Role of colony stimulating factor-1 in the establishment and regulation of tissue macrophages during postnatal development of the mouse

Colony stimulating factor-1 (CSF-1) regulates mononuclear phagocytic cells (Stanley et al., 1983; Tushinski et al., 1982) via a high affinity cell surface receptor which is the c-fms proto-oncogene product (Sherr et al., 1985). Although derived from a common precursor cell (reviewed by Van Furth, 1992), the mononuclear phagocytic system (MPS) is composed of a cell population that is heterogeneous in terms of its tissue localization (Gordon, 1988; Morris et al., 1991b), functional activity (Gordon, 1988) and expression of surface molecules (De Jong, 1990). One mechanism contributing to this heterogeneity within the MPS may be the exposure of cells to different tissue microenvironments, due to differences in the local production of growth factors, such as granulocyte-macrophage colony stimulating factor (GM-CSF) (Witmer-Pack et al., 1987; Kaplan et al., 1992) and CSF-1 (Bartocci et al., 1986; Pollard et al., 1987). CSF-1 can be locally presented either in a bio-logically active, membrane-spanning form on the surface of the cells that synthesize it (Rettenmier et al., 1987; Price et al.,1992) or alternatively, as the predominant secreted (proteoglycan) form (Price et al., 1992), which may be targeted to a tissue-specific extracellular matrix via its glycosaminoglycan moiety. Differences between tissue microenvironments may also arise due to differences in cellular accessibility to circulating growth factors. For example, CSF-1 is found in the cir-culation (Stanley, 1979) and certain macrophage populations, such as Kupffer cells (Bartocci et al., 1987), have access to circulating CSF-1, while others, such as peritoneal macrophages (Chen, 1991), do not.

The CSF-1-less osteopetrotic (op/op) mutant mouse exhibits impaired bone resorption associated with a paucity of osteoclasts (Marks and Lane, 1976), is deficient in bone marrow macrophages, blood monocytes and serosal cavity macrophages (Wiktor-Jedrzejczak et al., 1982, 1990; Felix et al., 1990b) due to an inactivating mutation in the coding region of the CSF-1 gene (Wiktor-Jedrzejczak et al., 1990; Yoshida et al., 1990; Felix et al., 1990b; Pollard et al., 1991). Administration of CSF-1 to newborn op/op mice cured their osteopetrosis and substantially corrected the deficiencies in osteoclasts, bone marrow cellularity and blood monocytes, but not the deficiencies of the serosal resident macrophage populations (Kodama et al., 1991; Felix et al., 1990a; Wiktor-Jedrzejczak et al., 1991), suggesting that some macrophage populations are dependent on circulating CSF-1, while others require locally produced CSF-1. Thus, the op/op mouse is an ideal model in which to investigate how CSF-1 regulates the anatomical distribution of the MPS. Indeed, preliminary descriptions of how this mutation affects the distribution and morphology of macrophages detected by the expression of the macrophage specific antigen, F4/80 (Austyn and Gordon, 1981), indicated that their number was reduced in the majority of tissues examined in the adult op/op mouse (Naito et al., 1991; Wiktor-Jedrzejczak et al., 1992b). Furthermore, these cells generally exhibited a morphology consistent with their incomplete differentiation.

We here describe the distribution of tissue macrophages in normal (+/+ or +/op) and mutant (op/op) mice during the first three months of postnatal development. Furthermore, we have investigated the role of circulating CSF-1 in tissue mononu-clear phagocyte regulation by analysis of the redistribution of tissue macrophages in op/op mice when circulating CSF-1 levels are restored during this period.

Osteopetrotic op/op mice and littermate controls (+/+ or +/op) were bred and maintained in isolated units of the Albert Einstein College of Medicine animal house and of the Department of Pathophysiology at the University of Berne as described previously (Wiktor-Jedrzejczak et al., 1990; Felix et al., 1990b; Pollard et al., 1991). Mice were fed ad libitum with powdered chow and infant milk formula (Enfamil). At birth, the op/op homozygotes were radiologically distinguished from the phenotypically normal siblings based on the dense aspect of bone due to the absence of a distinct medullary cavity (Felix et al., 1990a). At 10 days of age, they were distinguished by the absence of incisors and by a domed skull (Marks and Lane, 1976). Three month-old op/op mice weigh 65% as much as littermate control mice.

Highly purified recombinant human CSF-1, a generous gift from Chiron Corporation (Emeryville, California), was suspended in physiological saline (2×10⁷ units, equivalent to 0.24 mg protein/m; Stanley et al., 1972) and stored at −20°C. Unless otherwise indicated, the op/op mice were subcutaneously inoculated with 50 μl of this solution (10⁶ units per mouse) daily from 3 days of age. Littermate control mice were injected daily with 50 μl of physiological saline. Mice were killed at 3-4 months of age.

