Expression of EAAT2 in Neurons and Protoplasmic Astrocytes during Human Cortical Development

Abstract

The major regulators of synaptic glutamate in the cerebral cortex are the excitatory amino acid transporters 1–3 (EAAT1–3). In this study, we determined the cellular and temporal expression of EAAT1–3 in the developing human cerebral cortex. We applied single- and double-label immunocytochemistry to normative frontal or parietal (associative) cortex samples from 14 cases ranging in age from 23 gestational weeks to 2.5 postnatal years. The most striking finding was the transient expression of EAAT2 in layer V pyramidal neuronal cell bodies up until 8 postnatal months prior to its expression in protoplasmic astrocytes at 41 postconceptional weeks onward. EAAT2 was also expressed in neurons in layer I (presumed Cajal-Retzius cells), and white matter (interstitial) neurons. This expression in neurons in the developing human cortex contrasts with findings by others of transient expression exclusively in axon tracts in the developing sheep and rodent brain. With western blotting, we found that EAAT2 was expressed as a single band until two postnatal months after which it was expressed as two bands. The expression of EAAT2 in pyramidal neurons during human brain development may contribute to cortical vulnerability to excitotoxicity during the critical period for perinatal hypoxic-ischemic encephalopathy. In addition, by studying the expression of EAAT1 and EAAT2 glutamate transporters it was possible to document the development of protoplasmic astrocytes.

INTRODUCTION

The complex clinical abnormalities resulting from prematurity are now viewed as a consequence of combined gray and white matter insults, termed the encephalopathy of prematurity (Pierson et al., 2007, Volpe, 2008, Kinney and Volpe, 2009, Volpe, 2009). Neuroimaging studies indicate reduced cortical volumes in premature infants surviving into childhood (Counsell et al., 2003, Inder et al., 2003, Miller et al., 2003, Hamrick et al., 2004). In preterm infants with periventricular leukomalacia, a 38% reduction in the density of Layer V pyramidal neurons compared to controls was found (Andiman et al., 2010). Cortical gray matter injury due to excitotoxicity may underlie cognitive impairments in survivors of prematurity with or without cerebral palsy. Yet, despite the critical need to understand the underlying basis of excitotoxic damage in the developing human cortex, virtually nothing is known about the normative developmental profile of its glutamate transporters—the key regulators of synaptic glutamate. The objective of the following study was to determine the cellular and temporal expression of glutamate transporters in the normative cerebral cortex in the preterm and neonatal infant in relationship to the anatomic basis and maturational time-table of vulnerability to perinatal injury. We tested the hypothesis that changes occur in the developmental profile of glutamate transporters during the period of risk for perinatal excitotoxic injury such that immaturity in this profile may contribute to the cortex’s susceptibility to excitotoxicity during this time-frame. Additionally, since little is known about astrocytic development in the human cortex over the same time-frame, we analyzed the expression of EAAT1 and EAAT2, the major astrocytic transporters, as markers of astrocytic phenotypes.

At least five subtypes of excitatory amino acid transporters (EAATs), i.e., EAAT1-EAAT5, are distributed throughout the brain in cell-, regional- and species-specific patterns (Danbolt, 2001). In the following study, three members of the glutamate transporter gene family were studied: 1) EAAT1, the human counterpart of GLAST that was isolated from the rat (gene name: SLC1A3; HGNC: 10941); 2) EAAT2, the human counterpart of GLT1 also isolated from the rat (gene name: SLC1A2; HGNC: 10940); and 3) EAAT3, the human counterpart of EAAC1 that was first isolated from the rabbit (gene name: SLC1A1; HGNC: 10939) (Pines et al., 1992, Storck et al., 1992, Arriza et al., 1994, Fairman et al., 1995). EAAT4 (gene name: SLC1A6; HGNC: 10944)(Fairman et al., 1995) and EAAT5 (gene name: SLC1A7; HGNC: 10945)(Arriza et al., 1997) were not analyzed in this study since they are expressed predominantly in the cerebellum and retina, respectively. The nomenclature for human glutamate transporters is used here, as this study concerns the human cerebral cortex. We utilized single- and double-label immunocytochemistry and western blot analysis in the normative frontal and parietal (associative) cerebral cortex of fetuses and infants ranging in age from midgestation through infancy. This study indicates that the expression of glutamate transporters, but particularly EAAT2, is developmentally regulated in both neurons and astrocytes in the cerebral cortex, with implications for an important role for EAAT2 in cortical development and maturational susceptibility to excitotoxicity in early life.

MATERIALS AND METHODS

Clinical Database

Cerebral cortical samples were analyzed from human fetuses and infants from the frontal and parietal regions. Human brain samples were accrued from the Departments of Pathology, Brigham and Women’s Hospital and Children’s Hospital Boston, MA, with parental and Institutional Review Board approval. The cortical samples were obtained from the brains of cases without significant neurological disease or neuropathologic findings based upon reviews of the autopsy and neuropathologic reports and examination of microscopic sections of brain and spinal cord. The age of the cases is expressed in postconceptional (gestational plus postnatal) weeks. Only samples with a postmortem interval of less than 24 hours were used for analysis.

Antibody Characterization

Table 1 provides a summary of the antibodies used in this manuscript.

Table 1.

Antibodies used for single- and double label immunocytochemistry and western blot

Antibody	Source	Host and Type	Antigen	Dilution
GFAP (SMI 22)	Sternberger Monoclonals Lutherville, MD	mouse mono IgG1 (mixture)	purified bovine GFAP	1:3000
MAP2 (M9942)	Sigma	mouse mono IgG	rat brain MAP2	1:500
synaptophysin (S5768)	Sigma	mouse mono IgG	rat retina synaptosome	1:500
S100 (612376)	BD Transduction Laboratories	mouse mono IgG	mouse S100B	1:500
GLT1a	Jeff Rothstein’s Laboratory, Johns Hopkins University, MD	rabbit poly IgG, affinity purified	Rat peptide of the carboxyl-terminus containing aa 559–573 (S-A-D-C-S-V-E-E-E-P-W-K-R-E-K)	1:500
GLT1a (AB1783)	Millipore	guinea pig poly IgG, serum	Synthetic peptide of the carboxyl-terminus containing aa 554–573 (A-A-N-G-K-S-A-D-C-S-V-E-E-E-P- W-K-R-E-K)	1:400
nGLT1	Paul Rosenberg’s Laboratory, Children’s Hospital Boston, MA	rabbit poly IgG, affinity purified	Rat peptide of the amino-terminus containing aa 1–15 (M-A-S-T-E-G-A-N-N-M-P-K-Q-V-E)	1:100
EAAT1 (sc-15316)	Santa Cruz	rabbit poly IgG, serum	Recombinant protein of the human amino terminus of EAAT1 containing aa 1–50 (M-T-K-S-N-G-E-E-P-K-M-G-G-R-M-E-R-F-Q-Q-G- V-R-K-R-T-L-L-A-K-K-K-V-Q-N-I-T-K-E-D-V-K-S-Y-L-F-R-N-A-F)	1:100
EAAT3 (sc-25658)	Santa Cruz	rabbit poly IgG, serum	Recombinant protein of human carboxyl-terminus containing aa 455-524 (D-A-F-G-T-G-I-V-E-K-L-S-K-K-E-L-E-Q-M-D-V-S-S-E-V-N-I-V-N-P-F-A-L-E-S-T-I-L-D-N-E-D-S-D-T-K-K-S-Y-V-N-G-G-F-A-V-D-K-S-D-T-I-S-F-T-Q-T-S-Q-F)	1:100

Abbreviations: poly = polyclonal; mono = monoclonal; aa = amino acids; GFAP = glial fibrillary acidic protein; nGLT1= rat glutamate transporter homologue/EAAT2 in human

Anti-EAAT1

The anti-EAAT1 antibody (Santa Cruz, Biotechnology, CA) recognizes a 67 kDa band on an immunoblot of cultured astrocytes (Vallejo-Illarramendi et al., 2005), HEK293 cells transfected with EAAT1 (Vallejo-Illarramendi et al., 2005, Liang et al., 2008) or human gray matter lysates (Bauer et al., 2010). Immunocytochemical analysis demonstrated staining of astrocytes from mice (Langley et al., 2009).

