High-density oligonucleotide microarray analysis was carried out for total RNA from brains of Nfia^-/- mice and age-matched, wild-type littermate controls. Analyses were performed with independent samples from three Nfia^-/- and three wild-type (Nfia^+/+) animals each for E18 and P16. All animals were F₁ hybrids of C57BL/6 and 129S6 mice, ensuring a survival rate of 38.5% until P16. A total of 356 genes were identified as being differentially expressed in the Nfia^-/- animals at P16 (197 upregulated and 159 downregulated), taking a cutoff of a 1.2-fold change and a significance of P < 0.05 in expression relative to the wild-type control (see Additional data file 1). Among these, 53 genes were found to exhibit a greater than 1.5-fold change in expression (39 downregulated [74%] and 14 upregulated [26%]; Table 1).

Within this latter group of strongly dysregulated genes, a total of 11 genes exhibit greater than twofold dysregulation, with nine genes downregulated and two upregulated. The downregulated genes include those encoding the following: angiotensinogen (Agt); aldehyde dehydrogenase family 1, subfamily A1 (Aldh1a1); folate hydrolase (Folh1); GABA-A receptor, subunit α6 (Gabra6); gap junction membrane channel protein β6 (Gjb6); lecithin cholesterol acyltransferase (Lcat); myelin and lymphocyte protein (Mal); myelin-associated oligodendrocytic basic protein (Mobp); and neurotensin receptor 2 (Ntsr2). The upregulated genes encode fatty acid binding protein 7 (FABP7) and the transcription factor SRY-like HMG-box containing 11 (Sox11).

At the late embryonic stage (E18), fewer genes were significantly dysregulated in the Nfia^-/- mutant relative to the wild-type animals when compared with the postnatal stage (P16). A total of five genes was identified as being significantly dysregulated with changes of more than 1.2-fold (Table 2). One of the three downregulated genes encodes a yet uncharacterized protein, whereas the two others encode phosphatidylinositol-4-phosphate 5-kinase, type II, γ (Pip5k2c) and synaptotagmin binding, cytoplasmic RNA interacting protein (Syncrip). mRNAs for synaptotagmin 1 (Syt1) and pleiomorphic adenoma gene-like 1 (Plagl1) were expressed at elevated levels in the Nfia^-/- animals. At E18, no gene was differentially regulated more than 1.5-fold in the Nfia^-/- mutants relative to the wild-type controls. Pleiomorphic adenoma gene-like 1 (Plagl1) is the only gene that exhibits a 1.2-fold up-regulation in Nfia^-/- mice at both developmental stages.

In P16 Nfia^-/- mice, a total of 356 individual genes, represented by 395 probe sets, were dysregulated in comparison with the wild-type control group. Among these, 35 genes were represented by more than one probe set on the microarray (Additional data file 1, blue labels). In all cases, probe sets representing the same gene showed the same direction of dysregulation when comparing Nfia^+/+ with Nfia^-/- mice. For instance, all of the four probe sets for Mobp identified a weaker signal in Nfia^-/- mice than in Nfia^+/+ mice. Moreover, it is likely that at least four probe sets represent more than one transcript (Additional data file 1, red labels). Therefore, the number of dysregulated genes in Nfia^-/- mice could even be higher than 356.

To investigate overall gene expression profiles, microarray data were analyzed using the robust multi-array average algorithm 26. Correlations between the expression profiles of individual samples were calculated with all samples, using one Nfia^+/+ E18 mouse brain ('E18WT1') as a reference for this search (Figure 1). As expected, the greatest similarity in mRNA levels was found between E18WT1 and the two other Nfia^+/+ mice at E18, namely E18WT2 and E18WT3. Broadly speaking, similarity in the gene expression profile relative to E18WT1 increased in the following order: P16 Nfia^+/+ < P16 Nfia^-/- < E18 Nfia^-/- < E18 Nfia^+/+.

Figure 2a shows the overall gene expression pattern in both E18 Nfia^+/+ and Nfia^-/- mice. The expression patterns are much more similar between Nfia^+/+ and Nfia^-/- mice than those seen at P16, and very few genes were changed significantly between Nfia^+/+ and Nfia^-/- in the E18 animals. In contrast, at P16 many changes in gene expression levels are observed (Figure 2a) between Nfia^+/+ and Nfia^-/- animals. Most genes that are expressed at a higher level in E18 become less strongly expressed in P16 mice and vice versa, both in Nfia^+/+ and Nfia^-/- mice. However, comparison of P16 Nfia^-/- with P16 Nfia^+/+ animals revealed many significant changes in gene expression (Figure 2b).