For immunostaining with the macrophage specific rat monoclonal antibody F4/80 (Austyn and Gordon, 1981), op/op mutants and their normal siblings at the age of 2 days, 2 weeks, 2 months and 3 months and the CSF-1 injected op/op mice were perfused in vivo under ether anesthesia through the left ventricle (Hume and Gordon, 1983) with periodate-lysine-2% paraformaldehyde-0.05% glutaraldehyde, pH 7.4 (PLPG) (McLean and Nakane, 1974). Tissues were then excised and fixed for 6 hours at 4°C in the same fixative. The tibia, including the knee-joint, was decalcified for 48-72 hours in several changes of acid/citrate buffer (13% sodium citrate in 2% formaldehyde, pH 4.7 with formic acid) (Hume et al., 1984). The tissues were then dehydrated and embedded in polyester wax. Sections of 5 μm were cut and air dried on gelatin coated slides.

For immunostaining with the rat monoclonal antibody MOMA-1, specific for marginal metallophilic macrophages of the spleen (Kraal and Janse, 1986), frozen sections of spleens from op/op mutant and their normal siblings at the age of 2 months and from CSF-1 injected op/op mice were fixed in hexazotized pararosaniline for 1 minute as described (Kraal and Janse, 1986; De Jong et al., 1991), stored at − 70°C and immunostained within 1 week.

Immunostaining was performed according to the indirect peroxidase-conjugated streptavidin procedure (Hume and Gordon, 1983; Felix et al., 1990b), using rat gamma globulin (5 μg/ml) as a control.

At each age, at least two op/op and two normal mice were examined. For tissues with linearly arrayed macrophages (synovial membrane, epidermis and periosteum), F4/80+ cell densities were determined using an ocular micrometer and expressed in cells/mm; otherwise the densities were determined by scoring at least 5-10 fields and expressed in cells/mm². In heterogeneous tissues, such as bone marrow, or in tissues with an uneven parenchymal distribution of F4/80+ cells, such as spleen and lymph node, the F4/80+ cell density was determined with the aid of the Zeiss I integrating eyepiece (Meunier and Courpron, 1973) and expressed in cells/mm² of the specific tissue of interest. F4/80+ cell densities were averages derived from 2 animals and standard deviations for multiple counts (n>5) were <10% of the means. Most tissues were examined in the two participating institutions and the results independently verified.

Sections of tissues obtained from 2-day-to 3-month-old phenotypically normal (+/+ or +/op) mice were immunostained for F4/80 antigen in order to follow the postnatal development of their macrophage populations. Tissues have been classified into groups (Table 1) depending on the behavior of their F4/80+ populations in order to simplify the description and discussion of results. Here and subsequently in the results section, the text refers only to the topology and morphology of stained cells. The reader is referred to the tables for the changes in cell density.

In muscle and tendon, macrophages occur in regions of tissue remodeling, for example in areas of osteoinsertion and generally they are spindle-shaped with the major axis parallel to the fibers of both muscle and tendon. Macrophages of the dermis are generally round, except those in proximity to the basal epidermal layer, which are aligned with it as spindle-shaped cells. This is particularly evident at 2 days of age (Fig. 1A). At 2 weeks of age they are sometimes found to surround the developing hair bulbs. The F4/80+ cells in neonatal liver (Kupffer cells) are evenly distributed within the liver parenchyma. They are highly dendritic and intimately associated with islands of hematopoietic cells (Fig. 2A). By 2 months, their density is dramatically decreased in the centrilobular regions but maintained around the portal triads and they are much less dendritic (Fig. 2C). By this time, hematopoietic cells are no longer visible. At 2 days of age, dendritic F4/80+ cells in the retina, with characteristics of microglia, are restricted to the inner nuclear layer and to the developing inner plexiform layer (Fig. 3A), as previously reported (Hume et al., 1983a). By 2 weeks these cells are absent from the above sites, but similar dendritic cells appear in the outer plexiform layer. At 2 and 4 months there are no detectable F4/80+ cells in the retina.

At 2 days of age, F4/80+ cells in the stomach are exclusively located in the subglandular region of the lamina propria (not shown). They are few in number, but by 14 days their density is greatly increased in this region, they are more closely associated with the chief cells and they are much more dendritic. F4/80+ cells are also apparent along the glandular columns in association with parietal and mucous cells. In addition, by 14 days they are present in the lamina propria of the non-glandular regions and at a much lower density in the muscularis externa. At 2 and 4 months, the distribution and morphology of the F4/80+ cells is essentially unchanged. At 2 days of age, the F4/80+ cells of the small intestine are characteristically rounded and present predominantly in the lamina propria towards the base of the villi (Fig. 4A). Their density in this region increases dramatically over the ensuing 12 days and the cells adopt a more spindle-shaped morphology. At later ages they are predominantly found within the villi, closely associated with the capillaries and the base of the columnar cell layer. Occasional cells are also found lining the inner surface of the serosa (Fig. 4C). Similar postnatal changes in the morphology and distribution of F4/80+ cells are observed in the large intestine, where the cells are also located in the lamina propria.

