit antisera to rat skin collagen and isolated alpha2 chains were used. Iodine-labeled alpha chains and CNBr-produced peptides served as test antigens, and native collagen, alpha chains, and CNBr peptides were employed as inhibitors in the assay. The alpha1 and alpha2 chains were immunologically distinct. Although the alpha1 chain was not immunogenic, antibodies to alpha1 were detected in antisera to the intact collagen molecule. The major antigenic determinant of the alpha1 chain was located in alpha1-CB6 which constitutes the carboxy-terminal region of the chain. The alpha2 chain contained two non-cross-reacting antigenic determinants, one in the amino-terminal region (alpha2-CB1) and the other in the carboxy-terminal region (alpha2-CB5) of the chain. The native collagen molecule was less effective than isolated alpha chains in inhibiting binding of labeled peptides to antisera, indicating that antigenic determinants were less accessible in the triple helical molecule. These immunologic studies are consistent with preliminary comparative biochemical data which indicate that interspecies structural differences in collagen predominate at both the amino- and carboxy-terminal ends of the chains.

Heparin and related glycosaminoglycans modulate the secretory phenotype of vascular smooth muscle cells.

Previous studies have established a role for heparin-like molecules in the regulation of vascular smooth muscle cell growth and migration in vitro. We present data indicating that the secretory phenotype of cultured rat aortic smooth muscle cells can be modulated by exogenous soluble heparin, heparan sulfate, and dermatan sulfate glycosaminoglycans. In the presence of these molecules, smooth muscle cells secrete increased amounts of two noncollagenous proteins (Mr 37,000 and 39,000). This effect can be mimicked by iota carrageenan and dextran sulfate but not by hyaluronic acid, chondroitin-4-sulfate, or chondroitin-6-sulfate. The inductive effect of heparin was dose-dependent and occurred rapidly (within 1 h) with maximal induction (three- to fivefold over controls) occurring after 10-12 h of treatment. The effect was rapidly reversible (within 1 h) and was not altered in the presence of actinomycin D, suggesting regulation at a posttranscriptional level. These data indicate that the biosynthetic expression of specific smooth muscle cell proteins may be determined, at least in part, by components of the smooth muscle cell extracellular matrix.

Heparin regulates the collagen phenotype of vascular smooth muscle cells: induced synthesis of an Mr 60,000 collagen.

The effect of heparin on the biosynthetic phenotype of rat vascular smooth muscle cells (SMC) was investigated in vitro. Addition of heparin to the culture medium of early passage rat SMC resulted in a marked (3-15-fold) increase of a cell layer-associated Mr 60,000 protein that was sensitive to digestion by purified bacterial collagenase and contained significant amounts of hydroxyproline. Pulse-chase analysis of heparin-treated SMC revealed that the Mr 60,000 collagen was a primary and abundant product of these cells and was not processed extracellularly to a smaller form. The inductive effect of heparin could be mimicked by iota carrageenan or dextran sulfates but not by hyaluronic acid, dermatan sulfate, or chondroitin sulfates. The induction was concentration dependent with a maximal effect observed at a heparin concentration of 10 micrograms/ml. Synthesis of the Mr 60,000 collagen increased 18-24 h after addition of heparin to the cultures. Following induction and subsequent removal of heparin, synthesis of the protein remained maximal for at least 12 h and required 72 h to return to a basal level. These data demonstrate that the biosynthetic phenotype of vascular SMC in vitro can be controlled, at least in part, by heparin and related polyanions and suggest a role for similar molecules endogenous to the vessel wall in the regulation of SMC function.

The elastic fiber. I. The separation and partial characterization of its macromolecular components.

The two morphologically different constituents of the mature elastic fiber, the central amorphous and the peripheral microfibrillar components, have been separated and partially characterized. A pure preparation of elastic fibers was obtained from fetal bovine ligamentum nuchae by extraction of the homogenized ligament with 5 M guanidine followed by digestion with collagenase. The resultant preparation consisted of elastic fibers which were morphologically identical with those seen in vivo. The microfibrillar components of these elastic fibers were removed either by proteolytic enzymes or by reduction of disulfide bonds with dithioerythritol in 5 M guanidine. The microfibrils solubilized by both methods were rich in polar, hydroxy, and sulfur-containing amino acids and contained less glycine, valine, and proline than the amorphous component of the elastic fiber. In contrast, the amino acid composition of the amorphous component was identical with that previously described for elastin. This component demonstrated selective susceptibility to elastase digestion, but was relatively resistant to the action of other proteolytic enzymes and to reduction. These observations establish that the microfibrils consist of a different connective tissue protein (or proteins) that is neither collagen nor elastin. During embryologic development the microfibrils form an aggregate structure before the amorphous component is secreted. These microfibrils may therefore play a primary role in the morphogenesis of the elastic fiber.