A total of 356 genes that were dysregulated in P16 Nfia^-/- mice were assigned to biologic functions based on Gene Ontology (GO) Biological Process categories (Additional data file 2 provides a list of assignments). Among the 197 upregulated genes, a large percentage (34.6%) of the candidate genes is involved in transcriptional and translational regulation. These groups include genes that encode RNA binding proteins, transcription factors and ribosomal proteins, and comprise only 10.4% of probe sets on the microarray. Figures 3 and 4 show the distribution of gene functions in the upregulated and downregulated groups, respectively. Within the group of upregulated genes, 33 probe sets (16.8%) were in the category of 'protein biosynthesis', as compared with 1.5% of probe sets on the complete microarray. Twenty probe sets (10.2%) could be assigned to 'regulation of transcription, DNA dependent' (8.2% on the complete microarray), and 15 gene products (7.6%) are involved in mRNA processing (0.7% on the complete microarray). Among the 159 downregulated genes, a significant effect on ion transport related genes can be observed; 14 genes (9.0% of downregulated genes in Nfia^-/- mice) fall into this category, which is an 'over-representation' compared with the complete array, in which ion transport related genes account for only 1.6% of probe sets.

Interestingly, a number of genes encoding myelin-related proteins exhibited significantly reduced expression in brains of Nfia^-/- mice at P16 (Figure 4). Because central nervous system myelin is formed by oligodendrocytes, we suspected that NFI-A could influence oligodendrocyte differentiation. We therefore interrogated our list of dysregulated genes for markers of either mature oligodendrocytes or immature oligodendrocyte precursor cells. We found that five genes typically expressed in oligodendrocyte precursors exhibited a higher expression level in Nfia^-/- mice than in Nfia^+/+ animals, whereas eight genes that are markers of mature oligodendrocytes exhibited decreased expression in Nfia^-/- mouse brains (Table 3). As shown in Figure 5, the oligodendrocyte precursor markers Sox2, Sox4, and Sox11 are expressed in both Nfia^+/+ and Nfia^-/- animals at E18. However, the decrease in gene expression between E18 and P16 is less pronounced in Nfia^-/- animals than in Nfia^+/+ animals, causing an apparent mRNA overexpression of these genes at P16.

Moreover, agenesis of the corpus callosum in animals lacking NFI-A suggests that this transcription factor plays a role in regulating axonal growth. It is tempting to assume that NFI-A could do so by influencing the expression of genes that encode growth promoting or growth repelling proteins. For this reason, we also attempted to identify molecules that have already been shown to stimulate or inhibit neurite growth among those differentially expressed in Nfia^-/- mouse brains. For 22 genes that were either upregulated or downregulated in the NFI-A mutant animals, reports from the literature indicate that the respective gene product is involved in regulating neurite growth (Table 4). Among these genes, 12 encode proteins whose expression is favorable for neurite growth (for instance acidic fibroblast growth factor (aFGF), melanoma cell adhesion molecule (MCAM) and neural cell adhesion molecule (NCAM), whereas five encode proteins that function in axon repulsion (for example, Ephrin B2 and collapsin response mediator protein-1 (CRMP1)). Five genes encode proteins that influence neurite growth in a cell type or presentation dependent manner, including tenascin-C and CD24.

Quantitative real-time polymerase chain reaction (qRT-PCR) was performed on 15 genes, selected according to their biologic relevance from the dysregulated genes in P16 Nfia^-/- mice (Figure 6). On the one hand we chose oligodendrocyte precursor genes such as Sox2 and Sox11, but we also selected markers of mature oligodendrocytes, such as Car2 (carbonic anhydrase 2) and Mobp (myelin oligodendrocyte basic protein). In addition to genes relevant to oligodendroglial differentiation, we also chose further markers of immature (Dcx [doublecortin]) or mature (Gfap [glial fibrilliary acidic protein] and Gabra6) neural cells. All PCR analyses were performed in triplicate, using eight independent Nfia^-/- and six independent Nfia^+/+ brain samples, and confirmed differential expression of all the genes selected for qRT-PCR validation, indicating the significance of our microarray analysis findings (Figure 6). Importantly, both genes exhibiting a strongly differential expression pattern on the microarrays (for instance, Dcx, which exhibited 1.87-fold upregulation) and genes with a much lower fold change (such as Sox2, which was 1.28-fold upregulated) were confirmed to be differentially expressed.