Anti-EAAT3

The antibody for EAAT3 (santa cruz) recognizes a 57 kDa band on an immunoblot of lysates prepared from C6 glioma cells and rat brain hippocampus (Bianchi et al., 2010) as well as rat spinal cord (Crema et al., 2009).

Anti-nGLT1

There are several amino and carboxyl terminal variants of EAAT2 (Danbolt, 2001, Chen et al., 2002, Schmitt et al., 2002, Rozyczka and Engele, 2005, Peacey et al., 2009, Lee and Pow, 2010). This study uses an N-terminal antibody (n-GLT1), which detects c-terminal variants of GLT1. The anti-nGLT1 antibody was generated based on the published sequence for the N-terminus of rat GLT1 (aa 1–15; GenBank accession number: AF451299), which is identical to the human EAAT2 sequence (NM_004171). The nGLT1 antibody stained a monomer band at 67 kD and a band at 120 kDa, corresponding to the dimer form, in adult rat brain lysates (Chen et al., 2002, Chen et al., 2004). Peptide preabsorption experiments demonstrated complete block of immunoreactivity. There was absence of immunostaining using the nGLT1 antibody in the GLT1 knock-out mouse by light microscopy (LM) and electron microscopy (EM) (Chen et al., 2004). To test the specificity of the anti-nGLT1 antibody in human brain tissue, the immunoreactivity of purified anti-nGLT1 antibody was compared with that of the preimmune serum (DeSilva et al., 2007). Furthermore, the ability of the peptide against which the antibody was generated to block immunoreactivity was tested in human tissue (DeSilva et. al., 2007).

Anti-GLT1a

This study also uses an antibody that was generated against the C-terminus of rat GLT1a (gift of Dr. Jeff Rothstein, Johns Hopkins, MD) and has the same sequence as human EAAT2a. This anti-GLT1a antibody detects two bands on an immunoblot of adult rat brain lysate that corresponds to the monomer and dimer forms of GLT1. These same bands were detected on an immunoblot of COS7 cells transfected with GL1a cDNA, but not COS7 cells transfected with GLT1b cDNA (Chen et al., 2002). Immunoreactivity was blocked by peptide preabsorption (Chen et al., 2002). No staining was observed in the GLT1 knock-out mouse by LM and EM using the anti-GLT1a antibody (Chen et al., 2004). Immunohistochemical experiments were performed with the anti-GLT1a antibody in addition to the anti-nGLT1 antibody to confirm the expression of EAAT2. Similar immunostaining results for EAAT2 using antibodies directed toward the C-terminus (anti-GLT1a) and the N-terminus (n-GLT1) demonstrates that the labeling of EAAT2 in this study is not due to expression of an immunogenic peptide in a protein other than EAAT2.

Double-label immunocytochemistry for EAAT1 and EAAT2

For immunocytochemical double-labeling of EAAT1 and EAAT2, an anti-GLT1 antibody generated in guinea pig (Millipore, MA) was used. The anti-GLT1 guinea pig antibody was used for colocalization with EAAT1 since the anti-nGLT1 antibody (Rosenberg), the anti-GLT1a antibody (Rothstein), and the EAAT1 antibody (Santa Cruz) were generated in rabbits. The anti-GLT1 guinea pig antibody was generated against the C-terminus of EAAT2. This antibody detects a 67 KDa band in adult rat brain lysates that corresponds to the monomer form of GLT1 (Zhang et al., 2007).

MAP-2

To identify the cellular specification of EAAT1–3 in neurons, fluorescent double-labeling immunocytochemistry was performed with the neuronal marker microtubule-associated protein2 (MAP-2; Sigma, St Louis, MO). The MAP-2 antibody recognizes three forms of MAP2 (MAP2a,b,c) by western immunoblotting (Teng et al., 2001). No staining was observed in the MAP2 knockout using this antibody for immunoblotting assays (Teng et al., 2001). Light microscopy and electron microscopy studies demonstrated MAP2 staining in neuronal cell bodies, dendrites, but not in axons or non-neuronal cells (Bernhardt and Matus, 1984).

Synaptophysin

Immunocytochemical staining to detect synapses was performed with the synaptic marker, synaptophysin (Sigma, St. Louis, MO). This monoclonal antibody detects one band at 38 kDa on an immunoblot of rat brain extracts and purified synaptosomes (Jahn et al., 1985, Wiedenmann and Franke, 1985, Barnstable et al., 1988). Immunocytochemical studies using synaptophysin demonstrated staining in presynaptic vesicles of neurons in brain, spinal cord, and retina as well as neuromuscular junctions (Wiedenmann and Franke, 1985).

GFAP

To detect astrocytes a monoclonal antibody cocktail against glial fibrillary acidic protein (GFAP) derived from the Bigner-Eng clones MAb1B4, MAb2E1 and MAb4A11 was used. These clones were validated by (McLendon et al., 1986) using indirect radioimmunoassay against fixed cell monolayers of a GFAP-positive human glioma cell line and also by competitive radioimmunoassay with radiolabelled GFAP and by competitive immunoradioassay with radiolabelled antibody. They have been further characterized by immunoblots of GFAP and by immunoperoxidase histochemistry (McLendon et al., 1986). This antibody stains astrocytes and astrocytic processes as well as Bergman glia in a wide variety of species; human, sheep, cow, dog, pig, rat, guinea pig, rat, mouse and chicken (Pegram et al., 1985).

S100

S100 (BD Transduction Laboratories, San Jose, CA) was used to detect astrocytes. This monoclonal antibody recognizes a single band at 8 kDa on an immunoblot of mouse cerebellum lysate. Immunocytochemical staining demonstrates staining in cell bodies and processes of astrocytes. Staining in rat glioma and human glioblastoma has also been documented (Jiang et al., 1993, Scotto et al., 1999, Takata et al., 2011).