Fetal calf ligament fibroblasts in culture secrete a low molecular weight collagen with a unique resistance to proteolytic degradation.

A highly unusual collagen was secreted by fibroblasts cultured from 150- and 270-d-old fetal calf nuchal ligaments. Purification revealed that this protein (which may be synthesized in a higher molecular weight form) was precipitated at unusually high concentrations of ammonium sulfate and was also eluted from DEAE-cellulose at greater salt concentrations than were types I and III procollagens. On SDS PAGE, the collagenous protein exhibited an Mr of approximately 12,750 that was not altered in the presence of reducing agent. The low molecular weight collagen (FCL-1) was sensitive to bacterial collagenase and had a [3H]glycine content comparable to that found in type I procollagen, although the [3H]Hyp to [3H]Pro ratio was 0.43. FCL-1 was not cleaved by human skin collagenase, mast cell protease, trypsin, Staphylococcal V8 protease, or proteinase K at 37 degrees C. The collagen was susceptible to trypsin, but not to V8 protease, only after heating at 80 degrees C for 30 min. Preliminary structural studies indicate that FCL-1 was resistant to cleavage by CNBr but exhibited limited proteolysis with pepsin. Both 150- and 270-d-old fibroblasts produced comparable levels of interstitial (types I and III) procollagens, which comprised approximately 70% of the total protein secreted into the culture medium. However, 270-d-old (term) fibroblasts secreted approximately 50% more FCL-1, as percent of total culture medium protein, in comparison to the cells from the earlier gestational stage. This collagen may therefore play a role in the development of the nuchal ligament.

Interactions of thrombospondin with extracellular matrix proteins: selective binding to type V collagen.

Thrombospondin (TS), a protein first described in platelets, was recently shown to be synthesized and secreted by endothelial cells, fibroblasts, and smooth muscle cells. The presence of TS in the extracellular matrix of cultured cells has prompted us to examine the associations of this protein with matrix macromolecules. Interactions of TS with both matrix and serum proteins were tested using an enzyme-linked immunosorbent assay. With this assay we assessed the binding of TS in solution to proteins adsorbed to polystyrene microtiter plates. Among collagens, platelet TS bound to type V but not to types I, III, or IV. This selective interaction was confirmed in experiments using proteins linked to cyanogen bromide-activated Sepharose. TS released from platelets in response to thrombin activation, as well as that secreted by endothelial cells in culture, bound to type V but not to type I collagen-Sepharose. No binding was observed to denatured type V collagen-Sepharose. The binding region for type V collagen was located in a chymotrypsin-produced fragment of TS with chains of Mr = 70,000, after reduction. Interactions of TS with a number of other proteins, including fibronectin, fibrinogen, and laminin, could be demonstrated using the enzyme-linked immunosorbent assay technique but the interpretation of these findings is difficult since comparable binding to protein-Sepharose was not always observed. Our findings suggest that both the extravascular distribution and function of TS in vivo may involve an interaction with type V collagen.

Effects of antimicrotubular agents on the secretion of collagen. A biochemical and morphological study.

Embryonic chick cranial bone was cultured in the presence of the antimicrotubular agents, colchicine and vinblastine, and with a number of other compounds known from previous studies to affect the cellular handling of collagen. Secretion of procollagen, quantitated by light microscope autoradiography, was correlated with the extent of conversion of procollagen to collagen and with rates of collagen and noncollagen-protein synthesis. Colchicine inhibited procollagen secretion and conversion to collagen and specifically inhibited collagen synthesis. Cells exposed to colchicine revealed an increased number of dilated Golgi-associated vacuoles and vesicles, some of which contained parallel aggregates of filamentous structures. These observations suggest that the pathway of at least a fraction of procollagen secretion by osteoblasts includes the Golgi complex. Disruption of microtubules may interfere with the movement of Golgi-derived vesicles, and the resulting accumulation of collagen precursors in the Golgi complex may lead secondarily to an inhibition of synthesis. Although vinblastine also inhibited both procollagen secretion and conversion to collagen, the observed reduction in general protein synthesis and striking changes in the ultrastructure of the rough endoplasmic reticulum complicated interpretation of the effects. Interpretation of the effects of cytochalasin B was limited by the finding that the cellular response in cranial bone was markedly heterogeneous and that, contrary to some previous reports, the drug caused an inhibition in the incorporation of radiolabeled amino acids into both collagen and noncollagen protein.

