We constructed 10 SAGE libraries from the cerebral cortex of E15.5, E17.5 and 4- to 6-month-old adult C57BL/6 mice (N = 10; Table 1). The data from two additional SAGE libraries generated from E15.5 and P1.5 cerebral cortices from Gunnersen et al. 27 were also incorporated into our analysis. These SAGE libraries represent key stages of cerebral corticogenesis and are accessible from the Gene Expression Omnibus (GEO) website [GEO:

GSE15031

To identify differentially expressed tags/transcripts (DETs), we considered only those 25,165 unique tags with a count >2 across all libraries. Under stringent analyses (Table S1 in Additional data file 1), we identified a total of 561 DETs in various comparisons between developmental stages (Table 2; Figure S3 in Additional data file 1). A full list of DETs with assigned IDs is provided in Additional data file 2. Greater numbers of DETs are observed when the interval of two comparative developmental stages becomes wider. We find the largest number of DETs (approximately 59% or 328 DETs) in the embryonic versus adult stages (E versus Ad) followed by P1.5 versus Ad (approximately 34% or 192 DETs), E15.5 versus P1.5 (approximately 6% or 36 DETs) and E15.5 versus E17.5 (approximately 7% or 38 DETs) comparisons. These indicate distinctive transcript signatures during cerebral cortex development. Comparisons between rostral and caudal E15.5 (R versus C), and left and right adult cerebral cortices (L versus Ri) are described in a different section below.

Approximately 69% of DETs have been assigned to known genes and 6% to expressed sequence tags (ESTs), while 10% of DETs have multiple matches (tag matching multiple gene identifiers) and 8% having ambiguous matches (tag matching the same gene identifier at multiple chromosome loci). Approximately 8% of DETs have no matches and it is most likely that these DETs belong to transcripts from less defined regions, such as centromeric and telomeric areas of a chromosome or assembly gaps of the mouse genome.

To identify co-regulated genes, all 561 DETs were hierarchically grouped into 12 clusters based on the calculation of the Euclidean distance of logged normalized counts (Figure 1). Clusters 1, 5 and 6 consist of embryonic-specific DETs that exhibit the highest expression during embryonic development of the cerebral cortex. DETs in cluster 1 are expressed throughout all stages of development but exhibit the lowest expression in the adult cortex. Expression of DETs in cluster 5 ceases prior to birth, whereas DETs in cluster 6 are expressed up to early postnatal stage. On the other hand, clusters 4, 8 and 10 consist of adult specific DETs, showing very similar temporal expression profiles, but with different magnitudes (for example, highest expression in cluster 10). Clusters 2 and 7 are termed 'gene-switching' clusters as they show interesting expression-switching profiles. Cluster 2 shows an expression switch between P1.5 and adult stages whilst cluster 7 shows an expression switch between embryonic and adult stages. Clusters 3 and 9 consist of DETs showing region- (caudal region of E15.5 cerebral cortex) and stage-specific (P1.5 only) expression. Clusters 11 and 12 were excluded from subsequent analyses because they contained very few annotated tags. DETs within the same cluster may be co-regulated and/or involved in similar biological functions during cerebral corticogenesis.

We performed a systematic gene ontology functional clustering using the standardized Gene Ontology term analysis tool DAVID (Database for Annotation, Visualization and Integrated Discovery) 30 (Additional data file 3). Functional analysis of these gene clusters reveals that they have distinctive roles during cerebral corticogenesis. Embryonic-specific gene clusters (1, 5 and 6) are dominated by genes associated with cellular protein and macromolecule metabolic processes or biosynthesis, and nervous system and neuron development. These results match with the expected functional ontologies during embryonic cerebral cortex development in which neuronal migration, differentiation and axonogenesis events are at their peak. In contrast, adult-specific gene clusters (4, 8 and 10) consist of genes related to biological processes in the mature cerebral cortex, such as ion homeostasis, synaptic transmission and regulation of neurotransmitter level. In addition, these gene clusters are also enriched for ribonucleotide/nucleotide binding activity and components of cytoplasmic membrane-bound vesicles. These functional ontologies are in accordance with adult cerebral cortex function, which features synaptogenesis and nerve impulse transmission at synapses. Gene-switching clusters 2 and 7 are enriched with gene ontologies that are similar to both the embryonic- and adult-specific gene clusters. In addition, these gene clusters are also enriched for microtubule cytoskeleton organization and biogenesis, nucleotide biosynthesis and regulation of mRNA translation processes.

