Hematopoiesis takes place within the extravascular compartment of bone marrow from which blood cells reach the circulation by passing through the thin wall of venous sinusoids (Branemark, 1959; DeBruyn et al., 1970; Campbell, 1972; Weiss, 1965, 1967, 1970; Chamberlain and Lichtman, 1978; Weiss and Sakai, 1984). Arising from progenitor cells situated near the interface with surrounding bone, the developing blood cells become congregated around segments of sinusoids in more central regions of the bone marrow (Lord et al., 1975; Shackney et al., 1975; Jacobsen et al., 1989; Jacobsen and Osmond, 1990; Osmond et al., 1992). Upon reaching the appropriate stage of maturity, the various blood cells pass into the sinusoidal lumen (Chamberlain and Lichtman, 1978; Osmond and Batten, 1984; Jacobsen and Osmond, 1990). Certain hemopoietic cells can also pass from sinusoidal lumen into the bone marrow parenchyma (Yoshida and Osmond, 1978; Dick et al., 1985; Spangrude et al., 1988; Iscove and Nawa, 1997). The sinusoidal wall is thus highly unusual in being the site of a continuous, large-scale, selective transport of cells, and may itself play an important role in facilitating and regulating cell traffic between bone marrow and the blood stream.
The sinusoid wall is composed of three layers. An inner endothelium is surrounded by a basement membrane and an outer layer of adventitial cells (Weiss, 1965, 1970, 1976; Campbell, 1972; Jacobsen et al., 1996). The endothelium is a continuous layer in which adjoining cells often interdigitate and are bound to one another through maculae occludentes. In contrast, adventitial cell processes only partially abut the sinusoidal wall, covering approximately 60% of the wall in rat bone marrow (Campbell, 1972; Weiss, 1970). Migration of blood cells from bone marrow into sinusoids takes place by a transendothelial cell route, rather than intercellularly, a mode of transportation limited to hematopoietic tissues. Prior to cell emmigration, adventitial cells retract to clear the surface of the endothelium, allowing cells to pass through. The retraction of adventitial cell processes is accompanied by loss of associated areas of basement membrane (Campbell, 1972). It remains to be determined whether the basement membrane of bone marrow sinusoids is structurally modified to facilitate frequent cell transit.
In recent years, individual components of basement membranes in various tissues have been examined ultrastructurally at high resolution in our laboratories. The findings have been correlated with immunohistochemical labeling of basement membranes as well as with the ultrastructural features of basement membrane constituent molecules forming aggregates in vitro (Inoue, 1991, 1994, 1995; Inoue and Leblond, 1988; Leblond and Inoue, 1989; Sawada and Inoue, 1994; Inoue et al., 1983, 1989, 1997) (Table 1). The results have shown that certain molecular components of basement membranes are unique in their ultrastructural morphology and size, and that they can be identified by these characteristic features.
Table 1. Ultrastructural conformation of extracellular (basement membrane) components as observed by high resolution examinations after immunohistochemical staining/labeling and by in vitro replication*
Although a basement membrane has been described and its discontinuous nature noted in bone marrow sinusoid walls (Weiss, 1965, 1970; Campbell, 1972), previous observations have been limited by inadequate tissue preservation, and detailed ultrastructure has not been elucidated (Weiss, 1976). In the present study, the basement membrane of the sinusoidal wall in mouse bone marrow has been examined by electron microscopy at high resolution after preserving the bone marrow under optimal conditions by perfusion-fixation in vivo. An attempt has been made to clarify its molecular organization, identifying individual components by their ultrastructural features. The results show that the basement membrane of bone marrow sinusoids is unusual in organization and composition, consistent with the specialized role of the sinusoidal wall in transmural transportation of hematopoietic cells.