Western Blot Analysis of the Developing Human Cerebral Cortex

Lysates were made from human brain tissue as previously described (DeSilva et al., 2007). Tissue was homogenized in a frosted glass homogenizer in 1% SDS containing a protease inhibitor cocktail with EDTA (Roche, Germany). After homogenization, samples were dispersed in an ultrasonic bath for approximately 15 minutes until solution was clear. Protein concentration was measured using the Lowry assay with bovine serum albumin as the standard. Samples (40 µg/lane) were run on an 8–18% polyacrylamide gel and electroblotted onto a polyvinylidene fluoride (PVDF) membrane (PerkinElmer, Wellesley, MA). PVDF membranes were incubated with anti-nGLT1 at 1µg/ml overnight at 4°C in TBST buffer (50 mM Tris, 150 mM NaCl, 0.01% Triton, pH 7.4) containing 5% nonfat milk. Blots were then washed 3 times in TBST buffer followed by one hour incubation with HRP-conjugated goat anti-rabbit IgG (Amersham Life Science, NJ). For protein detection, membranes were incubated in Western Lightning Chemiluminescence Reagent (Perkin Elmer, Boston, Ma) and exposed on X-Omat Blue XB-1 film (Kodak, NY). Densitometric analysis was performed on film using the MCID elite version 7.0 software published by Imaging Research (Ontario, Canada). A human adult standard lysate containing parietal white matter from 3 pooled cases (ages: 55, 65 and 75 years) was run as an internal standard on each blot. Density of the individual bands obtained from the human developmental series of cerebral cortex was calculated and plotted as a percent of a human adult standard prepared by us, as performed in previous studies (DeSilva et al., 2007). An ANCOVA was performed to test for differences in EAAT2 protein levels as a function of age. A p-value <0.05 was considered significant. A regression analysis of the effect of postmortem interval (PMI) on EAAT2 expression showed no statistically significant effect of postmortem interval (data not shown). A total of 27 cases were used ranging in age from 19 weeks to 21 years.

Single- and Double-Label Immunocytochemistry

Standard methods were applied to deparaffinized human tissue sections for immunocytochemical analysis. After the tissue sections were deparaffinized, they were placed in 10 mM citrate buffer (pH 6.0) and boiled for 15 minutes to enhance antigenicity. Primary antibodies diluted in blocking buffer [phosphate-buffered saline (PBS) with 5% goat serum and 0.01% Triton-X] were placed on the tissue sections following a one hour incubation with blocking buffer at room temperature. Primary antibodies were incubated with the tissue sections overnight at 4°C. The secondary antibody (DakoCytomation, Carpenteria, CA) was then applied for one hour at room temperature, and the reaction was visualized with the chromagen 3,3’-diaminobenzidine (DAB). Negative controls included omitting the primary antibodies. A total of 14 cases was stained with DAB for each transporter, EAAT1, EAAT2, and EAAAT3.

Double-labeling immunocytochemistry experiments were performed sequentially by staining the sections first with the transporter antibodies followed by the respective cell specific markers (e.g., MAP-2, synaptophysin, and GFAP, Table 1). Primary antibodies were visualized with a secondary red fluorescent antibody (Alexa 594, Invitrogen, CA), or a fluorescent green (Alexa 488, Invitrogen, CA) antibody. For the primary antibodies GFAP, MAP2, and synaptophysin a secondary anti-mouse IgG heavy and light chain was used [Alexa 488 (A-11001) or Alexa 594 (A11032)]. For the primary antibodies against GLT1a (Rothstein lab), nGLT1 (Rosenberg lab), EAAT1, and EAAT3 a secondary anti-rabbit heavy and light chain was used [Alexa 488 (A11008) or Alexa 594 (A11037)]. For the primary antibody against GLT1a from Millipore a secondary anti-guinea pig heavy and light chain was used [Alexa 488 (A-11073) or Alexa 594 (A11076)]. Specificity of secondary antibodies was confirmed by processing control sections without primary antibodies. Digital imaging was performed on a Zeiss Axioscop equipped with a Spot advanced camera. Confocal imaging was performed on a Zeiss LSM 510 MetA microscope. Images were captured using Zeiss LSM software. Photomicrographs used in this publication were produced using Adobe Photoshop. Images from individual channels were merged and the brightness and contract was adjusted with Adobe Photoshop.

RESULTS

Clinical Database

Human samples of the cerebral cortex were examined for EAAT1, EAAT2, and EAAT3 staining using DAB, single and double-label immunofluorescence, and western blot analysis. For DAB staining, 16 cases were stained for each respective transporter EAAT1, EAAT2, and EAAT3 ranging in age from 23 gestational weeks to 2.5 postnatal years. A subset of 12 of these cases was used for double-label immunocytochemical studies to identify EAAT1, EAAT2, and EAAT3 in specific cell types. A summary of the human cases used for immunolabeling is included in Table 2. Neuropathologic examination of the microscopic sections stained with hematoxylin-and-eosin/Luxol-fast-blue in each case did not reveal pathologic changes in the cerebral cortex or underlying white matter; there was no evidence of periventricular leukomalacia. The mean postmortem interval for the immunocytochemical cases was 17±3 hours. For western blot analysis, 17 samples of the cerebral cortex were analyzed from cases aging from 19 gestational weeks to 21 postnatal years. The mean postmortem interval was 15 hours ± 2. A regression analysis of PMI on EAAT2 expression showed no statistically significant effect of the postmortem interval.

Table 2.

Summary of human cases used for DAB and immunofluorescent staining

#	age	PMI	brain site	GLAST DAB	nGLT1 DAB	GLT1a DAB	EAAC1 DAB	GLAST ICC	nGLT1 ICC	GLT1a ICC
1	23 wks	28	parietal		X	X	X	X	X	X
2	27 wks	*	frontal		X	X	X
3	27wks	6	parietal		X	X		X	X	X
4	33 wks	32	frontal	X	X	X	X	X	X	X
5	35 wks	6	frontal		X	X		X	X	X
6	37 wks	11	parietal		X	X	X	X	X	X
7	38 wks	48	frontal		X	X
8	40 wks	1	frontal	X	X	X	X	X	X	X
9	41 wks	26	parietal	X	X	X		X	X	X
10	43 wks	24	parietal	X	X	X	X	X	X	X
11	52 wks	22	frontal	X	X	X	X	X	X	X
12	72 wks	8	frontal	X	X	X	X	X	X	X
13	92 wks	17	frontal	X	X	X	X	X	X	X
14	112 wks	9	frontal	X	X	X	X	X
15	144 wks	17	frontal		X	X		X	X	X
16	170 wks	12	frontal	X	X	X	X	X	X	X

^*not reported; PMI, post-mortem interval in hours; frontal is posterior frontal

Glutamate Transporter Expression in the Cerebral Cortex in the Human Fetus and Infant by Immunocytochemical Analysis

We analyzed the cellular expression of EAAT1-EAAT3 in the glial limitans, cortical layers I–VI, and white matter neurons directly underneath the cortex (site of the subplate proper). EAAT1 was expressed exclusively in astrocytes, and EAAT3 exclusively in neurons. EAAT2, on the other hand, was initially expressed in neuronal cell bodies before undergoing a transition around the end of the first postnatal year. At this point EAAT2 expression was confined to glial cells with the morphology of mature protoplasmic astrocytes (Table 3).

Tables 3.