To ascertain the robustness of the SAGE datasets, we selected 136 candidate DETs and two additional genes of interests (ATPase, Cu++ transporting, alpha polypeptide, Atp7a, and cullin-associated and neddylation-dissociated 2, Cand2) for validation after considering both stage-to-stage and hierarchical based analyses (Table S2 in Additional data file 1). The selected DETs are transcription regulators, chromatin modifiers or post-translational regulators, such as ubiquitination pathway related molecules. Seventeen DETs are ESTs, which have been identified in brain-related mouse cDNA libraries or transcriptomes. Independent quantitative RT-PCR (RT-qPCR) validation was carried out using three biological replicates of unpooled cerebral cortex total RNA for each developmental stage. We validated 70 DETs (including 10 ESTs) from SAGE profiles of comparisons between two developmental stages (Additional data file 3) after considering various stringent criteria and cutoffs. The RT-qPCR results for all the 70 validated DETs, Atp7a and Cand2 are presented in Tables 3, 4, 5 and 6 (Additional data file 4). To validate the expression profiles of gene clusters from the hierarchical clustering analysis, we performed additional RT-qPCR analyses on 65 validated DETs (based on the analysis of two developmental stages), Atp7a and Cand2 by including other developmental stages (Additional data file 5). The analysis validated 62 out of 65 DETs (Figure 2). Of these, 22 are from embryonic gene clusters, 26 from adult gene clusters, and 14 from gene-switching clusters. We assigned Atp7a and Cand2 to clusters 2 and 5, respectively, based on their RT-qPCR expression profiles across all the developmental stages. See Tables 3, 4, 5 and 6 for complete list of DETs and their full gene names.

According to the comparison between two developmental stages, the most abundantly expressed and validated DETs in the E versus Ad analysis are Camk2a, Egr1 and Plp1 (Table 3). In the adult cerebral cortex, the expression of these DETs is more than 100-fold greater than that at E15.5. Other validated adult-specific DETs with expression levels of approximately tenfold or greater than those in the embryonic brain are Camk2n1, Cryab, Nrgn, BQ176089, CD802535, Sncb and Nptxr. Conversely, Sox11, an embryonic specific DET, is expressed in the E15.5 and E17.5 cerebral cortices with an expression level of at least 100-fold greater than that in the adult. Other validated embryonic specific DETs are Dcx, Zfp57, Ezh2, Sfrp1 and Cdk4, which have expression levels approximately tenfold or greater than those in the adult.

In the P1.5 versus Ad analysis, Dcx is expressed at a level approximately 80-fold greater in P1.5 compared to the adult cerebral cortex (Table 4). Other P1.5 enriched and validated DETs are (in descending order of enrichment) Zfp57, Csrp2, AA122503, Cdk4, Sox4, Marcks, Actb, BQ177889, Hmgb3, Rps4x, Actl6b, Zswim4 and Dr1, whose expression ranges from 33- to 1.4-fold greater than in the adult. On the other hand, the Plp1 is expressed at a level 40 times greater in the adult cerebral cortex compared to P1.5. Other validated genes that are enriched in the adult cerebral cortex include (in descending order) Hprt1, Calm1 and Mbp, with a 2.3- to 1.8-fold greater expression than the P1.5 cerebral cortex. Comparison between E15.5 and P1.5 shows that Mapt has a 1.5-fold greater expression level in the P1.5 cerebral cortex while Sox11 expression is 3.3-fold lower (Table 5).

We were unable to validate all 17 DETs from L versus Ri regions, suggesting the left and right hemispheres of the adult mouse cerebral cortex are highly similar and indistinguishable at the molecular level. SAGE and RT-qPCR analyses for R versus C regions of E15.5 are discussed in a separate section below.

