Transient calbindin-D28k-positive systems in the telencephalon: ganglionic eminence, developing striatum and cerebral cortex

Calbindin-D28k (calbindin) is a member of the superfamily of calcium- binding proteins implicated in the regulation of intracellular calcium. In the mature brain, calbindin is widely expressed in neurons of the forebrain and the hindbrain, and in the telencephalon calbindin-like immunoreactivity is particularly strongly expressed by medium-sized neurons of the striatum and by certain other neurons in the cortex and subcortex. We have traced the development of calbindin expression in the forebrain of the rat, and report here that in addition to the steady development of these calbindin-positive neuronal systems, transient waves of calbindin expression occur in cells of the ventricular zones of the basal ganglia and cortex and in cells of the telencephalic regions derived from these ventricular zones including radial glia of the developing striatum. In the striatum and its ventricular zone (the ganglionic eminence, or GE) we identified four transient calbindin-positive systems in the perinatal period. First, calbindin-immunoreactive cells began to appear in the GE by embryonic day (E)18, and by E20 an extensive dorsal and lateral part of the GE was marked by dense calbindin-like immunoreactivity in the ventricular zone. This calbindin system peaked at postnatal day (P)0-P3 and disappeared by P15. Its presence suggests that the GE is divisible on a molecular basis into lateral and medial districts that may correspond to derivatives of the lateral and medial ventricular ridges. Second, a system of calbindin-positive processes appeared in the dorsal and lateral caudoputamen with temporal and spatial distributions matching the germinal zone system. Many of these processes could be traced from calbindin-positive cells in the ventricular zone of the GE, including processes stretching across the full width of the dorsal caudoputamen. Double-staining experiments demonstrated that these radial processes were Rat.401-positive, suggesting that they form a subset of radial glia in the developing telencephalon. These findings demonstrate that during development calbindin is expressed in glial as well as neural cells. They further suggest that the radial glia associated with the GE form heterogeneous populations, the transient calbindin-positive radial glia being associated with the lateral ridge of the GE and its derivatives. Third, a scattered population of calbindin-positive cells with morphologies different from the common medium-sized calbindin- immunoreactive neurons of the striatum appeared in the dorsal and lateral striatum from about E20 to P15. Some of these cells were close to the transient calbindin-positive radial processes in the same region, but others were not.(ABSTRACT TRUNCATED AT 400 WORDS)

is a member of the superfamily of calcium-binding proteins implicated in the regulation of intracellular calcium. In the mature brain, calbindin is widely expressed in neurons of the forebrain and the hindbrain, and in the telencephalon calbindin-like immunoreactivity is particularly strongly expressed by medium-sized neurons of the striatum and by certain other neurons in the cortex and subcortex. We have traced the development of calbindin expression in the forebrain of the rat, and report here that in addition to the steady development of these calbindin-positive neuronal systems, transient waves of calbindin expression occur in cells of the ventricular zones of the basal ganglia and cortex and in cells of the telencephalic regions derived from these ventricular zones including radial glia of the developing striatum. In the striatum and its ventricular zone (the ganglionic eminence, or GE) we identified four transient calbindin-positive systems in the perinatal period. First, calbindin-immunoreactive cells began to appear in the GE by embryonic day (E)18, and by E20 an extensive dorsal and lateral part of the GE was marked by dense calbindin-like immunoreactivity in the ventricular zone. This calbindin system peaked at postnatal day (P)O-P3 and disappeared by P15. Its presence suggests that the GE is divisible on a molecular basis into lateral and medial districts that may correspond to derivatives of the lateral and medial ventricular ridges. Second, a system of calbindin-positive processes appeared in the dorsal and lateral caudoputamen with temporal and spatial distributions matching the germinal zone system. Many of these processes could be traced from calbindinpositive cells in the ventricular zone of the GE, including processes stretching across the full width of the dorsal caudoputamen.
Double-staining experiments demonstrated that these radial processes were Rat.401 -positive, suggesting that they form a subset of radial glia in the developing telencephalon.
These findings demonstrate that during devel-opment calbindin is expressed in glial as well as neural cells. They further suggest that the radial glia associated with the GE form heterogeneous populations, the transient calbindinpositive radial glia being associated with the lateral ridge of the GE and its derivatives.
Third, a scattered population of calbindin-positive cells with morphologies different from the common medium-sized calbindin-immunoreactive neurons of the striatum appeared in the dorsal and lateral striatum from about E20 to P15. Some of these cells were close to the transient calbindinpositive radial processes in the same region, but others were not. Often they formed small clusters. The absence of such cells at maturity suggests that the developing dorsolateral striatum contains a population of cells that either migrate through the region, change phenotype, or undergo cell death.
Fourth, from about PO, a system of transient calbindinpositive neuropil patches appeared, then became very prominent by P3, and finally disappeared by P15. These transient patches were in spatial register with tyrosine hydroxylasepositive "dopamine islands," forerunners of the developing striosomal system of the striatum. They were frequently, but not invariably, associated with small groups of the transient calbindin-immunoreactive cells. Thus, with respect to the neurochemical striosome/matrix compartments of the striaturn, calbindin-positive systems of two sorts are present during development: a transient one associated with dorsal striosomes, and a permanent one associated with the matrix. In the developing cortical primordium and its ventricular zone, calbindin was expressed with a prominent gradient in which the strongest early expression was dorsal and medial. Calbindin-positive cells first appeared in the ventricular zone and then appeared in the single-cell thick plexiform plate. This plate subsequently split into calbindin-positive subplate and marginal zones, known to contain transient cell populations. These bands could be followed as distinct calbindinenriched layers into the early postnatal period as calbindinpositive neurons began to inhabit the developing cortical layers in between. Thus, in the cerebral cortex, as in the developing striatum, early waves of calbindin expression in the ventricular zone are followed by the emergence of transient calbindin-immunoreactive systems that ultimately give way to the permanent calbindin-positive neuronal systems of maturity.
We conclude that the early calbindin-positive telencephalic systems identified here could participate in phases of cell proliferation, migration, and differentiation as well as in the early development of compartments in the striatum and layers in the cerebral cortex. Given the calcium-binding capacity of calbindin, it is possible that such functional participation involves the control and redistribution of calcium ions.