At 2 days of age, lymph node F4/80+ cells are uniformly distributed throughout the still primitive parenchymal structure. At 2 weeks, F4/80+ cells are present in both the cortex and the medulla. They are abundant adjacent to the sub-capsular sinus and concentrated within the medullary sinuses. The cells are stellate with some scattered reticular macrophages in the germinal centers of the lymphoid follicles. By two months, the number of F4/80+ cells is decreased, especially in the subcapsular area. However, they are still well rep-resented in the medullary region. F4/80+ cells in the 2-day thymus are found predominantly in the cortical area, where they are slightly spread and surround developing thymocytes. Less spread F4/80+ cells are also found, more sparsely distributed, in the medulla. By 2 weeks, slightly dendritic F4/80+ cells can be seen along the connective tissue septae associated with the arterioles and around the corticomedullary junction. Scattered, less dendritic, F4/80+ cells are also apparent in the medulla and surrounding the Hassel’s corpuscles. By 2 months, the number of positive cells is decreased, especially in the subcapsular region. They are still well represented in the medulla.

Within the first 2 weeks the F4/80+ cells in periosteum are associated with sites of active bone modeling during rapid bone growth. Subsequently, they are predominantly found at the metaphyseal junction. As described by others, the elongated F4/80+ cells of the kidney are more dense in the medulla and primarily line the medullary and cortical tubules (Hume and Gordon, 1983) and in cross sections of the skin, the Langerhans cells in the epidermis are elongated and intercalated with cells of the basal layer (Hume et al., 1983b, Fig. 1).

Two morphologically distinct F4/80+ cell types were resolved in the bone marrow. The monocyte type are lightly stained, rounded cells whereas the macrophage type, described by Hume as resident bone marrow macrophages (Hume et al., 1983b), stain heavily and are highly dendritic (Fig. 5). At 2 days of age, they are smaller and less dendritic than in older mice (Fig. 5A). By 14 days of age, they assumed the typical adult stellate morphology (Felix et al., 1990b) (Fig. 5B). These cells are present in the diaphysis and penetrate the primary spongiosa to the level of the capillary invasion front where they assume a spindle shape and are spread on the outer walls of the vessels.

F4/80+ cells are almost absent from 2-day bladder. By 14 days, they are found as spindle-shaped cells in the lamina propria, occasionally in close relationship to the basal cell layer of the transitional epithelium. They are also found scattered within the smooth muscle layer. By 2 months of age, their density in both locations is substantially increased. Significantly, the cells within the lamina propria acquire a much closer relationship with the basal cell layer and are much more dendritic. Some F4/80+ cells are associated with the capillaries and extend processes into the epithelium. By 4 months of age, the vast majority of F4/80+ cells are associated with the basal layer of the lamina propria while their density in the smooth muscle layer is reduced.

In the spleen, F4/80+ cells are initially rounded (day 2) but by 2 weeks they become stellate. They are exclusively located in the red pulp where they represent the major sub-population of splenic macrophages (De Jong, 1990). The F4/80+ cells of the adrenal gland at 2 days, prior to the development of the specialized architecture of this gland are round and randomly distributed in the parenchyma. At 2 weeks, they are stellate and preferentially located in the outer cortex. By 2 months, their numbers have increased in the outer cortex and they are also found in the inner cortex and in the periphery of the medulla. In the inner cortex, F4/80+ cells are smaller than in the outer cortex, while in the medulla they assume a more spindled shape.

The F4/80+ cells of the submandibular and sublingual salivary glands are found in both subepithelial and interstitial locations. In the interstitium they are scattered and less spread. In the submandibular gland they are closely associated with the tubuloalveoli and are spindle-shaped. A similar morphology of F4/80+ cells is observed in sublingual salivary glands where they are located on the epithelial side of the basement membrane and associated with the secretory duct.

Between 2 and 14 days of age, the density of F4/80+ cells lining part of the synovial membrane (synovial ‘A-cells’) increases significantly and remains constant thereafter (Fig. 6). At 2 days, however, many F4/80+ cells are also uniformly scattered within the synovial cushion. By 14 days, consistent with a redistribution from the synovial cushion to the synovial membrane, the density of synovial cushion macrophages is decreased. During the first 2 weeks of life, the monocytes of the bone marrow (Fig. 5) are mainly localized in the diaphysis and are scarce in the metaphysis, being almost totally absent within 300 μm of the capillary invasion front. By 2 months of age, when longitudinal bone growth has substantially slowed, the monocytes are more evenly distributed between the meta-physeal and diaphyseal regions. In the diaphyseal region, the monocytes are more often found closely associated with the dendritic macrophages.

A significant proportion of the developmental changes in the mouse mononuclear phagocytic system occurs postnatally (see above and Morris et al., 1991b). To determine the role of CSF-1 in this process, the postnatal changes in F4/80+ cells in the CSF-1-less op/op mutant were compared with those described above for normal littermate control mice (Tables 1 and 2). Tissues were divided into the groups a-d (below), according to the behavior of their F4/80+ populations in op/op mice.

