Investigating gradients of gene expression involved in early human cortical development


Dr Nadhim Bayatti, Institute of Human Genetics, Newcastle University, International Centre for Life, Central Parkway, Newcastle upon Tyne NE1 3BZ, UK. T: +44 191 2418705; F: +44 191 2413022; E:


The division of the neocortex into functional areas (the cortical map) differs little between individuals, although brain lesions in development can lead to substantial re-organization of regional identity. We are studying how the cortical map is established in the human brain as a first step towards understanding this plasticity. Previous work on rodent development has identified certain transcription factors (e.g. Pax6, Emx2) expressed in gradients across the neocortex that appear to control regional expression of cell adhesion molecules and organization of area-specific thalamocortical afferent projections. Although mechanisms may be shared, the human neocortex is composed of different and more complex local area identities. Using Affymetrix gene chips of human foetal brain tissue from 8 to 12.5 post-conceptional weeks [PCW, equivalent to Carnegie stage (CS) 23, to Foetal stage (F) 4], human material obtained from the MRC-Wellcome Trust Human Developmental Biology Resource (, we have identified a number of genes that exhibit gradients along the anterior–posterior axis of the neocortex. Gene probe sets that were found to be upregulated posteriorally compared to anteriorally, included EMX2, COUPTFI and FGF receptor 3, and those upregulated anteriorally included cell adhesion molecules such as cadherins and protocadherins, as well as potential motor cortex markers and frontal markers (e.g. CNTNAP2, PCDH17, ROBO1, and CTIP2). Confirmation of graded expression for a subset of these genes was carried out using real-time PCR. Furthermore, we have established a dissociation cell culture model utilizing tissue dissected from anteriorally or posteriorally derived developing human neocortex that exhibits similar gradients of expression of these genes for at least 72 h in culture.


The areal development of the neocortex involves the proliferation and differentiation of the neuroepithelial cells of the cortical primordium, resulting in a complex multi-layered structure with regionally diverse cytoarchitecture. A large body of evidence in rodents has demonstrated that initially this process is controlled by a number of genes with overlapping gradients of expression in the cortical primordium, leading to the establishment of a cortical map. Later extrinsic influences such as incoming thalamocortical afferents refine the areas of gene expression and promote the formation of cytoarchitectural differences, resulting in the final map. In humans, damage to the neocortex can occur during the pre- or perinatal period [the motor cortex and corticospinal tract are most commonly affected (Eyre, 2007)], but the developing human brain is plastic in nature, and substantial reorganization of connectivity can take place (Basu et al. 2010). Studies both on human regional identity and on characterizing differentiation and origin of neurones in different regions are necessary to understand the mechanisms underlying these reorganizations. This knowledge should make assessments of the probable outcomes after developmental brain injury more accurate and help to develop new therapies.

The vast majority of published work has focused on rodent development (Muzio et al. 2002; O’Leary et al. 2007; Rakic et al. 2009) and although many mechanisms involved are shared in common with humans, the human neocortex is composed of diverse and more complex local area identities reflecting differences in structure and function (Monuki & Walsh, 2001; Molnar et al. 2006). Animal models are therefore not sufficient and appropriate human studies and human model systems are necessary.

Our previous work has highlighted important differences in human gene expression patterns and localization compared to previously published data in rodents (Bayatti et al. 2008a,b). The two best characterized transcription factors presumed to play a role in rodent neocortex regionalization are Pax6 and Emx2 (Muzio et al. 2002). Mutant mice experiments have shown evidence for their roles in controlling anterior and posterior cortical area specification, respectively (Bishop et al. 2000). We analysed their gradients of expression and observed similarities and differences with previously published rodent data (see Table 1). At 8 post-conceptional weeks [PCW; equivalent to Carnegie stage (CS) 23] they exhibited complementary gradients as described in rodents. However, from 9 PCW [equivalent to Foetal (F) stage 1] PAX6 failed to exhibit the high anterior/lateral, low posterior/medial gradient (seen at 8 PCW, or CS23), whereas the EMX2 gradient (high posterior/medial, low anterior/lateral) persisted. The loss of the PAX6 gradient at 9 PCW (F1) is still early in cortical development (cortical plate starts to form at 7.5 PCW, or CS21) and even layer VI has not formed fully (Bayatti et al. 2008a). Although in rodents Pax6 has been found to be expressed in a gradient throughout neurogenesis (Manuel & Price, 2005), our observation supports recent evidence in rodents suggesting that Pax6 is not directly involved in regionalization (Manuel et al. 2007; Pinon et al. 2008) and therefore studies must be carried out to identify potential anterior markers involved in regional identity. The observation of reciprocal gradients at early stages of human cortical development implies that regionalization events may be occurring relatively early in the developing human brain and that cells in the developing neocortex at this stage may exhibit a degree of intrinsic identity information (Rakic, 1991; Rakic et al. 2009).