The validated DETs of embryonic, adult and gene-switching clusters were functionally characterized using proprietary software, Ingenuity Pathway Analysis (IPA) from Ingenuity Systems^®, to identify enriched molecular networks and canonical pathways. Given a list of input genes (known as focus genes), IPA mapped these genes to a global molecular network developed from information contained in the Ingenuity knowledge base (a manually curated database of experimentally proven molecular interactions from published literature). Networks of these focus genes were then algorithmically generated based on their connectivity. IPA determined the most significantly enriched biological functions and/or related diseases by calculating the P-value using Fisher's exact test. Using similar methods, significantly represented canonical pathways in a set of focus genes were also determined using IPA (Section C in Additional data file 1).

From the embryonic-specific gene clusters, we identified two statistically significant molecular networks (made up of 19 focus genes and 47 associated nodes; networks 1 and 2 in Figure 3; Figures S4 and S5 in Additional data file 1). The networks are interconnected through two genes, Marcks and Neurod1. In general, these networks are associated with cell cycle and DNA replication, recombination and repair processes. Three statistically significant (using P < 0.05 as a cutoff) canonical pathways are enriched in these networks (Figure 3); Wnt/β-catenin signaling (Sox4, Sox11 and Sfrp1), P53 signaling (Cdk4 and Pcna) and tight junction signaling (Cdk4 and Actb) pathways. Validated DETs such as Btg1, Cdk4, Cdkn1c, Csrp2, Ezh2, Neurod1, Pcna, and Rps4x are associated with cell cycle control whereas Actb, Ezh2, Als2cr2, Marcks, Robo1 and Dcx are associated with cellular assembly and organization. These processes are important in the formation of filopodia, membrane blebs and growth cones during neuronal growth, migration and axonogenesis. Known human neurological disorders associated with the networks, particularly network 2, include X-linked lissencephaly (Online Mendelian Inheritance in Man [OMIM:300067]; DCX), juvenile onset dystonia ([OMIM:607371]; ACTB) and Beckwith-Wiedemann syndrome ([OMIM:130600]; CDKN1C). All the DETs implicated in these networks are expressed in the cortical plate with the exception of Pcna (Table S5 in Additional data file 1) 313233343536373839404142.

In adult-specific gene clusters, two molecular networks (18 focus genes and 50 associated nodes; networks 3 and 4 in Figure 3) were identified and interconnected via a single gene, Mbp (Figures S6 and S7 in Additional data file 1). These molecular networks enrich for nine statistically significant canonical pathways (P < 0.05) such as synaptic long-term potentiation and depression, calcium signaling, B cell receptor signaling, cAMP-mediated signaling, GM-CSF signaling, amyotrophic lateral sclerosis signaling, G-protein-coupled receptor signaling and xenobiotic metabolism signaling pathways. Validated DETs such as Camk2a, Gria3, Itpr1, Egr1, Rgs4, Gstm1 and Ppp3ca are associated with cell-to-cell communication in the adult cerebral cortex. Biological functions such as cell proliferation, movement and death processes are also linked to these networks through the genes Camk2a, Cryab, Egr1, Gstm1, Itpr1, Mbp, Nrgn, Ppp1r1b, Ppp3ca, Rgs4, Sept4, Sirpa and Sncb. Known human neurological disorders associated with Network 3 include Pelizaeus-Merzbacher disease ([OMIM:312080]; PLP1) and dementia with Lewy bodies ([OMIM:127750]; SNCB), whereas X-linked mental retardation ([OMIM:300210]; TSPAN7 or TM4SF2) is associated with network 4. MBP, found in both network 3 and 4, is associated with genetic susceptibility to multiple sclerosis [OMIM:126200]. All the DETs implicated in these networks are expressed in layers I to VI of the cerebral cortex, with the exception of Gstm1 (layers II to V), Sirpa and Ppp3ca (layers II to VI), and Nptxr (no expression data available) (Table S5 in Additional data file 1) 32333443444546474849505152.

