THE .JOURNAL OF COMPARATIVE NEUROLOGY 325:257-270 (1992)

Neuroepithelial Cells in the Rat Spinal Cord Express Glutamate Decarboxylase Immunoreactivity In Vivo and In Vitro WU MA, TOBY BEHAR, DRAGAN MARIC, IRINA MARIC,

AND JEFFERY L. BARKER Laboratory of Neurophysiology, National Institute of Neurological Disorders and Stroke, National Institutes of Health, Bethesda, Maryland 20892

ABSTRACT It is unknown whether neuroepithelial cells in the mammalian central nervous system express neurotransmitter-synthesizing enzymes. In this study, expression of glutamate decarboxylase (GAD), the y-aminobutyric acid (GABAI-synthesizing enzyme, was examined in proliferative cells and postmitotic neuroblasts in embryonic rat spinal cord. Immunostaining coronal sections of the embryonic spinal cord with K2 antiserum, which recognizes GAD proteins encoded by the GADs7gene, revealed intensely stained neuroepithelial cells in the basal plate a t embryonic day (EI 13, in the intermediate plate between E 13-16, and last seen in the alar plate at E 16. Nissl counterstaining demonstrated that a small number of these GAD-immunoreactive cells adjacent to the neural tube lumen were mitotic. The ventral-todorsal gradient of GAD expression in precursor cells and postmitotic neuroblasts correlates anatomically and temporally with the sequential generation of motoneurons, commissural neurons, and interneurons in the dorsal horn. Some of these GAD-immunoreactive neuroepithelial cells may re-enter the mitotic cycle, while others are postmitotic neuroblasts presumably migrating to the intermediate zone to differentiate into young neurons. Double-immunostaining cells acutely dissociated from E 11-18 spinal cords with K2 and anti-bromodeoxyuridine antisera, following a bromodeoxyuridine pulse in vivo, revealed considerable numbers of DNA-synthesizing cells immunoreactive for GAD. The absolute number of double-stained cells peaked during E 12-15, coinciding with terminal cell division in most spinal neurons. These observations suggest that spinal neuronal precursors can synthesize GAD-related proteins prior to, or during, the terminal cell cycle. Although GAD immunoreactivity revealed by K2 antiserum was detected in proliferative cells and in migrating postmitotic neuroblasts, GABA immunoreactivity was never detectable in these cells. These early embryonic GADimmunoreactive neuroepithelial cells may either synthesize levels of GABA that cannot be detected immunocytochemically, and/or express enzymatically inactive GAD-related proteins. I 1 ~ 9 2Wiley-Liss, Inc. Key words: immunocytochemistry, neuroepithelium,mitosis, bromodeoxyuridine, neurofilaments

During early fetal life, the wall of the recently closed neural tube consists primarily, if not exclusively,of neuroepithelial cells, which replicate and differentiate into neurons and glia. The neurons eventually differentiate various permutations and combinations of neurotransmitter expression and coexpression. However, the point in development a t which neurons first express neurotransmitters or their synthesizing enzymes and the relationship of these early expressions to the transmitter phenotype ultimately detected in adults remain to be resolved. In the peripheral nervous system, neurotransmitter-synthesizing enzymes are detectable in neuronal precursor cells during replication (Rothman et al., '80). In contrast, neurotransmitter expression in the central nervous system (CNS) appears to be O 1992 WILEY-LISS, INC.

detectable immunocytochemically only after the neurons cease to divide, or after they have migrated from the proliferative zone to their final position (Specht et al., '81; Wallace and Lauder, '83; Lauder et al., '86; McLean and Shipley, '88; Phelps et al., '90). Recent studies, however, have shown that some proliferative cells acutely dissociated from the spinal cord express glutamate decarboxylase (GAD) immunoreactivity (Behar et al., '90). GAD is the ratelimiting enzyme in the synthesis of y-aminobutyric acid Accepted July 11,1992 Address reprint requests to Wu Ma, Laboratory of Neurophysiology, National Institute of Neurological Disorders and Stroke, NIH, Bldg. 36, Room 2C02, Bcthcsda, MD 20892.

258 (GABA),an amino acid present throughout the adult CNS, where it is thought to mediate a relatively ubiquitous form of fast synaptic transmission. Presumably, the differentiation of precursor cells into GABAergic neurons involves expression of GAD, which catalyzes GABA production from glutamate. However, previous studies reported that the immunocytochemical appearance of GAD lags far behind the appearance of biochemically and immunocytochemically detectable GABA in embryonic brain and spinal cord regions (Oertel et al., '81; Wolff et al., '84; Van Eden et al., '89; Hokoc et al., '90). One explanation for the discrepancy is that GABA may be synthesized through a polyamine pathway (Shank et al., '83; Seiler and Sarhan, '83; Hokoc et al., '90). Alternatively, previous studies may not have used antisera that recognize an embryonic form of GAD. In fact, two GAD genes have been recently identified in the rat. These genes encode at least two forms of GAD proteins that have different molecular weights, co-factor requirements, and intraneuronal distributions (Erlander et al., '91; Kaufman et al., '91). The two forms of GAD can be detected by different antisera on Western blots and in fixed tissue sections (Erlander et al., '91; Kaufman et al., '91). The anti-GAD antiserum designated as K2 antiserum (Kaufman et al., '91) recognizes GAD proteins encoded by the GADG7gene, while anti-GAD antiserum designated as 1440 antiserum preferentially binds to GAD proteins encoded by the GADs5 gene (Oertel et al., '81; Kaufman et al., '91). Recent evidence indicates that during embryogenesis the gene encoding 67 kDa GAD proteins undergoes alternative splicing (Bond et al., '90). An exon containing a stop codon is found to be spliced into embryonic GADe7 RNA transcripts. The embryonic GAD message encodes truncated GAD proteins, which are also stained by K2 antiserum on Western blots of developing rat spinal cord tissue (Schaffner et al., '90; Behar et al., '91). These findings suggest that there are two GAD genes encoding GAD proteins of different sizes, which emerge during development and play important roles in regulating GABA production. Here we sought to determine whether precursor cells in the neuroepithelium express proteins encoded by the GADG7 gene and to examine iminunocytochemically the expression of GAD proteins in proliferative cells from E l l , when the neural tube closes and significant neurogenesis begins, to E19, when neurogenesis ends (Altman and Bayer, '84). Immunocytochemical staining was carried out with the recently characterized K2 antiserum (Kaufman et al., '91; Martin et al., '91; Erlander et al., '91) together with cresyl violet counterstaining to identify GAD-immunoreactive (GAD-IR)cells with mitotic figures in vivo. GAD immunocytochemistry was also combined with simultaneous incorporation of bromodeoxyuridine (BrdU) into proliferative cells synthesizing new DNA to identify S-phase cells containing GAD proteins in cell suspensions acutely dissociated from the spinal cord. Since the generation times of neuronal and glial precursors are essentially the same (Korr, ' 8 0 ) , neuronal and precursor traits of GAD-IR cells were examined by using double- or triple-immunostaining with K2, anti-neurofilament, and anti-rat401, which recognizes an intermediate filament protein encoded by the nestin gene, expressed in neuroepithelial stem cells of rat CNS (Hockfield and McKay, '85; Frederiksen and Mckay, '88; Lendahl et al., '90). Our results indicate that GADimmunoreactivity can consistently be detected in proliferative elements in the absence of immunocytochemically detected GABA.