Calcium ions act as the targets of second messenger systems in neurons and as first and second messengers themselves (Rubin et al., 1985). They participate in critical biochemical events in mature neurons including signal transduction, neurotransmitter synthesis and release, ion channel opening, and activation of kinases, proteases, and other enzymes (for review, see Evered and Whelan, 1986;Hidaka et al., 1988). The regulation of intracellular calcium levels is thus essential for normal cellular function in the nervous system, and indeed, prolonged increases in Ca2+ have been implicated in cell death in the nervous system (Mayer and Westbrook, 1987;Choi, 1988). Calcium ions also play important roles in the molecular events underlying neural development such as cell proliferation, migration, outgrowth of growth cones, and cellular differentiation (Walaas and Naim, 1985;Takeichi, 1988;Rasmussen and Means, 1989;Rater and Mills, 199 1).
Little is yet known about how calcium ions are regulated during maturation of nervous system, but a natural approach to this problem is to trace the developmental expression of members of the superfamily of calcium-binding proteins, candidate molecules for control of intracellular calcium levels at adulthood (for review, see Welsh, 1988;Rogers, 1989). Many of these calcium-binding proteins, which include calbindin-D,,, (calbindin), calmodulin, oncomodulin, parvalbumin, S-100 proteins, troponin C, and calretinin, are expressed in regionally specific distributions in the brain and spinal cord at maturity and during ontogeny, suggesting that they function in relation to regional specifications in neurotransmission and development (Welsh, 1988).
In the study reported here, we determined the development of calbindin expression in the telencephalon, and, in particular, in the striatum and cerebral cortex. In both systems, calbindin expression characterizes subsets ofneurons at maturity. We were especially interested in the striatum, where calbindin has a focus for study because of two unique properties of its distribution. First, calbindin expression distinguishes the common mediumsized spiny neurons, the main cell type in the striatum, from other striatal neurons (DiFiglia et al., 1989). Second, the expression of calbindin distinguishes between neurons in the two main neurochemical compartments of the striatum, the striosomes and the matrix (Graybiel and Ragsdale, 1983): calbindin is constitutively expressed by medium-sized neurons in most of the striatal matrix, but it is expressed by very few striosomal neurons (Gerfen et al., 1985).
For the striatum, we found that the expression of calbindin is characterized not only by a gradual appearance of calbindin in medium-sized neurons of the striatal matrix (Liu and Graybiel,199 l), but also by four transient calbindin-positive systems including (1) a subset of cells of the germinal zone of the basal ganglia, the ganglionic eminence (GE); (2) a subset of radial glial fibers extending from the calbindin-positive part of this germinal zone into the striatum; (3) a population of aspiny cells in the caudoputamen; and (4) patches of calbindin-positive neuropil corresponding to dopamine islands, which mark the sites of future striosomes. For the developing neocortex, we show that there are also waves of calbindin expression in its ventricular zone and that calbindin is expressed by cells of the transient subplate and marginal zones as well as in a gradually emerging set of cells in the cortical plate. These patterns of expression strongly suggest that calbindin could participate in developmentally regulated cellular events in the striatum and neocortex during phases of development ranging from neurogenesis and migration to the formation of compartments and layers.

Materials and Methods
Embryos and pups from nine time-pregnant Sprague-Dawley rats (Taconic Farm) were used for brain tissue harvesting. The day of sperm positivity was counted as embryonic day (E)l and the day of birth as postnatal day (P)O. Prenatal specimens were obtained from pregnant rats deeply anesthetized by an intraperitoneal injection of sodium pentobarbital (Nembutal), and brain tissue was obtained by immersion fixation of heads of embryonic day (E) 13 embryos (n = 3) or by perfusing embryos (E14, II = 3; E15, n = 3; E16, n = 4; E18, I? = 4; E20, n = 5) through the transverse sinus or the heart. For all cases the fixative was ice-cold 4% paraformaldehyde containing 5% sucrose in 0.1 M phosphate-buffered saline (PBS) (pH 7.4). For postnatal materials, PO (n = 7) P3 (n = 12) P7 (n = 6) and P15 (n = 3) rat pups and adult rats (n = 2) were anesthetized by cooling on ice (PO, P3) or by intraperitoneal iniection of sodium nentobarbital (P7, P15) and were then peerfUsed transcardially with the same fixative. Heads (E13-E18) or brains (E20-P15) were postfixed in the same fixative at 4°C for 2-12 hr and then cryoprotected at 4°C for 24-36 hr in 20% sucrose in PBS. Brains or whole heads were cut on a freezing microtome at 40 pm (El 3-E20) or 20-30 pm (PO-adult) in the coronal plane. The adult brain tissue (n = 4) was obtained from another study (Graybiel et al., 1990).
Immunostaining was performed by peroxidase-antiperoxidase (Stemberger, 1979) or avidin-biotin-peroxidase method (Hsu et al., 1981;Vector Laboratories). Free-floating sections were pretreated with 31 H,O, and 10% methanol in 0.1 M Tris buffer saline (TBS) (pH 7.4), then rinsed sequentially in 25%, 40%, 25%, and 10% ethanol diluted in TBS. Sections were incubated in 3.3Oh or 5% normal goat serum (NGS) in TBS for 30-60 min and washed in TBS several times before being placed into primary antisera. The concentrations of primary antisera were as follows: 1: lb00-4000, 1:500, and 1:800, respectively, for rabbit polyclonal anti-calbindin-D,,, antisera kindly donated by Drs. P. C. All primary antisera were diluted with TBS containing 1% NGS and 1% normal rat serum (NRS) with or without 0.2% Triton X-100. Sections were incubated in primary antisera at 4°C for 24-48 hr. For peroxidase-antiperoxidase immunocytochemistry, sections were washed several times in TBS and incubated in 1:50 goat anti-rabbit IgG (Antibodies Incorporated) or goat anti-mouse IgG (ICN Immuno-Bioloaicals) containine 1% NGS and 1% NRS in TBS for l-2 hr and 1:30-50 rabbit or mo;se peroxidase-antiperoxidase (Stemberger-Meyer) containing 1% NGS and 1% NRS in TBS for l-2 hr, with several intervening rinses in TBS. For older tissue (PO, P3, P7), a double-bridge procedure was followed. Sections were developed in 0.05% diaminobenzidine (DAB) in 0.1 M phosphate buffer (PB) by adding H,O, to a final concentration of 0.0024%. For avidin-biotin-peroxidase complex immunocytochemistry, sections were rinsed in the same buffer, incubated in 1:500 biotinylated goat anti-rabbit IgG or horse anti-mouse IgG (Vector Laboratories) containing lob NGS or 1% normal horse serum and 1% NRS in TBS for l-2 hr. rinsed several times in TBS. and incubated in avidin-biotin complex (6 ~1 A and 6 ~1 B/ml) (Vector Laboratories) in TBS for l-2 hr. Sections were developed in 0.05% DAB containing 0.2% @(+)glucose and 0.04% NH,Cl in 0.1 M PB by adding glucose oxidase (Sigma) to a final concentration of 0.0004%. Alternating sets of serial sections from PO, P3, and P7 brains were processed for calbindin and TH immunostaining. Some sections of prenatal brains were stained with cresylecht violet to demonstrate cytoarchitecture.