The macrophage populations of striated muscle, tendon and kidney failed to develop in op/op mice. Although present at a small fraction of their normal density, some macrophages of the dermis were always present throughout life. They were found exclusively in the region immediately below the epidermal layer and exhibited a morphology similar to those found in the same region of normal mice. In contrast, the macrophages in the deeper region of the dermis were virtually absent, except at 2 weeks of age, when rounded macrophages were found in association with the developing hair bulb, as in normal mice. Interestingly, in op/op mutants this deep region of the dermis was thinner (hypotrophic and hypoplastic) than age-matched normal littermates at all the ages examined (Fig. 1).

Although extremely reduced in density during the first 2 weeks of postnatal life, some macrophages developed in the periosteum by 2 and 3 months of age, but they were exclusively concentrated at the remodeling site, located at the meta-physeal-diaphyseal junction. True synovial ‘A-cells’ were always absent and occasional F4/80+ cells could only be found immediately below the synovial membrane, never adjoining it. Noteworthy was the hypoplasia and, in extreme cases, the atrophy of this membrane, especially at the later ages examined. A peculiar situation was observed in the retina where, in contrast to normal littermate control mice, F4/80+ cells could not be detected in 2-day (Fig. 3B) or 2-week-old op/op mice. However, by 2 months of age, dendritic F4/80+ cells became apparent in the inner nuclear and plexiform layers of the retinas of op/op mice with a density matching their density in 2 day control mice. They persisted with a slightly lower density at 3 months. While the appearance of these cells in op/op mice was delayed and their presence less transient, their morphology was indistinguishable from the microglia of the 2-day-old control mice.

In the adrenal glands, by 2 weeks, F4/80+ cells are less stellate than in control mice and are found preferentially in the outer cortex and between the cortex and medulla. The perivascularly located cells disappear with age. In the 2-week-old bladder, F4/80+ cells are scarce, less spindle-shaped than in control mice and located in the lamina propria. By 2 months, compared with control mice, F4/80+ cells are less dendritic and not so closely associated with the transitional epithelium. In salivary glands, by two weeks, F4/80+ cells are flatter, less spindle-shaped and smaller than in control mice. They are more interstitially localized and less associated with the tubuloaveoli (submandibular glands) or secretory ducts (sublingual glands). In contrast to control mice, throughout the postnatal period in op/op mice, bone marrow macrophages are confined exclusively to the diaphyseal region. There was a progressive enlargement of the marrow space with age. By 2 weeks, the percentage of the diaphyseal region occupied by marrow was 46% and this increased to 72% by 2 months of age. Within the first 2 weeks the strongly F4/80+ bone marrow op/op macrophages were much less dendritic than those of littermate control mice (Fig. 5B,D). By 2 months, only approximately 50% of the F4/80+ cells had adopted the typical dendritic shape of cells from control mice (Fig. 5F).

In liver, there was no apparent difference in the morphology of the Kupffer cells at any stage but there was a precocious loss of their centrilobular localization in op/op mice (Fig. 2B,D). In op/op mice, the distribution of macrophages at day 2 in stomach, small (Fig. 4B) and large intestine was normal. However, at later times, in contrast to control mice, in the stomach they were virtually absent from the lamina propria surrounding the glandular columns. In the intestine the villous cores (Fig. 4D) and the connective tissue surrounding the crypts of Lieberkuhn were similarly devoid of cells. F4/80+ cells of the gastrointestinal tracts of op/op mice tended to be smaller, more rounded and reminiscent of these populations in normal day-2 mice.

There was no significant difference in the distribution of F4/80+ cells in the spleen. However, their morphology was different at birth and 2 weeks of age, in that cells that were less dendritic and more rounded than in normal mice appeared unevenly distributed in the red pulp. Differences in morphology at the later time points could not be detected. In order to detect populations of splenic macrophages not recognized by the F4/80 antibody, in particular the marginal metallophilic macrophages (Hume et al., 1983b; Witmer and Steinman, 1984), frozen sections from two-month-old op/op and control mice were immunohistochemically stained with monoclonal antibody MOMA-1. Interestingly, the population of MOMA-1+ marginal metallophilic macrophages, found in control sections at the junction between the red and white pulp, was totally absent in the sections from two day and two month (Fig. 7) old op/op mice.

No difference in the distribution or morphology of epidermal Langerhans cells(Fig. 1) and of F4/80+ cells in thymus and lymph node could be discerned, except in the case of the bone marrow monocytes, which tended to closely associate within the larger marrow spaces.

CSF-1 is normally found (approx. 12-18 ng/ml) in the circulation (Wiktor-Jedrzejczak et al., 1990). In order to assess whether macrophage populations required locally presented, rather than circulating CSF-1 for their development, circulating CSF-1 in op/op mice was restored to normal levels or above with human recombinant CSF-1 from 3 days of age. In 8-week-old op/op mice that were injected daily as described, the con-centration of circulating human CSF-1 at 24 hours after their last injection was approximately the concentration of circulating CSF-1 in normal control mice (18 ng/ml) (Wiktor-Jedrzejczak et al., 1990), so that on average the circulating CSF-1 concentration in the injected mice was above normal. According to their response to injected CSF-1 (Table 3), tissue macrophages could be classified into those that apparently required circulating CSF-1 and those that did not.