Both networks 5 and 6 (Figure 3) are linked to gene-switching clusters. These two molecular networks (14 focus genes and 52 associated nodes) are associated with cellular morphogenesis, amino acid metabolism, cell death processes and developmental disorders (Figures S8 and S9 in Additional data file 1). These networks are connected by Ctsd and Ywhae and are implicated in amyloid processing (App and Mapt), aryl hydrocarbon receptor signaling (Ctsd and Nedd8) and cell cycle G2/M DNA damage checkpoint regulation pathways (Ywhae). App, Atp7a, Hprt1, Mapt, Ctsd and Clcn2 are involved in cell morphogenesis, assembly and organization. Among all the networks in the study, network 5 has the greatest number of DETs associated with known human neurological disorders with six out of ten focus genes being associated with known neurological disorders such as neuronal ceroid lipofuscinosis ([OMIM:610127]; CTSD), Alzheimer's disease ([OMIM:104300]; APP), Menkes disease ([OMIM:309400]; ATP7A), X-linked cutis laxa ([OMIM:304150]; ATP7A), epilepsy ([OMIM:606904, 607628 and 607631]; CLCN2), Lesch-Nyhan syndrome ([OMIM:300322]; HPRT1), frontotemporal dementia ([OMIM:600274]; MAPT), progressive supranuclear palsy ([OMIM:601104]; MAPT) and Pick disease of the brain ([OMIM:172700]; MAPT). All the DETs implicated in these networks are expressed in the cortical plate during embryonic development with the exception of Ube2e3. In the adult, all DETs are expressed in layers I to VI of the cerebral cortex except for Bcl11a, Clcn2, Hprt1 (layers II to VI), Atp7a and Ube2e3 (no expression was detected in the adult cerebral cortex) (Table S5 in Additional data file 1) 3233343546535455565758596061.

To refine the functional analysis to a cellular level, we grouped all 40 RT-qPCR validated DETs from networks 3, 4, 5 and 6 (adult specific and gene-switching gene clusters) into three groups according to where they are expressed: only in cortical neurons (N group); only in cortical glia (G group); and in both cortical neurons and glia (B group). None of the DETs from networks 1 and 2 (embryonic-specific gene clusters) were analyzed because most of the cells within the cerebral cortex are committed to the neuronal lineage at E15.5. The methods used to tabulate and group validated DETs are detailed in Section D in Additional data file 1. DETs classified as part of the N group are Bcl11a, Calm1, Camk2a, Camk2n1, Hprt1, Itpr1, Mapt, Nedd8, Nrgn, Ppp1r1b, Ppp3ca, Rbm9, Rgs4, Sncb and Ywhae, whereas those in the G group include Cryab, Mbp, Nptxr, Plp1, Ppap2b, Sept4 and Tspan7. The B group consists of App, Atp7a, Chgb, Ctsd, Egr1, Gria3, Gstm1 and Sirpa. DETs without any cellular expression information, such as Actl6b, Clcn2, Hmgb3, Ube2e3, AK138272, AK139402, AK140219, AK154943, AU258168 and BQ176089, were placed into group B to facilitate downstream analysis. Of the 40 DETs, only 35 were mapped to the IPA knowledgebase and subjected to further analysis. Functional analysis of these 35 DETs show that the neuron-based DETs (N and B groups) are enriched for various human neurological disorders, such as schizophrenia (9 DETs; P = 4.36E-8), Huntington's disease (8 DETs; P = 1.03E-5), X-linked mental retardation (4 DETs; P = 3.78E-7), Parkinson's disease (3 DETs; P = 6.99E-3) and Alzheimer's disease (3 DETs; P = 1.52E-2) (Figure S10 in Additional data file 1). Thirteen DETs associated with these neurological disorders are also implicated in processes related to cell death. Of these, eight DETs are expressed in neurons only, two in glia only and three in both neurons and glia.

From the above IPA analysis, Cand2 (embryonic gene clusters), Camk2n1 (adult gene clusters), Hmgb3 (gene-switching clusters) and 10 ESTs (BQ176089, CD802535, AK139402, AU258168, AK138272, AK140219, AK154943, AA122503, NM_001033273 and BQ177886) are not currently connected to any networks.