Double immunostaining for calbindin and Rat.401 was carried out on sections from E20, PO, and P3 brains by first immunostaining for calbindin (Emson) with the avidin-biotin-peroxidase method as described above and then, after extensive rinsing in TBS, incubating in Rat.40 1 primary antiserum at 4°C for 18-24 hr followed by incubating in 1:200-400 goat anti-mouse IgG conjugated to Texas Red (Molecular Probes Inc.) at room temperature for 2 hr. After incubation, sections were rinsed (TBS) and mounted on slides with glycerol/PB buffer containing DABCO (5 mg/ml). Controls for double immunostaining for Rat.40 1 and calbindin were carried out by omitting the secondary antibodies in the staining procedure as described above. Cellular colocalization of calbindin and Rat.401 antigen (nestin) was studied with the aid of a confocal microscope (Bio-Rad) by comparing the calbindin staining pattern viewed in laser-Nomarski light field optics with the Rat.401 immunofluorescence pattern observed with a rhodamine filter (Bio-Rad).

Results
Calbindin-like immunoreactivity in the forebrain during early telencephalic development (E 13-E15) Calbindin-like immunoreactivity was already distributed in the ventricular zone of the El 3 telencephalon along pronounced mediolateral, dorsoventral, and caudorostral gradients ( Fig.  l&t'). The highest levels of calbindin expression were in the mediodorsal part of the ventricular zone, where the entire epithelium was intensely stained. Dorsolaterally, calbindin-immunoreactive cells in the ventricular zone were piled up in rows perpendicular to the ventricular surface, and a few tangentially oriented calbindin-immunoreactive cells began to form a thin layer (plexiform primordium or preplate; Uylings et al., 1990). Farther ventrally in the ventricular zone, calbindin-like immunoreactivity was very weak.
In the germinal zone of the GE at El 3 and E14, only a few sporadic cells (Fig. 20) and fibers ( Fig. 2A,C,D) were as strongly immunostained as cells in the more dorsal part of the ventricular zone. By E 14, the medial and lateral ridges of the GE had formed, and a few chains of weakly calbindin-positive cells were found stretching from the ventricular and subventricular zones toward the base ofthe lateral GE (Fig. 2B). These rare calbindin-positive columns were not obviously associated with other calbindinimmunoreactive elements. The lateral GE had perceptibly more diffuse immunostaining than the medial GE, but both were very pale.
Just lateral to the developing lateral GE, intensely immunoreactive cells began to appear in the developing cortical plate, and as early as E 13 the ventral part of the preplate had swollen to form a prominent group of calbindin-positive cells (Fig. 1&4:&B'). This group probably corresponds to the anlage of the pyriform cortex. Ventral to the medial GE, another prominent group of calbindin-positive cells. piled up (Fig. 1P). This medial group lay near the base of the third ventricle, apparently in the hypothalamic primordium. Some of the cells associated with this group were extremely large and highly ramified (Figs. lA', 2E). Such calbindin-positive giant cells were found as early as E13.
Starting from El 5, many tangentially oriented calbindin-immunoreactive fibers appeared in the intermediate zone of the cortical primordium, and increasing numbers of calbindin-immunoreactive fibers could be traced from the lateral cortical anlage toward the growing striatal anlage (Fig. lC,C'). When these fibers reached the striatal anlage associated with the lateral GE, they became intermingled with a group of fusiform calbindin-immunoreactive cells that were present there (Fig. 2P). A number of the calbindin-immunoreactive fibers appeared to turn at acute angles to invade the subventricular zone of the lateral GE ( Emergence of calbindin expression in the GE and in radial processes in the striatal anlage A sharp increase in expression of calbindin-like immunoreactivity in the ventricular zone of the GE began after E 15. At E 16 ( Fig. 3A), most of the GE still had very little calbindin, but a transition to heightened expression began to appear at the dorsal margin of the lateral GE. In the following days a wave of calbindin expression spread from this dorsal zone farther ventrally along the ventricular and subventricular zones of the GE, maintaining strong dorsoventral and rostrocaudal gradients in the GE as described below. Ventrally, at the foot of the lateral ventricle, a second transient wave of calbindin expression developed in the germinal epithelium lining the medial side of the lateral ventricle. A striking dorsoventral gradient of calbindinlike immunoreactivity was present in the ventricular zone of GE at El8 (Fig. 3B). By this time, lateral and medial bulges could no longer be distinguished, but in the dorsal part of the ventricular zone of the GE, throughout its anteroposterior extent, intensely calbindin-immunoreactive cells appeared singly and in vertically aligned pairs or, rarely, in multicellular radial columns (Fig. 3C). The dorsal cells of this calbindin-immunoreactive population were associated with weakly stained calbindin-immunoreactive processes that extended from the ventricular zone through the subventricular zone, across the striatal anlage, and beyond it into the deep white matter separating the developing striatum from the cortical primordium ( Fig. 3B,C).
These calbindin-immunoreactive fibers gave the impression of forming a sharp border between the pallial epithelium and white matter above, and the GE and striatal anlage below. Along more ventral parts of the ventricular zone of the GE, there were progressively fewer calbindin-positive cells, and they were progressively more restricted to the superficial part of the ventricular zone. Very little calbindin-like immunoreactivity was present in the ventral one-third to one-half the ventricular zone ( Fig.   30).