While macrophage densities of spleen red pulp returned to normal, F4/80+ cell densities in the livers (Fig. 2E) and kidneys of the CSF-1 treated op/op mice were several fold higher than in the age-matched, littermate controls (Table 3). It was not possible to discern a difference in the morphology of splenic F4/80+ cells from op/op mice, normal mice and injected op/op mice at three months of age. In the liver, the cells were more concentrated around the portal triads and around the central vein and adopted a more elongated morphology than in the uninjected mice. Similarly in the kidney, the F4/80+ cells were more spindle-shaped than in the unin-jected op/op mice and were concentrated in the medulla along the tubules and within the juxta-glomerular complex, as is the case of normal kidney (see above and Hume and Gordon, 1983). A dramatic restoration of the density and morphology of the MOMA-1+ marginal metallophilic macrophages of the spleen in their normal location between the white and red pulp was observed in the CSF-1 injected op/op mice (Fig. 7C). The density of microglia in the inner nuclear and plexiform layers of the retinas of op/op mice was approximately twice their density in age-matched op/op mice that did not receive CSF-1, and approximately 1.5 times their density in normal 2-day-old mice. These cells also had longer and more delicate projections than their counterparts in uninjected mice.

In these tissues, the F4/80+ cell densities increased from 1.5-6.5 times the densities observed in untreated op/op mice, but at best attained 85% of the densities of the age-matched, littermate controls (Table 3b). In the bladders of CSF-1-treated op/op mice, the cells were found preferentially in the lamina propria surrounding the vessels and they were less spindle-shaped and less associated with the basal layer of the transitional epithelium than the F4/80+ cells in untreated mice. The F4/80+ cells within the smooth muscle layer were not restored by CSF-1 treatment. In the salivary glands, the F4/80+ cells were more equally distributed between the glandular and non-glandular parenchyma and were smaller and less dendritic than in untreated mice. The data in Table 3 suggesting that there is partial restoration of F4/80+ cell densities in the stomach and intestine is misleading if one considers subpopulations within these tissues. While there was partial restoration of the density of the cells in the lamina propria at the base of the mucosa, the cell density in the lamina propria surrounding the glandular columns of the stomach, the Crypts of Lieberkuhn and the villi was unaffected by CSF-1 treatment (Fig. 3E). The morphology of the F4/80+ cells at the base of the mucosa was unaffected by CSF-1 treatment. A difference in the local distribution of the macrophages induced by CSF-1 was also evident in the dermis. There was a clear reconstitution of the macrophage density in the deep region of the dermis, but these cells were larger and more irregular than their normal counterparts. In contrast, in the region immediately below the epidermis, the macrophage density was virtually unchanged compared with untreated op/op mice. The cell density and morphology of macrophages of the bone marrow were both partially restored by CSF-1 treatment. Only one-third of these cells acquired the typical, highly dendritic morphology exhibited in normal mice. They were distributed in both diaphysis and metaphysis, but in the metaphysis they never reached the capillary invasion front.

This group includes F4/80+ cells in striated muscle, tendon, periosteum, synovial membrane and adrenals, whose densities remain very low throughout postnatal to adult development in op/op mice. It also includes bone marrow monocytes, Langerhans cells, thymus and lymph node macrophages whose densities were approximately normal in op/op mice. In neither group was there significant effect of restoration of circulating CSF-1 on the morphology or distribution of F4/80+ cells.

The densities of F4/80+ cells in the livers and kidneys of the CSF-1-injected mice were several-fold higher than their densities in age-matched control mice (Table 3), possibly reflecting the choice of a CSF-1 dose that ensured that the injected mice possessed at least normal circulating concentrations of the growth factor at all times. Because of the sensitivity of the macrophages in these organs to circulating CSF-1, they were chosen to examine the CSF-1 dose-response. As shown in Table 4, a dose of 3×10⁴ units or 0.36 μg per day was sufficient to maintain the F4/80+ cell densities found in the livers and kidneys of normal littermates. Higher concentrations were needed, however, to maintain the splenic population.

In general, the maximal expression of F4/80+ cells during normal postnatal development correlated with the period of maximum organogenesis and/or cell turnover. For example, muscledevelopment peaks immediately prior to birth, and this is reflected in the rapid postnatal decline in F4/80+ cells (Rugh, 1991). On the other hand, tissues such as tendon, which is subjected to extensive remodeling as the osteoinsertion changes with longitudinal bone growth, and dermis, which continues to expand after birth, have a significantly slower decrease in F4/80+ cell densities. The changes observed in the microglial population of the retina are also consistent with scavenger and trophic roles of F4/80+ cells during development. As previously noted (Hume et al., 1983a), the coincidence of macrophage invasion of the different layers of the retina with neuronal cell death, a characteristic of the development of central nervous tissue (Oppenheim, 1981), corroborates their scavenger function. In addition, their subsequent localization in the plexiform layers may also be ascribed to their trophic role in promoting neurite extension (Perry et al., 1987). Consistent with the possible role of the stromal macrophages of the hemopoietic tissues in regulating hematopoiesis (Gordon et al., 1986; Crocker et al., 1988; Morris et al., 1991a,b), the postnatal decrease in F4/80+ cells in liver may be related to its transition from a fetal hematopoietic to adult parenchymal organ.