Early regionalized development is an important event that could lay the foundations for adult arealization of cerebral functions. To identify genes with regionalized expression profiles, we compared SAGE libraries generated from rostral and caudal regions (equivalent to anterior and posterior regions of the human cerebral cortex) of the E15.5 cerebral cortex. We identified 44 DETs and selected 25 DETs (22 known genes, 1 EST and 2 ambiguous genes; Additional data file 2) and 2 genes of interest (bladder cancer associated protein, Blcap, and ankyrin repeat and zinc finger domain containing 1, Ankzf1) for RT-qPCR validation and further detailed region-based analysis. This was done using independent biological triplicates of clearly defined regions/quadrants, such as rostro-lateral (RL), rostro-medial (RM), caudo-lateral (CL) and caudo-medial (CM) of the E15.5 cerebral cortex (see Materials and methods). Two positive controls with known regionalised expression were included in the RT-qPCR: RAR-related orphan receptor beta (Rorb) and nuclear receptor subfamily 2, group F, member 1 (Nr2f1). Rorb is highly expressed in the rostral region whereas Nr2f1 is highly expressed in the caudal region of the cerebral cortex 162.

An initial RT-qPCR analysis of combined RL and RM (rostral) as well as CL and CM (caudal) regions shows upregulation of Rorb and Nr2f1 in rostral and caudal regions, respectively (based on fold change direction and magnitude of 1.3 times). The same analysis also confirmed the SAGE data for 3 out of 25 DETs (Actb, Tmsb4x and BC025816) and Blcap (Table 6; Additional data file 6). Both Actb and Tmsb4x have greater expression in the rostral region whereas Blcap and BC025816 are greater in the caudal region. To identify expression profiles in a more refined area and prevent regional compensation due to combined quadrant analysis, we performed a quadrant versus quadrant multiple regions comparison. The largest number of DETs were found in the RL versus CM comparison, as they are the two developmentally most distinct regions within the cerebral cortex at E15.5 compared to others; RL versus CL > RM versus CL > RL versus RM = RM versus CM > CL versus CM. The region-specific expression profiles were plotted for each of the DETs (Figure 4) and grouped into two categories: RL-specific DETs, such as Actb, Tmsb4x and cytochrome b-245, beta polypeptide (Cybb); and CM-specific DETs, including Blcap, EST BC025816, Ankzf1 and cytochrome c oxidase I, mitochondrial (Cox1) (Additional data file 6).

To visualize the regionalized expression profiles, we performed in situ RNA hybridization (ISH) on all DETs validated by quadrant versus quadrant RT-qPCR analysis, Blcap, Ankzf1 as well as the positive controls Rorb and Nr2f1. We performed ISH on sagittal and coronal sections (from rostral to caudal regions) of the E15.5 mouse brain (Figure 5). Under dark-field microscopic examination, we confirmed the regionalized expression of Rorb (at the rostral cortical plate; Figure 5a) and Nr2f1 (at the caudal ventricular zone; Figure 5f). From the analysis, Actb is highly expressed at the cortical plate and the subplate (Figure 5b). Tmsb4x is highly expressed at the cortical plate and the intermediate zone (Figure 5c-e).

ISH analysis showed that both Blcap and Ankzf1 are caudal specific. Serial coronal sections from rostral to caudal regions of the brain (Figure 5g-i) show Blcap is weakly expressed in the rostral cerebral cortex, particularly at the intermediate zone, subplate and the cortical plate, but is highly expressed in the hippocampus and thalamus (Figure 5i). Ankzf1 is expressed specifically in the ventricular zone as well as the cortical plate towards the caudal region of the cerebral cortex (Figure 5j-m). The distinctive expression of Ankzf1 in both the ventricular zone and cerebral cortex prompted us to extend our ISH analysis to various developmental stages starting from E11.5 to adulthood (Figure 5n-t). Ankzf1 is expressed in the primordial plexiform and the ventricular zone layers of the telencephalon at E11.5 (Figure 5n). By E13.5, Ankzf1 is weakly expressed in the ventricular zone, but is highly expressed in the cortical plate (Figure 5o). From E17.5 to P1.5, the expression of Ankzf1 is maintained in the cerebral cortex (Figure 5p, q). In the adult, Ankzf1 expression is obvious in the piriform cortex, hippocampus and cerebellum (Figure 5r-t).