Platelet-derived growth factor and heparin-like glycosaminoglycans regulate thrombospondin synthesis and deposition in the matrix by smooth muscle cells.

Platelet-derived growth factor (PDGF), a smooth muscle cell (SMC) mitogen, and heparin-like glycosaminoglycans, known inhibitors of SMC growth and migration, were found to regulate thrombospondin synthesis and matrix deposition by cultured rat aortic SMC. The synthesis and distribution of thrombospondin was examined in growth-arrested SMCs, in PDGF-stimulated SMCs, and in heparin-treated SMCs using metabolic labeling and immunofluorescence techniques. Thrombospondin synthesis in response to purified PDGF occurred within 1 h after addition of growth factor to growth-arrested SMCs, peaked at 2 h, and returned to baseline levels by 5 h. The induction of synthesis of thrombospondin by PDGF was dose dependent, with a maximal effect observed at 2.5 ng/ml. Actinomycin D (2 micrograms/ml) inhibited thrombospondin induction by PDGF, suggesting a requirement for new RNA synthesis. In the presence of heparin and related polyanions, the incorporation of thrombospondin into the SMC extracellular matrix was markedly reduced. This effect was dose dependent with a maximal effect observed at a heparin concentration of 1 microgram/ml. Heparin did not affect the ability of SMCs to synthesize thrombospondin in response to PDGF. We interpret these data to suggest a role for thrombospondin in the SMC proliferative response to PDGF and in the regulation of SMC growth and migration by glycosaminoglycans.

Amniotic fluid fibronectin. Characterization and synthesis by cells in culture.

A glycoprotein immunologically related to plasma cold-insoluble globulin (CIG) and fetal skin fibroblast fibronectin has been purified from second-trimester human amniotic fluid. This protein (amniotic fluid fibronectin) migrated more slowly than CIG on sodium dodecyl sulfate gel electrophoresis and showed greater polydispersity which could result, at least in part, from heterogeneity in glycosylation. Cloned human amniotic fluid epithelioid and fibroblastic cells synthesized and secreted a protein with similar properties into the culture medium. Fibronectin was shown to be associated with the pericellular and extracellular matrix of cultured amniotic fluid cells by immunofluorescence, lactoperoxidase-catalyzed iodination, and labeling with ferritin-conjugated antibodies. The kinetics of secretion of the protein were consistent with its role as a matrix protein. We anticipate that amniotic fluid fibronectin will prove to be the same protein which elsewhere in the body is incorporated into connective tissues and basement membranes. Amniotic fluid could, therefore, serve as a convenient source of in vivo synthesized fibronectin for biological and structural studies.

Biochemical and ultrastructural studies of proteoheparan sulfates synthesized by PYS-2, a basement membrane-producing cell line.

The mouse teratocarcinoma-derived cell line, PYS-2, has been shown to produce laminin, a basement membrane-specific glycoprotein. In these studies we demonstrate that PYS-2 cells synthesize and secrete into the culture medium a proteoglycan which contains only heparan sulfate as its sulfated polysaccharide side chains, as well as type IV procollagen and laminin. The apparent molecular weights of the proteoglycan and its heparan sulfate side chain were estimated to be 400,000 and 25,000, respectively, by gel chromatography. A proteoheparan sulfate with properties closely similar, if not identical, to those of the proteoglycan in the medium, together with two heparan sulfate single chains of different molecular size, were extracted from the cell layer with 2% SDS in the presence of protease inhibitors. Ultrastructurally, a fine fibrillar intercellular matrix was recognized which contained discrete 100-200 A diameter ruthenium red-positive granules interspersed throughout the filamentous meshwork. The PYS-2 cultures were shown by immunofluorescence to react with antibodies against the heparan sulfate-containing proteoglycan isolated from the mouse EHS sarcoma (Hassell, J. R., P. G. Robey, H. J. Barrach, J. Wilczek, S. I. Rennard, and G. R. Martin. 1980. Proc. Natl. Acad. Sci. U. S. A. 77:4494-4498). Immunoelectron microscopic examination, using the same antibodies, revealed that the proteoheparan sulfate was located not only at the edges but also within the interstices of the matrix. These findings indicate that PYS-2 cells synthesize and secrete a proteoglycan with properties similar to those of basement membrane proteoglycan. These cells may therefore serve as a useful model system for the study of the biosynthesis and structure of basement membranes.

