Site-Specific Distribution of CD68-Positive Microglial Cells in the Brains of Human Midterm Fetuses: A Topographical Relationship with Growing Axons

Using 5 fetuses of gestational age (GA) of 15-16 weeks and 4 of GA of 22–25 weeks, we examined site- and stage-dependent differences in CD68-positive microglial cell distribution in human fetal brains. CD68 positive cells were evident in the floor of the fourth ventricle and the pons and olive at 15-16 weeks, accumulating in and around the hippocampus at 22–25 weeks. At both stages, the accumulation of these cells was evident around the optic tract and the anterior limb of the internal capsule. When we compared CD68-positive cell distribution with the topographical anatomy of GAP43-positive developing axons, we found that positive axons were usually unaccompanied by CD68-positive cells, except in the transpontine corticofugal tract and the anterior limb of the internal capsule. Likewise, microglial cell distribution did not correspond with habenulointerpeduncular tract. Therefore, the distribution of CD68-positive cells during normal brain development may not reflect a supportive role of these microglia in axonogenesis of midterm human fetuses.


Introduction
Microglia are the parenchymal mononuclear phagocytes of the central nervous system (CNS) and progressively populate the CNS during the fetal period of development in man, most conspicuously during the second trimester of life. During development, microglia appear in developing fiber tracts throughout the white matter before myelinogenesis in vivo [1]. Although microglia use white matter tracts as migratory pathways to the cerebral cortex [2], the main phase of migration does not occur until the second trimester in the telencephalon [3]. Although microglial hot spots are densely distributed in the subplate at 10-12 weeks, microglial cells at [19][20][21][22][23][24] weeks are restricted to (1) the ditelencephalon fissure, (2) being around the thalamus, (3) the corona radiata around the developing putamen, (4) the midline septal area, and (5) the optic tract [4,5]. These findings suggest that microglial cells phagocytose specific transient axons such as parts of thalamocortical projections. Indeed, microglia were reported to eliminate exuberant transcallosal projections during the development of the cat brain [6]. However, because of stages much earlier than myelination, myelin phagocytosis [7] is unlikely to occur in midterm fetuses.
The cerebral white matter during the last half of human gestation expresses high levels of GAP-43, indicating active outgrowth of axons during this mid-to late gestation stage [8]. Therefore, it is possible that a transient elevation of activated microglia during the time frame of active axonal outgrowth may reflect either a supportive role of these microglia in axonogenesis or a role in "pruning" overabundant axons by activated microglia. We used two immunological markers to assess the density of activated microglia, namely, CD 68 and major histocompatibility complex-class II (MHC-II). CD68positive microglial cells have frequently been observed in human fetal white matter [9][10][11]. We hypothesized that, between 12 and 18 weeks, sites of elimination of nerve fibers by microglial cells move from the subplate of the telencephalon to the diencephalon and more caudal areas. Müller and O'Rahilly [12] described key tracts of the embryonic human brain from stage 8 to stage 23, such as the habenulointerpeduncular, preoptic-hypothalamotegmental, accessory optic, and mammillotegmental tracts. These tracts are characterized by the early development but become less evident in the later stages. The habenulointerpeduncular tract or fasciculus retroflexus provides major monoaminergic input to the midbrain and is well developed in adult fish and rats [13]. Although these tracts were described in sagittal sections [12], excellent atlases by Bayer and Altman [14] showed that the habenulointerpeduncular tract can be identified also in frontal sections, starting from the medial aspect of the thalamus and running inferiorly and caudally between the red nucleus and the oculomotor complex. These well-developed tracts at these stages, in and around the diencephalon, are likely to be

Materials and Methods
The study was performed in accordance with the provisions of the Declaration of Helsinki 1995 (as revised in Edinburgh to encourage donation. Because the samples were collected without personal identifiers, it was not possible to trace any of the families concerned. The donated fetuses were fixed in 10% w/w neutral formalin solution for more than 1 month (3 months at maximum). After division into the head and neck, the thorax, the abdomen, the pelvis, and the four extremities, all parts were decalcified by incubating at 4 ∘ C in 0.5 mol/L EDTA (pH 7.5; decalcifying solution B; Wako, Tokyo) for 1-3 days, depending on the size of the body part. At 50 or 100 micrometer intervals, depending on size, specimens of the head were processed for sagittal or horizontal sections, each 5 micrometers thick. These sections included not only the brain but also any surrounding structures, including the eye, the ear, and the nose.
Most sections were stained with hematoxylin and eosin (HE), whereas others were used for immunohistochemical assays. The primary antibodies used were (1)  The samples were not autoclaved prior to treatment because of the loose nature of the fetal tissues. Secondary antibody (Dako Chem Mate Envison Kit) was labeled with horseradish peroxidase (HRP), and antigen-antibody reactions were detected via the HRP-catalyzed reaction with diaminobenzidine. All samples were counterstained with hematoxylin. We tried to perform a pair of different immunohistochemistry (such as CD68 and GFAP) using adjacent sections, but we often used near (not adjacent) sections due to a failure of the sectioning and/or immunostaining. The anterior commissure as well as the roots of the cranial nerves contained no or few positive cells. The cerebellum, including the eosinophilic lamina dissecans, did not contain any positive cells, but the developing cerebellar peduncles near the pons contained a cluster of positive cells. The microglial cells in the brain were round, fiber-like, or amoeboid in shape (Figures 1 and 2), whereas the pia and choroid plexus contained large, round cells expressing CD68 (Figure 1(c)).