We performed genomic clustering analysis of the SAGE tags to determine any actively transcribed chromosomal loci throughout cerebral corticogenesis. Probabilities for chance occurrences of two, three, four, and five DETs being clustered within a window of ten adjacent tags present within each chromosomal location, irrespective of genetic distance, were calculated. This analysis was based on the DET lists described above (Additional data file 2). The analysis showed two overrepresented chromosome loci at Sox4 and Sox11, which derive from embryonic-specific gene clusters.

At both loci, we observed multiple SAGE tags with both sense and antisense orientations, which signify alternative polyadenylation sites, differential splicing and overlapping antisense transcription. As an initial validation of the antisense messages, we performed strand-specific RT-PCR (Figure 6) using cDNA synthesized from equally pooled total RNAs (three mice from each of E15.5, E17.5, P1.5 and adult stages). Three primer sets were used for each locus (designed at the middle, 5'- and 3'-ends of the canonical transcripts). For both Sox4 and Sox11 loci, the analysis showed positive signals from all primer sets used that were complementary to the antisense strand, therefore confirming the presence of one or more antisense transcript(s) that span the canonical transcript. Hydroxymethylbilane synthase gene (Hmbs) served as a negative control and there were no positive bands in the antisense strand RT-PCR, confirming the absence of antisense transcripts.

We mapped SAGE tags from the genomic cluster to the Sox4 gene locus using the University of California Santa Cruz (UCSC) genome browser (Figure 7a; Additional data file 7) 63. Only tags within and around the Sox4 canonical transcript are shown. Evidence of mapped mouse mRNAs within this locus further justifies the existence of multiple SAGE tags in addition to the canonical Sox4 transcript. Subsequent validation of the genomic clusters was solely based on the SAGE tags situated within the canonical transcript. Based on the SAGE tag information, 6 out of 12 tags are DETs. These include four DETs within the canonical transcript (Figure 7b), with sox4_tag10 and sox4_tag15 having greater expression in P1.5 compared to P150 (adult stage), whereas sox4_tag12 and sox4_tag16 are both abundantly expressed in the caudal region of the E15.5 cerebral cortex. Based on the RT-qPCR analysis (Figure 7c), sox4_tag10 has greater expression in E15.5 compared to E17.5 (1.76-fold change), P1.5 (3.72-fold change) and Ad (43.67-fold change). For sox4_tag12, sox4_tag15 and sox4_tag16, differences are seen only in P1.5 (-2.00-fold, -1.57-fold and -1.84-fold changes, respectively) and Ad (-82.08-fold, -68.92-fold and -69.41-fold changes, respectively) when compared to E15.5. RT-qPCR analysis on the same tags did not find any differences between rostral and caudal cerebral cortices of E15.5. The differences in fold change between DETs suggest irregular overlapping of various transcript variants at different Sox4 gene loci.

To further validate the expression profiles of the multiple Sox4 DETs, we performed 3' rapid amplification of cDNA ends (RACE)-Southern analysis using pooled adaptor-oligo-d [T]₁₅ synthesized cDNAs from three mice at each developmental stage. Based on this method, we were able to semi-quantitatively and accurately measure the expression levels of individual SAGE tags at the locus. To show that the amplification was cDNA specific, we performed PCR by using the same primer sets on mouse genomic DNA under the same conditions. In all cases, no amplification was observed (data not shown). This analysis confirmed the presence of four out of seven alternative transcripts for Sox4 (Figure 7d). Corresponding tags were determined by estimating the amplicon sizes between the strand-specific primers used, the next downstream AAUAAA/AUUAAA polyadenylation signal (if any) and succeeding CATG sequence or SAGE tags. Figure 7d(1)-d(3) confirms the existence of sox4_tag10, sox4_tag12 and sox4_tag15. Of these tags, SAGE expression profiles of sox4_tag10 and sox4_tag15 were validated (embryonic-specific and reduced expression after P1.5) but not sox4_tag12 (E15.5 caudal region-specific). 3' RACE-Southern analysis using a sense probe detected bands in the rostral and caudal regions of E15.5, E15.5 and E17.5 cerebral cortices and, therefore, confirmed the existence of the Sox4 antisense transcripts (Figure 7d(4)). Even though none of these tags were differentially expressed in between these regions based on the SAGE analysis, our findings show distinctive regionalization for sox4_tag14 expression at the E15.5 rostral cerebral cortex. Proteasome (prosome, macropain) subunit, beta type 2 gene (Psmb2) and Hmbs were used as controls and no antisense or alternative transcripts were identified at these gene loci (Figure 7d(5)-d(8)).