W. MA ET AL.

MATERIALS AND METHODS Experimental animals Timed pregnant Sprague-Dawley rats were obtained from Taconic Farms (Germantown, NY). Embryonic day 1 ( E l ) was defined as the day ofconception established by the presence of a vaginal plug. Because there are discrepancies in the designation of E l , the crown-rump lengths (CRL) of embryos within each litter were measured and used to assign each embryo a gestational age corresponding to the CRL range for each gestational day (Hebel and Stromberg, '86).

Tissue section preparation Pregnant rats were anesthetized with chloral hydrate (400 mgikg body weight, ip). Embryos (E13-19) were removed from the mother individually by Caesarean section and perfused through the heart with 4% paraformaldehyde in 0.1 M phosphate buffer (pH 7.0) following phosphatebuffered saline (PBS). Embryos were postfixed in the same fixative for 4 hours at 4°C and soaked in a 30% sucrose phosphate buffer solution at 4°C for 2 days. Coronal sections (10-12 km thick) through the neural tube were serially cut on a cryostat and placed onto poly-l-lysinecoated slides before immunocytochemical staining.

Cell suspension preparation Pregnant rats were pulse-labeled with BrdU (Sigma Chemical Go., St Louis, MO; 50 mgikg body weight, ip). Two hours later the animals were anesthetized with chloral hydrate (400 mgikg body weight, ip), and the embryos were immediately dissected out and placed in cold Earle's Balanced Salt Solution (EBSS) medium (Biofluids, Rockville, MD). Spinal cords were carefully separated from each embryo, the meninges and dorsal root ganglia teased off, and the tissues enzymatically dissociated with papain (20 Uiml, Worthington Biochemical Gorp., Freehold, NJ) according to the method described by Huettner and Baughman ('86). Single cell suspensions were obtained by gently triturating the enzyme-treated tissue. The cells were then carefully layered onto a preformed two-step discontinuous Percoll (Sigma Chemical Co., St. Louis, MO) gradient ~~

Fig. 1. A is a fluorescence photomicrograph of a coronal section through the brachial spinal cord of embryonic day (El13 rat after incubation with K2 antiserum. B, C, and D are three exposures of a single field of the basal plate in section A. B and C illustrate doublestaining with K2 (B) and anti-rat401 (C) antisera. D shows cresyl violet counterstaining. In A K2 antiserum stains many neuroepithelial cells. These glutamate decarboxylase immunoreactive (GAD-IR) cells are horizontally arrayed and project laterally within the basal plate (bp, open arrow) and the intermediate plate (ip, arrow) of the ventricular zone. Cells within the roof plate (rp, curved arrow), and putative motoneurons and other neurons within the presumptive ventral horn (VH), as well as fibers within the ventral commissure (vc), ventral funiculus (vr?, and the dorsal root entrance zone (dre) are also immunoreactive for GAD. An arrowhead points to a GAD-IR relay cell with a long, contralaterally projecting process. In C , anti-rat401 stains mostly cell processes that span the entire medial-lateral extent of the spinal cord. There are fewer rat401-IR processes detectable in the ventral horn. Arrows in B, C, and D and Figure 2 point to the same cell immunoreactive for both GAD and rat401. Small arrows in B and C indicate a double-stained epithelial cell process for GAD and rat401. In B, the GAD-IR cells in circles marked with letters a 4 aligned in the inner surface of the ventricular zone (VZ) correspond to the same cells marked a-d in D and in Figure 2. 1, lumen of the neural tube; ap, alar plate of the neuroepithelium. Scale bar in A: 60 pm; in B-C: 60 pm.

GAD IN PROLIFERATIVE CELLS OF EMBRYONIC SPINAL CORD

Figure 1

259

W. MA ET AL.

260 containing 3 ml of 50% (viv) and 4 ml of 15% (viv) Percoll-EBSS medium solutions, and centrifuged in a Sorvall swing-out rotor at 800g for 15 minutes at 18°Cwith no braking on deceleration. Cell debris, separated from the intact cells on the interface between the crude cell suspension and the 15% Percoll-EBSS step, was discarded by aspiration. The remaining spinal cord cells were carefully collected at the interface between 15% and 50%, PercollEBSS gradient steps with a Pasteur pipette, washed two times in fresh EBSS medium, two times in PBS (pH 7.41, and finally resuspended in isotonic saline solution at a density of lo6 cells/ml.

Immunocytochemistry Antisera. K2 antiserum, a polyclonal rabbit antiserum raised against a recombinant GAD fusion protein, was used for immunocytochemistry. The antiserum, shown to recognize specifically a 67 kDa GAD protein on Western blots of feline brain, was purchased from Chemicon (Temecula, CA) (Kaufman et al.,'86; Houser et al., '89). This antiserum has also been used in immunocytochemical studies and specifically labels GABAergic cells in rat and mouse tissue sections (Kaufman et al., '86, '91) and acutely dissociated cells (Behar et al., '901, without prior treatment with colchicine or other special fixation procedures. The secondary antiserum was affinity-purified rhodamine-conjugated donkey anti-rabbit IgG (Jackson Immunological Research, West Grove, PA). An IgM mouse monoclonal anti-NF, which recognizes all three neurofilament subunits used for immunostaining tissue sections, was a gift from Dr. C. Gibbs. Affinity-purified FITC-conjugated rat anti-mouse IgM or 7-amino-4-methylcoumarin-3-acetate (AMCAI-conjugated strepavidin and biotinylated donkey anti-mouse IgM were used as secondary antisera. A monoclonal mouse anti-160 kDa neurofilament subunit (NF160), used for immunostaining dissociated cell suspensions, was purchased from Boehringer Mannheim (Indianapolis, IN). Dr. S. Hockfield contributed a monoclonal antiserum to rat401, which recognizes epitopes specifically expressed in bipotential precursor cells of rat CNS. The gene encoding the proteins recognized by this antibody has been recently designated as the nestin gene (Lendahl et al., '90). The secondary antiserum, FITC-conjugated goat anti-mouse IgG, was purchased from Jackson Immunological Research (West Grove, PA). FITC-conjugated anti-BrdU monoclonal antiserum was purchased from Becton-Dickson, San Jose, CA. Immunostaining tissue sections. The staining procedure employed for tissue sections was performed with indirect immunofluorescence as previously described (Behar et al., '90). Briefly, sections were rinsed in three changes of PBS and incubated with K2 antiserum (diluted 1 5 0 in PBS), at 4°C for 48 hours. The sections were then rinsed three times in PBS and incubated with rhodamineconjugated donkey anti-rabbit IgG at a final dilution of 1:30 at room temperature for 45 minutes to 1 hour. Sections analyzed by double-immunofluorescence staining for GAD and neurofilament proteins were incubated overnight at 4°C in a mixture of one part K2 antiserum (diluted 150) and one part anti-NF hybridoma tissue culture supernatant (diluted 1:8). After PBS rinsing, the sections were incubated with a mixture of rhodamineconjugated donkey anti-rabbit IgG (diluted 1:30)and FITCconjugated goat anti-mouse IgG (diluted 1:30) antisera for 1 hour at room temperature. To double-immunostain for GAD and nestin, sections were stained sequentially with K2

Fig. 2. Higher magnification of the basal plate area shown in Figure 1D demonstrating the mitotic figures (in circles) at different stages of the cell cycle. a and c, telophase; b, prophase; d, anaphase. Cresyl violet counterstaining. Scale bar: 35 km.

and anti-rat401 antisera. GAD-stained sections were rinsed in PBS three times and incubated with full-strength tissue culture supernatant of mouse anti-rat401 for 48 hours at 4°C. The sections were rinsed in PBS again and incubated with FITC-conjugated rat anti-mouse IgG for 1 hour at room temperature.