Table 1.   Summary of the observed cortical expression patterns of EMX2 and PAX6 (gradients and laminar localization) in the human brain as compared to previous studies in rodent. Laminae highlighted in bold indicate localization in human, but not in rodents.
 GradientLaminar localization
  1. Compiled by comparing studies reported in O’Leary et al. (2007) and Bayatti et al. (2008b). CP, cortical plate; SVZ, subventricular zone; VZ, ventricular zone.

↓Posterior/medial at 8 GW, no gradient afterwards

To identify potential anterior markers during human cortical development and assess whether any putative markers may be intrinsically regulated, we have analysed gradients of gene expression in developing cortical tissue between 8 and 12.5 PCW (CS23 and F4) using an Affymetrix whole genome chip array. Expression differences for a number of genes were confirmed by real-time PCR in cortical tissue. In addition, an in vitro human cell culture model of regionalization was established by dissecting tissue from anterior and posterior neocortex, dissociating and culturing in isolation. Gene expression patterns previously confirmed by real-time PCR for tissue were also analysed in this model, and a number of these exhibited differential regulation when comparing between anterior- and posterior-derived cultures, suggesting that cells out of their environment also maintain intrinsic identity.

Materials and methods

Human tissue

Brains were dissected from human foetal and embryonic terminations of pregnancies obtained from the MRC-Wellcome Trust Human Developmental Biology Resource at Newcastle University (HDBR, Tissue from ages between 8 and 12 PCW (CS23–F4, total n = 10, 8 PCW, CS23, n = 3; 10–10.5 PCW, F2, n = 4; 12 PCW, F4, n = 3) for real-time(rt) PCR, between 8 and 12.5 PCW (CS23–F4, total, n = 6, 8–12.5 PCW; 8–9 PCW, CS23–F1, n = 2, 10–10.5 PCW, F2, n = 2, 11–12.5 PCW, F3–F4, n = 2) for Affymetrix Gene Chip Analysis. For dissociated cell cultures, tissue aged 11 PCW, F3 (n = 4) was used. All tissues were used with maternal consent and approval of the Newcastle & North Tyneside 1 Research Ethics Committee. Age was estimated from measurements of foot length and heel to knee length and compared with a standard growth chart (Hern, 1984). Right-sided cortical hemispheres were dissected, subcortical structures, temporal lobes and meninges of the brains were removed, and cortices were cut with a scalpel into 5-mm coronal slices along the anterior–posterior axis for Affymetrix Gene Chip Analysis and rtPCR (Supporting Information Fig. S1). RNA was extracted from two slices per brain, the anterior-most and posterior-most slices, for comparison of RNA expression levels. For cell culture, after removal of meninges, cortices from each hemisphere were divided into three equally sized slices along the anterior–posterior axis and cells were isolated from the anterior-most and posterior-most slices and cultured separately.

Dissociated cell cultures

Short-term dissociated human foetal cell cultures were initiated according to an established rodent protocol with minor modifications (Franke et al. 2000). In brief, anterior-most and posterior-most cortical slices were dissected under sterile conditions in Ca2+- and Mg2+-free Dulbecco’s phosphate-buffered saline (DPBS; Invitrogen, Paisley, UK). Cell culture was only carried out with dissected tissue aged 11 PCW (F3). The tissues were cut into small pieces and incubated for 20 min in Hank’s balanced salt solution (HBSS; Invitrogen) containing 0.05% trypsin-ethylenediamine tetra-acetic acid (EDTA; Invitrogen). Trypsin action was terminated by transferring tissue pieces to HBSS supplemented with 10% heat-inactivated foetal calf serum (FCS; Invitrogen). The tissue was gently dissociated by trituration through a plastic pipette and passed through a 70-μm-pore nitex mesh (Becton Dickinson, Oxford, UK). Cells were pelleted at 400 g for 5 min and resuspended in minimum essential medium (MEM; Invitrogen) supplemented with 10% FCS. Cells were plated at 200 000 cells cm−2 into 12-well culture dishes (Falcon, Franklin Lakes, NJ, USA) coated with poly-l-lysine (0.01% solution; molecular weight 70–150 kDa; Sigma-Aldrich, Poole, UK). After 24 h [1 day in vitro (DIV)], cells were switched to serum-free MEM:F12 supplemented with B27 (Invitrogen) for a further 48 h. RNA was isolated from cells after a total of 3 DIV and subjected to reverse transcription for rtPCR or cells were fixed with 4% paraformaldehyde (PFA; Sigma-Aldrich)/PBS and subjected to immunocytochemistry.

RNA isolation and reverse transcription reaction

Total RNA was isolated from the anterior- and posterior-derived neocortex and dissociated cell cultures using PeqGOLD RNAPure reagent (Peqlab, Fareham, UK) according to the manufacturer’s instructions. Total RNA concentration was determined by spectrophotometric absorbance at 260 nm using NanoDrop 8000 (Fisher Scientific, Loughborough, UK). A total of 2 μg of RNA was reverse transcribed using 200 U μL−1 of SuperScript Reverse III Transcriptase Kit (Invitrogen) and 2 μg random hexadeoxynucleotide primers (Promega, Southampton Science Park, UK) in a final volume of 50 μL following the manufacturers’ instructions. The concentration of the transcribed cDNA template was further diluted twofold prior to rtPCR.