Summary diagram of the expression of EAAT1-3 in the cerebral cortex and subcortical white matter

GFAP expression

The key cellular changes in EAAT1 and EAAT2 expression in the developing human cerebral cortex relative to the expression of GFAP were studied (Figure 1). We observed the two well-established morphologies of astrocytes in the developing human brain (Berry et al., 2002): 1) fibrous astrocytes with long, thin processes in a star-like pattern found predominantly but not exclusively in white matter, which are labeled by an anti-GFAP antibody (white brackets, Figure 1) and; 2) protoplasmic astrocytes with shorter, coarser processes with a patch-like pattern (arrows; Figure 1 D, F) that was found exclusively in the grey matter and that did not label with the GFAP antibody. GFAP-immunopositive astrocytes found in the hippocampus demonstrated the same morphology as that of the EAAT2-immunopositive patches in the cortex (Figure 2 A, B), suggesting that EAAT2 patches are astrocytes, and, in the cortex, do not express GFAP (see Discussion). In addition, GFAP immunopositive cells (Figure 2 D, red) were found around blood vessels in the cerebral cortex; in the same fields, however, GFAP was not detected in EAAT2 immunopositive protoplasmic astrocytes (Figure 2 C, E green). These observations demonstrate that GFAP is immunoreactive in the cerebral cortex, but not in the protoplasmic astrocytes. Furthermore, the coarse short processes of the EAAT2 patches appear morphologically similar to those of protoplasmic astrocytes (Oberheim et al., 2009; Marin-Padilla et al., 1995), reinforcing the concept that EAAT2 is expressed in GFAP negative protoplasmic astrocytes. Intralaminar astrocytes were also observed in this study, as defined by the position of their cell bodies (white boxes, Figure 1 A, C) in the upper cortical layers and their extension of processes (carrots; Figure 1 E) into the cortical layers II–III later in development (Oberheim et al, 2009).

Figure 1. Montage of double-labeling with GFAP and EAAT2.

Immunocytochemistry of the glutamate transporter EAAT2 compared with staining of the astrocytic marker GFAP (glial fibrillary acidic protein). GFAP is not detected in protoplasmic astrocytes; in contrast, EAAT2 appears to be expressed within them (white arrows; D,F). GFAP does stain the underlying white matter at all ages of development which serves as a positive internal control (white brackets). GFAP is expressed in the glial limitans, interlaminar astrocytes (white boxes; A,C) as well as GFAP positive astrocytes surrounding blood vessels (circles; C). In layers I–III GFAP is also expressed in fibrillary processes (arrow heads; A,C) and later in development it is observed in long fibers (carrots; E). Scale bar, 50 µm.

Figure 2. GFAP-immunopositive astrocytes with similar course morphology to that of EAAT2 immunoreactive patches are noted but GFAP and EAAT2 immunostaining do not co-localize.

A, GFAP immunopositive patches are observed in the hippocampus and have a coarse morphology, consistent with EAAT2 immunoreactive patches observed in the cortex and hippocampus. However, GFAP immunopositive patches in the hippocampus do not colocalize with EAAT2 immunopositive patches in the hippocampus. In the cortex, there were no GFAP immunopositive patches at the ages studied. B, GFAP immunopositive cells (red) were found around a blood vessels in the cerebral cortex, however, in the same field, GFAP was not detected in an EAAT2 immunopositive protoplasmic astrocyte (green). Scale bar, 50 µm. Magenta-green images are available as supplementary figure 1.

Expression of GFAP-immunostained glial cells with the stellate-like morphology of fibrillary astrocytes and positive GFAP-immunostaining were identified in the glial limitans and white matter underlying the cerebral cortex at 23 gestational weeks, the earliest time-point examined by us; at this early age, they were not observed in cortical layers II–VI (data not shown). By 33 weeks, GFAP-immunostaining was present in the glial limitans, interlaminar astrocytes in layers I–II (white boxes, Figure 1 A), fibrillary processes in layers II–IV (white arrow heads, Figure 1 A, C), and numerous fibrous astrocytes in the underlying white matter (white bracket, Figure 1 A). At 41 weeks, GFAP-immunostaining was increased in the glial limitans (Fig. 1 C) and in fibrous astrocytes in the underlying white matter (white bracket, Figure 1 C) compared to 33 weeks (Figure 1 A). GFAP expression was also observed in the interlaminar astrocytes in layers I–II (white boxes, Figure 1 C), fibrillary processes in layers II–IV (white arrow heads, Figure 1 C), and in astrocytic cell processes abutting the walls of arterial blood vessels, consistent with astrocytic foot processes (circles, Figure 1 C and 2 D). With increasing age, the density of GFAP immunopositive, interlaminar astrocytes increased, as well as the number and length of GFAP positive fibers (carrots, Figure 1 E); these fibers extended through layers I–III, as illustrated in a representative case at 2.5 years (Figure 1 E). At this postnatal age, the processes of the fibrous astrocytes in the underlying cerebral white matter extended into the overlying layer VI (marked by asterisks, Figure 1 E). At no age were GFAP-immunopositive cells with the morphological features of protoplasmic astrocytes present in any cortical layer, only in the hippocampus (Figure 2 B). Immunostaining with S100, a second astrocytic marker, was technically unsuccessful in our tissue sections. In summary with increasing age, the density of GFAP immunopositive, interlaminar astrocytes increased, the number and length of GFAP immunopositive fibers in layer I–III increased, and the number and length of GFAP immunopositive fibers from the white matter extending into cortical layer VI increased.

EAAT1 Expression

EAAT1 was expressed exclusively in glial cells with the morphological features of astrocytes, including in the glial limitans. EAAT1 staining in the glial limitans co-localized with GFAP (Figure 3 A–C, arrows). EAAT1-immunopositive cells with the morphology of human interlaminar astrocytes (Oberheim et al., 2009) were identified in layers I–III and co-expressed GFAP (Figure 3 D–F, arrows). Before one year of age, finely particulate EAAT1 staining was also noted diffusely throughout the neuropil among scattered astrocytic cell bodies, which typically did not co-localize with GFAP (Figure 3 D – F). At approximately one year of age, this diffuse neuropil staining appeared in “patches” of labeling of coarse rather than fine processes (Figure 3 G, arrow heads), particularly in layers II and VI, which did not colocalize with GFAP (Figure 3 G – I). After one year of age, EAAT1 patches were scattered throughout all laminae (Figure 4 C – D). These EAAT1 patches did not co-localize with GFAP but they did co-localize with EAAT2, likewise a known astrocytic transporter (Figure 5). Interestingly, the “match” in immunostaining between the EAAT1 and EAAT2 patches was not precise, such that all EAAT2 patches co-localized with EAAT1 patches, but not all EAAT1 patches co-localized with EAAT2 (Figure 5 C). Late in gestation, GFAP immunopositive processes of the fibrous astrocytes in the underlying cerebral white matter extended into layer VI (Figure 3 H, arrow heads) where they were in close proximity to the EAAT1 patches but did not co-localize with them (Fig 3 G – I). EAAT1 patches were never found in the underlying white matter.

Figure 3. EAAT1 expression with a fine diffuse pattern co-localizes with GFAP before the occurrence of course patches at 8 postnatal months in the cerebral cortex.

A) EAAT1 staining in a 33 week old case co-localizes with GFAP in the glial limitans (arrow) (A–C) at 20× magnification. At higher magnification (40×) in the same case, EAAT1 staining also co-localizes with GFAP positive cells and diffuse neuropil staining in layers I–III (arrows; D–F). Expression of EAAT1 in course patches does not co-localize with GFAP positive fibers in layer 6 (G–I). Scale bar, 50 µm. Magenta-green images are available as supplementary figure 2.

Figure 4. Expression of EAAT1 in the developing human cortex (layer I–VI) at representative ages (×4).

From 23 weeks until term EAAT1 expression is diffuse (A). At 6 months of age, the expression of EAAT1 in patches is observed and by 1 year of age (B) these patches become prominent in layer II and VI. B. After 1.5 years of age (D,E) the expression of EAAT1 immunoreactive patches becomes confluent in all layers of the cortex. Scale bar, 50 µm.