GAD IN PROLIFERATIVE CELLS OF EMBRYONIC SPINAL CORD

261

Fig. 3. Fluorescence photomicrographs of GAD immunoreactivity in coronal sections through the basal plate of the brachial (A and B), upper thoracic (0,and lower thoracic (D) spinal cord of E l 3 rat, showing a rostral-to-caudal gradient in the number of GAD-IR cells in the neuroepithelium (open arrows). Putative young motoneurons and other neurons and their processes in the ventral horn are also stained.

Arrows in D point to immunoreactive cells in the floor plate. Small arrows in C indicate a stained fiber exiting into the ventral roots (vr).B is a substitution control in which the section adjacent to A is incubated with normal rabbit serum instead of K2, and GAD immunoreactivity is completely eliminated. Scale bar: 60 IJ-m.

Triple-immunostaining for GAD, neurofilament, and nestin proteins was carried out sequentially. Sections were first stained for GAD and nestin with procedures similar to those for double-staining GAD and nestin as described above. The double-stained sections were then incubated in the IgM mouse anti-NF for 48 hours. After being rinsed in PBS three times, t h e sections were incubated for 1 hour at room temperature in biotinylated donkey anti-mouse IgM (diluted 1:50), rinsed, and finally incubated in AMCAconjugated strepavidin (diluted 1:50). Sections were then rinsed three times in PBS and coverslipped with a mixture of glycerol and PBS (3:1), and examined with Leitz and Zeiss epifluorescence microscopes equipped with the appropriate filters to visualize FITC, RITC, and AMCA. In order to display the cytoarchitectural features and to identify mitotic activity of the spinal cord regions, coverslips were removed from some immunostained sections after photography, and the sections were rinsed in PBS, stained with 0.1% cresyl violet, dehydrated, cleared in xylene, and coverslipped.

Immunostaining acutely dissociated cell suspensions. Acutely dissociated cells were fixed in 70% ethanol for 30 minutes and stored at -20°C in 70% ethanol for up to 1 week before use. Before staining, cells were rinsed in PBS, permeabilized with 2 N HC1/0.5% Triton-X 100 (Sigma Chemical Co., St. Louis, MO) for 15 minutes at room temperature, and treated with 0.1 M Na2B407,pH 8.5, to neutralize the acid. The cells were then washed and incubated with one of the following antisera: a) K2 (rabbit anti-GADs7) antiserum (diluted 1:400 in 10% normal rat serum/0.5% TweeniPBS, pH 7.4); b) FITC-conjugated anti-BrdU monoclonal antibodies (diluted 1 : l O in 1% BSA/ 0.5% Tween 2O/PBS, pH 7.4); and c) mouse anti-NFIbo antiserum (diluted 1:5in 1%BSA/0.5% Tween BOIPBS, pH 7.4). I n double-staining experiments, the cells were simultaneously incubated with the following combinations of primary antisera at 4°C overnight: a) K2 with mouse antiBrdU antibodies and b) K2 with mouse anti-NFlho antibodies. Affinity-purified secondary antisera used for

W. MA ET AL.

262

Fig. 4. Fluorescence photomicrographs of coronal sections through the cervical spinal cord of an E l 4 rat, showing GAD immunoreactivity in the neuroepithelium (ne) (A) and substitution control (B).A cluster of epithelial cells (arrow) in the ventral intermediate plate (ip) are

stained. Another intensely stained structure is comprised of sensory nehron-derived fibers i n the dorsal root bifurcation zone (drb) or fasciculus ovalis. B is a section adjacent to section A and incubated with normal rabbit serum. Scale bar: 140 pm.

double-staining were rhodamine-conjugated donkey antirabbit IgG and FITC-conjugated goat anti-mouse IgG. The number of cells in a drop of stained cell suspension (a 30 +l aliquot) was counted on a Leitz epifluorescence microscope with a water immersion 25 x Planapo objective. The drop of cell suspension was placed on the slide and ten consecutive fields, each containing a n average 500 cells, were counted per slide; three slides were counted for each sample. Cells immunostained with K2 antiserum were counted with a rhodamine filter, and cells stained with anti-BrdU or anti-NFIGO with a fluorescein filter. The total number of cells in the same field was counted under phase contrast. The data in Figure 12 are expressed as a percentage obtained by dividing the number of single- or doublestained cells by the total number of cells counted in each field. The mean percentage of stained cells represents a n average of the results from at least three different experiments. The total number of cells dissociated from each spinal cord was calculated. The mean number of stained cells for each spinal cord was calculated by multiplying the total number by mean percentage. In control experiments for GAD immunoreactivity, frozen sections and dissociated cells of the spinal cord were incubated with normal rabbit serum or buffer without K2 antiserum. Control experiments resulted in absence of immunostaining (see Figs. 3, 4, 8 ) . I n addition, fluorochrome-conjugated secondary antibodies were screened for species cross-reactivity; none was detected.

cells of the ventricular zone (VZ) between E13-16. At E13, two distinct populations of immunoreactive cells were detected in the VZ. One was located in the basal plate (bp, open arrow) and the other in the intermediate plate (ip, arrow) of the neuroepithelium (see Figs. lA, 3A,C,D). Stained cells were sharply defined in a spindle shape and were horizontally oriented with laterally projecting processes (small arrows, Fig. IB). Immunoreaction products diffusely filled these cell bodies and processes. A significant number of GAD-IR cells were aligned along the inner surface of the VZ (Fig. 1B) and appeared, at least as revealed by GAD immunoreactivity, to have lost their basal attachment to the external surface of the neural tube. These cells had rounded u p toward the lumen where cell division usually occurred (Fig. 1B). Cresyl violet staining of the fluorescently stained sections demonstrated that some of these rounded immunoreactive cells (in circles, Fig. 1B) clearly contained mitotic figures. Most such replicating cells were in metaphase and anaphase (encircled images, Figs. 1D and 2). Other immunoreactive cells in theVZ situated at increasing distances from the lumen were attached with their long processes to both inner and external surfaces of the neural tube. Some of these cells appeared to be postmitotic neuroblasts and presumably were fixed while migrating toward the presumptive ventral horn (VH) of the intermediate zone, there destined to differentiate into young neurons in the VH (Figs. 1A,B and 3A,C,D). GAD-IR epithelial cells in the basal plate were more frequently found at the cervical level than at the lower thoracic level (Fig. 3), indicating that at E13, there is a rostral-to-caudal gradient in expression of GAD immunoreactivity in the ventricular zone. In the intermediate plate, quite a few GAD-IR cells were detected with morpholo& features similar to GAD-IR neuroblasts in the basal plate. Some of these intermediate plate cells were presumably migrating to become the contralaterally projecting commissural neu-