By E20 (Fig. 4A), there was intense calbindin-like immunoreactivity in the ventricular zone of the GE and a strong development of radially organized calbindin-immunoreactive processes stretching away from the epithelium into the striatal anlage. The dorsoventral gradient in calbindin expression in the germinal zone was still prominent. In the dorsal part of the ventricular zone, there were layers of intensely calbindin-immu-noreactive cells, but both the numbers and staining intensity of the epithelial cells and their depth distribution in the epithelium decreased ventrally. There was also a pronounced rostrocaudal gradient in calbindin expression in the ventricular zone. Rostrally the calbindin-positive cells were more piled up within the zone, and they extended farther ventrally than calbindin-positive cells at caudal levels.
The long calbindin-immunoreactive processes emerging from the ventricular zone were also best developed dorsally and rostrally. They appeared through most of the height of the rostra1 striatal anlage, but farther caudally they were restricted primarily to its dorsal half. In these dorsal regions, the immunostained processes stretched through the subventricular zone and across the dorsal striatal anlage, bending in parallel with the curving dorsal surface of the striatum (Fig. 4A). Some of these calbindin-immunoreactive processes could be traced from the ventricular epithelium as far as the external capsule. The dorsalto-ventral and rostral-to-caudal decline in the number of such calbindin-immunoreactive processes paralleled the decline in staining of cells in the ventricular zone. The calbindin-immunoreactive processes were best stained in sections treated with Triton X-100. Without Triton X-100 treatment, they tended to have vague outlines, could not be traced well in continuity, and had a reduced spatial extent.
The calbindin-immunoreactive radial processes were fully developed at PO. As shown in Figures 40 and 5, they formed dense parallel arrays and fascicles emerging from the ventricular epithelium, where there were numerous cells immunoreactive for calbindin, and some could be traced through the full width of the caudoputamen. The calbindin-positive germinal cells and radial processes were still most numerous in the rostra1 and dorsal caudoputamen. Even at levels where the processes were numerous, they did not appear in the ventral part of the caudoputamen in which a latticework of calbindin-immunoreactive medium-sized striatal cells was forming (Liu and Graybiel,199 1).
The calbindin-positive radial processes were still prominent at P3 (see Fig. 7A), and some occurred far enough ventrally to pass through the calbindin-positive band located at the ventrolateral edge of the rostra1 caudoputamen. At mid-striatal levels, however, only a few weakly stained calbindin-immunoreactive processes remained, and they disappeared altogether in the cau-da1 caudoputamen. The immunostaining for calbindin in the ventricular epithelium also progressively decreased from rostral to caudal and dorsal to ventral.
Demonstration that the transient calbindin-immunoreactive processes in the developing striatum form a subpopulation of radial glia To test whether the distributions of the calbindin-positive radial processes were the processes of radial glial cells in the developing striatum, we carried out double immunostaining of sections from E20, PO, and P3 brains with calbindin (provided by Dr. Emson) and Rat.401 antiserum, which is a marker for radial glial cells and precursor cells for neurons and glial cells (Hockfield and McKay, 1985;Frederiksen and McKay, 1988;Lendahl et al., 1990). Rat.401-positive radial processes emerged from the entire extent of the ventricular zone, and they were distributed throughout the developing striatum (Fig. 4B).
The double staining experiments showed unequivocally that a subpopulation of these Rat.40 1 -positive radial glia, primarily in the dorsal part of caudoputamen, also expressed calbindinlike immunoreactivity (Fig. 6). In these zones, almost all calbindin-positive processes could be shown to express Rat.401like immunoreactivity. In the zones of densest calbindin expression, the majority of Rat.40 1 -positive radial processes also expressed calbindin-like immunoreactivity, but not all Rat.40 1 -positive processes were calbindin-positive. For co-labeled processes it was not always possible to detect each antigen along the entire length of a given process. Presumably, this reflected a technical problem (e.g-incomplete antibody penetration or veering of the process out of the focal plane), but incomplete expression along the length of such processes cannot be discounted. The transient calbindin-positive cells distributed in the same regions as the calbindin-positive processes (see below) were not Rat.40 1 positive, nor were the occasional Rat.40 lpositive cells calbindin positive. These single-labeled cells served as internal controls for the double immunostaining of the radial A and A', At E13, there is a distinct dorsomedial-to-ventrolateral gradient of calbindin immunostaining in the cortical primordium (C in A'). Ventrally, two groups of calbindin-immunoreactive cells are present. One lies in the lateral cortical primordium (arrowhead, presumably pyriform cortical primordium); the second (2A arrow) is at the base of GE (see also Fig. 2-4). The GE itself has very little immunostaining. At caudal levels (A'), a group of large calbindin-positive cells with ramified processes (2E arrow) appears ventrally (presumably in the hypothalamic primordium; see also Fig. 2E). B and B', At E14, a cleft divides the GE into medial (GE@ and lateral (GEZ) bulges. The calbindin-positive pyriform cortical anlage (arrow) now appears lateral to the lateral GE. The mediolateral gradient of calbindin expression is still evident in the cortical primordium. C and C', Calbindin expression appears throughout the mediolateral extent of the El5 cortical primordium (see Fig. 8B), and many calbindin-immunoreactive fibers appear in the intermediate zone of the cortical primordium. Laterally, they extend ventrally away from the cortical primordium and appear just under the lateral GE (C'). The striatal anlage (St) now appears beneath the GE (in C). In this region, many calbindin-positive cells are intermingled with calbindin-immunoreactive fibers that appear to emerge from the cortical primordium above (see also Fig. 2fl. V, lateral ventricle; ssi, sulcus subpallii intennedius; ssd, sulcus subpallii dorsalis (see Lammers et al., 1980). Scale bar (for all photographs), 500 pm. Figure 2. Photomicrographs in A-D illustrate calbindin-immunoreactive fibers and cells in the GE. The ventricular zone is shown at the top in each photograph. A, High-magnification view of the zone indicated by arrow in Figure 1A to show a single calbindin-positive fiber (double arrows) stretching across the El 3 GE. Note its swollen end in GE. B, A column of calbindin-positive cells (curved arrows) appearing in the subventricular zone in the El4 lateral GE. C, A bifurcated calbindin-positive fiber can he traced from the ventricular zone (arrowhead) of the El5 lateral GE toward the striatal anlage that lies below. D, Two curving calbindin-positive fibers in the El 5 lateral GE in which a calbindin-immunoreactive cell (curved arrow) is also present. E, Calbindin-positive cells in the region indicated by an arrow in Figure 1A'. These calbindin-positive cells have large perikarya (-15 pm) and thick, highly ramified processes. F, High-magnification view of the zone indicated by arrow in Figure 1C  Transient appearance of calbindin-immunoreactive cells and calbindin-positive patches in the perinatal striatum As the development ofthe calbindin-positive GE cells and radial processes was reaching its peak, the two other transient systems of calbindin-immunoreactive elements in the striatal anlage appeared for the first time. The first of these was a set of calbindinimmunoreactive cells first detected at E20 primarily in the lateral striatal anlage (Fig. 4A) and found as late as P15. At E20, some of these cells seemed to be associated with the radial calbindin-positive processes described above. By PO, many calbindin-immunoreactive cells were scattered in the rostrodorsal caudoputamen and in the dorsal and lateral parts of the middle and caudal caudoputamen (Figs. 40, 5). Some cells were smallto medium-sized cells (diameter, -6-12 pm), whereas others had larger perikarya (diameter, -13-l 8 pm) with ramified immunoreactive processes lacking detectable spines. Few such larger cells were found in the ventral part of the caudoputamen. These calbindin-positive cells were scattered between and alongside the calbindin-immunoreactive radial processes, and some of the calbindin-positive cells were attached to the processes (Figs. 4C, 5). We could not, however, find a consistent relationship between the orientations of the processes of calbindin-immunoreactive cells and the calbindin-positive processes. These cells were prominent dorsally until P3, and then gradually disappeared.