Gene expression arrays

Gene expression studies of brain tissue were carried out using the Human Genome U133 Plus 2.0 Array (Affymetrix UK Ltd., High Wycombe, UK). Preparation of in vitro transcription products, hybridization and scanning using the GeneChip scanner 3000 were performed according to Affymetrix protocols using a minimum of 1 μg of total RNA to prepare antisense biotinylated RNA without a second round of amplification. The quality of hybridization was assessed in all samples following the manufacturer’s recommendations.

Quality control was assessed using the affy package for bioconductor (; RNA degradation plots and PM intensity charts) and GeneSpring GX 7.3 (box plots; Agilent Technologies, South Queensferry, UK) software, and GeneSpring GX 7.3 was used for expression level analysis. Expression data for each probe was normalized by gc-rma, and gene expression levels are presented as a fold change with a threshold of 1.75-fold (this threshold was chosen, as EMX2 was measured at 1.77-fold and had previously been confirmed to exhibit a gradient on the anterior–posterior axis by in situ hybridization in tissue sections (see Bayatti et al. 2008b)). Gene ontologies were analysed using the GO Ontology Browser within the GeneSpring GX analysis platform (Agilent Technologies). All regulated genes passed a filter on confidence set at P < 0.05. All raw data CEL files and experiment data have been submitted to MIAMExpress (; Experiment submission: ‘Investigating gradients of gene expression involved in early human cortical development’).

Real-time PCR (rtPCR)

The mRNA expression of various target genes in the anterior-most and posterior-most of the neocortex and dissociated cultured cells were determined by quantitative real-time PCR. The sequence-specific primer sets (Eurofins MWG Operon, Raynes Park, UK) and amplicon sizes for all target genes are listed in Table 2. Three housekeeping genes, β–ACTIN, GAPDH and SDHA, were used as internal references to normalize the cDNA template between different samples. Negative control was incorporated by replacing the cDNA template with Molecular Biology grade water (VWR International, Lutterworth, UK). The SYBR Green-based rtPCR assay was performed in the 7900HT Fast Real-Time PCR system (Applied Biosystems, Lincoln Centre Drive, Foster City, CA, USA). A total volume of 10 μL rtPCR reaction was set up in triplicates, containing 5 μL of 2× SYBR Green qPCR Master Mix (Invitrogen), 1 μL of the diluted cDNA template, 0.5 μL of each primer (10 pmol μL−1), and 3 μL of Molecular Biology grade water (VWR International). The thermal cycle protocol for SYBR Green-based rtPCR was 95 °C for 15 min, followed by 40 cycles of 95 °C for 15 s, 60 °C for 30 s, 72 °C for 30 s and 74 °C for 10 s. Amplification of a single PCR product was confirmed by dissociation curve analysis of PCR products (68 °C for 10 s followed by 99 °C for 10 s) after completion of each rtPCR reaction. Baseline was set automatically and threshold was set manually above all the background signals. The data obtained were analysed with qBase software (Hellemans et al. 2007). Normalized relative quantities (NRQ) of mRNA expression were calculated in qBase and NRQ data was presented with fold changes calculated for each gene comparing the two regions of the neocortex or cells ± SEM. Paired Student’s t-test was performed to determine the significance differences between expression levels from anteriorally and posteriorally derived tissue or cells (< 0.05).

Table 2.   Gene-specific primer sequences used for rtPCR in this study.
Primer set5′ to 3′ SequenceProduct size (bp)
  1. *Housekeeping genes.



After fixation with 4% PFA/PBS, cells were washed three times with 1× PBS. The cells were incubated with primary antibodies and 3% appropriate blocking serum (Vector Laboratory, Peterborough, UK) in 0.1% Triton X-100/PBS (PBST) in a moist chamber at 4 °C overnight with gentle agitation. Table 3 provides details of all the primary and secondary antibodies used, as well as blocking sera. After washing three times in 0.1% PBST, cells cultured in 12-well plates were incubated with appropriate biotinylated secondary antibodies (1 : 200 in 0.1% PBST; Vector Laboratory) at 4 °C for 2 h. Subsequently, cells were washed three times in 0.1% PBST. Cells were then incubated with streptavidin-horseradish peroxidase (1 : 200 in 0.1% PBST; Vector Laboratory) at 4 °C for 1 h. After incubation, cells were washed three times in 0.1% PBST then incubated with the 3, 3′-Diaminobenzidine Enhanced Liquid Substrate System (DAB; Sigma-Aldrich) for up to 10 min before washing in PBS. All cells were then incubated with DAPI (1 : 10000 in 0.1% PBST; Invitrogen) for 5 min at room temperature in the dark and subsequently washed three times in PBS. Cells in 12-well plates were kept in PBS at 4 °C in the dark until examination by microscopy. False-colour images of random fields of view were taken for each antibody stain (n = 4) and merged. DAB staining was assigned to the red channel, and DAPI assigned to green. The percentage number of stained cells was calculated from counting numbers of specifically stained cells (red + yellow) and comparing with total number of cells (red + green + yellow).