Expression

Of

RESULTS GAD xmmunoreactivity in viva

Neuroepithelial cells express GAD immunoreactiuit.y. Immunostaining with K2 antiserum was carried out in coronal sections through E 13-19 spinal cords. GAD immunoreactivity was consistently detected in neuroepithelial

GAD IN PROLIFERATIVE CELLS OF EMBRYONIC SPINAL CORD

263

rons (Fig. lA,B). However, GAD immunoreactivity was not prominent in the intermediate plate until E14, when the majority of immunoreactive cells were restricted to the ventral intermediate plate (Fig. 4A), the region giving rise to commissural neurons (Altman and Bayer, '84). GAD immunoreactivity in the receding VZ of the basal plate had almost disappeared. By E16, a small proportion of GAD-IR cells were detected in the receding intermediate plate (Fig. 5C,B) and alar plate (Fig. 5A) of the neuroepithelium. These laterally oriented epithelial cells gave rise to long processes projecting to the intermediate grey and formative dorsal horn, where some stained cells (arrows, Fig. 5A-C) were seen, presumably migrating from the neuroepithelium to their final positions. From E l 7 to E19, epithelial cells in the VZ were not stained. There was an apparent ventral-to-dorsal temporal gradient of GAD expression in the neuroepithelium between E l 3 to E l 6 (Fig. 7). In addition, some cells in the floor plate (Fig. 3D) and roof plate (Fig. 1A) were also clearly GAD-IR. Embryonic motoneurons and commissural neurons express GAD immunoreactiuity. Motoneurons. At E13, cells in the ventrolateral column of the spinal cord with their exiting ventral root fibers were intensely stained with K2 antiserum. In coronal sections they were distributed homogeneously throughout the presumptive ventral horn (Figs. 1A,B, 3). The morphology of immunoreactive cells, revealed by Nissl staining, was distinguishable from the GAD-IR spindle-shaped cells in the neuroepithelium. They had rounded nuclei and lightly stained cell bodies, which were parallel to the spinal cord surface (Fig. ID). From their position and morphologic features, these GAD-IR cells were identified as motoneurons that had presumably migrated radially from the proliferative zone to formative motor pools of the ventral horn. Commissural neurons. At E13, a few vertically oriented cells (arrowhead, Fig. lA,B) were stained near the dorsolatera1 wall of the spinal cord. The axons of the neurons coursed ventrally and traversed the space between the neuroepithelium and the ventricular zone, and then crossed contralaterally, forming the ventral commissure (vc) (Fig. lA,B). These decussating fibers joined the ventral funiculus (vf),which was intensely stained (Fig. lA,B). In addition, GAD immunoreactivity was also seen in fibers within paired oval-shaped bundles at sites called the dorsal root entrance zone (dre) and the dorsal root bifurcation zone (drb) or fasciculus ovalis. These zones consisted of dorsal root afferents and bifurcating fibers of the dorsal root ganglion (DRG) neurons forming the anlage of the dorsal funiculus (Figs. 1A and 4 ) . Colocalization o f GAD and nestin or GAD and NF proteins. Double-immunostaining coronal sections through E l 3 spinal cord with K2 and anti-rat401 antisera revealed coexpression of GAD and nestin proteins in the neuroepithelium. Rat401 antiserum stained most processes of epithelial cells in a palisade-like pattern throughout the

Fig. 5. Fluorescence photomicrographs of coronal sections through the upper thoracic spinal cord of E l 6 rat, showing clusters of horizontally oriented neuroepithelial cells immunoreactive for GAD in the intermediate (ip) and alar (ap) plates. Their processes project radially and can be traced into the intermediate grey (IG) and dorsal horn (DH). Some GAD-IR cell bodies are found in the intermediate zone and in apparent migration toward the IG and DH (arrows).Schematic drawing of a coronal section of E l 6 spinal cord at the top shows boxed areas indicating regions presented in A-C. The dashed line defines the border of the neuroepithelium (ne).CC, central canal; VH, ventral horn. Scale bar: 60 Fm.

264

W. MA ET AL.

Fig. 6. Triple-immunostaining of the same region within the basal plate neuroepithelium of E l 3 rat, showing the colocalization of neurofilament (A),GAD (B),and nestin (D)proteins in two cells (large arrows). Note the coexistence of three antigens in both cell bodies (large arrows) and in processes (small arrows). C is a coronal section through the E l 3

neural tube at the cervical level stained for the three antigens, showing immunostaining for nestin. Photomicrographs A, B, and D are taken from the region in a box of C. Scale bar in A, B, and D: 30 k m ; in C: 150 km.

spinal cord, but iminunoreactivity was not present in the ventral horn except a few processes running through the region (Figs. l C , 6D). In the ventricular zone, most GAD-IR cell processes were also immunoreactive for nestin, while

only a few GAD-IR cell bodies (arrow) were stained for nestin protein (Fig. lB,C), probably because rat401 antiserum preferentially stains processes rather than nuclei or cell bodies.

265

GAD IN PROLIFERATIVE CELLS OF EMBRYONIC SPINAL CORD

El3

El4

El6

Fig. 7. Diagrammatic summary of the distinctive emergence and anatomic distribution of GAD-IR cells in the neuroepithelium of the rat spinal cord during neurogenesis. The schematic drawings illustrate the ventral-to-dorsal gradient in the appearance of GAD-IR epithelial cells in the cervical cord of E13-16 rats. The solid circles represent GAD-IR

cells, the dashed line defines the border of the ventricular zone, and arrows indicate the likely migratory path of postmitotic neuroblasts immunoreactive for GAD. ap, alar plate; bp, basal plate; ip, intermediate plate; DH, dorsal horn; IG, intermediate grey; VH, ventral horn.

TABLE 1. Sources of Embryonic Rat Spinal Cord Used for Immunostaining Acutely Dissociated Cell Suspensions