Gut and kidney are known to undergo substantial postnatal development associated with acquisition of adult function and this is paralleled by the flux in macrophage population (Rugh, 1991). Increased F4/80+ cell expression in thymus, lymph node and epidermis (Langerhans cells) is correlated with the development of immune competence (Bier et al., 1981). The postnatal changes in bone marrow macrophages may be related to two separate physiological changes. During the first few weeks of postnatal life there is a progressive increase in bone marrow hematopoiesis, which is probably supported by the resident F4/80+ cell population (Gordon et al., 1986; Crocker et al., 1988; Morris et al., 1991b). Simultaneously, in conjunction with the peak in expression of periosteal macrophages, the maximum rate of longitudinal bone growth, demanding substantial bone modeling and remodeling, is achieved.

In tissues with a sustained increase in F4/80+ cells the correlation between expression of high F4/80+ cell density and tissue turnover is less apparent. While the function of macrophages within these tissues has not been studied in detail, it is clear that there is a persistently high cell turnover in adult bone marrow and spleen. The F4/80+ monocyte-like cells in the latter two organs probably represent the precursors of blood monocytes (Gordon et al., 1986) for which there is a continual demand in the adult.

This group (Table 2a) includes muscle, tendon and dermis in which the F4/80+ cell density was normally highest at birth (Table 1), indicating that their requirement for CSF-1 is prenatal. The F4/80+ cells in periosteum, synovial membranes, kidney and the inner nuclear and plexiform layers of the retina were normally well established at birth and continued to increase in number during the first 2 weeks of life (Table 1), consistent with a prenatal and early postnatal requirement for CSF-1. Osteoclasts, likely derived from mononuclear phagocyte precursors (Suda et al., 1992; Hofstetter et al., 1992), and similarly dependent on CSF-1, also normally develop during the prenatal period (Scheven et al., 1986). Indeed, in contrast to earlier statements (Yoshida et al., 1990; Naito et al., 1991), in op/op mice osteoclasts also fail to develop during the prenatal period, since the principal trait of the op mutation, the osteopetrosis, characterized by the absence of a distinct medullary cavity, is recognizable both radiologically at birth (Felix et al., 1990a) and histologically at day-17 of post-conceptional age in femurs and tibias which normally at this age are already invaded by marrow (M. G. C. and T. Morohashi, unpublished observations). Taken together, these observations indicate that significant expression of the op mutation occurs prenatally.

A first subgroup (Table 2b), adrenals, bladder, salivary glands and bone marrow macrophages, with the exception of bone marrow macrophages, normally appeared postnatally and thus their partial requirement for CSF-1 is exclusively postnatal. In the case of bone marrow macrophages, the partial requirement for CSF-1 is extended to both pre- and post-natal periods. A second subgroup, liver, stomach, gut and spleen (Table 2c), were initially independent of CSF-1, but by two weeks of age demonstrated a CSF-1 requirement. Since there was no major effect of the absence of CSF-1 at birth on these cells, they must either be independent of CSF-1 or regulated by maternal CSF-1 during the prenatal period (see below). However, all of the populations in this subgroup require CSF-1 for their postnatal development and maintenance. At two weeks of age, the spleen, normally at this time a major site of hematopoiesis, exhibits a normal density of F4/80+ cells in op/op mice. However, at this age, we have almost invariably observed a splenomegaly accompanied by a concomitant increase of less dendritic, more rounded F4/80+ cells, consistent with the reported accumulation of macrophage progenitors (Begg et al., 1993) and probably secondary to the failure of monocyte differentiation in and migration from the spleen.

A more detailed analysis of the F4/80+ cells exhibiting an absolute or partial requirement for CSF-1, reveals that some F4/80+ cell populations exhibit delayed development in op/op mice, either partially, as in the case of bone marrow and periosteal macrophages, or completely, as in the case of the macrophages of the retina. Concomitant with the increase in density of F4/80+ bone marrow macrophages in op/op mice, there is also an improvement of the osteopetrotic status, as indicated by an increase in the bone marrow area. Indeed, it has been shown that in older op/op mice (Begg et al., 1993), bone marrow cellularity returns to normal levels by 22 weeks of age and by 35 weeks, the frequency and total number of F4/80+ cells are normal. These changes in the bone marrow are accompanied by an increase in splenic granulocytopoiesis and megakaryocytopoiesis. This improvement of F4/80+ cell density is restricted to the hematopoietic organs active at these ages (spleen and bone marrow). In fact, in the liver there is no detectable improvement even at 40 weeks of age (M. G. D., S. M. and E. R. S., unpublished observations). An improvement of the bone remodelling at 45 days of age has been recently demonstrated (Wink et al., 1991). The mechanism leading to the spontaneous remission in these tissues is unknown. Irrespective of the mechanism involved, the maximum effect of the absence of CSF-1 in op/op mice is best studied within the first 4-6 weeks of age. The interpretation of morphological and functional studies, performed at later time points, especially those involving the hematopoietic organs (Naito et al., 1991; Wiktor-Jedrzejczak et al., 1992a; Wiktor-Jedrzejczak et al., 1992b), may have been influenced by this compensatory phenomenon.