MATERIALS AND METHODS
Four-week-old C57BL/6J mice (1 male; 1 female) (Jackson Laboratories, Bar Harbor, ME) were perfused via the left ventricle under chloral hydrate anesthesia (1.6% in 0.9% NaCl; 0.75 ml/25 g body weight), with a fixative containing 2.5% glutaraldehyde and 2% acrolein in 0.1 M sodium cacodylate buffer, pH 7.4, for 10 min, as described previously (Batten and Osmond, 1984; Jacobsen and Osmond, 1990). Femoral shafts were removed and processed for ultrastructural examination, as described (Jacobsen et al., 1994). Briefly, isolated femoral shafts were further fixed by immersion in fresh 2.5% glutaraldehyde/2% acrolein fixative overnight at 4°C, decalcified in 10% EDTA for 4 to 7 days at 4°C, and after washing with cacodylate buffer the tissue was cut into small blocks (10 blocks/animal). Blocks were postfixed with 1% osmium tetroxide and 1.5% potassium ferrocyanide in 0.1 M sodium cacodylate buffer (Karnovsky, 1971), dehydrated, and embedded in Epon. Thin sections (2–3 grids/block) were prepared and counterstained with uranyl acetate followed by lead citrate and examined in a JEM-2000FX electron microscope (JEOL, Ltd., Tokyo, Japan). Magnification was calibrated with a grating replica and a catalase crystal. Size was measured on prints by means of a hand magnifier equipped with a 0.1-mm scale.
Under electron microscopic examination at low magnifications, basement membrane material was observed around wide-bore venous sinusoids in bone marrow, localized to a uniformly narrow space approximately 60 nm wide between apposed endothelial cells and adventitial cell processes (Fig. 1A, SBM). The basement membrane was usually composed of irregular assemblies of apparently amorphous material varying in density. Rarely, the basement membrane was represented by a dark central layer with lucent spaces on both sides (Fig. 1B).
At higher magnification, the irregular masses of basement membrane appeared as sparse material that either extended across the entire width of the membrane space or was loosely concentrated on each side of the space (Fig. 2A). Occasionally, the material was more densely packed and contained round or irregular dark patches (Fig. 2B). At highest magnification, the basement membrane material was resolved into randomly arranged assemblies of several morphologically distinct components (Fig. 3A–F and line drawings of Fig. 4A–F).
Areas in which an optimum density of material filled the basement membrane space were first chosen for detailed examination to determine whether any basement membrane components could be identified by their characteristic ultrastructure (Figs. 3A, 4A). First, dense spots, or masses, scattered throughout the space were found to take the form of irregular coils or “accordion pleats” of fine filaments (Inoue, 1994; Leblond and Inoue, 1989) when their orientation was favorable (Figs. 3A, 4A, thick single arrow, and more typically in Figs. 3B, 4B). As presented in Table 1, this type of structure in the basement membrane has been identified as laminin (Inoue, 1994; Leblond and Inoue, 1989). The second component identified formed 3-nm-wide structures made up of two parallel dark lines separated by a lucent space, termed “double tracks” (Inoue, 1995) (Figs. 3A, 4A, paired arrows). This elongated ribbon-like structure is a form taken by chondroitin sulfate proteoglycan (CSPG) as identified by immunogold labeling (Inoue et al., 1997) as well as by in vitro reconstitution of very similar or identical double tracks from a preparation of purified CSPG (Inoue, 1995). Another noteworthy component was flexible filaments 1–1.5 nm in width (Figs. 3A, 4A, narrow single arrows). Their biochemical nature could not be established with certainty based on ultrastructural features alone. However, their appearances strongly suggest that they were made up of type IV collagen molecules because (1) the width of single and two-to-three laterally-associated molecules of type IV collagen measure 1.5, 2–2.5, and 3 nm, respectively (Table 1; Inoue et al., 1983; Inoue, 1994; Inoue and Leblond, 1988), and (2) immunohistochemical evidence has shown type IV collagen to be a major component of the sinusoidal basement membrane in bone marrow (Reilly et al., 1985; Nilsson et al., 1998).