Results
Vimentin-or GFAP-positive fiber bundles of the coronal radiata were much thicker at 22-25 weeks than at 15-16 weeks (Figure 3(a) versus Figure 1(a)). Sites of accumulation of CD68-positve cells were similar in the two sets of fetuses, with    (Figures 4(b) and 4(d)). The positive cells at 22-25 weeks were smaller than those at 15-16 weeks, with the former having a relatively simple, dot-like shape. All of the CD68-positive microglial cells, at both 15-16 and 22-25 weeks, including osteoclasts in the cranial base, were negative for HLA-class II DR. However, a few large, reticular cells in the occipital lymph nodes strongly expressed CD68. In contrast to Monier et al. [4], we did not find microvessels filled with CD68-positive cells (Figures 2(a), 3(a), 3(d), and 4(d)). Differences in CD68-positive cell distribution between stages are summarized in Table 1.
All of the 22-25-week samples were negative immunohistochemically for GAP43, partly due to the long preservation of these specimens in formalin solution. At 15-16 weeks, GAP43-positive fibers were evident in the subplate of the neocortex, including the frontal lobe ( Figure 5(a)), the fornix ( Figure 5(e)), the anterior and posterior commissures, and the cerebellum, especially its peduncles. However, positive fibers were not so abundant in the early form of the internal capsule or pyramidal tract ( Figure 6). GAP43-positive fiber bundles were also observed in and along the putamen (Figure 5(a)) and thalamus, in the entorhinal cortex at the bottom of a recess of the lateral ventricle along the ventral hippocampus ( Figure 5 and around the optic tract ( Figure 5(e)). A candidate for the habenulointerpeduncular tract, crossing the diencephalon on the dorsal side of the putamen, was also positive for GAP43 ( Figure 5(a)). However, with the exception of the corticofugal fibers passing through and along the pons and olive (Figures 2(d) and 2(e)), these GAP43 positive fibers did not contain any CD68-positive cells. Finally, immunohistochemistry of PCNA suggested that CD68-positive cells were in a proliferative state (Figure 7) although the analysis was conducted using semiserial sections (not serial).

Discussion
Although microglia are identified in the human CNS as early as embryonic development, the main phase of migration does not occur until the second trimester in the telencephalon [3], and it constitutively exhibits at least three types of morphology (i.e., amoeboid, ramified, and intermediate) [2]. The cerebral white matter during this period is potentially in a more "activated" state [11], because these cells may participate in developmental processes in the mid-to late-gestation organization of the brain, including vascularization [2], involution of the germinal matrix [17], programmed cell death [18], axonal development [8], and myelination [19]. The finding of a higher proportion of amoeboid microglia during the latter half of gestation suggests that the cerebral white matter during this period is potentially in a more "activated" state, a notion further supported by us with use of the antibody directed against CD68 to quantify activated microglial density. As with tissue injury, microglial upregulation of CD68 in response to programmed cell death during development prepares the cell to become phagocytic [20], whereas upregulation of MHC-II plays a role in immunity against foreign antigens in infection [21].
We observed differences in the distribution of CD68positive cells between fetal stages. CD68-positive cells were evident in the floor of the fourth ventricle and the pons and olive at 15-16 weeks, whereas they accumulated in and around the hippocampus at 22-25 weeks. The accumulation of these cells around the optic tract may not correspond to well-developed tracts at these stages but to a part of the commissural plate [12]. Likewise, only a limited part of the habenulointerpeduncular tract seemed to contain CD68positive cells. When we compared CD68-positive cell distribution with the topographical anatomy of GAP43-positive developing axons, we found that most of the GAP43positive axons were negative for CD68 expression, with a few exceptions, such as the transpontine corticofugal tract and the anterior limb of the internal capsule. The corticofugal fibers passing through the pons and olive may correspond to the early form of the pyramidal tract [14]. In contrast to our working hypothesis, discrepancies between GAP43 and CD68 immunopositivity were observed in the subplate of the neocortex, the anterior commissure, and peduncles of the cerebellum, all of which expressed GAP43 strongly but contained no or few CD68-positive cells. Hyaluronic acid has been found to facilitate cell movement in the fetal brain by weakening cell attachment to adhesive substrates and by creating hydrated pathways for migrating cells [22]. We recently reported the distribution of hyaluronan in the human fetal brain [23]. Although CD68-positve cells were apparently absent from hyaluronan-rich areas, such as the subplate and putamen, hyaluronan-poor areas did not always correspond to the localization of CD68-positve cells.
Another striking difference between stages was the size and shape of CD68-positive microglial cells, with the large, round, highly proliferative cells present at 15-16 weeks (Figlinebreak ure 7) but absent at 22-25 weeks. However, in contrast to Wierzba-Bobrowicz et al. [24,25] and Billiards et al. [11], all CD68-positive cells in the present study were consistently negative for HLA-DR. Thus the lack of MHC-II immunostaining of microglia in our cases reinforces classification of these cells as normative (i.e., not activated in response to infection and pathological inflammation).
We therefore question the basic concept that a specific pattern of distribution of CD68-postive cells corresponds to a functional state, such as the elimination of nerve fibers. The site-and stage-dependent distribution of CD68-positive cells may not reflect the supportive role of these microglia in axonogenesis in midterm human fetuses but may have a more specific role in histogenesis and modelling of the CNS during the development.