Table 3.   Primary and secondary antibodies used in this study.
Primary antibodySupplierDilutionSpecies/blocking serumSecondary antibody
EMX2Sigma-Aldrich E67801 : 1000Rabbit polyclonal/goatBiotinylated goat anti-rabbit IgG
GFAPSigma-Aldrich G38931 : 400Mouse monoclonal/horseBiotinylated horse anti-mouse IgG
MAP2Sigma-Aldrich M44031 : 1000Mouse monoclonal/horseBiotinylated horse anti-mouse IgG
NESTINChemicon, Watford, UK MAB53261 : 400Mouse monoclonal/horseBiotinylated horse anti-mouse IgG
PAX6Covance, Cambridge Bioscience, Cambridge, UK PRB-278P1 : 300Rabbit polyclonal/goatBiotinylated goat anti-rabbit IgG


Human genome microarrays identify anteriorally and posteriorally regulated genes in the early developing human neocortex

To search and identify genes that are expressed in gradients across the developing neocortex, tissue was dissected from embryonic and foetal human brains (between 8 and 12.5 PCW, CS23–F4; n = 6), from which RNA was extracted, in vitro transcribed and hybridized onto Affymetrix U133plus2 chips (n = 12 for analysis). Quality control tests were carried out (Supporting Information Fig. S1), gene expression levels were analysed in GeneSpring GX7.3.

Clustering studies, as analysed by condition tree analysis (Fig. 1A) and principal component analysis (Supporting Information Fig. S1) indicated that gestational age was a variable affecting total gene expression between samples. In the condition tree, close clustering was observed at ages 9–11 PCW (F1–F3), whereas samples aged 8 (CS23) and 12.5 PCW (F4) exhibited more variability. Using 1.75 as a threshold fold change, we were able to identify 383 Affymetrix probe sets that are more highly expressed anteriorally than posteriorally (Fig. 1B, Supporting Information Table S1), and 154 probe sets that exhibit higher expression posteriorally than anteriorally (Fig. 1B, Supporting Information Table S2). A condition tree analysis using a gene list comprising of genes listed in Supporting Information Tables S1 and S2 resulted in close clustering of samples according to location, i.e. discrete clustering of anterior samples and posterior samples (data not shown).

Figure 1.

 Affymetrix Chip Analysis of genes differentially expressed across the anterior–posterior axis of the developing human neocortex. Condition tree (A), calculated by GeneSpring GX and showing levels of all genes indicates that clustering of samples occurs in an age-dependent manner. At each indicated age, two chips corresponding to levels of genes from anteriorally and posteriorally derived tissue are indicated. Scatter plot (B) shows expression levels of genes posteriorally (X-axis) and anteriorally (Y-axis), plotted against each other. Genes marked in white correspond to those found to be differentially regulated along the anterior–posterior axis by at least 1.75-fold. Colours correspond to an arbitrary ‘heatmap’, indicating expression levels.

Functional annotation clustering analysis was carried out, using david ( and ontology categories listed in Tables 4 and 5 are specific to either anteriorally or posteriorally expressed genes respectively. Anteriorally upregulated genes included a large number involved in cell-signalling, cell adhesion, proliferation and morphogenesis, whereas a number of the genes expressed posteriorally during development were involved in ion transport.

Table 4.   Gene ontology categories enriched in anteriorally but not posteriorally upregulated probe sets.
TermCount%P value
  1. Carried out on david Bioinformatics database 6.7 beta ( Count refers to the number of genes present within the category; %, the proportion of genes within anteriorally upregulated probe set present in the GO category, and P-value corresponds to statistical reliability of the enrichment of this GO category within the probe set list.

GO:0007242–intracellular signalling cascade3311.871.63E-03
GO:0042127–regulation of cell proliferation3010.793.25E-06
GO:0007155–cell adhesion279.718.58E-06
GO:0022610–biological adhesion279.718.79E-06
GO:0009891–positive regulation of biosynthetic process279.719.01E-06
GO:0010604–positive regulation of macromolecule metabolic process279.711.26E-04
GO:0031328–positive regulation of cellular biosynthetic process269.352.23E-05
GO:0050877–neurological system process269.354.42E-02
GO:0045935–positive regulation of nucleobase, nucleoside, nucleotide and nucleic acid metabolic process258.991.28E-05
GO:0051173–positive regulation of nitrogen compound metabolic process258.992.16E-05
GO:0010557–positive regulation of macromolecule biosynthetic process258.992.56E-05
GO:0032989–cellular component morphogenesis248.634.61E-09
GO:0007267–cell–cell signalling248.631.85E-05
GO:0000902–cell morphogenesis227.911.31E-08
GO:0042981–regulation of apoptosis217.551.25E-02
GO:0043067–regulation of programmed cell death217.551.38E-02
GO:0010941–regulation of cell death217.551.43E-02
GO:0032990–cell part morphogenesis207.191.20E-09
GO:0008284–positive regulation of cell proliferation207.199.74E-06
GO:0010033–response to organic substance207.199.10E-03
GO:0031175–neurone projection development196.833.69E-08
GO:0048812–neurone projection morphogenesis196.832.11E-09
GO:0048667–cell morphogenesis involved in neurone differentiation196.832.44E-09
GO:0048858–cell projection morphogenesis196.834.00E-09
GO:0000904–cell morphogenesis involved in differentiation196.832.59E-08
GO:0019226–transmission of nerve impulse196.833.26E-06
GO:0009719–response to endogenous stimulus176.122.71E-04
GO:0007268–synaptic transmission165.762.97E-05
GO:0045165–cell fate commitment155.401.30E-08
GO:0016337–cell–cell adhesion155.405.18E-05
GO:0009725–response to hormone stimulus155.409.45E-04
GO:0007423–sensory organ development145.045.37E-05
GO:0044057–regulation of system process145.044.13E-04
GO:0040008–regulation of growth145.049.59E-04
Table 5.   Gene ontology categories enriched in posteriorally but not anteriorally upregulated probe sets.
TermCount%P value
  1. Carried out on david Bioinformatics Database 6.7 beta ( Count refers to the number of genes present within the category; %, the proportion of genes within posteriorally upregulated probe set present in the GO category. The P-value corresponds to statistical reliability of the enrichment of this GO category within the probe set list.