spinal cord of Ell-19 rats were stained with K2 antiserum to quantify the number of GAD-IR cells. K2 antiserum stained a subpopulation of spinal cells, both large and small (Fig. 8A). Based on counts of the percentage of immunoreactive cells over the period Ell-19, two distinct phases of GAD expression were found during embryogenesis. At E l 1, when many cells are proliferating, approximately 60% of all cells were GAD-IR (Fig. 12). The percentage of stained cells markedly declined to about 30% over the next two days. After E 13, the percentage of immunoreactive cells progressively increased to about 40-50% (Fig. 12). DNA-synthesizing cells express GAD immunoreactiuity. The possibility that some GAD-IR cells actively synthesized new DNA was studied by double-immunostaining acutely dissociated cells previously exposed to a 2-hour pulse of BrdU in vivo with K2 and anti-BrdU antisera. In doublestained cells, GAD immunoreactivity was distributed diffusely throughout cell bodies, whereas BrdU immunoreactivity was restricted to nuclei (Figs. 9, 10). BrdU-IR fluorescence intensity was nearly equal among the stained cells, probably due to the relatively short pulse. Stained cells were those that incorporated BrdU during the pulse. Twenty percent of the total population acutely dissociated from the entire spinal cord at E l l , the earliest age we examined, were immunoreactive for both GAD and BrdU; 60% of the cells were BrdU-IR at this age (Fig. 12). The percentage of double-stained cells in the total population declined thereafter, parallel to the decrease in the number of BrdU-IR cells. By E19, the last day examined, only 2.4% of the population were BrdU-IR, marking the nadir in the early period of neuronal proliferation (Fig. 12, Nornes and Das, '74; Altman and Bayer, '84). At this time, there were no double-stained cells. Although the percentage of doublestained cells was higher at E l l than E13-14, the absolute number of double-stained cells was lower at E l l since at E l l there were considerably fewer cells composing the spinal cord (Fig. 13). The total number of double-stained cells peaked at E12-15, coinciding with the time at which most spinal neurons withdraw from the terminal cell cycle (Nornes and Das, '74; Altman and Bayer, '84). Thereafter, the number of double-stained cells decreased and at E l 8 only 0.5%of the population were double-stained.

Number of animals Embryonic age (day) 11

12 13 14 15 16 17 18 19

Crown-rump length of embryos (mm)

Mothers

Embryos

5 5 6 2 5

30

2.0

32 52 24

3

26 13 17 13

4.5 6.5 8.5 11.6 14.0

2 2

1

62

16.5

20.5 25 n

Double-immunostaining coronal sections of E 13 spinal cord with K2 and anti-NF antisera resulted in a few double-stained cells in the ventricular zone and many double-stained cells in the VH. In the VZ, double-stained cells were mostly detected in the basal plate (Fig. 61, and in the VH, all GAD-IR cells were immunoreactive for neurofilament protein. Colocalization o f GAD, NF, and nestin immunoreactivity in neuroepithelial cells. GAD-IR cells were examined in vivo to determine if they expressed neurofilament proteins, which are reportedly found only in neurons, and nestin, a cell marker for neuroepithelial stem cells. Tripleimmunostaining with K2, anti-NF, and anti-rat401 was performed in coronal sections through the E l 3 cervical cord. A significant number of GAD-IR cells in the VZ were stained for both NF and nestin. Such triple-stained cells were detected only in the basal plate and were scattered throughout the mediolateral extent of the ventricular zone. They were wedge-shaped and possessed cytoplasmic processes extending from the lumen to the external limiting membrane (Fig. 6A,B,C). In addition, most GAD-IR cells and processes in the VH were immunoreactive for NF but not for rat401.

Quantifying GAD immunoreactivity in acutely dissociated cells Developmental changes in the relative number of GAD-IIi cells. Cell suspensions acutely dissociated from the whole

266

W. MA ET AL.

Fig. 8. Fluorescence photomicrographs of acutely dissociated cells derived from E l 5 spinal cord and incubated with K2 antiserum (A)and normal rabbit serum (B).A: K2 antiserum intensely stains both large (arrow) and small (curved arrow) spinal cells in a diffuse manner. B: Normal rabbit serum substituted for K2 antiserum does not lead to inimunofluorcscentcells. Scale bar: 20 pm.

To examine the relationship between the ratio of cells double-stained for BrdU and GAD to total cells stained for BrdU and developmental ages, a correlational analysis was undertaken to compare individual ages with the corresponding BrdU- and GAD-IRIBrdU-IR cells ratios. As seen in Figure 14, there was a very high degree of correlation between ages and ratios. The correlation coefficient obtained from linear regression analysis was 0.93. Thus, the ratio of GAD-IR cells in the proliferative population (BrdU-IR) increases in a linear manner, i.e., the older animals had corresponding increased BrdU- and GAD-IRI BrdU-IR cells ratios, although the number of cells doublestained for BrdU and GAD was lower. This may be attributed to a dramatic drop in the number of BrdU-IR cells and an increase in the number of GAD-IR cells in older rats (Fig. 12). Colocalization of GAD and NFlti@ Double-immunostaining cell suspensions with K2 and anti-NFIGO antisera demonstrated a subpopulation of GAD-IR cells immunoreactive for NFIBoprotein in all ages we examined (Fig. 11).The NFIBoimmunoreactivity was distributed in the cytoplasm

as a perinuclear ring with intensely stained neurites (Fig. 1lC). NFIBO immunoreactivity was first observed at E l l in approximately 2% of all spinal cord cells, while half the NFIGO-IR cells were immunoreactive for GAD. Between E l 2 and E19, the number of NFIGO-IRcells gradually increased from 4% to 20% of the total cells, while cells double-stained for NFIBoand GAD increased 5-fold from 2.1%,to 10.6% of the total. The double-stained cell bodies tended to be larger in diameter in older embryos.

Fig. 9. Double-immunostaining with a mixture of K2 (B) and anti-BrdU (Cj antisera of a cell acutely dissociated from E l 3 rat spinal cord, showing that GAD and BrdU immunoreactivity are distributed in different cellular compartments. A Phase-contrast micrograph of the double-stained cell reveals a large nucleus with a basal cytoplasmic cap (arrow). B: GAD immunoreactivity in the cell, visualized by using K2

antiserum and rhodamine-conjugated donkey anti-rabbit IgG, is diffusely distributed in the entire cell body and more intensely in the cytoplasmic cap. C: The same cell photographed with a fluorescein filter shows that BrdU immunoreactivity is restricted within the nucleus (Nu), which is more concentrated in the nucleolus (arrow). Scale bar: 20 pm.

DISCUSSION In this study, immunocytochemical localization of GAD proteins with K2 antiserum in frozen sections of embryonic spinal cord and quantitatively combined GADiBrdU immunocytochemistry of acutely dissociated spinal cells revealed that GAD proteins encoded by the GADB7gene are expressed in both mitotic and DNA-synthesizing cells of the rat spinal cord. These observations, together with the known temporal and spatial patterns of neurogenesis in the spinal cord, suggest that the neuronal precursor cells in the

GAD IN PROLIFERATIVE CELLS OF EMBRYONIC SPINAL CORD

267

Fig. 10. Representative cells double-stained for nuclear BrdU and cytoplasmic GAD immunoreactivity, showing the colocalization of the two immunoreactivities in cell suspensions acutely dissociated from E l 2 rat spinal cord. Three exposures of a representative field demon-

strate phase contrast (A) and immunostaining with K2 (B) and anti-BrdU (C) antisera. Note that three cells in the field are stained for GAD proteins and two of them are also stained for BrdU. Arrows in A indicate two double-stained cells. Scale bar: 30 pm.

rat spinal cord can synthesize GAD proteins prior to, or during, the terminal cell cycle. Immunostaining coronal sections through the E13-16 spinal cord revealed a detectable number of mitotic cells and considerable number of postmitotic neuroblasts that were GAD-IR. It appears that GAD proteins are expressed in cells before they leave the VZ. At E13, the immunostained, mitotic cells and neuroblasts were predominantly present in the basal plate. Since the basal plate is the source of ventral horn motoneurons (Altman and Bayer, '841,

these immunoreactive cells are most likely motor and other VH neuronal precursors. Results from birthdating experiments with "-thymidine autoradiography in the rat cervical spinal cord show that motoneuronal precursors actively synthesize DNA between days E l l and E l 3 (Nornes and Das, '74; Altman and Bayer, '84). Our results from immunostaining dissociated cells indicate a large proportion of spinal cells double-stained by K2 and anti-BrdU antisera. Many proliferative GAD-IR cells are likely to be motoneuronal precursors. In the intermediate plate, the region of

Fig. 11. Representative cells of E l 3 spinal cord showing colocalization of GAD and NFleoimmunoreactivity. Three micrographs of a single field demonstrate phase contrast (A)and immunostaining for GAD (B) and NFlso(C).Arrows in A indicate two double-stained cells. Scale bar: 30 km.