Figure 5. Co-localization of EAAT1 and EAAT2 in cortical patches at 170 wks.

All EAAT2 immunopositive patches co-localize with EAAT1; however, not all EAAT1 immunopositive patches co-localize with EAAT2 (C, asterisks). Scale bar, 50 µm. Magenta-green images are available as supplementary figure 3.

EAAT2 expression

At 33 weeks gestation, EAAT2 was expressed in the glial limitans (Figure 6 A – C) with co-localization in GFAP immunopositive astrocytes whose end feet attached to the external glial limiting membrane (Figure 6 A–C, arrow). EAAT2 staining was also present in astrocytic processes that extended from the glial limitans into layers I, II, and III (Figure 6 A – C). Before term, there was finely particulate EAAT2 immunostaining diffusely in the neuropil and predominately in layers II and III that did colocalize with GFAP (white circles, Figure 6 C). There were also EAAT2 immunopositive long processes in these layers that did not co-localize with GFAP (asterisk, Figure 6 C). At 41 weeks, coarse EAAT2 patches were first detected in the cortex, and by 8 postnatal months, they tended to concentrate in layers II and VI (arrow heads, Figure 6 D; 7 B). Thus, there was a distinctive change in the pattern of expression of EAAT2 in the neuropil from fine and diffuse around term (Figure 6 A–C; Figure 7 A) to coarse and restricted patches by 8 postnatal months (Figure 6 D–F; Figure 7 B – E). After 8 postnatal months, the density of EAAT2 patches visually increased in layers II–VI (Figure 7 C – D). Similar to the pattern of EAAT1 patches, EAAT2 patches were characterized by immunostaining around a group of 5–10 neurons, adjacent to regions of ≥ 5 neurons that were devoid of immunostaining (Figure 7 E). EAAT2 immunoreactive patches did not co-localize with GFAP, nor were there GFAP-immunopositive cell bodies located within the patches themselves. There were, however, sparse GFAP immunopositive patches in the hippocampus (Figure 2 B) with the same coarse morphology as EAAT1/EAAT2 (Figure 2 A), suggesting that the EAAT1/EAAT2 patches represented the coarse processes of protoplasmic astrocytes (see Discussion). To determine if EAAT2 was expressed at synapses, we performed double-label immunofluorescence with an antibody to synaptophysin. Confocal imaging failed to demonstrate co-localization of EAAT2 with synaptophysin (Figure 8).

Figure 6. EAAT2 is expressed in fine processes and co-localizes with GFAP before the appearance of EAAT2-labeled patches after term which, in contrast, do not express GFAP.

EAAT2 staining in a 33 week old case shows some co-localization with GFAP in the glial limitans (A–C). Expression of EAAT2 is present in a GFAP immunopositive astrocyte whose end feet are attached to the glial limitans (arrow, A–C). EAAT2 staining colocalizes with GFAP positive cells and diffuse neuropil staining in layer I–III of the cortex (white circles, C). There were also GFAP immunopositive long processes in these layers that did not colocalize with EAAT2 (asterisk, Figure 6C). The expression of EAAT2 in patches does not co-localize with GFAP positive fibers in layer II–III (arrow heads, D–F). Scale bar, 50 µm. Magenta-green images are available as supplementary figure 4.

Figure 7. EAAT2 expression in layer I–VI during human cortical development at representative ages (x4).

From 23 weeks until term, EAAT2 expression is diffuse throughout the neuropil (A). At 41 weeks of age, its expression appears in course patches, and by 8 months of age (B) these course patches become prominent in layer II and VI. B. After 1.5 years of age (C,D), the expression of EAAT2 immunopositive patches becomes confluent in layers II–VI. A high-powered view of EAAT2 immunopostive patches shows that they are located around a group of 5–10 neurons, adjacent to regions of ≥ 5 neurons that were not in contact with EAAT2 immunoreactive patches (E). Scale bar, 50 µm.

Figure 8. Lack of synaptophysin expression in the EAAT2 patches, as demonstrated in a representative case at term.

As observed with confocal imainging, synaptophysin staining (B) does not co-localize with EAAT2 immunopositive patches, highlighted by merging (C) (x63). Bisbenzamide stains cell nuclei (C). Scale bar, 50 µm. Magenta-green images are available as supplementary figure 5.

In addition to astrocytes, EAAT2 was expressed in several neuronal subtypes, i.e., layer I neurons, pyramidal neurons, and subcortical white matter neurons in the developing human cortex. EAAT2 expression was detected in the cytoplasm of layer I neurons at 23 gestational weeks, the earliest age studied, and persisted within this site until term (Table 3 and Figure 9). These EAAT2-mmunopositive layer I neurons had the morphology of Cajal-Retzius cells with abundant perikaryia and horizontal orientation (Figure 9). The neuronal specification of EAAT2 expression in these layer I neurons was confirmed by co-localization with MAP-2, a neuronal marker (Figure 9A). The expression of EAAT2 was also observed in the cytoplasm and apical dendrites of pyramidal cells in layer V from 23 gestational weeks until 8 postnatal months (Table 3 and Figure 10 A – D). To confirm specificity of the EAAT2 antibody in pyramidal neurons, the GLT1a antibody that detects EAAT2a was also used to stain pyramidal neurons in layer V (Figure 11, white arrows). Inset in upper left hand corner shows a higher power magnification (40x) of a pyramidal cell with EAAT2a staining in the cell body and apical dendrite (Figure 11). At 41 weeks, when EAAT2 patches first emerged, EAAT2 immunostaining was present in MAP-2 positive neurons (Figure 10 E – G). At 1.5 years of age, when patches became more confluent in layers II–VI, EAAT2 was no longer observed in pyramidal neurons (see Figure 7C and E). Neurons in the white matter at 23 weeks expressed EAAT2 until term (Figure 12). These neurons were unipolar (A, arrows) or multipolar (D, arrow head), and located in the subcortical and central white matter, and rarely observed in the deep (periventricular) white matter. The EAAT2 immunopositive neurons tended to be concentrated in the subcortical regions, i.e., the site of the subplate proper.

Figure 9. Expression of EAAT2 in layer I neurons consistent with Cajal-retzius cells, as defined by large perikarya and horizontal morphology.

EAAT2 (B) is co-localized in a MAP2 expressing neuron (A) at 33 weeks that has the morphological features of Cajal-Retzius cells. Scale bar, 50 µm. Magenta-green images are available as supplementary figure 6.

Figure 10. EAAT2 expression in layer V pyramidal neurons in the human cerebral cortex in a representative case at 27 gestational weeks.

DAB staining of EAAT2 expression in layer V pyramidal neurons (A). Immunoflourescent staining of the same case demonstrates that EAAT2 colocalizes with MAP2 immunopositive pyramidal cells in layer V (B–D). At 41 weeks, when EAAT2 patches first emerged, EAAT2 immunostaining was still present in MAP-2 positive neurons (E–G). Scale bar, 50 µm. Magenta-green images are available as supplementary figure 7.

Figure 11. EAAT2a expression in layer V pyramidal neurons in the human cerebral cortex in a representative case at 39 gestational weeks.

To confirm specificity of the EAAT2 antibody in pyramidal cells, the EAAT2a antibody (GLT1a) was also used to stain pyramidal neurons in layer V (white arrows). Insert in upper left hand corner shows a higher power magnification (40x) of a pyramidal cell with EAAT2 staining in the cell body and apical dendrite. Scale bar, 50 µm.