The second transient calbindin-immunoreactive system was made up of calbindin-immunoreactive patches approximately 75-150 pm wide dispersed in the same regions as the calbindinimmunoreactive cells. The patches were first visible at PO (Fig.  40, peaked at P3 (Fig. 7A), and disappeared after P15. They appeared to be made up largely of a fine-fibered calbindin-immunoreactive neuropil, but many of the scattered calbindinimmunoreactive cells were also associated with these patches (Fig. 4C). Some of the calbindin-immunoreactive radial processes ran through the calbindin-immunoreactive patches. Both the scattered calbindin-immunoreactive cells and the calbindinimmunoreactive patches were prominent in the dorsal and lateral parts of caudoputamen at P3. The calbindin-positive patches were clearly set off from the now intensely immunostained permanent calbindin-positive mosaic of medium-sized striatal cells developing ventrally (Fig. 7A, C). Altogether, the transient calbindin-positive patches formed a dorsolateral system that was roughly co-distributed with the region containing the transient calbindin-positive cells.
It was necessary to treat sections with Triton X-100 to obtain clear staining of the calbindin-immunoreactive patches (Fig.  7A). Without Triton X-100 treatment, the calbindin-immunoreactive patches were at most weakly stained and in some sections not visible at all.
Demonstration that the transient calbindin-positive patches correspond to dopamine islands The system of calbindin-positive patches was strongly reminiscent of the "dopamine island system" of the developing caudoputamen even in the detail of including the rim of calbindin-immunoreactivity along the dorsolateral margin of the caudoputamen. To compare the locations of the calbindin-immunoreactive patches and dopamine islands directly, we studied serially aligned pairs and triplets of sections consecutively stained for calbindin-like immunoreactivity and for TH-like immunoreactivity (PO, n = 7; P3, n = 12; P7, n = 6; P15, n = 3) or, in P3 and P7 cases (n = 1 for each age), for three other markers of dopamine islands (DARPP-32, SV48, and CCPK II immunoreactivity) (Foster et al., 1987;Newman-Gage and Graybiel, 1988). In the PO brain, the structure of the patches changed too rapidly in adjoining sections to allow secure comparisons, but serial-section analysis was possible beginning at P3. The calbindin-immunoreactive patches were closely aligned with THpositive patches in adjacent sections, and there were few THpositive patches that lacked corresponding calbindin-immunoreactive patches (Fig. 7A,B). Similar alignments were found for the calbindin patches and islands stained for DARPP-32, SV48, and CCPK II (data not shown).
TH-positive dopamine islands were still visible at P7, and Fiaure 4. A. A coronal section through an E20 forebrain, stained for calbindin. Cortical region shown at higher magnification in Figure 8F is indicated by arrow in A. There is a pr&tounced dorsoventral gradient of calbindin expression in the ventric& and subventricular zones of the GE. Dorsally, many radial calbindin-immunoreactive processes stretch from the ventricular zone into the dorsal part of developing striatum (St), some reaching as far as the external capsule (EC). These processes are fully developed by PO. Note that by E20, numerous calbindin-immunoreactive bundles appear in the ventral striatum (see arrowhead in A). II, Pattern of immunostaining for Rat.401 in the E20 telencephalon. In contrast to calbindin-positive cells in the GE (A), which are concentrated dorsally, Rat.40 1 -positive cells are distributed throughout the germinal epithelium. Rat.40 1 -positive radial glia also appear both dorsally and ventrally in the developing striatum. The Rat.40 1 -positive radial glia extend from the ventricular zone across the developing striatum to the external capsule. C, High-magnification photomicrograph showing a patch of calbindinpositive neuropil in the dorsal PO caudoputamen. The calbindin-immunoreactive radial processes (straight arrows) arch through the calbindinpositive patch. A few calbindin-positive cells (curved arrows) are associated with the patch. D, Calbindin expression pattern in the PO striatum at the rostra1 level shown, the calbindin-positive processes have invaded more ventral parts of the caudoputamen than at earlier developmental stages. AC, anterior commissure. Scale bars: A (for A, B, and D), 500 pm; C, 20 pm.  continued alignment of the calbindin-positive patches and dorsal islands was confirmed. By P15, however, TH-like immunoreactivity in the extra-islandic matrix had greatly increased so much that only rarely could matches be attempted between the nearly faded calbindin-immunoreactive patches and the island system.