Various sizes of dense, round (Fig. 3C, arrow; Fig. 4C, Agg) or irregularly shaped (lower center of Figs. 3A, 4A, Agg, and Figs. 3E, 4E) aggregates localized within the basement membrane space were examined in greater detail. A lighter print (Figs. 3D, 4D) of the area shown in Figure 3C, as well as Figures 3E and 4E, clearly showed that these aggregates were made up of random assemblies of 3-nm-wide double tracks denoting CSPG. In contrast, basement membrane-type heparan sulfate proteoglycan (HSPG), perlecan, was not present in the sinusoidal basement membrane. Perlecan is a major component of other basement membranes where it takes the form of a second type of double track that is distinctly wider (4.5–5 nm) than that of CSPG (3 nm) (Table 1; Inoue et al., 1989; Sawada and Inoue, 1994). After a thorough search of the basement membrane space in bone marrow sinusoids, HSPG-type double tracks were found only extremely rarely (Figs. 3F, 4F, arrows with H).
Finally, tiny pentagonal frames in two distinct sizes, 3.5 nm (Figs. 3A, 4A, large circles and Figs. 3D, 4D, circles) and 2.7 nm (Figs. 3A, 4A, small circles) were localized within bone marrow sinusoidal basement membrane. These ultrastructural characteristics have been shown to be those of the subunits (referred to as “pentosomes”) of amyloid P component (AP). These 2.7 nm- and 3.5-nm-wide pentagonal frames represent single pentosomes viewed at two different levels (Inoue, 1991).
Thus, the observations indicate that the irregular basement membrane of bone marrow sinusoids is made up of four major components (Fig. 5): (1) laminin, as densely stained irregular coils, (2) type IV collagen, as 1–1.5-nm-wide filaments, (3) CSPG, as 3-nm-wide ribbon-like “double tracks”, and (4) AP subunits (pentosomes), as 2.7- and 3.5-nm-wide pentagonal frames.
The four molecular components were present in the sinusoidal basement membrane as apparently random mixtures. No further structured organization of these components was observed. Basement membranes elsewhere are organized into a fine network of irregular anastomosing strands, referred to as “cords” (Inoue, 1989, 1994), built upon a three-dimensional network of either single, or two-to-three laterally associated molecules of type IV collagen (Inoue et al., 1983). To determine whether there was any evidence of initiation of such organization of structures in the sinusoidal basement membrane, 1–1.5-nm-wide filaments were further examined, particularly in areas where the amount of other components which might obscure the filaments was minimal. The filaments (Fig. 6A and line drawing of Fig. 7A) were associated with the surface of either endothelial or adventitial cells, but no networks or other organized structures were observed. Only rarely, could small network-like structures made up of the filaments be seen to span across the space between endothelial and adventitial cells (Figs. 6B, 7B, rectangle).
While the abundance of each of the four components of the sinusoidal basement membrane could not be accurately quantitated by the present techniques, visual inspection of numerous micrographs allowed a preliminary estimate to be made of their relative amounts (Table 2). While some aggregates of basement membrane material contained comparable numbers of laminin, type IV collagen, and CSPG profiles, other aggregates showed mainly CSPG double tracks, suggesting that the total amount of CSPG in the basement membrane was likely to be as much as twice the amount of laminin or type IV collagen. Only a small number of AP subunits (pentosomes) was present. By these criteria, the observations indicated that the predominant component of sinusoidal basement membrane in the bone marrow is CSPG (Table 2).
Table 2. Relative abundance of five extracellular components localized in basement membrane of sinusoidal wall of mouse bone marrow*
HSPG, heparan sulfate proteoglycan; CSPG, chondroitin sulfate proteoglycan; AP, subunits of amyloid P component (pentosomes).