GO:0006811–ion transport1311.821.58E-03
GO:0030001–metal ion transport87.271.80E-02
GO:0006812–cation transport87.274.06E-02
GO:0001501–skeletal system development76.361.11E-02
GO:0043009–chordate embryonic development76.361.60E-02
GO:0009792–embryonic development ending in birth or egg hatching76.361.68E-02
GO:0006816–calcium ion transport65.451.41E-03
GO:0015674–di-, tri-valent inorganic cation transport65.453.53E-03

A number of known posterior markers were identified, including FGFR3, COUPTFI and EMX2 (Simeone et al. 1992; Zhou et al. 1999; O’Leary & Nakagawa, 2002). Anteriorally, a number of potential markers of frontal and motor cortex as well as corticospinal tract neurones were observed including CNTNAP2 (Abrahams et al. 2007), PCDH17 (Kim et al. 2007), and S100A10 (Arlotta et al. 2005; Table 6). The pan-neuronal markers MAP2, GAP43 and Synaptophysin exhibited no gradients in gene expression when comparing levels in anteriorally vs. posteriorally derived tissue (data not shown).

Table 6.   Selected genes upregulated anteriorally or posteriorally in RNA extracted from human neocortex, hybridized to Affymetrix gene chip (U133plus2 human genome) and fold change variation analysed on GeneSpring GX software.
383 probe sets > 1.75-fold in anterior compared to posterior154 probe sets > 1.75-fold in posterior compared to anterior
  1. CST, corticospinal tract.

Protocadherin 172.84Motor cortex/arealFGFR315.57Areal
Cadherin 82.53Rostral/arealCOUPTFI/NR2F15.97Areal
S100A102.48Layer V/CSTEMX21.77Areal
Cadherin 62.34Cell adhesion   
ER812.28Layer V   
ROBO12.05Axon guidance/CST   
Cadherin 101.95Cell adhesion   
Protocadherin-γ A31.95Cell adhesion   
CTIP21.94Layer V/CST   
LMO41.94Layer V   
Protocadherin 191.88Cell adhesion   
Protocadherin 201.75Cell adhesion   

rtPCR confirmation of a subset of anterior/posterior-regulated genes in human cortical tissue

Confirmation of the differential expression of a subset of genes was carried out in embryonic and foetal human tissue dissected from brains aged between 8 and 12 PCW (CS23–F4, n = 10). Slices 5 mm thick were dissected from the anterior and posterior poles of the developing human brain, and after RNA extraction and reverse transcription, rtPCR was carried out with gene-specific primers (Table 2). After normalizing to three housekeeping genes (beta-actin, SHDA, and GAPDH), rtPCR confirmed the results of the Affymetrix gene chip analysis (Fig. 2). The putative anterior markers CNTNAP2 (2.33-fold), PCDH17 (4.59-fold) and S100A10 (2.24-fold) were found to be significantly higher in anteriorally compared with posteriorally dissected tissue. The previously identified rodent posterior markers FGFR3 (61.36-fold), COUPTFI (18.98-fold) and EMX2 (1.59-fold) exhibited significantly higher expression posteriorally than anteriorally. Averaged over 8–12 PCW (CS23–F4), PAX6 exhibited no statistical significance either anteriorally and posteriorally. FGFR2 (1.21-fold), was found to be expressed more highly posteriorally, and FGFR1 (1.21-fold) and MAP2 (1.18-fold) more highly anteriorally. These differences were small but statistically significant (Fig. 2B).