Postnatal decline of the transient calbindin-positive systems of the striatum The scattered calbindin-immunoreactive cells and patches lingered slightly longer than the calbindin-immunoreactive GE cells and radial processes. A few calbindin-immunoreactive cells and weak calbindin-immunoreactive patches were still present in the rostrodorsal caudoputamen and the dorsal and lateral parts of the middle and caudal caudoputamen at P7 (Fig. 7C9 and P15, but they were not detected in the mature caudoputamen. By P7, only a few calbindin-immunoreactive processes remained in the most rostra1 part of the caudoputamen. Calbindin-like immunostaining was also dramatically reduced in the ventricular epithelium by P7, even far rostrally. No calbindin-immunoreactive processes could be detected in any part of the caudoputamen by P15. Some of the cells of the ependyma were still stained, however, in regions adjoining calbindin-positive parts of the striatum. patches visible in A are aligned with TH-positive dopamine islands in the adjoining section (B) (e.g., see the corresponding asterisks). Note that the strong calbindin-positive patchwork in the ventrolateral striatum (double asterisks in A) is part of the developing permanent calbindin-positive mosaic system (see Liu and Graybiel,199 1). C, In the P7 brain, a few transient calbindin-positive processes, cells, and patches (e.g., asterisk) are only faintly stained. A typical calbindin-poor zone in this mosaic is indicated by the star. D, A section through the P3 caudoputamen demonstrating the specificity of the calbindin immunostaining obtained with the calbindin antiserum provided by Dr. P. C. Emson. The section was incubated with 1:4000 calbindin antiserum preabsorbed with calbindin protein purified from the rat kidney (Dr. S. Christakos). The preabsorption procedure almost completely eliminates the immunostaining in the striatum and the neocortex. AC, anterior commissure. Scale bars: El (for A and B), 500 pm; C, 500 pm; D, 500 pm.  Fig. 3B and was rotated 90" clockwise to permit comparison with other panels). A few calbindin-positive cells (example at the curved arrow in D) are present in the ventricular zone. In the same section, a clear separation of the calbindin-positive subplate and calbindin-positive marginal zones has occurred in the dorsolateral part of the cortical anlage, and many fusiform calbindin-positive cells are scattered between. A few calbindin-positive cells (example at the curved arrow in E) appear below the subplate. F, By E20, the separation between the calbindin-positive subplate and calbindin-positive marginal zones is even greater and many calbindin-positive cells with radially oriented processes are visible. Scale bar: A (for A-F), 100 pm.
Developmental segregation of the calbindin-positive cortical subplate and calbindin-positive marginal zones from a single calbindin-positive cortical primordium began to appear for the first time in the dorsal and lateral caudoputamen. They were not found in the mature caudoputamen.

Immunohistochemical controls
The wave of calbindin expression documented here for the ven-Controls for the specificity of immunostaining suggested that tricular zone of the striatum began as an earlier wave of calbin-calbindin-like immunoreactivity observed in the GE and in the din expression in the ventricular zone of the cortex subsided developing striatum and cortical plate reflected authentic cal-(see above and Fig. 1). During the E 13-E20 period, however, bindin-D,,, or a closely related antigen. Sections incubated there was intense calbindin-like immunoreactivity in the de-without primary antiserum had no immunostaining. The preabveloping cortical anlage, as documented in Figure 8, and this sorption control sections, taken from E20, PO, and P3 brains to calbindin expression appeared in transient patterns.
test the effects on each of the four transient calbindin-positive The expression of calbindin-like immunoreactivity in the cor-systems (Fig. 7D), showed that immunostaining in the GE, striatical anlage started at El 3 with a single-cell thick plexiform plate turn, and cortex was virtually abolished by the addition of calcontaining tangentially orientated calbindin-immunoreactive bindin-D 28K protein, as was all staining in the white matter (see cells (Fig. 8A). As this plexiform plate gradually expanded, more also below). Interestingly, calbindin-immunoreactive cells were calbindin-immunoreactive cells appeared, and by El 5 (Fig. 8B) still present in the basal forebrain, though compared to basal most if not all of the cells in this plate appeared immunostained.
forebrain neurons in noncontrol sections processed at the same At E16, the calbindin-positive plexiform plate remained as a time, these neurons were weakly stained. Finally, the intensity single layer dorsally (Fig. 8C'), but in the ventrolateral cortical of calbindin immunostaining in all regions decreased as the anlage it began to split into two layers packed with numerous concentration of primary antiserum decreased. calbindin-immunoreactive cells and separated by a layer with Controls for double immunostaining for calbindin and Rat.40 1 far fewer calbindin-positive cells (Fig. 3A).
showed no staining for Rat.401 in sections in which the sec-This separation process followed a ventrolateral to dorso-ondary antibody, goat anti-mouse conjugated with Texas Red, medial gradient that was very obvious at El 8. In fact, the only was omitted during sequential procedure of double staining. region lacking the split was the dorsomedial cortical anlage, Similarly, no calbindin immunostaining was present in sections where the two layers were still fused (Figs. 3B, 8D). The outer in which the secondary antibody, biotinylated goat anti-rabbit, and inner layers of calbindin-immunoreactive cells appeared to was omitted. correspond to the marginal zone and the subplate, respectively (Fig. 8E). Many fusiform calbindin-immunoreactive cells were Discussion distributed between these two layers. They did not seem to have Transient calbindin-like immunoreactivity in regionally a single (e.g., radial) orientation. The distance between the two specialized zones of GE calbindin-positive layers increased as the cortical anlage became Our findings establish that a wave of expression of calbindinincreased in size, but by E20 the intensity of calbindin immulike immunoreactivity occurs in the ventricular zone of the gannostaining in the subplate and marginal zones had decreased glionic eminence in the perinatal period. During the entire time (Fig. 8F), and it continued to decrease with age. A few calbindin-that calbindin-positive cells appear in this ventricular zone immunoreactive cells were present in these two zones in the (-El 8-P7), they are spatially concentrated in its dorsal and neocortex of adult rats, along with other calbindin-positive cor-rostra1 parts. The specificity of the immunostaining for calbintical neurons.