Figure 12. EAAT2 expression in subcortical white matter neurons in a representative case at 33 gestational weeks.

DAB staining of EAAT2 in unipolar subcortical white matter neurons (A, black arrows). Immunofluorescent staining of an adjacent section shows co-localization of EAAT2 and MAP2 in a bipolar subcortical neuron. Scale bar, 50 µm. Magenta-green images are available as supplementary figure 8.

EAAT3 Expression

The expression of EAAT3 was observed exclusively in neurons at all ages, with no demonstrable astrocytic immunostaining. EAAT3 was expressed in neurons in the subcortical white matter i.e., site of the subplate proper, at 27 weeks until term (Figure 13A). It was also expressed in a subset of pyramidal neurons in layers III and V at 27 weeks and persisted throughout all ages of our series, i.e., until approximately 2.5 postnatal years (Figure 13B). EAAT3 staining did not form discrete neuropil patches in any of the cases studied.

Figure 13. EAAT3 expression in subcortical white matter neurons and layer V pyramidal neurons.

DAB expression of EAAT3 in unipolar subcortical white matter neurons (A, arrows) in a 27 week old case. DAB expression of EAAT3 in layer V pyramidal neurons in a 2.5 year old case (B, arrows). Scale bar, 50 µm.

Western Blot Analysis of EAAT2 Expression in the Developing Human Cerebral Cortex

Seventeen human samples from the parieto-occipital cortex at the level of the atrium of the lateral ventricle were examined for EAAT2 expression. At 19 gestational weeks, the earliest age examined, a single band at 67 kDa was detected using the anti-nGLT1 antibody (Figure 14A). The density of this band did not vary with increasing age from midgestation to term (37–40 weeks) or through the neonatal and postnatal periods (Figure 14 A). At approximately 2 postnatal months (i.e., 49 postconceptional weeks), a second band appeared at 77 kDa and persisted to 3 postnatal years, the oldest time-point examined (Figure 14B). The density of the second band (77 kDa) did not statistically vary in density with age (Figure 14B). In addition, this second band was noted in a single adult (21 year old) case for comparison with the mature brain.

Figure 14. Densitometric analysis of western blot expression of EAAT2 in the human developing cerebral cortex.

Band 1 refers to the protein band at 67 kDa; Band 2 refers to the protein band at 77kDa. Band 1 is present at 23 gestational weeks, the earliest age examined, and remained constant through the second half of gestation. Band 2 appears around 2 postnatal months and persists until the oldest age examined 2.5 postnatal years (129 postconceptional weeks). A single case at 21 years of age also demonstrates the presence of Band 1 and Band 2. These data suggest a previously unrecognized post-translational modification of cortical EAAT2 after birth (see text).

DISCUSSION

Our studies document the transient expression of EAAT2 in the soma and dendrites of certain neuronal subtypes during human development, whereas previous studies in the rodent have shown transient neuronal expression of EAAT2 only in developing axons (Yamada et al., 1998; Northington et al., 1999; Furuta et al., 1997) and in a subset of presynaptic terminals in the adult (Chen et al., 2002, Furness et al., 2008, Melone et al., 2009, de Vivo et al., 2010). In light of these new findings and previous work showing predominant expression of EAAT2 in glia, the issue of antibody specificity must be considered. Both the anti-nGLT1 antibody and the anti-GLT1a antibody used in this study have been tested previously in null mice (Chen et al., 2004). In addition, the staining in the cell bodies and apical dendrites of layer V pyramidal neurons was observed using an N-terminal directed antibody (Fig.10), and a C-terminal directed antibody (Fig.11). These data indicate that the labeling in neurons (which also colocalized with MAP2) (Fig. 10), is not due to reactivity of the antibodies with epitopes on a protein other than EAAT2. EAAT2 expression was also found in the cell body and processes of Cajal-Retzius cells (Fig. 9) and subplate neurons (Fig. 12). This labeling also co-localized with MAP2. The expression of EAAT2 in these neuronal populations and its relationship to cortical vulnerability in perinatal hypoxic-ischemic encephalopathy will be addressed in the following discussion.

Previous studies have shown the expression of EAAT1 and EAAT2 in protoplasmic astrocytes in humans and rodents (Milton et al., 1997, Williams et al., 2005, Yang et al., 2011). Our study is novel in documenting the expression of EAAT1 and EAAT2 in developing human cortex, and its dissassociation from the expression of GFAP. This expression of glutamate transporters in developing protoplasmic astrocytes and the lack of detectable expression of GFAP in these cells indicates that the glial glutamate transporters are more reliable morphological markers of astrocytic phenotypes than GFAP. The complexity and diversity of cortical astrocytes is a distinguishing hallmark of the human brain (Oberheim et. al., 2006) underscoring the importance of documenting astrocyte development directly in human brain studies. Furthermore, our data show that EAAT2 is expressed at lower levels during fetal development and increases after term birth consistent with studies in human and rodent cerebral cortex (Sutherland et al., 1996b, Bar-Peled et al., 1997, Furuta et al., 1997a, Ullensvang et al., 1997). Finally, we demonstrate that EAAT2 is post-translationally modified after birth. EAAT3 was found in pyramidal neurons similar to previous findings (Furuta et al., 2005), and was also found in subplate neurons. In the following discussion, we highlight the significance of these findings in the context of known information about the role of glutamate transporters during cortical development.

Neuronal Expression of EAAT2 in the Developing Human Cerebral Cortex

Previous findings have demonstrated that EAAT1 is expressed in astrocytes (Torp et al., 1994, Chaudhry et al., 1995, Lehre et al., 1995, Schmitt et al., 1997, Berger and Hediger, 1998) and oligodendrocytes (Sutherland et al., 1996a, Domercq et al., 1999, Regan et al., 2007, DeSilva et al., 2009), and that EAAT3 is expressed in neurons (Kanai and Hediger, 1992, Rothstein et al., 1994) and oligodendrocytes (Domercq et al., 1999, Kugler and Schmitt, 1999, Pitt et al., 2003, DeSilva et al., 2009). Although EAAT2 is considered to be mainly expressed in astrocytes (Danbolt et al., 1992, Hees et al., 1992, Levy et al., 1993, Rothstein et al., 1994, Chaudhry et al., 1995, Lehre et al., 1995, Schmitt et al., 1996, Milton et al., 1997, Berger and Hediger, 2000) and oligodendrocytes (Sutherland et al., 1996a, Domercq et al., 1999, Regan et al., 2007, DeSilva et al., 2009), we report here the novel observation that it is expressed in the somata of layer V pyramidal neurons, layer I neurons (putative Cajal-retzius cells), and interstitial white matter neurons. Of note, EAAT2 expression varies depending upon the cell type, brain region, age, species, and methodology (Rauen and Kanner, 1994, Rothstein et al., 1994, Torp et al., 1994, Euler and Wassle, 1995, Rauen et al., 1996, Schmitt et al., 1996, Milton et al., 1997, Torp et al., 1997, Berger and Hediger, 1998, Chen et al., 2002, Chen et al., 2004, Berger et al., 2005). The paradox that mRNA encoding EAAT2 is expressed in adult neuronal populations without obvious protein expression has been attributed to the lack of translation, protein levels undetectable by immunocytochemistry using electron microscopy, and/or the failure of current antibodies to identify EAAT2 variants (Danbolt, 2001). While EAAT2 mRNA has been detected in neurons, the expression of EAAT2 protein in neurons was considered virtually nonexistent except for transient expression during development in axons in white matter regions preceding astrocytic expression in rodents and sheep (Yamada et al., 1998; Northington et al., 1999; Furuta et al., 1997). Transient expression of EAAT2 in axons and dendrites in the mouse somatosensory cortex has also been shown to precede astrocyte expression (Takasaki et al., 2008). More recently, however, EAAT2 protein has been reported in presynaptic terminals in the adult hippocampus and somatic sensory cortex in rats and mice (Chen et al., 2002, Furness et al., 2008, Melone et al., 2009, de Vivo et al., 2010). In our study, we observed expression of EAAT2 in the somata of pyramidal neurons in layer V at 23 gestational weeks up until 8 postnatal months of age. Expression of EAAT2 in protoplasmic astrocytes followed, first appearing at 41 postconceptional weeks, becoming highly expressed in layers I and VI at 8 postnatal months, and confluent in all layers of the cortex by 1.5 years. Layer V pyramidal neurons are glutamatergic and project to the basal ganglia, brainstem, and spinal cord, and also provide local excitatory input, thereby enhancing rich collateral networks that form an important part of the cortical circuitry (Shepherd, 2004). Since EAAT2 has been detected in adult presynaptic terminals (Chen et al., 2002, Furness et al., 2008, Melone et al., 2009, de Vivo et al., 2010), a possible explanation for the transient expression of EAAT2 in layer V pyramidal neurons is that its expression in the somata of developing neurons precedes its polarized expression in axon terminals as neurons mature. In this regard, high affinity glutamate uptake by EAAT2 in the striatum appears to be due to axon terminals deriving especially from the cerebral cortex (Divac et al., 1977, Streit, 1980, Fonnum et al., 1981, Storm-Mathisen and Wold, 1981, Gundersen et al., 1996, Bellomo et al., 1998, Suchak et al., 2003).