To investigate whether the differential gene expression observed at P16 is maintained at a later age, we also analyzed RNA from postnatal day 43 (P43) Nfia^-/- and Nfia^+/+ brains for expression of the selected genes mentioned above. As shown in Figure 7, differences in gene expression between Nfia^-/- and Nfia^+/+ animals generally decrease from P16 to P43. However, certain genes such as Gabra6 and Gfap exhibit pronounced downregulation at both ages.

NFI-A is a nuclear, DNA-binding protein. It plays a role in adenovirus DNA replication and in transcription of viral and cellular genes. Therefore, we assumed that at least some of the dysregulated genes found in our study might be direct transcriptional targets of NFI-A. In order to identify such potential targets, we conducted a promoter analysis of all genes exhibiting a significant decrease or increase in transcript level in Nfia^-/- relative to Nfia^+/+ mice. This in silico analysis aimed to detect potential NFI-A binding sites within 2 kilobases (kb) upstream of the respective gene's transcription start site. The palindromic nucleotide sequence TTGGC(N)₅GCCAA has been demonstrated to be the optimal binding motif for members of the NFI family. However, most of the NFI binding sites experimentally identified thus far do not contain the complete motif, and even half sites can be physiologically relevant 3. This reflects the fact that transcription factors have a certain degree of freedom in their sequence recognition. For this reason, matrices are used that give the different nucleotides various weightings depending on their importance for transcription factor binding 27. In addition to the use of such a matrix for NFI binding motifs, we also considered the phylogenetic conservation of these binding sites by comparing the mouse, rat, and human orthologs of the respective genes. We supposed that motifs with a high degree of interspecies conservation are those that are most likely to have physiologic relevance.

Using these criteria, we were able to identify more than 70 genes among our microarray candidate molecules bearing a conserved NFI recognition site in their promoter region (Table 5 and Additional data file 3). This group of genes includes Gfap and Gabra6, whose promoter activity can be regulated by NFI proteins, according to previous studies 142829. Interestingly, according to our analysis, Ncam, Vcam1, Mcam and Mag, four genes that encode adhesion molecules of the immunoglobulin superfamily, also possess conserved NFI motifs in their promoter sequences.

When the list of genes changed at P16 was analyzed using Gene Ontology (GO) terms, a total of 34.6% of all upregulated genes in Nfia^-/- mice fell into the functional groups of transcriptional and translational activities. This is a significantly higher percentage compared with the representation of these groups on the complete microarray (10.4%). In particular, many transcripts encoding ribosomal proteins exhibit elevated levels in these mutants. This upregulation of messages encoding ribosomal proteins suggests increased translational activity in Nfia^-/- brains.

Interestingly, the expression of several genes associated with immature stages of the nervous system is upregulated in postnatal Nfia^-/- mice. In particular, elevated mRNA levels of Dcx (which encodes doublecortin) and Nnat (encoding neuronatin) were observed. Doublecortin is expressed primarily in migrating and differentiating neurons during embryonic development 31. It is essential for cortical layer formation, most probably because of its role in neuronal migration 32. Neuronatin is strongly expressed in late fetal and early postnatal brain, but it disappears at later developmental stages 33. Remarkably, Plagl1, the only gene expressed at elevated levels both at E18 and P16 in Nfia^-/- mice, encodes a transcription factor synthesized preferentially by neural precursor cells 34. On the contrary, mRNAs for Gabra6 and Gfap, genes associated with terminal differentiation of neural cells, are found at lower levels in NFI-A deficient mice at P16. The α6 subunit of the GABA-A receptor (encoded by Gabra6) is expressed by differentiated neurons in the cerebellum 14. Like Gabra6, Gfap (which encodes the glial fibrillary acidic protein, expressed by differentiated astrocytes in the central nervous system) has been identified as a direct target of NFI-A 29 and is downregulated in Nfia^-/- mice.