din-D,,, is suggested by the immunohistochemical controls, and Changing patterns of calbindin-like immunoreactivity in the the pronounced regional pattern of immunostaining was confiber bundles of the caudoputamen firmed in all brains analyzed. This differential distribution bears emphasis because it sug-During the time that the patterns of calbindin-like immunoreactivity changed in the developing cortex and striatum, a marked change in calbindin expression was apparent also in the fiber fascicles in the cortical white matter, many of which penetrated the striatum from its lateral side and then organized to form the dispersed fiber bundles characteristic of the internal capsule of the mature rodent (see Figs. 3A,B; 4A,D). Intensely calbindin-immunoreactive fibers sweeping between the developing cortical plate and the deepest part of the corona radiata became incorporated in the lateral striatum by E20. These calbindin-immunoreactive fascicles aggregated to form calbindinimmunoreactive bundles squeezed into the medial and ventral gests that the dilIerentia1 antigen expression may be related to the division of the GE into medial and lateral eminences (striatal ridges, ventricular ridges). These ridges are thought to contribute to different structures in the mammalian forebrain (Smart and Sturrock, 1979;Sidman and Rakic, 1982;Bayer and Altman, 1987;Miiller andO'Rahilly, 1988, 1990). They thus might represent lineage restriction compartments analogous to those observed in the hindbrain (Lumsden and Keynes, 1989;Wilkinson and Krumlauf, 1990; see also McMahon and Bradley, 1990;Thomas and Capecchi, 1990). Interpreted in this light, our results would point to one pattern of gene expression that might be selective for the lateral (dorsal and lateral) compartment. part of developing striatum (Fig. 4A). Curiously, the most dorsal We could not directly test this possibility in the present study of these calbindin-immunoreactive bundles, seen in cross section, because the cleft visible between the medial and lateral ridges, consisted of calbindin-immunoreactive rings around calbindin-although striking at El4 and El 5, becomes obscured before poor centers. From PO (Fig. 40) to P7, the calbindin-positive calbindin is strongly expressed in the GE. Nevertheless, the internal capsule remained visible and was mainly confined to consistent dorsolateral location of the calbindin-positive zone the ventral part of the caudoputamen, but the calbindin-like of the GE in the immediately following developmental period immunostaining had disappeared by P15. By contrast, at P15 suggests that the calbindin-rich part of this germinal zone may many weakly stained small calbindin-immunoreactive bundles correspond to the lateral ganglionic ridge.
The entire region of the striatum as well as other structures of the basal ganglia (the pallidum and the amygdala) and certain other structures of the forebrain are thought to derive from the GE. The striatum itself is thought to arise predominantly from the lateral ridge (Mtiller and O'Rahilly, 1990). Morphological observations further suggest that the GE of later embryonic stages originates by fusion of earlier-appearing medial and lateral ridges, and that these may have separate origins (Lammers et al., 1980;but see KgillCn, 195 1). The more medial ridge, according to this view, derives from the basal part of the telencephalic wall, probably near the junction of the diencephalon and telencephalon, whereas the lateral eminence derives from the lateral wall of the telencephalon. Interestingly, calbindin was expressed in the ventricular zone of the cerebral cortex as well as in the lateral part of the GE. The joint expression of calbindin could reflect their common derivation. Yet the waves of calbindin expression that we observed in these two regions were temporally and spatially distinct, and they were marked by a clear boundary at the sulcus subpallii dorsalis separating the two telencephalic germinal zones.
The idea that the restricted expression of calbindin in the GE could reflect gene-specific compartmentation of this germinal zone fmds some indirect support in two other sets of findings.
First, the homeobox gene, Distal-less, has been shown to be expressed in the GE (including both medial and lateral parts), but not in adjoining ventricular epithelia, during development in the rodent (Price et al., 1991). This specificity demonstrates unequivocally that there is gene-specific compartmentation in the germinal epithelia of the forebrain. Second, molecular specialization of the dorsal (dorsolateral) part of the GE has been noted in previous developmental studies. Both the JONES and the D 1.1 monoclonal anti-ganglioside antibodies stain the dorsal, but apparently not the ventral, part of the GE of the El8 rat brain (Mendez-Otero et al., 1988). These antibodies are thought to mark ganglioside molecules important in cell adhesion patterns during development, possibly in relation to migrations guided by radial glia. Our evidence that a transient subset of calbindin-positive radial glia is associated with the dorsolateral GE suggests that the calbindin-positive radial glia could be related to specific migratory paths of cells derived from this zone, including those of the striatum.
In the regions of densest calbindin expression, the entire ventricular zone was intensely immunostained, and no unstained cells could be detected. It is unlikely, therefore, that the transient calbindin expression was restricted to radial glia or their precursors. Very early during the onset of calbindin expression, however, did we see isolated columns ofcalbindin-positive cells in the ventricular layer of the GE. These recall the small foci of proto-oncogene expression reported by Johnston and van der Kooy (1989).
The question of what functions such intense but transient calbindin expression serves in the germinal epithelium clearly is not settled. One intriguing possibility is that during ontogeny, calbindin may participate in differential regulation of calcium ions in subdistricts of the GE and, according to our observations on the developing cortex, the germinal epithelium of the cortical plate as well. As calcium can modulate the mitotic process (Rasmussen and Means, 1989;Silver, 1990;Whitaker and Patel, 1990), calbindin, with its calcium-binding ability, may in turn differentiately regulate cell proliferation in subregions of the GE. Interestingly, S-1 OOfi protein, another member of the superfamily of calcium-binding proteins, has been shown to be a mitogen for glial cells (Selinfreund et al.,199 1). Transient population of calbindin-positive radial glia in the developing striatum The concurrent appearance and disappearance of calbindin-positive cells in the GE and radial processes in the striatal anlage, and their matching dorsorostral distributions, suggested that the epithelial cells and radial elements might be related. The fact that these calbindin-positive processes expressed Rat.40 1 strongly suggests that they are radial glia. The Rat.401 antigen nestin is an intermediate-filament protein expressed by radial glia as well as by neuronal and glial precursor cells (Hockfield and McKay, 1985;Frederiksen and McKay, 1988;Lendahl et al., 1990). Thus, our results strongly suggest that the radial glia associated with the GE are heterogeneous, falling into at least calbindin-positive and calbindin-negative populations. The physiological function of expression of calbindin in radial glia is unknown. It would be of great interest to compare the migratory patterns of cells along calbindin-positive and calbindinnegative radial glia (e.g., Gasser and Hatten, 1990). Radial glia in the embryonic brainstem and spinal cord have been shown to be positive for S-100 protein as mentioned above (Gomez et al., 1990).