268

W. MA ET AL.

-

....0..GAD-IR

.

0.9-

BrdU-IR BrdU- and GAD-IR

0.8-

T

0.70.60 .5 0.4-

0.3-

0.20.1 0 1

10

1

I

I

I

I

I

1

I

I

11

12

13

14

15

16

17

18

19

Embryonic age (days)

10 1 1 12 1 3 1 4 1 5 16 17 18 19 20

Embryonic age (days) Fig. 12. Developmental changes in the mean percentage of cells immunoreactive for GAD, BrdU or for both in cell suspensions acutely dissociated from day EllLE19 rat spinal cords. Each point represents the mean 5 S.D. from the number of animals shown in Table 1.

10

11

12

13

14

15

16

17

18

19

Embryonic age (days) Fig. 13. Developmental changes in the number of GAD-IR proliferating cells demonstrated by double-irnmunostaining with K2 and anti-BrdU antisera of cell suspensions acutely dissociated from E l 1E l 9 spinal cord. The absolute number of double-stained cells is most frequent during days E l 2 and E15.

commissural neuron derivation, GAD-IR cells are detected between E13-16, which coincides with the production time of commissural neurons in the intermediate zone (Altman and Bayer, '84). GAD-IR cells in the alar plate are the last to be detected at E16, when dorsal horn neuron production ends (Altman and Bayer, '84). These observations indicate that over Ell-16, GAD-IR expression in the VZ occurs along a ventral-to-dorsal gradient. These findings in sections are complemented by our results from combined immunostaining cell suspensions acutely dissociated from the whole spinal cord for GAD and BrdU. Although a high proportion of GAD-IR cells contain-

Fig. 14. The percentage of GAD-IR cells in BrdU-IR population increases linearly during Ell-18. Correlation between the ratio of cells double-stained for BrdU and GAD to cells stained for BrdU and developmental ages is significant. Correlation coefficient (r) is 0.93;P < 0.005.

ing BrdU-labeled nuclei are found at day E 11, the greatest number of such double-stained cells are detected between E l 2 and E l 5 (Fig. 131, which coincides with the peak time of neuronal production in the spinal cord (Sims and Vaughn, '79; Altman and Bayer, '84). Moreover, at E l l , as many as 60% of spinal cells are immunoreactive for GAD proteins. Since the majority of cells in the neural tube at this age are replicating, GAD-IR daughter cells may re-enter the mitotic cycle, indicating that GAD proteins are expressed in cells prior to their terminal division. We conclude that neuronal precursor cells in the rat spinal cord synthesize GAD proteins encoded by the GAD67 gene prior to, or during their terminal cell cycle. Coexpression of GAD and NF identifies the subpopulation of GAD-IR spinal cord cells as presumptive neurons. However, since the neuroepithelium surrounding the lumen of the neural tube has a mosaic organization specified to produce different classes of cells; GAD-IR cells in this region may not all be neuronal. Some GAD-IR cells in the roof and floor plates (Figs. lA, 3D) may be precursors of neuroglial cell types, rather than neurons (Altman and Bayer, '84). Furthermore, the coexpression of GAD and nestin indicates that GAD proteins can be expressed in the neuroepithelial stem cells, which will ultimately differentiate into specific neuronal and glial cell types. Colocalization of GAD, NF, and nestin proteins in the neuroepithelial cells revealed by triple-immunostaining shows that a subset of GAD-IR cells simultaneously express NF and nestin proteins. This is not surprising in light of reports by others that both NF and nestin are expressed in precursor cells (Bennett and DiLullo, '85; Frederiksen and McKay, '88).By using immunocytochemistry, 3H-thymidine autoradiography, and cell sorting, neuronal precursor cells in the rat neural tube have been shown to express nestin (Frederiksen and McKay, '88). Previous studies of neurotransmitter expression in the CNS have demonstrated a considerable variation in delay between the time of the final cell division and the commitment to a specific neurotransmitter cell type, which depends on the transmitter cell types and brain regions. For

GAD IN PROLIFERATIVE CELLS OF EMBRYONIC SPINAL CORD example, dividing cells in the spinal cord are not yet committed to the cholinergic phenotype since choline acetyltransferase is not detected until 2-3 days after onset of their generation (Phelps et al., '90). Tyrosine hydroxylase (TH) is detected only after the cells have completed their final cell division in the olfactory bulb (McLean and Shipley, '88). On the other hand, catecholamine biosynthetic enzymes in the sympathetic chain are expressed while neurons are in mitosis but not in the CNS (Rothman et al., '80). Here we have found that GAD proteins encoded by GADs7 gene, different from these transmitter-synthesizing enzymes, can be expressed before or during neuronal withdrawal from the terminal cell cycle. GAD proteins are not the only ones detected in precursors of the CNS; other cell-specificmarkers, such as neurofilament proteins (Bennett and DiLullo, '85) and glial fibrillary acidic protein (Levitt and Rakic, '80; Levitt et al., '811, are also found in some dividing cells of the VZ. Neurofilament proteins are immunocytochemically detected in significant numbers of mitotic cells of embryonic chick spinal cord (Bennett and DiLullo, '85). These immunoreactive mitotic cells are present in the mid-to-dorsal basal plate and the dorsalmost alar plate during E3 and E4, coinciding with the production of lateral column motoneurons in the basal plate (Hollyday and Hamburger, '77). This suggests that a subpopulation of ventral spinal cord neurons expresses neurofilament proteins during their terminal cell cycle (Bennett and DiLullo, '85). A surprising result is that during the earliest period ( E l l ) of neurogenesis in the spinal cord when most epithelial cells in the neural tube are replicating and prior to the appearance of significant numbers of postmitotic neuroblasts, more than half of the cells in the neural tube at the cervical level are already immunoreactive for GAD proteins (Fig. 12). The number of GAD-IR cells decreases over Ell-13, suggesting that there is an early transient appearance of GAD proteins in neuroepithelial cells. We do not know the significance of the first wave of GAD expression (Ell-E13), but a similar phenomenon is also found in neurofilament expression within the forebrain, hindbrain, and optic vesicle of embryonic chicks, where all neuroepithelial cells express NF-M (145-160 kDa) proteins at very early stage, prior to the presence of postmitotic neuroblasts. These immunoreactive epithelial cells become negative during the next 1-2 days until postmitotic neuroblasts appear (Bennett and DiLullo, '85).In the petrosal ganglion of embryonic rats, TH immunoreactivity is first detected in a large number of precursor cells at E12.5, and then declines sharply to near zero by E15.5. However, a second wave begins a t E16.5, and by postnatal day 1, adult numbers of TH-IR cells are present. The two temporally distinct waves of TH expression suggest that differentiation of two separate populations of petrosal ganglion neurons with only the second population of TH-IR cells persisting into the adult (Katz and Erb, '90). The two waves of GAD expression in the embryonic spinal cord may be involved in two temporally separate populations of GAD-IR neuroepithelial cells. The first population is entirely transient, while a subpopulation in the later wave differentiate into the GABAergic neurons. This study has demonstrated GAD immunoreactivity in many if not all embryonic motoneurons of the presumptive ventral horn. GAD and GABA have not been reported to occur in motoneurons of the adult rat spinal cord (McLaughlin et al., '75; Hunt et al., '81; Vaughn et al., '81; Barber et

