Reelin, a 420 kDA glycoprotein secreted by Cajal-Retzius cells of the marginal zone in the developing cortex, is essential for the correct positioning of cortical plate neurons (for review, see Curran and D'Arcangelo, 1998; Frotscher, 1998; Tissir and Goffinet, 2003). This protein is mutated in the reeler mouse, leading to brain cytoarchitecture disorganization. More recently, studies on reelin function suggested also specific roles postnatally: secreted reelin has been suggested to provide a signal for pyramidal cells dendritic spine plasticity (Rodriguez et al., 2000; Costa et al., 2001), motor end-plate maturation, proper nerve-muscle connectivity, and synapse elimination (Quattrocchi et al., 2003).
Few studies examined extraneuronal reelin expression. Smalheiser et al. (2000) reported reelin plasma levels thought to be produced by liver cells; they also detected reelin immunoreactivity in some chromaffin cells and in the pituitary pars intermedia. Ikeda and Terashima (1997), analyzing reelin mRNA localization during mouse development, reported some transient or permanent expressions in different peripheral organs, i.e., liver, kidney, testis, and ovary.
In the present work, we reexamined extraneuronal reelin immunolocalization and report reelin expression in lymphatic endothelial cells and liver stellate cells in both rat and human and discuss these results in correlation with phenotypic results in reeler mouse and human reelin mutations.
MATERIALS AND METHODS
Reelin was detected and its expression analyzed in developing rat fetuses, adult rat organs, and human fetal organs.
Tissue Preparation in Fetal and Adult Rat
For rat fetal analysis, inbred Wistar rats of our animal facilities were used. Vaginal smears were controlled 5 days a week. Three- to 4-month-old females were paired with one male for mating during the night preceding estrus. The day of observation of a copulatory plug was designed as gestation day 0 (E0). Timed pregnant rats were anesthetized with sodium pentobarbital and fetuses were removed by cesarean section; the head (E15, 16, 17, 18) or the fetus (E13, 13.5, 14) was fixed in Bouin Holland fixative for histological and immunohistochemical analysis. Serial sagittal or frontal 5 μm paraffin sections were cut; two sections every 75 μm were stained with Groat hematoxylin and eosin for histological examination; two sections every 75 μm were processed for immunohistochemistry. Four fetuses at each stage were used.
Adult organs were treated using the same procedure: 2- or 3-month-old female rats were anesthetized with sodium pentobarbital and the following organs were removed: ovaries, lung, spleen, gut, striated muscle, and liver. To localize reelin immunoreactivity in liver at the ultrastructural level, two additional rats were used for immunocytochemistry on free-floating sections. The animals were perfused with 4% PFA; the liver was postfixed overnight in the same fixative; 50 μm thick sections were cut on a vibratome and processed for immunocytochemistry.
Tissue Preparation in Human Fetus
The organs of nine human embryos/fetuses (6 to 20 gestational weeks old) came from medical (three fetuses) or legal abortions (six embryos-fetuses) under the control of the local ethics committee. Tissues were fixed in formol or Bouin Holland and 5 μm paraffin sections were cut. Microscopic examination following hematoxylin-eosin staining showed no abnormalities and adjacent sections were further processed for immunocytochemistry. No differences in immunostaining were observed between the two fixative schedules.