The observed pattern of changes in gene expression suggests a delay in brain development in NFI-A mutants. In the absence of NFI-A, genes that are normally expressed during embryonic development and around birth remain at high levels of expression, leading to an overexpression at P16. By contrast, genes whose expression usually increases during the course of terminal differentiation after birth appear not to be activated adequately in Nfia^-/- mice, leading to their reduced mRNA level in the P16 mutants. Investigation of Nfia^-/- and Nfia^+/+ animals at P43 showed that, for most of the selected genes, dysregulation decreased in comparison with P16 or even disappeared. This observation further supports the idea of a delayed expression program in Nfia^-/- mice.

To identify cell types that are probably affected by the expression delay suggested above, we examined further the function of dysregulated genes. A significant number of myelin-related proteins exhibited decreased expression in Nfia^-/- mouse brains, prompting us to analyze our results with regard to oligodendrocyte differentiation. Oligodendrocytes produce central nervous system myelin, thereby facilitating rapid impulse conduction 35. Myelinating oligodendrocytes mainly accumulate after birth in rodents, whereas their progenitors are already apparent in the ventricular zone at around embryonic day 12.5 (E12.5) 36. Thus, in E18 brains oligodendrocyte progenitor cells are predominant, whereas at P16 mature oligodendrocytes have developed. In Nfia^-/- brains at P16, we observed that several genes expressed by mature oligodendrocytes, namely MAG, Mal, Mobp, Mog, Ugt8, Cldn11, Plp1, and Car2, exhibited reduced transcript levels compared to Nfia^+/+ animals. In contrast, mRNAs encoding Sox2, Sox4, Sox11, tenascin-C and Hmgb2, which are typically expressed by precursor cells rather than by mature oligodendrocytes, are upregulated in Nfia^-/- brains relative to Nfia^+/+ brains at this stage. Moreover, Dio2 exhibits reduced expression levels in P16 Nfia^-/- brains. Dio2 encodes the iodothyronine deiodinase II, which catalyzes the conversion of the hormone thyroxine to tri-iodothyronine 37. Tri-iodothyronine is known to trigger terminal oligodendrocyte differentiation 38, and so a reduction in Dio2 levels could delay oligodendrocyte maturation indirectly via reduced tri-iodothyronine levels. Furthermore, we detected slightly increased transcript levels of Myef2 in Nfia^-/- mice. My-EF2, the corresponding protein, represses expression of myelin basic protein 39, a major component of myelin 3640. Higher My-EF2 levels could therefore slow down differentiation of oligodendroglia. Moreover, the tremor exhibited by rare NFI-A deficient mice surviving until adulthood 20 would be in accordance with a myelin compromised phenotype. Interestingly, L1 has also been implicated in myelination 41. Although we did not observe a dysregulation of L1 in our microarray analysis, one cannot exclude an induction in glial cells being obscured by the high expression level of L1 in neurons. In contrast to the peripheral nervous system, glial cells of the central nervous system do not express L1 at any age investigated.

To summarize, our data suggest a role for NFI-A in regulating terminal differentiation of oligodendrocytes, both by repressing expression of progenitor specific gene products and by enhancing expression of genes that are relevant to mature oligodendrocyte function (Figure 8). It is noteworthy that the time window between E18 and P16, during which these changes in expression pattern emerge, fits nicely to the main period of oligodendrocyte differentiation. In agreement with the observations presented here, a crucial role for NFI-A in spinal cord gliogenesis was recently shown 42, which lends further support to the idea that NFI-A mediates oligodendrocyte differentiation.

The absence of the corpus callosum in Nfia^-/- mice suggests that NFI-A could participate in controlling expression of molecules relevant for axonal growth and guidance. For this reason, we analyzed whether dysregulated genes detected in our microarray study had been reported to be involved in these processes. For at least 22 of the genes differentially expressed in Nfia^-/- brains, previous studies indicated a participation of their gene products in growth and guidance of neuronal processes. This is quite a large number, which strengthens the view that NFI-A could contribute to brain wiring during the early postnatal period by regulating the transcription of genes that encode neurite growth promoting or inhibiting proteins. A more detailed look at the dysregulated genes reveals that growth-promoting molecules, such as clusterin, aFGF, and Ndrg2, are expressed at a rather lower level in Nfia^-/- mice, whereas an upregulation of the repulsive guidance cues such as ephrin B2 and CRMP1 can be observed. However, there are both upregulated growth promoting and downregulated growth inhibiting molecules. Moreover, some molecules can either favor or inhibit growth, depending on their interaction partners or on the way in which they are presented by neurons and glial cells. Therefore, it will be necessary to study NFI-A dependent expression of these proteins at the cellular level in order to shed more light on the role played by NFI-A in axonal growth and guidance.