269

al., '82; Magoul et al., '87; Todd and Sullivan, '901, although nonspecific GAD-immunoperoxidase staining was found in motoneurons and was caused by excessive concentration of glutaraldehyde in the fixative (Vaughn et al., '81). Thus, there is a transient expression of GAD proteins in motoneurons during embryonic spinal cord development, the significance of which is unknown. However, our observations indicate that onset of GAD protein expression coincides with the period of spinal neuronal proliferation and differentiation. There is a ventral-to-dorsal gradient of GAD expression in neuroepithelial cells, which correlates anatomically and temporally with the sequential generation of motoneurons, commissural neurons and interneurons in the DH (Altman and Bayer, '84). Furthermore, GAD expression occurs concomitantly with neurite outgrowth, including immunoreactive fibers within the ventral roots derived from motoneurons, the ventral commissure from lateral projecting neurons, and the dorsal root bifurcation zone from the DRG cells (Figs. lA, 3, 4). It is also significant to note that GAD immunoreactivity is present in the floor plate cells (Fig. 3D), which have been suggested to control neuronal differentiation along the dorsal-ventral axis of the neural tube (Yamada et al., '91). This implies that GAD proteins detected during embryogenesis may promote and regulate the generation and differentiation of neurons in ventricular and intermediate zones. Recent evidence has shown that GAD immunoreactivity is found outside the nervous system, such as endocrine cells of the oxyntic and pyloric mucosa of the rat stomach and in islet cells of the rat and mouse pancreas, corresponding only to insulin-secretingcells (Gilon et al., '91). GAD and GABA can be stored around synaptic-like microvesicles and released from pancreatic p-cells (Reetz et al., '91). In these cells GAD and GABA may have functions other than fast communication at synapses. There has been some discrepancy about the timing of the initial appearance of GAD immunoreactivity in the rat CNS. Most previous immunocytochemical studies involving GAD expression during embryogenesis have used 1440 antiserum (Oertel et al., '81; Wolff et al., '84; Van Eden et al., '89; Hokoc et al., '90). These studies have reported that GABA immunoreactivity is detectable prior to the appearance of GAD immunoreactivity in many brain regions and the spinal cord. Here, we have demonstrated that GAD immunoreactivity, revealed by K2 antiserum, is widely distributed in cells of the neuroepithelium as early as E l l , while GABA immunoreactivity cannot be detected in cell bodies until E l 4 (Ma et al., accompanying paper). The GAD proteins revealed in precursor cells and neuroblasts by K2 antiserum may synthesize levels of GABA that cannot be detected immunocytochemically, or the GAD proteins may be low-molecular weight, non-functional products encoded by embryonic transcripts of the GADs7 gene (Bond et al., '90). In fact, we have found that the K2 antiserum stains low-molecular weight GAD proteins in Western blots, which may be truncated products of the GADs7gene (Behar et al., '91). Because the stop codon in the GAD67 gene lies upstream from the cofactor-binding domain of the enzyme, truncated GAD proteins are presumably enzymatically inactive. Thus, neurons in the developing spinal cord may sequentially express several forms of GAD proteins with low molecular weight GAD proteins being expressed first. This may explain why the early embryonic neurons are immunoreactive for K2 antiserum, but not for GABA antiserum. Alternatively, GABA may be synthesized by polyamine pathways, which may generate immunocy-

270

W. MA ET AL.

tochemically undetectable levels of GABA. Thus, the later appearance of GABA at E l 4 in cells of the ventral portions of the cervical spinal cord may mark the emergence of full-length, functional GADs7gene product,s.

ACKNOWLEDGMENTS The authors thank Dr. James Vaughn for his critical appraisal of the manuscript, Lisa Chang for her skillful technical assistance, and Devera Schoenberg for editing the manuscript.

LITERATURE CITED Altman, J., and S.A. Bayer, eds. (1984) The Development of the Rat Spinal Cord. Berlin: Springer Verlag. Barber, R.P., J.E. Vaughn, and E. Roberts (1982) The cytoarchitecture of GABAergic neurons in rat spinal cord. Brain Res. 238~305-328. Behar, T., L. Hudson, S. Komoly, W. Ma, and J.L. Barker (1991) Embryonic GAD stop codon message disappears coincident with the emergence of GABA immunoreactivity in developing rat CNS. Soc. Neurosci. Abs. 17:304.10. Behar, T., A.E. Schaffner, and J.L. Barker (1990) Immunocytochemical analysis of GAD expression in cells of the developing rat spinal cord. J. CellBioLAbs. 111:1361. Bennett, G.S., and C. DiLullo 11985) Expression of a neurofilament protein by the precursors of a subpopulation of ventral spinal cord neurons. Dev. Biol. I07:94-106. Bond, R.W., R.J. Wyborski, and D.I. Gottlieb (1990) Developmentally regulated expression of an exon containing a stop codon in the gene for glutamate acid decarboxylase. Proc. Natl. Acad. Sci. U.S.A. 87:8771-8775. Erlander, M.G., N.J.K. Tillakaratne, S. Feldblum, N. Patel, and A.J. Tobin (1991) Two genes encode distinct glutamate decarboxylase. Neuron 7:91100. Frederiksen, K., and R.D.G. McKay 11988) Proliferation and differentiation of rat neuroepithelial precursor cells in vivo. J. Neurosci. 8:1144-1151. Gilon, P., M. Tappaz, and C. Remacle 11991) Localization of GAD-like immunoreactivity in the pancreas and stomach of the rat and mouse. Histochemistry 96355-365. Hebel, R., and M.W. Stromberg (1986) Anatomy and Embryology of the Laboratory Rat. Worthsee: BioMed Verlag. Hockfield, S., and R.D.G. Mckay (1985) Identification of major cell classes in the developing mammalian nervous system. J. Neurosci. 53310-3328. Hokoc, J.N., A.L.M. Ventura, P.F. Gardino, and F.G. De Mello (1990) Developmental immunoreactivity for GABA and GAD in the avian retina: Possible alternative pathway for GABA synthesis. Brain Res. 532197-202. Hollyday, M., and V. Hamburger (1977) An autoradiographic study of the formation of'the lateral motor column in the chick embryo. Brain Res. I32:197-208. Houser, C.R., J.E. Miyashiro, D.L. Kaufman, and A.J. Tobin 11989)Immunocytochemical studies using a new antiserum against bacterially produced feline glutamate decarboxylase. Soc. Neurosci. Abstr. 15488. Huettner, J.E., and R.W. Baughman 11986) Primary culture of identified neurons from visual cortex of postnatal rats. J. Neurosci. 6-3044-3060. Hunt, S.P., J.S. Kelly, P.C. Emson, J.R. Kimmel, R.J. Miller, and J.-Y. Wu (1981) An immunocytochemical study of neuronal populations containing neuropeptides or y-aminobutyric acid within the superficial layers of the dorsal horn. Neuroscience 6: 1883-1989. Katz, D.M., and M.J. Erb (1990) Developmental regulation of tyrosine hydroxylase expression in primary sensory neurons of the rat. Dev. Biol. 137:233-242. Kaufman, D.L., C.R. Houser, and A.J. Tobin 11991) Two forms of the y-aminobutric acid synthetic enzyme glutamate decarboxylase have distinct intraneuronal distributions and cofactor interactions. J. Neurochem. 56:720-723. Kaufman, D.L., J.F. McGinnis, N.R. Krieger, and A.T. Tobin 11986) Brain glutamate decarboxylase cloned in lambda gl-11: fusion protein produces y-aminohutyric acid. Science 2321138-1140. Korr, H. 11980) Proliferation of different cell types in the brain. Adv. Anat. Embryol. Cell Biol. 6l:l-72. Lauder, J.M., V.K.M. Han, P. Henderson, T. Verdoorn, and A.C. Towle (1986) Prenatal ontogeny of the GABAergic system in the rat brain: An immunocytochemical study. Neuroscience 2465-493.