Type IV collagen
Basement membranes are extracellular sheet-like layers, no wider than 200 nm, which separate certain tissues including epithelia, endothelia, muscle fibers, and the nervous system, from connective tissue compartments (Kefalides et al., 1979; Inoue, 1989). Typical, “common” basement membranes are usually described as being composed of three layers (Kefalides et al., 1979). The lamina densa, the main layer of the basement membrane, is separated from the surface of adjacent cells by a lucent layer or space, the lamina lucida, while the third layer, the lamina (or pars) fibrorecticularis, forms a transitional zone between the lamina densa and connective tissue.
“Double” basement membranes, such as the glomerular basement membrane of the kidney and the alveolar-capillary basement membrane of the lung, are formed by the fusion of two common basement membranes. Thus, two laminae densae fuse to form a single combined lamina densa, on both sides of which the laminae lucidae appear to persist while the laminae fibroreticulares are absent (Thorning and Vracko, 1977; Huang, 1979; Reeves et al., 1980). In recent years, however, the lamina lucida of both common and double basement membranes has been shown to be an artefact produced while preparing tissues by conventional methods for ultrastructural studies (Goldberg and Escaig-Haye, 1986; Chan and Inoue, 1994). In double basement membranes observed in tissues prepared by more advanced methods, including cryofixation and freeze substitution, which minimize alterations to the tissue, the entire “basement membrane space” is filled with lamina densa alone (Chan et al., 1993; Chan and Inoue, 1994).
Seen in the present study, the most commonly observed morphology of the sinusoidal basement membrane in mouse bone marrow is that of irregular basement membrane material, which either partially or completely fills the space between apposing endothelial and adventitial cells. Rarely, the basement membrane also appears as a straight lamina densa-like line with lamina lucida-like zones on both sides. This morphology generally resembles the appearances ascribed to double basement membranes (Laurie et al., 1984). A similar three-layered morphology has also been reported in bone marrow of mice, rats, and guinea pigs as a “well-defined basal lamina” (Campbell, 1972). However, these lamina densa-like lines appear to be rather imperfect as compared with a solid layer of lamina densa seen in typical double basement membranes. The adventitial cells are not surrounded by basement membrane (Campbell, 1972), an observation confirmed in the present study. However, this observation does not exclude the possibility that adventitial cells may synthesize basement membrane components that could contribute to the formation of a double basement membrane at the basal portion of the cell by adding to the basement membrane of the adjacent endothelium. If so, the basement membrane of bone marrow sinusoid would be another addition to the limited number of known double basement membranes. In any case, why the sinusoidal basement membrane of bone marrow takes on the two different morphologies, of both irregular assemblies of basement membrane components and a double basement membrane-like arrangement, is not clear. In view of the artefactual layers of glomerular basement membrane produced by conventional preparative methods (Chan et al., 1993; Chan and Inoue, 1994), the possibility that artefactual layering of material may have occurred in limited areas of the sinusoidal basement membrane during tissue preparation, must be considered.
Ultrastructural studies at high resolution have shown that the basic structure of both common and double basement membranes is a three-dimensional network of approximately 4 nm-wide irregular anastomosing strands, referred to as “cords,” the size of the network openings being 8–14 nm (Inoue, 1989, 1994). Each cord is composed of a core filament of type IV collagen around which other basement membrane components are clustered. The cord network has been observed as the basic structure in all basement membranes so far examined, except for a discontinuous basement membrane around spongio-trophoblast cells in rat placenta (Laurie, 1985).
The basement membrane of bone marrow sinusoids, in contrast with other basement membranes, shows no cord network, even in areas of compact packing of basement membrane material. One reason for the failure of cord network formation in the irregular sinusoidal basement membrane appears to be the lack of HSPG. This molecule binds to other basement membrane components through heparan sulfate chains as well as its core protein (Battaglia et al., 1992). A well-established role of HSPG is that of maintaining the integrity of basement membranes (Hassell et al., 1985; Fujiwara et al., 1984; Laurie, 1985). In the rat placental basement membrane, the only other currently available example of a discontinuous basement membrane, immunoperoxidase staining has demonstrated a total lack of HSPG (Laurie, 1985). Similarly, using immunofluorescence labeling, a lack of HSPG in sinusoidal basement membrane has been reported in rat bone marrow (Hamilton and Campbell, 1991). The present ultrastructural study confirms the lack of HSPG in sinusoidal basement membrane of mouse bone marrow by the absence of characteristic 4.5–5-nm-wide HSPG double tracks.