There was a clear complementarity between the dorsal and rostra1 distribution of the transient calbindin-immunoreactive radial processes and the ventral and caudal location of the growing mosaic of calbindin-positive medium-sized cells. The only exception was that some calbindin-immunoreactive processes did pass through a calbindin-positive "lateral band system" (Liu and Graybiel, 1991) that lay in the ventrolateral part of the rostra1 striatum. By contrast, the distribution of the calbindinpositive radial processes matched that of the scattered transient calbindin-positive cells in the dorsal part of the developing caudoputamen. It was not possible to determine whether the instances of close contact between the processes and cells actually represent a functional association (Rakic, 197 1, 1972(Rakic, 197 1, , 1988. However, the identification of the calbindin-positive processes as Rat.401 positive is consistent with a guidance function for these radial processes. A transient population of calbindin-immunoreactive cells is present in the developing striatum Our findings demonstrate that there is a transient population of calbindin-positive cells in the developing striatum visible from E20-P15. These appeared to have neuronal morphology, but we did not succeed in obtaining double-staining evidence for this phenotype. Some cells had quite large perikarya with ramified processes, and in general they appeared to be aspiny and otherwise phenotypically distinct from the calbindin-positive neurons in the calbindin-positive mosaic developing in the ventral parts of the caudoputamen (Liu and Graybiel,199 1). The dorsal location of these transient calbindin-positive cells included the region in which little calbindin-like immunoreactivity is expressed in the mature striatum and, as noted above, largely corresponds to the region penetrated by transient calbindin-immunoreactive radial processes. Calbindin-positive cells with similar phenotypes were only very rarely identified farther ventrally, but it is conceivable that the increasing immunoreactivity of the medium-cell mosaic hid others from view.
What accounts for the disappearance of the transient calbindin-immunoreactive cells remains an open question. They may be selectively removed during the period of neuronal death in the first postnatal week (Fentress et al., 198 1;Fishell et al., 1987) or undergo a transition of neurochemical phenotype, or they may be cells migrating through the striatum during development. This last possibility is plausible, because the phenotype of the larger of these calbindin-immunoreactive cells was similar to that of calbindin-positive cells found in the adjoining developing neocortex, and in some sections the transient calbindinpositive cells seemed to be piled up along the lateral edge of the caudoputamen.
Transient calbindin-immunoreactive patches aligned with dopamine islands in the developing striatum A surprising finding in the present study is that there is a transient system of calbindin-positive patches in the striatum corresponding to dopamine islands. The calbindin-positive patches were primarily located in the dorsal and lateral caudoputamen and were visible from PO to P15. The close correspondence between these transient calbindin-immunoreactive patches and dopamine islands was clear from serial-section comparisons with all four of the immunocytochemical markers of the islandic system that we used (antisera to TH, DARPP-32, SV48, and CCPK II) (Foster et al., 1987;Newman-Gage and Graybiel, 1988).
Dopamine islands mark the sites of developing striosomes (Graybiel, 1984;Moon Edley and Herkenham, 1984;Murrin and Ferrer, 1984;van der Kooy, 1984). Calbindin-like immunoreactivity is selectively expressed in neurons and neuropil of the matrix compartment of the mature rat's striatum except for a dorsolateral zone in which very little calbindin is present (Gerfen et al., 1985). Thus, on two counts the developmental-regulated calbindin-positive patchwork is inconsistent with the later pattern of expression of calbindin: the transient calbindinpositive patches are at the sites of future striosomes, not future matrix, and they are best developed dorsolaterally, where calbindin expression will be lowest. Moreover, at the same time that the transient system of calbindin-immunoreactive patches is developing dorsally, calbindin expression develops ventrally in a steadily increasing proportion of medium-sized neurons of the matrix (Liu and Graybiel, 1991). Thus, the calbindin-immunoreactive elements of the developing caudoputamen can be divided into two systems with respect to the neurochemical compartmentation of the striatum. One is a transient calbindinpositive system located in developing striosomes, and the other one is a permanent calbindin-positive system located in the matrix.
Two possible origins of the calbindin-immunoreactive fibers in the dopamine islands were suggested by our immunohistochemical findings. First, some of the transient calbindin-immunoreactive cells scattered through the dorsolateral caudoputamen were associated with these patches and so could contribute at least part of the neuropil of the patches. The coordinate time of disappearance of the calbindin-immunoreactive cells and patches (around P15) is consistent with this idea. The slightly earlier appearance of the transient calbindin-positive cells (E20 as opposed to PO) would also fit if the cells, once in dopamine island locations, then generated processes. They might, however, simply have more antigen in the cell body than in coexisting processes.
A second possibility is that the source of the calbindin-immunoreactive neuropil in the patches is extrinsic to the striatum; for example, they could derive from calbindin-immunoreactive cells of the developing neocortex. We observed calbindin-immunoreactive cells in the developing cortical plate and the immediately adjoining subplate. In the mature rat, neurons in the deep layers ofneocortex project predominantly to the striosomal (patch) compartment of the striatum (Gerfen, 1989). Moreover, in the ferret, subplate neurons have been shown to project to subcortical regions earlier than do neurons in the deep cortical layers (McConnell et al., 1989). Later in development, neurons of the subplate are thought to undergo programmed cell death (Luskin and Shatz, 1985;Chun and Shatz, 1989). If the calbindin-immunoreactive neuropil in the dopamine islands is derived from axons of calbindin-immunoreactive subplate neurons, a disappearance of the calbindin-positive patches would be predicted: as calbindin-immunoreactive subplate neurons die during development, so would the calbindin-immunoreactive patches disappear. Calbindin-immunoreactive subplate neurons apparently are not projection neurons in the ferret (Antonini and Shatz, 1990); given our findings, it would be interesting to look for this connectivity in the rodent.
Cortical subplate and marginal zones share the property of expressing intense calbindin immunostaining following an early wave of calbindin expression in the ventricular zone Our findings demonstrate that a single-cell layer of calbindinimmunoreactive cells appears in the cortical plexiform plate as early as E 13 and that, at about E 16, this single calbindin-positive plexiform plate begins to split into two calbindin-positive zones that become the subplate and marginal zones. This developmental separation is in good accord with the developmental sequence observed by Luskin and Shatz (1985) with 3H-thymidine neuronography in the cat's visual cortex. These authors proposed that cells in the cortical subplate and marginal zones are transient cells first cogenerated in a single zone, that they appear earlier than other cells of the cortex, and that the original single zone is subsequently separated into two bands by the insertion of later-born cells constituting the cortical plate. Our results suggest that the early-forming cells share the property of calbindin immunoreactivity and that this chemospecificity is retained during an extended period of cortical development that culminates finally in a major wave of cell death in the calbindinpositive cell populations in the marginal and subplate zones. Moreover, we have shown that this developmental sequence of calbindin expression in the cortical anlage is preceded by a period of intense calbindin expression in the underlying ventricular zone. Thus, our observations collectively suggest that in the developing cortex, as in the developing striatum, there are consecutive waves of transient calbindin expression in precursor cells of their germinal epithelia and in a restricted set of cells derived from these epithelia.