Lendahl, U., L.B. Zimmerman, and R.D.G. McKay 11990) CNS stem cells express a new class of intermediate filament protein. Cell 60:585-595. Levitt, P., M.C. Cooper, and P. Rakic (1981) Coexistence of neuronal and glial precursor cells in the cerebral ventricular zone of the fetal monkey: An ultrastructural immunoperoxidase study. J. Neurosci. 1.27-39. Levitt, P., and P. Rakic (1980) Immunoperoxidase localization of glial fibrillary acidic protein in radial glial-cells and astrocytes of the developing rhesus monkey brain. J. Comp. Neurol. 193:815-840. McLaughlin, B.J., R. Barber, K. Saito, E. Roberts, and J.-Y. Wu (1975) Immunocytochemical location of glutamate decarboxylase in rat spinal cord. J . Comp. Neurol. 164:305-322. McLean, J.H., and M.T. Shipley (1988) Postmitotic, postmigrational expression of tyrosine hydroxylase in olfactory bulb dopaminergic neurons. J. Neurosci. 8:3658-3669. Ma, W., T. Behar, and J.L. Barker (1992) Transient expression of GABA immunoreactivity in the developing rat spinal cord. J. Comp. Neurol. 325:271-290. Magoul, R., B. Onteniente, M. Geffard, and A. Calas (1987) Anatomical distribution and ultrastructural organization of the GABAergic system in the rat spinal cord. An immunocytochemical study using anti-GABA antibodies. Neuroscience 20: 1001-1009. Martin, D.L., S.B. Martin, S.J. Wu, and E. Espina (1991) Regulatory properties of brain glutamate decarboxylase (GAD): The apoenzyme of GAD is present principally as the smaller of two molecular forms of GAD in brain. J. Neurosci. 11:2725-2731. Nornes, H.O., and G.D. Das (1974) Temporal pattern of neurogenesis in spinal cord of rat. I. An autoradiographic study-time and sites of origin and migration and settling patterns of neuroblasts. Brain Res. 73:121-138. Oertel, W.H., D.E. Schmechel, E. Mugnaini, M.L. Tappaz, and I.J. Kopin 11981) Immunocytochemical localization of glutamate decarboxylase in rat cerebellum with a new antiserum. Neuroscience 6:2715-2735. Phelps, P.E., R.P. Barber, L.A. Brennan,V.M. Maines, P.S. Salvaterra, and J.E. Vaughn 11990) Embryonic development of four different subsets of cholinergic neurons in rat cervical spinal cord. J. Comp. Neurol. 291:9-26. Reetz, A,, M. Solimena, M. Matteoli, F. Folli, K. Takei, and P. De Camilli 11991) GABA and pancreatic B-cells: Colocalization of glutamate acid decarboxylase (GAD) and GABA with synaptic-like microvesicles suggests their role in GABA storage and secretion. EMBO J. 10:1275-1284. Rothman, T.P., L.A. Specht, M.D. Gershon, T.H. Joh, G. Teitelman, V.M. Pickel, and D.J. Reis 11980) Catecholamine biosynthetic enzymes are expressed in replicating cells of the peripheral but not central nervous system. Proc. Natl. Acad. Sci. U.S.A. 77:6221-6225. Schaffner, A.E., T. Behar, and J.L. Barker (1990) Immunoblot analysis of GAD proteins in the developingrat spinal cord. J. Cell Biol. Abs. 111:1360. Seiler, N., and S. Sarhan (1983) Metabolic routes of GABA formation in the chick embryo brain. Neurochem. Int. 5:625-633. Shank, R.P., and G.L. Campbell (1983) Metabolic precursors of glutamate and GABA. In L. Hertz, E. Krammer, E.G. McGeer and A. Schousboe (eds): Glutamine, Glutamate and GABA in the Central Nervous System. New York: Alan R. Liss, pp. 355-369. Sims, T.J., and T.E. Vaughn (1979) The generation of neurons involved in an early reflex pathway of embryonic mouse spinal cord. J. Comp. Neurol. 183.707-719. Specbt, L.A., V.M. Pickel, T.H. Joh, and D.J. Reis 11981) Light microscopic immunocytochemical localization of tyrosine hydroxylase in prenatal rat brain. I. Early ontogeny. J. Comp. Neurol. 199,233-253. Todd, A.J., and A.C. Sullivan (1990) Light microscope study of the coexistence of GABA like and glycine-like immunoreactivities in the spinal cord of the rat. J. Comp. Neurol. 296:496-505. Van Eden, C.G., L. Mrzljak, P. Voorn, and H.B.M. Uylings (1989) Prenatal development of GABAergic neurons in the neocortex of the rat. J. Comp. Neurol. 289:213-227. Vaughn, J., R. Barber, C. Ribak, and C. Houser (1981) Methods for the immunocytochemical localization of proteins and peptides involved in neurotransmission. In J.E. Johnson led): Current Trends in Morphological Techniques, Val. 111. Florida: CRC Press, Boca Raton, pp. 41-43. Wallace, J.A., and J.M. Lauder (1983) Development of the serotonergic system in the rat embryo. Brain Res. Bull. 10:459-479. Wolff, J.R., H. Bottcher, T. Zetzsche, W.H. Oertel, and B.M. Chronwall 11984) Development of GABAergic neurons in rat visual cortex as identified by glutamate decarboxylase-like immunoreactivity. Neurosci. Lett. 47207-212. Yamada, T., M. Placzek, H. Tanaka, J. Dodd, and T.M. Jessell 11991)Control of cell pattern in the developing nervous system: Polarizing activity of the floor plate and notochord. Cell 64535-647.