Four antibodies were used: a monoclonal antibody against reelin, G10, a generous gift from Professor A. Goffinet (De Bergeyck et al., 1998); a monoclonal antibody against CD 31 (a pan endothelial cell marker; Dako); a rabbit polyclonal antibody against GFAP (Dako); and a rabbit polyclonal antibody against von Willebrand factor (Dako). Microwave demasking of antigens was followed by incubation in blocking buffer; endogenous peroxidase was blocked and primary antibody incubation was made overnight; a secondary biotinylated antibody incubation for 2 hr was followed by avidin-biotin complex (Vecastain Elite kit, Vector Laboratories) incubation. The peroxidase reaction product was revealed by VIP (Vector Laboratories) as a chromogen for paraffin sections and DAB for vibratome sections. Paraffin sections were counterstained with methyl green. Some floating vibratome sections were mounted on slides to be observed directly at the optic microscope. The others were further processed for electron microscopic observation. They were further fixed for 2 hr in 2.5% glutaraldehyde and postfixed in 1% osmium tetroxyde; after dehydration, they were embedded in epoxy resin. Semithin sections were counterstained with toluidine blue to select immunostained areas. Ultrathin sections counterstained with uranyl acetate and lead citrate were observed in a transmission electron microscope (Siemens). Double staining for GFAP and reelin on free-floating adult rat liver sections was visualized with secondary antibodies labeled with rhodamine (GFAP) and fluorescein (reelin). Negative controls had omission of the primary antibody, resulting in no staining.
RESULTS AND DISCUSSION
As expected, reelin immunolocalization in the developing cortex in rat and human fetuses was restricted to Cajal-Retzius cells of layer 1 (Fig. 1A). At E13, E14 in the rat, the jugular lymphatic sac surrounded by small clefts was strongly stained; small clefts and elongated isolated cells could be stained in the mesenchyme of the whole body. No staining of endothelial cells of the aorta, main veins, and heart was observed. At E18, where only the head was examined, numerous empty clefts, near blood vessels, were strongly positive, especially in the mesenchyme of the mandible and in the tongue. Based on morphological criteria and von Willebrand factor immunostaining that intensely stains endothelial cells of blood vessels, but not lymphatics (Erhard et al., 1996; Akishima et al., 2004), the positive reelin, von Willebrand factor-negative stained clefts were identified as lymphatic vessels. No endothelial cells of blood vessels were stained (Fig. 1B and E). In normal adult rat tissues, staining was restricted to a few lymphatic vessels in lung and kidney. By contrast, staining was strong in lymphatics of the medulla of the ovary (Fig. 1C and F), of the lactating mammary gland, and in lymphatics of both small and large intestine. In human fetuses, similar observations were made: reelin immunoreactivity was detected in isolated mesenchymal cells or small clefts that contained no erythroblasts in the mesenchymal tissues, especially in lungs. As in rats, no arterial or venous blood vessels were stained (Fig. 1D and G).
Our observations showed that reelin is expressed in the developing lymphatic endothelium. Previously, Ikeda and Terashima (1997) reported reelin mRNA in some endothelial cells in kidney and spleen but not in thymus, lung, heart, or intestine. By contrast, we never observed reelin immunoreactivity in blood vessels of rats or human fetuses, but only in lymphatic clefts. The fact that reelin was detectable during development and in adult tissues where remodeling occurs (mammary gland and ovary) may suggest a role for reelin in lymphangiogenesis. We can only speculate about the role of reelin in lymphatics during development and adulthood. Reelin is a secreted glycoprotein. Secretion may occur at the luminal side of endothelial cells and reelin may then reach the blood circulation for an unknown role. A circulating pool of reelin has been detected in adult mammals; hepatocytes (Smalheiser et al., 2000) and plasma cells (Underhill et al., 2003) have been shown to be peripheral sources of circulating reelin; lymphatics may be another one.
On the other hand, reelin may be secreted at the abluminal side of the endothelial cells and then influence surrounding connective tissue organization. Reelin is a serine protease of the extracellular matrix, which degrades rapidly fibronectin and laminin (Quattrocchi et al., 2002). It is well known that lymphatic capillaries differ from blood capillaries by several features, among them the absence of pericytes, a very poorly developed basal lamina, and very leaky junctions. Reelin may then contribute to this special phenotype both during lymphangiogenesis and in mature lymphatics. Alternatively, it is known that reelin binds to receptors such as apolipoprotein E receptor 2, very-low-density lipoprotein receptor or integrins in the central nervous system (D'Arcangelo et al., 1999; Hiesberger et al., 1999; Dulabon et al., 2000), and promoting cell migration (Hack et al., 2002). Such a role in lymphatic endothelial cell precursors migration could be hypothesized. Lymphatic endothelial cell cultures of reeler mouse may help to elucidate the role of the presence of reelin in these cells. Although no special lymphatic phenotype has been reported in the reeler mouse (Ikeda and Terashima, 1997), some patients with mutations in the reelin gene had lymphoedema (Hong et al., 2000). These patients showed at birth lissencephaly, a severe brain development disorder with impaired neuronal migration resulting in a smooth unfolded cortex and congenital lymphoedema.