Since 3' RACE-Southern analysis was dependent on oligo- [dT]₁₅ priming, we could not rule out the possibility of amplicons that were generated by false priming on homopolymer-A stretches. Therefore, Northern analyses were performed on equally pooled total RNA extracted from the cerebral cortices of seven mice at E15.5, E17.5 and P150 (negative control). By using a double-stranded DNA probe at the 3' untranslated region (UTR) of Sox4 (Additional data file 8), we identified six bands ranging from approximately 2 kb to approximately 4.7 kb (Figure 7e). Sox4 sense transcripts are weakly expressed in the adult, but highly expressed in the embryonic cerebral cortex. The number of bands observed is similar and corresponds to overlapping mouse mRNAs as well as publicly available paired-end diTag (PET) sequences downloaded from Ensembl (Figure 7a; Table S6 in Additional data file 1) 64. Taken together, the analysis confirmed the existence of multiple overlapping variants of Sox4 sense and antisense transcripts at this gene locus.

To confirm the rostro-caudal expression of Sox4 sense transcripts, we performed ISH on sagittal sections of mouse brains using a Sox4 antisense riboprobe that spanned across the sox4_tag10, sox4_tag12, sox4_tag15 and sox4_tag16 SAGE tags. Sox4 showed regionalized expression at E15.5 and E17.5 (Figure 7f). At E15.5, Sox4 sense transcripts are expressed more in both the rostral- and caudal-end regions of the cortical plate compared to the intermediate region between them (Figure 7f(1)). By E17.5, expression of Sox4 sense transcripts is obvious in the rostral cortical plate (Figure 7f(2)). At both stages of development, Sox4 sense transcripts are uniformly expressed in the intermediate zone of the cerebral cortex. These findings correspond to the SAGE tag counts for E15.5 rostro-caudal regions of the cerebral cortex (Figure 7f). These observations explain the averaged total tag count per 100,000 tags for different Sox4 sense transcripts, which are predominantly expressed in both rostral and caudal regions of the cerebral cortex. The regionalized expression of Sox4 in the cortical plate is obvious only at E15.5 and E17.5, but not at other stages of development (Figure S11 in Additional data file 1).

To further ascertain the antisense expression of Sox4 in a spatio-temporal manner, we performed ISH on coronal sections obtained from E11.5, E13.5, E15.5, E17.5, P1.5 and P150 mouse brains. Sense and antisense RNA probes were generated from the same clone used in the Northern analysis. At E11.5, Sox4 sense transcripts are confined to the primordial plexiform layer (Figure 8a). From E13.5 to P1.5, the sense transcripts are expressed throughout the cortical plate (Figure 8b-e). Expression of sense transcripts in the subventricular zones is observed at E17.5 and P1.5 only (Figure 8d, e). There is no observable sense expression in the adult stage (Figure 8f). Sox4 antisense transcripts are expressed throughout the telencephalon at E11.5 (Figure 8g). From E13.5 to P1.5, Sox4 antisense expression is confined to the cortical plate only (Figure 8h-k). There is no obvious antisense expression in the cerebral cortex in the adult stage (Figure 8l). A microscopic examination at high magnification showed that Sox4 antisense transcripts are predominantly localized in the nucleus whereas Sox4 sense transcripts are found in both the nucleus and cytoplasm (Figures S12, S13 and S14 in Additional data file 1). We used hemoglobin alpha, adult chain 1 (Hba-a1) of the corresponding brain region and time-point as a control in the analysis (Figure 8m-r).