Figure 2.

 rtPCR confirmation of a subset of differentially regulated genes during early human neocortical development (8–12 PCW). Table indicating fold changes (A), and graphical representation (B) of NRQ of a subset of genes determined by rtPCR from RNA extracted from anterior and posterior regions of developing human neocortex aged between 8 and 12 PCW (n = 10). Genes exhibiting a high anterior to low posterior gradient include CNTNAP2, PCDH17, and S100A10. Genes exhibiting a high posterior, low anterior gradient include EMX2, FGFR3, and COUPTFI. No gradients were detected for PAX6. *P < 0.05, n.d., not detected; A, anterior, P, posterior.

Characterization of human cortical dissociated neuronal cultures

A putative in vitro model of human regionalization was established by initiating human cortical cultures using dissected and dissociated anterior and posterior slices of human neocortex aged 11 PCW (F3; Fig. 3A). Cell cultures were maintained in MEM 10% FCS for 24 h followed by 3 DIV in serum-free MEM:F12 medium supplemented with B27. Cell cultures were initially characterized by testing immunoreactivity for GFAP, Nestin, and MAP2, markers of astrocytes, neural progenitors and neurones, respectively (Fig. 3A). Cell counts were also carried out and indicate that MAP2-positive neurones account for the majority of cells cultured (75.1 ± 6.1%), while presumably overlapping populations of Nestin- and GFAP- positive cells accounted for 26.6 ± 4.8% and 19.2 ± 2.75%, respectively. There were no observable differences in the levels of these genes when comparing anteriorally and posteriorally derived cultures (data not shown).

Figure 3.

 Characterization of human neuronal cortical cultures. Neuronal cell cultures were initiated from 11 PCW human neocortex and maintained for 3 DIV. Representative photographs in (A) depict cells under phase contrast, and stained with antibodies for GFAP, NESTIN and MAP2. Cell counts from random fields of vision (n = 4) indicate that MAP2-positive neurones account for most of the cells in culture (75.1 ± 6.1%), while GFAP- and Nestin-positive cells account for 19.2 ± 2.75% and 26.6 ± 4.8%, respectively. In each case, immuno-positive cells are indicated by filled arrowheads, and examples of negatively stained cells are indicated by empty arrowheads. Scale bar: 20 μm.

In addition to general markers, cultures were analysed for immunoreactivity to EMX2 and PAX6, putative regionalization markers (Fig. 4A). Cell counts from random fields of view (n = 4 for each) indicated differences in levels of EMX2 but not PAX6 when comparing anteriorally and posteriorally derived cultures (Fig. 4B). The percentage of cells immunoreactive for EMX2 was significantly higher in cell cultures derived from the posterior neocortex (80.7 ± 3.0%) as compared with cultures initiated from the anterior neocortex (37.9 ± 18.3%) indicating a 2.13-fold increase (posterior vs. anterior) between the two cultures. PAX6, however, exhibited no significant differences between cultures (anterior, 49.37 ± 3.6%; posterior, 57.4 ± 0.9%, 0.86-fold, anterior vs. posterior).

Figure 4.

 Anteriorally and posteriorally derived human cortical cultures exhibit differences in EMX2 but not in PAX6 expression. Immunohistochemistry was carried out for EMX2 and PAX6 on cultures derived from tissue dissected from the anterior and posterior poles of 11 PCW aged neocortex (A). Cell counts carried from random fields of vision (n = 4 per anterior and posterior) show a significant increase in numbers of cells expressing EMX2 (filled arrowheads) posteriorally than anteriorally. However, PAX6 (filled arrowheads) showed no differences between cultures. EMX2- or PAX6- negative cells are indicated by empty arrowheads. Scale bar: 20 μm. *P < 0.05.

Real-time PCR confirmation of a subset of regulated markers in human neocortical neuronal tissue culture

To analyse whether cortical cells maintain their genetic/areal identity when removed from their natural physiological and anatomical environment and maintained in vitro, we carried out rtPCR on reverse transcribed RNA extracted from cell cultures initiated from 11 PCW (F3) neocortex and maintained for 3 DIV (Fig. 5). Examining the same subset of genes as characterized in vivo, a similar pattern of results to that of tissue was obtained when comparing expression levels in anteriorally and posteriorally derived cultures. CNTNAP2 (2.67-fold), PCDH17 (3.92-fold) and S100A10 (2.76-fold) were found to be significantly higher in anteriorally than posteriorally derived cultures, whereas the posterior markers FGFR3 (59.28-fold), COUPTFI (10.18-fold), EMX2 (1.73-fold) exhibited significantly higher expression when compared to anteriorally derived cultures (Fig. 5A,B). Three of the genes, FGFR2, MAP2 and PAX6, exhibited no statistically significant difference in expression either anteriorally or posteriorally. FGFR1 (1.32-fold) was found to be marginally but significantly expressed at higher levels anteriorally (Fig. 5B).

Figure 5.

 rtPCR confirmation for a subset of differentially regulated genes during early human neocortical development in cultures derived from 11 PCW human neocortex. Table indicating fold changes (A), and graphical representation (B) of NRQ of a subset of genes determined by rtPCR from RNA extracted from anteriorally and posteriorally derived cultures of human neocortex aged 11 PCW. CNTNAP2, PCDH17 and S100A10 exhibited a high anterior to low posterior gradient. EMX2, FGFR3 and COUPTFI exhibited a high posterior, low anterior gradient. No gradients were detected for PAX6, FGFR2 or MAP2. *P < 0.05, n.d., not detected; A, anterior, P, posterior.