The specific expression of EAAT2 in layer V pyramidal neurons may also have important pathological implications. The release of excessive glutamate by reversal of glutamate transport is known to occur under conditions of energy failure resulting in excitotoxicity (Roettger and Lipton, 1996, Seki et al., 1999, Fern and Möller, 2000, Rossi et al., 2000, Deng et al., 2003). Energy failure as a consequence of hypoxia-ischemia is thought to be an underlying mechanism in perinatal brain injury and may contribute to the vulnerability of layer V pyramidal neurons shown to be damaged in premature infants with periventricular leukomalacia (Andiman et al., 2010). Glutamate receptors are expressed on pyramidal neurons in other layers of the developing human cortex, raising the question of why only layer V pyramidal neurons are lost in periventricular leukomalacia. We hypothesize that the coincident expression of EAAT2 in layer V neurons creates a suicide feedback loop by which these neurons express both the transporters for releasing glutamate into the extracellular space and the receptors that are hyperactivated by excess ambient glutamate leading to cell death (Fern and Möller, 2000, Talos et al., 2006).

EAAT2 is expressed in two other neuronal populations that are relatively transient, i.e., layer I neurons and subplate neurons (Huntley and Jones, 1990, Del Rio et al., 1996, Marin-Padilla, 1998, Soriano and Del Rio, 2005). Glutamatergic Cajal-Retzius cells are postulated to provide tonic excitatory activity to developing pyramidal neurons whose apical dendrites extend into layer I (Del Rio et al., 1996, Marin-Padilla, 1998). Ablation of Cajal-Retzius cells also inhibits neuronal migration (Super et al., 2000). These data suggest a role for EAAT2 in glutamatergic interactions critical for maturation of pyramidal neurons and for neuronal migration. Very early in development, subplate neurons are essential for cortical targeting by in-growing thalamic axons as well as for lamina formation (Ghosh and Shatz, 1993). Approximately one-half of the subplate neurons that project to the cerebral cortex are glutamatergic (Finney et al., 1998) and a large proportion of subplate neurons receive synaptic excitatory input from other subplate neurons (Hanganu et al., 2002, Hirsch and Luhmann, 2008) suggesting an important role for the regulation of glutamate by glutamate transporters in subplate neurons during cortical development. In addition, the expression of GLT1 in these cells also puts them at risk for excitotoxic injury. The relevance of this pathogenic mechanism is implied by the vulnerability of subplate neurons to excitotoxicity in animal models of neonatal hypoxia-ischemia(McQuillen et al., 2003, Nguyen and McQuillen, 2010).

EAAT1 and EAAT2 Expression in Patches in the Developing Human Cerebral Cortex

A major observation in this study is that EAAT1 and EAAT2 immunostaining occurs in distinct patches during cortical development consistent with protoplasmic astrocytes. While GFAP is effective in immunostaining fibrillary astrocytes, it is relatively ineffective in staining protoplasmic astrocytes (Maxwell and Kruger, 1965, Connor and Berkowitz, 1985, Bushong et al., 2002). In fact, transgenic mice with GFAP promoter-driven enhanced green fluorescent protein (eGFP) were found to have a number of cells that were eGFP positive under the expression of the GFAP promoter, but did not show any staining with the GFAP antibody (Matthias et al., 2003). This study also demonstrated glutamate transporter currents in some but not all eGFP expressing cells in the hippocampus. EAAT2 immunoreactivity in cells with the morphological features of astrocytes has been described previously in the adult human cortex (Milton et al., 1997, Williams et al., 2005), however, co-localization with GFAP was not demonstrated. Recently, however, GFAP staining of protoplasmic astrocytes was observed in the adult human cortex in surgical samples of freshly resected temporal lobes from patients with epilepsy (Oberheim et al., 2009). This success in GFAP immunostaining of protoplasmic astrocytes in the surgically excised human cortex was attributed by the investigators to: 1) rapid tissue fixation in recently resected specimens, as opposed to fixation in autopsy specimens with unavoidably long postmortem intervals; and 2) the use of 4% paraformaldehyde, as opposed to the traditional formalin fixative. In our study, we sought to label putative protoplasmic astrocytes with GFAP as well as another astrocytic marker, S-100, but were likewise unsuccessful. Nevertheless, the co-expression of EAAT2 immunostaining with EAAT1, a well-recognized astrocytic marker (Danbolt, 2001), underscores the astrocytic phenotype of these cells and suggests the possibility that human protoplasmic astrocytes first express glutamate transporters before or to the exclusion of GFAP.

The gliogenesis of protoplasmic astrocytes is a late event in the human fetal period which occurs after the presumed completion of radial neuronal migration (approximately 20 weeks gestation) (Marin-Padilla, 1995). Protoplasmic astrocyte precursors have a late migration course through the white matter (Figure 15–1) and accumulate in layer I to form the subpial granular layer of Ranke (SGLR) (Ranke, 1909) (Figure 15–2). Cells from the SGLR form the external glial limiting membrane (Figure 15–3); some cells in contact with the pia in the EGLM migrate into the underlying cortical parenchyma where they establish contacts with plexus capillaries and differentiate into mature protoplasmic astrocytes (Marin-Padilla, 1995). Thus, after astrocytes loosen their attachment to the external glial limiting membrane, their long descending processes indicative of an immature protoplasmic astrocyte (Figure 15–4) progressively shorten and they acquire the coarse stellate morphology of mature protoplasmic astrocytes (Figure 15–5) (Marin-Padilla, 1995). Based upon these Golgi observations combined with our observations with the astrocytic markers EAAT1 and EAAT2, a more complete understanding of the progression of protoplasmic astrocyte development emerges. In essence, we find co-localization of EAAT2 with GFAP immunopositive astrocytes whose end feet are attached to the external glial limiting membrane (Figure 6 C; Figure 15–3). We also found EAAT1/EAAT2 immunopositive cells that do not colocalize with GFAP in layer 1 (Figure 6 A – C) suggesting the existence of immature protoplasmic astrocytes that express EAAT1/EAAT2 but not GFAP (4) before acquiring the patchy stellate morphology of mature protoplasmic astrocytes that do not express GFAP (Figure 6 D – F) (5).Our data suggest that EAAT1 and EAAT2 are expressed in protoplasmic astrocytes during their development whereas GFAP is not expressed by such cells.