Numerous theories have been proposed for the development of lymphatics. According to the centrifugal theory, lymph sacs develop from the neighboring veins and then give rise, by endothelial sprouting, to the peripheral lymphatic system; initial connections with the veins are lost and only maintained at the jugulosubclavian junction. In the centripetal theory, precursors of lymphatic endothelial cells in the mesenchyme arise close but independently from the veins and fuse into a lymphatic network. An integrating model has been proposed, with sprouting from the veins and in situ differentiation of mesenchymal precursors (for review, see Jeltsch et al., 2003; Oliver, 2004; Scavelli et al., 2004). The reelin positive scattered mesenchymal cells we observed may then be peripheral lymphatic precursor cells.
A second site of intense extraneuronal reelin immunoreactivity was observed in the developing rat and human liver and in adult rat liver (Fig. 2A and B). The stained cells lined the sinusoids (Fig. 2C and D). They appeared spindle-shaped or stellate (Fig. 2B–D) and their number decreased with age, relative to the number of hepatocytes (compare Fig. 2A and C). We identified these cells as stellate (Ito) cells. Ito cells lie within the space of Disse and have several functions: they store vitamin A in lipid droplets; they are responsible for the production of the extracellular matrix; when activated, they transform into potent fiber-producing cells and play a pivotal role in liver fibrosis (for review, see Senoo, 2004); they express GFAP (Buniatian, 1997). On adult rat liver sections, we observed quite similar expression of reelin and GFAP (Fig. 2E, F, and H); the cells contained large lipid droplets. Double immunofluorescence labeling showed that reelin immunoreactivity was colocalized with GFAP (results not shown). By immunoelectron microscopy, we observed reelin expression in stellate, lipid containing cells lying in the space of Disse, between endothelial cells and hepatocytes (Fig. 2I and J).
In rodents and human, during early organogenesis, mesenchymal cells in the septum transversum are entrapped in the growing liver, become localized in the subendothelium of the sinusoids, and turn into stellate cell progenitors (Enzan et al., 1997). Here, we show that reelin is expressed early during liver organogenesis in both human and rat stellate cells and that expression continues during adult life. Previously, Ikeda and Terashima (1997) reported reelin mRNA localization in sinusoid endothelial cells. Here we clearly localized the reelin protein only in stellate cells and not in endothelial cells. This lack of reelin localization in sinusoid endothelial cells is in agreement with the lack of reelin staining in other blood vessels. Kobold et al. (2002) previously localized reelin mRNA to stellate cells and hepatocytes in the adult rat liver; but in the present study, we did not find reelin protein in hepatocytes with the antibody we used. This strong expression of reelin in Ito cells remains unexplained. Indeed, reeler mice have no obvious hepatic phenotype, nor do they respond in a different way from wild-type mice to hepatic damage (Kobold et al., 2002). However, we cannot exclude a role for reelin in the migration of Ito cell precursors and therefor in liver organogenesis, as previously discussed for lymphatic endothelial cell precursors.
The role of extraneuronal reelin remains elusive. In a recent work, Maurin et al. (2004), by coculturing odontoblasts and trigeminal neurons, showed that odontoblasts express reelin, which may promote adhesion between dental nerve and odontoblasts. A role in guiding, adhesion, or anything else has yet to be discovered for lymphatic endothelial and hepatic stellate cell reelin.
The authors thank Professor A. Goffinet for his generous gift of reelin antibody and are grateful to P. Bos, A.L. Burry, R. Bury, and J. Meder for their excellent technical assistance. They are also grateful to Dr. B. Gasser for giving them the opportunity of studying human fetal tissues.