Utilizing gene chip arrays, rtPCR and cell culture, this study identifies and confirms a number of genes that are expressed in gradients across the developing human neocortex between 8 and 12 PCW (CS23–F4), an important time of corticogenesis. This corresponds to the period of development when the cortical plate is forming [at 7.5 PCW (CS21), Meyer et al. 2000] and during the early stages of neurogenesis (ten Donkelaar, 2000; Bystron et al. 2008) and regionalization, well before the ingrowth of thalamocortical fibres that may exert extrinsic influences on the differentiating progenitors (Kostovic & Rakic, 1990).

Genes that control the establishment of the early cortical map in rodents are presumed to be expressed in gradients throughout the period when neurogenesis is occurring in the developing neocortex (Cecchi & Boncinelli, 2000). Therefore we hypothesized that by assessing tissue derived from brains with an age range of 8–12 PCW (CS23-F4), we would uncover robust consistent gradients. When carrying out hierarchical clustering by condition tree analysis including all genes, it was interesting to note that samples clustered according to their age. This was also confirmed by principal component analysis, as age appears to be a major component of variance. Samples derived from anterior and posterior neocortex at each age range clustered closest to each other, the samples between 9 and 12.5 PCW (F1–F4) formed a closely clustered group, while samples from 8 PCW (CS23) were least related. These findings indicate that at 8 PCW (CS23), and probably before, different development mechanisms may be occurring, resulting in the differing gene expression profiles. This observation is supported by previous studies in human that demonstrate that a developmental switch may occur between 8 and 9 PCW (CS23 and F1) resulting in the deregulation of the PAX6 (high-anterior/lateral, low-posterior/medial) gradient, which is maintained in rodents throughout corticogenesis (Manuel & Price, 2005; Bayatti et al. 2008b). Recently, evidence has accumulated that the Pax6 gradient in rodents is not directly involved in regionalization processes (Manuel et al. 2007; Pinon et al. 2008), as over-expression of Pax6 in mutant mice does not alter the expression of Emx2 or other areal markers. Over-expression of Emx2, however, causes shifts in the position and increases the size of primary areas located posteriorally when compared to anterior structures (Hamasaki et al. 2004). Therefore there is a clear major regionalization influence in the developing neocortex which comes posteriorally from genes such as EMX2 and COUPTFI. Their expression is repressed by soluble factors such as FGFs-8, -15 and -17 released from signalling centres located anteriorally (O’Leary & Nakagawa, 2002; O’Leary et al. 2007; Rakic et al. 2009). There may be counter gradients of as yet undiscovered regionalization genes actively responsible for anterior specification, but these are still to be identified. There are numerous genes known to be expressed in counter gradients (Rakic et al. 2009); however, a distinction must be made between whether these genes affect regionalization per se or arise by being regulated by genes which would induce a region-specific or areal expression pattern. Classically such regionalization genes are expressed in the proliferative zones, while areal genes are expressed downstream in specific layers of the developing cortical plate (O’Leary et al. 2007).

Functional ontological analysis revealed expression profiles consistent with an enrichment of neural developmental processes occurring in both anterior- and posterior-derived tissue (Tables 4 and 5, respectively). However, genes classified within cell-signalling, adhesion, proliferation and cell death categories were upregulated in anterior compared to posterior-derived cells, suggesting that greater interactions between cells are taking place anteriorally than posteriorally. Conversely, posteriorally derived samples exhibit high expression levels of genes involved in ion transport, including ionotrophic neurotransmitters and transporters. Gene Ontology (GO) categorizations are useful for inferring global functional information but sometimes are unable to identify specific groups of genes that may be inter-linked either due to specificity or due to unknown or as yet unassigned gene function (e.g. GO category cerebral cortex regionalization GO:0021796 contains EMX1 and EMX2 but not COUPTFI). Therefore by searching manually through the gene list we were able to identify a number of anteriorally regulated genes which are potentially involved in regionalization at the anterior pole of the developing neocortex. These include cell-adhesion molecules (such as cadherins and protocadherins) and transcription factors that are markers of motor cortex or of corticospinal neurones, either at the site of origin of cortical motor neurones in layer V or expressed in their outgrowing fibres (Molnar & Cheung, 2006).

A subset of these genes was chosen for confirmation by rtPCR in dissected tissue. Of the anterior markers identified, CNTNAP2 is a frontal marker (Abrahams et al. 2007), PCDH17 is an areal marker corresponding to the motor cortex in rodents (Kim et al. 2007), and S100A10 is a layer V marker for corticospinal tract neurones (Arlotta et al. 2005); all exhibited robust and consistently higher levels in anteriorally derived tissue. EMX2, COUPTFI (Simeone et al. 1992; Zhou et al. 1999) and FGFR3 (O’Leary & Nakagawa, 2002) are reported regionalization genes and exhibited higher expression in posteriorally derived tissue. Additional members of the FGFR family that are highly expressed in the brain (Ford-Perriss et al. 2001) exhibited small but significant gradients of expression. FGFR1 exhibited higher expression anteriorally, whereas FGFR2 was higher posteriorally. PAX6 expression exhibited insignificant differences between anterior and posterior tissue. However, the neuronal marker MAP2 exhibited a small but significant increase in anterior tissue.