Figure 15. Summary diagram of the postulated developmental progression of protoplasmic astrocytes in the human cerebral cortex based upon observations with EAAT1 and EAAT2 immunostaining.

Protoplasmic astrocytes originate from a late migration of precursors from the subventricular zone through the white matter (1) which form the subpial granular layer of Ranke (SPLR) in layer 1 (2). SPLR is the source of astrocytes that form the EGLM as well as protoplasmic astrocyte precursors (3) that descend into the neocortex as immature protoplasmic astrocytes (4) before acquiring the stellate morphology of mature astrocytes (5) (Marin-Padilla, 1995). We found co-localization of EAAT2 (blue) with GFAP (red) immunopositive astrocytes whose end feet are attached to the EGLM. We also found EAAT1/EAAT2 immunopositive cells that do not colocalize with GFAP in layer 1 (Figure 6 A–C) suggesting the existence of immature protoplasmic astrocytes that express EAAT1/EAAT2 but not GFAP (4) before acquiring the patchy stellate morphology of mature protoplasmic astrocytes that do not express GFAP (Figure 6 D–F) (5).

Indeed, mature protoplasmic astrocytes are known to modulate cortical function by synchronizing neuronal firing (Fellin et al., 2004, Kozlov et al., 2006) coordinating neuronal networks. Three-dimensional reconstructions of cortical astrocytes and neurons in the mouse demonstrate that cortical astrocytes occupy non-overlapping territories, and that a single astrocyte enwraps 4–8 neuronal somata similar to human protoplasmic astrocytes. However, human protoplasmic astrocytes have a three-fold larger diameter and tenfold more processes than those in rodents allowing them to encompass approximately 2 million synapses compared to 100,000 in rodents (Halassa et al., 2007, Stevens, 2008, Halassa et al., 2009). Glutamate transporters move glutamate from the synaptic cleft into astrocytes and bind glutamate on a fast timescale suggesting that their pervasive expression on protoplasmic astrocytes, which surround synapses, as well as their fast binding kinetics modulate synaptic transmission (Asztely et al., 1997, Kullmann and Asztely, 1998, Tzingounis and Wadiche, 2007). As demonstrated in Figure 5 C, however, there are areas in the cerebral cortex that are not immunopositive for EAAT2 or EAAT1consistent with previous findings (Williams et al., 2005). Transgenic mice with GFAP promoter-driven enhanced green fluorescent protein (eGFP) also demonstrated glutamate transporter currents in some but not all eGFP expressing cells in the hippocampus (Matthias et al., 2003). Taken together these data suggest that there are many synapses in the cortex that are located in regions in which glial glutamate transporters are not expressed in astrocytes. Furthermore, ultrastructural studies demonstrate that not all excitatory synapses have an adjacent astrocytic process (Ventura and Harris, 1999). Since glutamate needs to be cleared from all excitatory synapses, these observations suggest a role for expression of glutamate transporters located pre-synaptically (Chen et al., 2002, Furness et al., 2008, Melone et al., 2009, de Vivo et al., 2010) and post-synaptically (Scimemi et al., 2009).

Developmental Changes in EAAT2 Protein Expression by Western Blot Analysis

We found that EAAT2 protein is expressed as a single band at 67 kDa, present in the human cerebral cortex as early as 19 gestational weeks, the earliest age examined by us, and remains relatively constant throughout the second half of gestation into infancy. Interestingly, the banding pattern on immunoblot changes after birth (at about 2 postnatal months) with the appearance of a second band at 77 kDa. This change suggests a post-translational modification of cortical EAAT2 after birth. A similar molecular weight difference has been previously reported and attributed to glycosylation (Furuta et al., 1997b, Furuta et al., 1997a, Danbolt, 2001). A mutation that reduces glycosylation has been shown to decrease glutamate transporter function in patients with amyotrophic lateral sclerosis suggesting that the presence of glycosylated GLT1 may represent an increase in glutamate transporter function after birth (Trotti et al., 2001). We did not find the higher molecular weight mass band in western blot analysis of human cerebral white matter during the same time-frame with the same antibody (DeSilva et al., 2007); suggesting that this modification is specific to the cerebral cortex. Given that the western blot analysis provides information about EAAT2 expression in homogenates, it is not known if this post-translational modification is restricted to one or more cell types. However, it appears that the time of expression of EAAT2 in protoplasmic astrocytes, which occurs after term birth, coincides with the time that the second EAAT2 protein band appears in western blot analysis. Further research into the molecular basis of the developmental changes in EAAT2 protein expression is needed.

Conclusions

In conclusion, the differential expression of EAAT1, EAAT2, and EAAT3 undergoes maturational changes during late gestation through infancy, a time-span that coincides with the period of high vulnerability of the developing brain to excitotoxic insults. Reversal of glutamate transport is known to occur under conditions of energy failure resulting in excitotoxicity (Roettger and Lipton, 1996, Seki et al., 1999, Fern and Möller, 2000, Rossi et al., 2000, Deng et al., 2003). Therefore, transient expression of EAAT2 may contribute to the selective vulnerability of layer V pyramidal neurons in human brains with periventricular leukomalacia (Andiman et al., 2010). Additionally, transient expression of EAAT2 in subplate neurons, susceptible to excitotoxicity in an animal model of neonatal hypoxic-ischemic injury (McQuillen et al., 2003, Nguyen and McQuillen, 2010), may also contribute to their loss. Of additional major interest is our finding that EAAT1 and EAAT2 are both expressed in protoplasmic astrocytes. The gliogenesis of protoplasmic astrocytes in the cerebral cortex is a relatively late developmental event in the second half of gestation and into early infancy, i.e., the time-frame of perinatal brain injury. While current attention focuses upon the role of neuronal injury in cortical dysfunction in long-term survivors, our data suggest that injury to astrocytes may contribute to impaired cognitive processing in preterm survivors due to the emerging role of protoplasmic astrocytes in neuron-astrocyte signaling. We speculate that injury to immature astrocyes which do not yet express GFAP and are destined for the cerebral cortex via late-gestation migration in the damaged white matter of periventricular leukomalacia may result in an (as yet unidentified) under-population of mature protoplasmic astrocytes in the cerebral cortex and secondary neuronal-astrocytic unit dysfunction and impaired cognition. Changes in astrocytes during perinatal brain injury is supported by the presence of reactive astrocytosis and an increase in the expression of GLT1 in reactive astrocytes in human brains with periventricular leukomalacia (Desilva et al., 2008). Thus, our study raises important issues for further analysis in understanding the role astrocytes, glutamate transporters, and neuronal-astrocytic interactions in the development and pathology of human cortex.