Cell-specific expression of alpha 1(I) collagen-hGH minigenes in transgenic mice.

Sequences within the first intron of the alpha 1(I) collagen gene have been implicated in the regulation of expression of alpha 1(I) collagen-reporter gene constructs in cultured cells. However, the physiological significance of these intronic elements has not been established. We have used in situ hybridization to examine whether a cell-specific pattern of expression of human alpha 1(I) collagen-human growth hormone minigenes exists in transgenic mice. Our results indicate that transgenes which contained 2,300 bp of promoter/5' flanking sequence and an intact first intron were well expressed by fibroblasts in dermis and fascia, whereas transgenes lacking the intronic sequence, +292 to +1440, were not expressed in dermis and poorly expressed in fascia. Analysis of transgene expression in cultured fibroblasts obtained from dermal explants of transgenic animals confirmed the requirement for these intronic sequences in the regulation of the alpha 1(I) collagen gene. In contrast, transgenes with or without the intronic deletion were expressed equally well in tendon and bone, in a manner comparable to the endogenous mouse alpha 1(I) collagen gene, and expression of neither transgene was detected in skeletal muscle or perichondrium. These data support a model in which cis-acting elements in the first intron, and their cognate DNA-binding proteins, mediate transcription of the alpha 1(I) collagen gene in some cells, such as dermal fibroblasts, but not in tendon cells or osteoblasts. Moreover, regions of the gene not included in the sequence, -2300 to +1440, appear to be required for transcription in tissues such as skeletal muscle and perichondrium.

Thrombospondin: synthesis and secretion by cells in culture.

Thrombospondin, a high molecular weight glycoprotein secreted by platelets in response to activation by thrombin, has been identified by immunofluorescence in bovine aortic endothelial cells, human foreskin fibroblasts, and human aortic smooth muscle cells. Immunofluorescence patterns were found to be similar using antisera raised to thrombospondins purified either from bovine aortic endothelial cells or from human platelets. Radioimmune precipitation of pulse-labeled cellular proteins confirmed the presence of thrombospondin in positively stained cells. A sensitive quantitative enzyme-linked immunosorbent assay (ELISA) was developed and used to determine that the accumulation of secreted thrombospondin was similar for endothelial cells and fibroblasts but was higher for smooth muscle cells. The presence of thrombospondin in a variety of cells suggests that its function may not be limited to an involvement in platelet interactions.

Type VIII collagen has a restricted distribution in specialized extracellular matrices.

A pepsin-resistant triple helical domain (chain 50,000 Mr) of type VIII collagen was isolated from bovine corneal Descemet's membrane and used as an immunogen for the production of mAbs. An antibody was selected for biochemical and tissue immunofluorescence studies which reacted both with Descemet's membrane and with type VIII collagen 50,000-Mr polypeptides by competition ELISA and immunoblotting. This antibody exhibited no crossreactivity with collagen types I-VI by competition ELISA. The mAb specifically precipitated a high molecular mass component of type VIII collagen (EC2, of chain 125,000 Mr) from the culture medium of subconfluent bovine corneal endothelial cells metabolically labeled for 24 h. In contrast, confluent cells in the presence of FCS and isotope for 7 d secreted a collagenous component of chain 60,000 Mr that did not react with the anti-type VIII collagen IgG. Type VIII collagen therefore appears to be synthesized as a discontinuous triple helical molecule with a predominant chain 125,000 Mr by subconfluent, proliferating cells in culture. Immunofluorescence studies with the mAb showed that type VIII collagen was deposited as fibrils in the extracellular matrix of corneal endothelial cells. In the fetal calf, type VIII collagen was absent from basement membranes and was found in a limited number of tissues. In addition to the linear staining pattern observed in the Descemet's membrane, type VIII collagen was found in highly fibrillar arrays in the ocular sclera, in the meninges surrounding brain, spinal cord, and optic nerve, and in periosteum and perichondrium. Fine fibrils were evident in the white matter of spinal cord, whereas a more generalized staining was apparent in the matrices of cartilage and bone. Despite attempts to unmask the epitope, type VIII collagen was not found in aorta, kidney, lung, liver, skin, and ligament. We conclude that this unusual collagen is a component of certain specialized extracellular matrices, several of which are derived from the neural crest.

