Hepatic lipocytes were isolated from normal rat liver and established in culture. A virtually pure isolate was obtained by fractionating enzymatically digested liver on a discontinuous gradient of arabinogalactan. Isolated cells displayed prominent rough endoplasmic reticulum and typical cytoplasmic droplets containing vitamin A. Lipocytes in pri-

Expression of the group of cytokines known as transforming growth factor-,f (TGF-pl, -.82 and -p3) is increased during liver regeneration induced by a 70% partial hepatectomy. The origin of these changes was examined in purified isolates of hepatocytes, sinusoidal endothelial cells, Kupffer cells (liver macrophages), and lipocytes (Ito or stellate cells) from normal and regenerating liver. In normal liver, TGF-131 and -132 levels were relatively high in sinusoidal endothelial cells and Kupffer cells. After partial hepatectomy, an early peak of TGF-182 and -,B3 was present in all four cell types, followed by a sustained increase in mRNA for TGF-131, -132, and -133 primarily in the hepatocyte population. The specificity of these changes was established by examining a mechanistically different injury model, fibrosis induced by ligation of the biliary duct. In this model, TGF13 mRNA was increased only in lipocytes and the increase was progressive over a 7-d period of observation. Secretion of TGF13 protein was examined in cell isolates placed in short-term primary culture and generally reflected the corresponding mRNA level. The TGFP released by hepatocytes was entirely in the latent form, whereas the individual nonparenchymal cell isolates released 50-90% active TGF13. Hepatocyte-conditioned culture medium, after treatment to activate latent TGFP, inhibited hepatocellular DNA synthesis as did the authentic factor. The data indicate that after injury TGFI3 increases selectively in the cells that are the target of the factor, i.e., in hepatocytes after partial hepatectomy and in lipocytes in inflammation and fibrosis. We conclude that the effects of TGFP in liver regeneration and fibrogenesis are predominantly, if not exclusively, autocrine. (J. Clin. Invest. 1995. 96:447-455.)

The subendothelial space of normal rat liver contains the constituent proteins of a basal lamina, as judged by immunohistochemical study oftissue sections. However, it is unknown whether these proteins constitute a complex with effects on hepatocellular function. We have examined this question, using normal rat hepatocytes cultured on substrata of matrix proteins as a model of the interaction between cells and basal lamina in vivo. In cultures on a type I collagen substratum, albumin secretion decreased progressively after 2 d. By contrast, when cells were cultured on a laminin-rich gel matrix, albumin secretion was stable for at least 3 wk; other functions and ultrastructural morphology were similarly maintained. None of the individual matrix proteins effectively substituted for the gel matrix, suggesting that full support of hepatocellular function requires a complex of matrix proteins. We speculate that a cause of hepatocellular dysfunction in acute inflammation is disruption of this matrix and alteration of its interaction with the hepatocyte plasma membrane.

We report a proteinase that degrades basement-membrane (type IV) collagen and is produced by the liver. Its cellular source is lipocytes (fat-storing or Ito cells). Lipocytes were isolated from normal rat liver and established in primary culture. The cells synthesize and secrete a neutral proteinase, which by gelatin-substrate gel electrophoresis and gel filtration chromatography, has a molecular mass of 65,000 D. The enzyme is 'secreted in latent form and is activated by p-aninophenylmercuric acetate but not by trypsin. Enzyme activity in the presence of EDTA is restored selectively by zinc and is unaffected by serine-protease inhibitors. In assays with radiolabeled soluble substrates, it degrades native type IV (basement membrane) collagen but not interstitial collagen types I or V and exhibits no activity against laminin or casein. At temperatures causing partial denaturation of soluble collagen in vitro, it rapidly degrades types I and V. Thus, it is both a type IV collagenase and gelatinase. The enzyme may play a role im initiating breakdown of the subendothelial matrix in the Disse space as well as augmenting the effects of collagenases that attack native interstitial collagen.

Affinity-purified rabbit antibodies specific for two large noncollagenous glycoproteinslaminin and fibronectin-were used to study the distribution of these proteins in normal murine kidneys. Immunofluorescence staining of conventional frozen sections demonstrates fibronectin within mesangial areas of the glomerulus . Laminin is also found in mesangial areas . However, it also appears to be distributed in typical basement membranelike patterns on glomerular and tubular basement membranes and Bowman's capsule. At the ultrastructural level, by labeling 600-800-Á thick frozen sections with a three-stage procedure consisting of specific antibodies, biotinyl sheep antirabbit IgG, and avidin-ferritin conjugates, fibronectin is present only in the mesangial matrix and is specifically localized to areas immediately surrounding mesangial cell processes. Laminin, on the other hand, is found uniformly distributed throughout tubular basement membranes, the mesangial matrix, and Bowman's capsule . In glomerular basement membranes, laminin labeling is restricted to the lamina rara interna and adjacent regions of the lamina densa.Basement membranes are complex structures containing collagenous and noncollagenous glycoproteins and proteoglycans. The collagenous components of basement membranes include several type-IV cotlagens (3,6,21,22) . In addition, the presence of type V (AB2) collagen has been demonstrated within basement membranes of lung, kidney, and selected vascular beds (18). Characterization of the noncollagenous basement membrane components has been hampered by many of the same problems observed in characterizing the collagenous moieties, namely low solubility and yield due to possible cross-linking, protease sensitivity, and difficulty in obtaining sufficient amounts of relatively homogenous basement membranes as starting material. Recently, Timpl et al. (20) described the isolation and purification of a large noncollagenous glycoprotein, laminin, from a transplantable murine sarcoma that produces an extracellular matrix of basement membranes. In immunofluorescence studies, antibodies to laminin were shown to react with basement membranes ofnormal tissues, suggesting that laminin or an immunologically related protein is a component of basement membranes in tissues (1,20) . Fibronectin, another large noncollagenous glycoprotein, has been found by some investigators to be associated with glomerular basement membranes (GBM) (12,15) . Although other investigators have found fibronectin associated with some basement membranes, they have not found it associated with the GBM (7,8,19). In this report, we demonstrate the differing ultrastructural localizations ofthese two noncollagenous glycoproteins in the kidney ofthe mouse, using ultrathin frozen-sectioning techniques coupled with immunoferritin reagents . Our results support a base-682