The presence of a small number of pentosomes, AP subunits, in the sinusoidal basement membrane (Table 2) may result from an influx of circulating AP. The localization of both laminin and type IV collagen in bone marrow sinusoidal basement membrane seen in the present study accords with previously reported immunofluorescence studies of basement membranes demonstrating a close co-localization of these two major components (Nilsson et al., 1998; Reilly et al., 1985). In addition to the lack of HSPG, certain observations in this study remain to be correlated with the observed inability to form typical basement membrane structures. Type IV collagen normally self-assembles (Yurchenco et al., 1992) to form a network, constituting a base onto which the cord network of the basement membrane is then built. Type IV collagen self-assembly takes place according to information inherent within the molecule itself (Paulsson, 1992; Timpl, 1996). The reason why type IV collagen filaments within sinusoidal basement membrane should remain free, without becoming organized into a network, is unknown. In bone marrow, the presence of isoforms of laminin has been reported (Gu et al., 1999; Siler et al., 2000). Such isoforms may be unable to self-associate and thus to contribute to the formation of an organized basement membrane structure.
The loosely organized sinusoidal basement membrane in bone marrow appears to be well suited for the particular functions of the sinusoidal wall. Being made up only of random assemblies of basement membrane components, as judged by ultrastructural criteria, in contrast with the usual organized cord network, it would appear to be well able rapidly to disassemble and reassemble, in unison with the movement of adventitial cell processes as they respond to the needs of transporting blood cells. To determine whether the unusual features of the basement membrane of bone marrow sinusoids are primarily related to cell trafficking, other sites of intense physiological cell transport may be similarly examined. Of particular interest will be the basement membrane in the wall of high endothelial venules in peripheral lymphoid tissues through which recirculating lymphocytes pass continuously in large numbers.
In bone marrow diseases such as myelofibrosis, pathologically altered sinusoids are associated with thickened or continuous basement membranes (Apaja-Sarkkinen et al., 1986; Lisse et al., 1991; Bussolino et al., 1994). However, it is not clear from published reports whether this pathological increase in sinusoidal basement membrane material represents simply a random accumulation of basement membrane components, or the formation of organized structures, such as a cord network.
Besides its easy assembly-disassembly, allowing the passage of blood cells, the role of the sinusoidal basement membrane of bone marrow is unclear. Other basement membranes provide mechanical support or anchoring, and also act as selective filters. Possibly, the discontinuous, irregular basement membrane can provide an adequate mechanical support for the wide-bore, low pressure, venous sinusoid wall while still allowing active cell traffic.
Basement membranes can also function as molecular reservoirs that temporarily store substances such as growth factors vital for cell development and differentiation. Unusually abundant CSPG (Table 2), in the form of either scattered or densely aggregated 3-nm-wide double tracks, was a most striking finding in the basement membrane of the sinusoidal wall in the present study. CSPG plays a vital role in hematopoiesis (Spooncer et al; 1988; Verfaillie et al., 1994b). This component, recognized as the CD44 receptor (Verfaillie et al., 1994a), is a major proteoglycan at the surface of hematopoietic progenitors in both human and murine bone marrow (Kolset and Gallagher, 1990; Miyake et al., 1990; Verfaillie et al., 1994a). In long-term bone marrow cultures, chondroitin sulfate, in addition to hyaluronic acid, has been observed to be associated with cells as well as being in the culture supernatant (Gallagher et al., 1983; Wight et al., 1986). The basement membrane of the sinusoidal wall may thus serve as a reservoir for chondroitin sulfate in the form of a proteoglycan for use in hematopoiesis.