To establish whether the regulation of these genes observed within the brain is dependent on their physiological environment, or whether such information is held intrinsically, we generated dissociated neuronal cultures from neocortical tissue dissected from anterior and posterior poles of 11 PCW (F3) neocortices. The protomap hypothesis states that progenitors possess areal intrinsic information by the time they undergo their final asymmetric division in the ventricular zone and their progeny maintain this information, so that by the time they have migrated to the cortical plate, they may phenotypically look like all other cortical neurones but are genetically different to their neighbours (Rakic, 1991). The cell culture model employed here involves maintenance of the cells in a differentiating-inducing medium that results in the vast majority of cells reacting positively for the neuronal marker, MAP2. The detection of GFAP- and Nestin-positive cells suggests that radial glia/astrocytes and other progenitors are also present. Cell counts from these markers showed no significant differences between anterior or posterior cultures indicating no differences in the differentiation processes occurring in the cells after 3 DIV. Cells were also stained with antibodies for PAX6 and EMX2, and whereas PAX6-positive cell counts from anterior- and posterior-derived cultures exhibited no significant differences, EMX2-positive cells were significantly more numerous in posteriorally derived cell cultures. This supports our previous evidence that a PAX6 gradient is not detectable in the developing neocortex at this time point (Bayatti et al. 2008b). Furthermore, EMX2, which has been reported to be expressed in the proliferative zones during rodent development (Cecchi & Boncinelli, 2000), was observed to be expressed in the majority of cells in culture posteriorally, indicating that the protein must be expressed in post-mitotic neurones, again supporting recent evidence indicating that EMX2 is expressed within the cortical plate during development (Bayatti et al. 2008b). This observation raises the interesting question as to whether EMX2 exerts an areal influence in the cortical plate in addition to having a regionalization function in the proliferative zones in human. The relative levels of EMX2 in culture after 3 DIV (posterior vs. anterior) mirror the relative EMX2 levels measured in tissue (posterior vs. anterior) and this may indicate a degree of intrinsic information contained within expressing cells. Real-time PCR confirmed that expression levels of EMX2 and the other members of the subset of genes tested in cultures showed similar posterior/anterior differences as detected by Affymetrix chip analysis and rtPCR using tissue RNAs. Robust and consistent anterior > posterior levels of CNTNAP2, PCDH17 and S100A10 were detected, whereas levels of EMX2, COUPTFI and FGFR3 were found to be posterior > anterior.

We have therefore identified early anterior spatial clustering of a number of markers using whole genome Affymetrix chip analysis. These include potential motor cortex/cortical motor neurone/corticospinal tract markers. As this occurs relatively early in cortical development, this implies that cells may have this information imparted to them at an early stage, possibly as cortical progenitors. This evidence also suggests that the anterior neocortex at this stage may be the site of origin of the later developing motor cortex and corticospinal tract neurones in a similar manner to the anterior location of the motor cortex in rodents (O’Leary & Nakagawa, 2002). In addition to these anterior markers, we have confirmed a number of known regionalization genes clustering posteriorally. These include FGFR3, a gene postulated to be involved in regionalization but reported to be involved in regulating brain size (Inglis-Broadgate et al. 2005). This observation is intriguing as this implies that progenitors or differentiating neurones located anteriorally exhibit a different balance in expression of FGFRs (high FGFR1, low FGFR3) than those located posteriorally (high FGFR2/3), which could lead to differential signalling activity (Fortin et al. 2005). Thus FGF-8, a signalling molecule released from an anterior signalling centre and a high affinity ligand for FGFR3, could potentially activate different signalling pathways in anteriorally compared to posteriorally located progenitors. Further work should identify putative differences in cell-signalling between these populations of cells as well as down-stream effects, i.e. proliferation vs. differentiation.

Differences in expression between human and rodent genes are mirrored in the cell culture model presented here. Examples include detection of EMX2 expression in differentiated neuronal cells, and lack of an observable PAX6 gradient. These findings highlight the usefulness of this culture system as a model of regionalization in humans. When combined with other in vitro approaches, e.g. purifying specific-subpopulations of developing cortical cells (e.g. radial glia) in culture (Mo et al. 2007; Mo & Zecevic, 2008), levels of gene expression in progenitors (regionalization) and neuronal cells (areal) genes can be compared during the early stages of corticogenesis.


The human embryonic and foetal material was provided by the Joint MRCWellcome Trust Human Developmental Biology Resource ( at the IHG, Newcastle-upon-Tyne, UK. We thank the consenting women who made this study possible, A. Farnworth who gained consent on our behalf and Dr Lee Clough for technical help. This study was funded by the Wellcome Trust and the Anatomical Society of Great Britain and Ireland.