Some of these F4/80+ cell populations normally developed prenatally and postnatally (Table 2d), while others developed exclusively postnatally (Table 1). F4/80+ cells of lymph node and bone marrow monocytes developed postnatally and are clearly completely independent of CSF-1 while Langerhans cells and cells of the thymus could be supported prenatally by maternal CSF-1, which has recently been shown to cross the placenta (P. Roth and E. R. S., unpublished observations). However, their substantially normal postnatal development and maintenance in the op/op mouse reported here and elsewhere (Takahashi et al., 1992, 1993; Witmer-Pack et al., 1993) is consistent with their total independence of CSF-1. Indeed, the in vitro requirement of granulocyte-macrophage CSF for Langerhans cell viability and function (Witmer-Pack et al., 1987), together with the in vivo stimulation of Langerhans cell recruitment by local injection of this growth factor (Kaplan et al., 1992), confirms that this population is regulated by factors other than CSF-1. Interestingly, bone marrow monocytes were the only F4/80+ population to exhibit above normal cell densities in op/op mice. Their accumulation in the marrow of 2-week-old op/op mice (Table 2d), occurs at the time of maximum development of this population in normal mice (Table 1c) and was temporally correlated with a severe deficiency of bone marrow macrophages. These observations are consistent with the failure of these monocytes to differentiate to macrophages in the absence of CSF-1.

Tissue F4/80+ cells in op/op mice showed three distinct patterns of response to postnatal restoration of circulating CSF-1 (Table 3).

In the case of liver and spleen, in which the F4/80+ cell densities were approximately normal in op/op mice at birth and decreased with increasing age, CSF-1 administration prevented the postnatal decline. The fact that CSF-1 can cross the placenta and that the op/op mice used in this study were offspring of +/op mothers, strongly suggests that their F4/80+ cell densities are normal at birth because in these two organs they are prenatally regulated by maternal CSF-1. These results confirm an earlier study (Hume et al., 1988), which indicated that tissue macrophages of liver and spleen were two major target cell populations affected by in vivo administration of recombinant human CSF-1 in normal mice. Kidney F4/80+ cells, which pre- and postnatally are markedly dependent on CSF-1, exhibited a dramatic response to injected CSF-1, consistent with their regulation by the circulating growth factor. The reasons for the failure of circulating maternal CSF-1 to support F4/80+ cells in the fetal kidney compared with fetal liver (P. Roth and E. R. S., unpublished observations) are at present unknown.

In the op/op spleen, CSF-1 administration completely reconstituted the MOMA-1+ marginal metallophilic macrophages with exactly the same morphology, location and pattern seen in normal spleen. Similar specific histological localization after restoration of circulating CSF-1 is also observed for the F4/80+ cells in the inner nuclear and plexiform layers of the retina and for osteoclasts (Felix et al., 1990a). As it seems unlikely that such cells have selective access to circulating CSF-1, their development is probably dependent on both circulating CSF-1 and a cytokine that is locally synthesized or presented at specific anatomical sites. The function of the splenic marginal metallophilic macrophages is not clear. It has been suggested that they are involved in antigen presentation (Kraal and Janse, 1986), the detoxification of endotoxins (Eikelenbloom, 1978) and the direction of lymphocyte traffic in the spleen (Brelinska and Pilgrim, 1982).

In dermis, bladder, salivary glands, stomach and gut, the F4/80+ cells were associated with the epithelia. With the exception of the salivary glands, restoration was limited to those regions where this association was weak, e.g. in the deep dermal regions and in the lamina propria at the base of the mucosa. Interestingly, this is the same region where these cells, which may have been recruited prenatally by maternal CSF-1, are found at birth. The failure of the epithelially associated F4/80+ cells to develop in CSF-1 injected op/op mice (e.g. Fig. 4E), strongly suggests that their postnatal development is regulated by local presentation of CSF-1. Bone marrow macrophages failed to be completely restored by CSF-1 administration, suggesting that locally presented CSF-1 is also necessary for their complete restoration. In contrast to the other F4/80+ cells of this group, they developed strongly during the postnatal period even without administration of CSF-1 (above, Table 3) (Begg et al., 1993). Importantly, both in the case of spontaneous or induced recovery, these stromal macrophages are poorly reconstituted in the region of the primary spongiosa and this further suggests that local presentation of CSF-1 at this particular region is critical. The recent observation that the proteoglycan form of CSF-1 binds bone-derived collagens and is extractable from bone matrix (Ohtsuki et al., 1993) suggests that the glycosaminoglycan moiety may mediate selective localization of CSF-1, as originally hypothesized (Price et al., 1992).