Location of procollagen in chick corneal and tendon fibroblasts with ferritin-conjugated antibodies.

Three distinct antiprocollagen preparations were characterized and used in immunocytochemical staining of chick embryo corneal and tendon cells. The several ferritin-conjugated antibody preparations permitted similar location of procollagen in the cisternae of the rough endoplasmic reticulum and in Golgi elements in both cell types. The ability to demonstrate and interpret specific ferritin staining was dependent on the extent of membrane breakage in each of those organelles, coupled with adequate retention of cell morphology. Corneal fibroblasts appeared to suffer more extensive intracellular membrane damage under controlled conditions of homogenization than tendon fibroblasts, facilitating the identification of procollagen in Golgi vacuoles of these cells. None of the labeled material appeared to by cytoplasmic in origin since ferritin was observed in the cytoplasm only in the vicinity of Golgi elements that were extensively broken. This study extends previous immunological evidence for the presence of procollagen in the Golgi complex and calls attention to the problems to be encountered in locating the antigen in small Golgi vesicles and lamellae.

Mice that lack thrombospondin 2 display connective tissue abnormalities that are associated with disordered collagen fibrillogenesis, an increased vascular density, and a bleeding diathesis.

Thrombospondin (TSP) 2, and its close relative TSP1, are extracellular proteins whose functions are complex, poorly understood, and controversial. In an attempt to determine the function of TSP2, we disrupted the Thbs2 gene by homologous recombination in embryonic stem cells, and generated TSP2-null mice by blastocyst injection and appropriate breeding of mutant animals. Thbs2-/- mice were produced with the expected Mendelian frequency, appeared overtly normal, and were fertile. However, on closer examination, these mice displayed a wide variety of abnormalities. Collagen fiber patterns in skin were disordered, and abnormally large fibrils with irregular contours were observed by electron microscopy in both skin and tendon. As a functional correlate of these findings, the skin was fragile and had reduced tensile strength, and the tail was unusually flexible. Mutant skin fibroblasts were defective in attachment to a substratum. An increase in total density and in cortical thickness of long bones was documented by histology and quantitative computer tomography. Mutant mice also manifested an abnormal bleeding time, and histologic surveys of mouse tissues, stained with an antibody to von Willebrand factor, showed a significant increase in blood vessels. The basis for the unusual phenotype of the TSP2-null mouse could derive from the structural role that TSP2 might play in collagen fibrillogenesis in skin and tendon. However, it seems likely that some of the diverse manifestations of this genetic disorder result from the ability of TSP2 to modulate the cell surface properties of mesenchymal cells, and thus, to affect cell functions such as adhesion and migration.

Collagen: relatively invariant (helical) and variable (nonhelical) regions.

The structural identity of certain helical regions of collagen from human and rat skin equals or exceeds that of other homologous proteins. In contrast, the short nonhelical sequences in the two proteins, although homologous, differ appreciably in structure. The requirements of the collagen helix and the numerous intermolecular interactions characteristic of collagen may restrict the number of functionally acceptable amino acid replacements occurring during evolution.

Procollagen: conversion of the precursor to collagen by a neutral protease.

An enzymatic activity (procollagen peptidase), capable of converting the biosynthetic precursor procollagen to collagen at neutral pH, has been identified in rat and chick calvarial bone. Limited proteolysis of procollagen with chymotrypsin resulted in a similar transformation. The activity in bone can be demonstrated in vitro despite inhibition of new collagen synthesis by cycloheximide. Preservation of the collagen precursor in preparations extracted with acetic acid results from inhibition of the enzymatic activity at low pH.