Furthermore, Sox4 antisense expression occurs in the piriform cortex layer II (Figure 9a-c) and dentate gyrus (Figure 9g-i) in the adult brain; however, no sense expression is observed in these regions. At P1.5, we identified complementary expression between Sox4 sense and antisense transcripts in the olfactory bulb (Figure 9d-f). Sox4 sense expression was confined to the granular and glomerular layers of the olfactory bulb whereas antisense expression was found only in the outer plexiform layer. We used either Sox11 or Hba-a1 of the corresponding brain region and time-point as a control in the analysis (Figure 9c, f, i).

SAGE tags, which represent multiple overlapping sense and antisense transcript variants at the Sox11 genomic cluster, were validated using 3' RACE-Southern analysis as described above. See Section F in Additional data file 1 for a full description of the Sox11 results. ISH analysis did not confirm the expression of antisense transcripts of Sox11, but the presence of PETs spanning three out of five antisense tags confirmed the existence of Sox11 antisense transcripts (Table S8 and Figures S16, S17 and S18 in Additional data file 1). The discrepancy between ISH and RT-qPCR or 3' RACE-Southern analysis suggests that Sox11 antisense transcripts might be expressed at low levels or at specific locations of the cerebral cortex, and hence can be detectable only by using serial sections or whole mount ISH.

We screened various adult brain regions (olfactory bulb, cerebellum, medulla, hippocampus, thalamus and cerebral cortex) and selected mouse organs (E15.5 whole brain, heart, kidney, liver, skeletal muscle, skin, spleen, stomach, testis and thymus) for the expression of Sox4 and Sox11 antisense transcripts by strand-specific RT-qPCR. Within the adult brain, Sox4 sense and antisense transcripts are expressed in all regions, with the highest level found in the olfactory bulb, which is approximately four- to nine-fold greater than those in other brain regions (Figure 10a). Expression of Sox4 antisense transcripts occurs in all mouse organs, with the highest level in the thymus followed by E15.5 whole brain, testis and skin (Figure 10b). Sox4 sense and antisense expression profiles are similar throughout the entire series of samples screened, with the sense transcripts being consistently expressed at a greater level than the antisense transcripts (approximately 1.7-fold in various brain regions and approximately 2- to 14-fold in various organ comparisons).

Sox11 sense transcripts are expressed at the highest level in the olfactory bulb, approximately two- to seven-fold greater than those in other brain regions (Figure 10a). Sox11 antisense transcripts, on the other hand, are expressed in all brain regions screened and at a comparable level in the olfactory bulb, hippocampus, thalamus and cerebral cortex. In comparison to other adult mouse organs, Sox11 sense and antisense transcripts are highly expressed in the E15.5 whole brain, with Sox11 sense transcript levels at least 100-fold greater than those in other mouse organs (Figure 10b). On the other hand, Sox11 antisense expression is observed only in the E15.5 whole brain, skin and stomach. Notably, Sox11 sense transcripts are expressed more highly than antisense transcripts in the E15.5 whole brain and skin (23- and 4-fold, respectively).

Since both Sox4 and Sox11 are implicated in neuronal differentiation and glial maturation processes 6566, we examined both Sox4 and Sox11 sense and antisense transcript expression in proliferating and differentiating P19 (embryonal carcinoma cells) and in embryonic NSPCs grown as neurospheres. Both Sox4 sense and antisense transcripts are upregulated during P19 cell differentiation (approximately 5.7- and 1.6-fold upregulation, respectively; Figure 11a) and neurosphere differentiation (approximately 1.9- and 1.8-fold upregulation, respectively; Figure 11b). For Sox11, both sense and antisense transcripts are upregulated in the differentiating compared to the proliferating P19 cells by approximately 2.3- and 4.2-fold, respectively (Figure 11c). Both the Sox11 sense and antisense transcripts are, however, downregulated in the differentiating neurospheres (approximately 2.6- and 1.5-fold, respectively; Figure 11d).