A subgroup of macrophages colonizing the dense connective tissues (muscle, tendon, dermis, periosteum and synovial membrane), which did not benefit from transplacental CSF-1 and failed to develop during the prenatal period, was also unresponsive to injected CSF-1. These tissue macrophages are most likely to be dependent on locally presented CSF-1 (see above).

The densities of F4/80+ macrophages in the thymus, lymph node, epidermis (Langerhans cells) and of the bone marrow monocytes were not decreased in op/op mice and were not affected by CSF-1 injection. This observation, together with the relatively normal early development of splenic F40/80+ macrophages are in agreement with recent immunological studies in which it has been shown that op/op mice possess normal in vivo phagocytic function, normal delayed type hypersensitivity and normal humoral and cellular immune responses to sheep red blood cells (Wiktor-Jedrzejczak et al., 1992a).

CSF-1 has been shown to stimulate the spreading of cultured macrophages (Boocock et al., 1989) and there are distinct differences in the morphology of macrophages grown in the presence of CSF-1 and those grown in the presence of granulocyte-macrophage CSF (GM-CSF), which are more rounded (Falk and Vogel, 1988; Akagawa et al., 1988). Previous studies have reported that uterine, splenic and liver macrophages from op/op mice (Naito et al., 1991; Pollard et al., 1991) are more round and possess less developed organelles than their counterparts in normal mice. In the present study a smaller, more rounded appearance of F4/80+ cells in op/op mice compared with littermate control mice was also noted in the spleen, liver, kidney, adrenal gland, bladder, salivary glands, bone marrow and gastrointestinal tract. In most situations where CSF-1 administration elicited a recovery of the F4/80+ population there was at least some reversion towards normal macrophage morphology. In the case of the MOMA-1+ marginal metal-lophilic macrophages and retinal microglia there was complete or even exaggerated reversion. The functional significance of the more rounded morphology of the CSF-1-dependent macrophages in op/op mice is not clear but is possibly related to their failure to differentiate terminally in the absence of CSF-1. Interestingly, there was no detectable difference in the morphology of F4/80+ cell populations whose densities were unchanged in op/op mice.

The approach used here to analyze the requirement of tissue mononuclear phagocyte populations for local and humoral CSF-1 has several limitations. Neither the mouse glycoprotein or proteoglycan forms of CSF-1, which naturally occur in the circulation, were used because human recombinant CSF-1 was the only form of CSF-1 available in sufficient quantities. This may mean that in some situations where lack of regulation by circulating recombinant CSF-1 occurred, local regulation by differential localization of circulating glycosylated CSF-1 might actually occur. Secondly, because of the difficulty in determining what dose of CSF-1 to inject, we chose to use higher concentrations than normal in order to ensure that excess rather than limiting concentrations of circulating CSF-1 would be available. This has helped our interpretation of the data in situations where F4/80+ populations were unaffected since they were unaffected at very high concentrations. However, since the doses used were significantly higher than those required for regulation of the responding populations (Table 4), it is possible that some populations that were partially or completely corrected are in fact not normally regulated by circulating CSF-1. An additional limitation is associated with the difficulties of studying the fetal op/op F4/80+ populations from fetuses carried by op/op mothers. Because of the very poor fertility of op/op×op/op mating pairs (Pollard et al., 1991), this was not possible in the present study.

A simple interpretation of the data is presented in summary form Table 5, where tissue macrophages are grouped depending on the nature of their requirement for CSF-1 and whether it is required prenatally and/or postnatally. In general, macrophages requiring CSF-1 for their establishment or maintenance appear to have trophic or scavenger roles important in organogenesis and tissue remodelling (physiological processes), whereas CSF-1-independent macrophages are primarily involved in immune and inflammatory responses (pathological processes). Further exploration of the functions of the CSF-1-dependent macrophages, utilizing the op/op mouse, is warranted in view of the aspects of the op phenotype, such as reduced weight, poor fertility, hypoplastic and hypotrophic dermis and synovial membrane and neuromuscular disfunction that might result from their absence.

We thank F.-C. Chuan, J. Portenier, R. Rubli, U. Mausli and R. Zadeh for excellent technical assistance and Dr Phillip Roth for critically reviewing the manuscript. This work was supported by Swiss National Science Foundation grants 3.894.0.88 and 32.31272.91., by American Cancer Society Grant DB-28 and a Monique Weill-Caulier Award (to J. W. P.), by NIH grant CA 32551 and a grant from the Lucille P. Markey Charitable Trust (to E. R. S,) and the Albert Einstein Core Cancer Grant P30-CA 13330.