TOM22, a core component of the mitochondria outer membrane protein translocation pore, is a mitochondrial receptor for the proapoptotic protein Bax.

The association of Bax with mitochondria is an essential step in the implementation of apoptosis. By using a bacterial two-hybrid assay and crosslinking strategies, we have identified TOM22, a component of the translocase of the outer mitochondrial membrane (TOM), as a mitochondrial receptor of Bax. Peptide mapping showed that the interaction of Bax with TOM22 involved the first alpha helix of Bax and possibly two central alpha helices, which are homologous to the pore forming domains of some toxins. Antibodies directed against TOM22 or an antisense knockdown of the expression of TOM22 specifically inhibited the association of Bax with mitochondria and prevented Bax-dependent apoptosis. In yeast, a haploid strain for TOM22 exhibited a decreased expression of TOM22 and mitochondrial association of ectopically expressed human Bax. Our data provide a new perspective on the mechanism of association of Bax with mitochondria as it involves a classical import pathway.

Thrombospondin-1, an inhibitor of angiogenesis, is regulated by progesterone in the human endometrium.

Thrombospondin-1 (TSP1), a multifunctional extracellular matrix glycoprotein, has been shown to suppress the angiogenic response in vivo and in vitro. We hypothesized that TSP1 might play a role in the inhibition of capillary morphogenesis during the endometrial cycle and examined its expression in 46 human endometrial specimens. Our results show that the expression of TSP1 in the endometrium is (a) cycle-dependent, (b) associated with periods of low capillary growth, and (c) regulated by progesterone. TSP1 protein was identified in the basement membrane of capillaries of the functional endometrium during the secretory phase. Abundant expression of TSP1 mRNA in the secretory phase was also detected by in situ hybridization, in contrast to the low levels seen in the proliferative phase. These findings were confirmed by Northern analysis of proliferative and secretory endometrium. Transcripts for TSP1 were observed predominantly in stromal cells, but signal was also detected in some endothelial and smooth muscle cells. Since the proliferation of endometrial tissue is regulated by steroid hormones, we tested the effects of estrogen and progesterone on TSP1 expression by stromal cells isolated from human endometrium. We found that levels of TSP1 mRNA and protein were increased after incubation with progesterone. Maximal stimulation of mRNA was observed after 8 h of treatment with 10-50 microM progesterone, and the effect was suppressed by the progesterone antagonist RU-486. Induction by progesterone was cell-specific and equivalent to the stimulation mediated by PDGF. Finally, the levels of TSP1 present in progesterone-stimulated cultures were sufficient to inhibit the migration of endothelial cells in vitro; this effect was nullified by anti-TSP antibodies. We therefore propose that the production of TSP1 at later stages of the endometrial cycle is linked to the inhibition of vessel formation and that TSP1 expression is progesterone-dependent in this tissue.

Two different cis-acting regulatory regions direct cell-specific transcription of the collagen alpha 1(I) gene in hepatic stellate cells and in skin and tendon fibroblasts.

The expression of the collagen alpha 1(I) gene in activated stellate cells plays an important role during liver fibrogenesis. To identify the critical cis-elements of the collagen alpha 1(I) gene in stellate cells, we used transgenic animals bearing various collagen alpha 1(I) regulatory regions directing the expression of either a human growth hormone minigene or the bacterial beta-galactosidase gene. We found that collagen alpha 1(I)-human growth hormone transgene expression was constitutively high in tendon and skin, provided the transgene contained the -2.3 to -0.44 kb collagen regulatory region. However in the liver, expression was stimulated several-fold, as was the endogeneous gene, by the fibrogenic hepatotoxin carbon tetrachloride. This stimulation occurred whether the collagen 5' regulatory region extended -2.3, -1.6 or -0.44 kb, and in the presence or absence of much of the first intron (+292 to +1607 bp). In addition, the -0.44 kb 5' region was sufficient for high-level transgene expression in stellate cells, following their activation by culture on plastic. In contrast, in skin and tendon, high-level transcription of the collagen alpha 1(I) gene required the -2.3 to -0.44 kb 5' flanking region. Thus, two different cis-regulatory regions direct cell-specific transcription of the collagen alpha 1(I) gene in stellate cells and in skin and tendon.