Because NFI-A is a transcription factor, we assumed that it regulates the transcription of at least some of the differentially expressed genes found in our study. In order to identify such potential targets, we performed an in silico analysis of potential, phylogenetically conserved NFI-A binding sites within 2 kb upstream of the respective genes' transcription start sites. We could identify more than 70 genes among our microarray candidate molecules bearing a conserved NFI recognition site in their promoter. Because of their high identity to the NFI consensus sequence combined with phylogenetic conservation, these motifs are likely candidates for transcriptional regulation by NFI-A. Consistent with this assumption, two established direct target genes of NFI-A, Gfap 29 and Gabra6 14, were among the candidates detected in our promoter analysis. Two possible reasons for why only about 70 genes carrying conserved NFI motifs were identified from more than 300 genes dysregulated in total are as follows.

First, NFI-A might alter the expression levels of other transcription factors that affect target gene expression (indirect influence). For example, expression of the transcription factor NeuroD1, which is downregulated 1.68-fold in Nfia^-/- mice, can be increased by GABAergic excitation 43. This effect is mediated via GABA-A receptors, and Gabra6, the gene encoding the α6 subunit of the GABA-A receptor, is a direct target of NFI-A 14, which is in accordance with the results of our promoter analysis. Thus, it is tempting to speculate that the reduced expression of this GABA-A receptor subunit in NFI-A deficient mouse brains could contribute to the observed reduction in NeuroD1 expression.

Second, our promoter analysis probably does not yield a complete list of direct NFI-A targets within the group of dysregulated genes investigated. The limitation to the 2 kb region upstream of the transcription start excludes all potential NFI sites that might be located either further upstream or within the first intron, and even NFI half sites, which are not considered by our stringent screening method, can mediate NFI protein binding under certain conditions. Therefore, the number of direct NFI-A targets among the differentially expressed genes might be even higher than 70.

Most importantly, however, this promoter screen provides a molecular basis for the role of NFI-A in regulating brain maturation; several genes associated with maturation of neurons or glial cells are not only dysregulated in postnatal Nfia^-/- mice, but also possess NFI binding sites in their promoter, enabling direct transcriptional regulation by NFI-A. Among them, there are both markers of oligodendrocyte precursors (Sox11 and Tenascin-C) and genes expressed by mature oligodendrocytes (Car2 and Mag). Therefore, it is likely that NFI-A exerts its influence on postnatal brain development at least partly by directly regulating the transcription of the genes identified in our promoter analysis.

Nfia^-/- mice were bred and maintained as described previously 20. Animals were killed by exposing them to CO₂ at embryonic stage E18.5 (E18), postnatal day 16 (P16), or postnatal day 43 (P43). Brains were dissected and snap frozen in liquid nitrogen. The respective Nfia^-/- and Nfia^+/+ animals used for analysis at E18, P16, or P43 were littermates. All of the animals were F₁ hybrids of C57BL/6 and 129S6 animals. In the initial studies of the Nfia^- allele in a mixed 129S6-C57BL/6-Black Swiss background, we found that about 80% of Nfia^-/- mice died at P0 and about 10% survived to P16. In contrast, 0% of Nfia^-/- mice survive past P0 in an inbred C57BL/6 background, precluding studies of postnatal animals. In the inbred 129S6 background about 30% of Nfia^-/- mice survive to P16, whereas in F₁ hybrids of C57/BL6 males and 129S6 females about 60% of Nfia^-/- mice survive to P16. Many of these animals die after weaning from their mothers at P24 and the remainder are severely runted (Gronostajski RM, unpublished observations). Thus, we chose P16 129S6-C57BL/6 F₁ hybrid animals for analysis because they could be obtained in sufficient numbers, and the Nfia^-/- mice were genetically identical to their Nfia^+/+ littermates. Within these matings, males had a B6 and females had a 129S6 background. In a large scale test, 38.5% of the offspring survive until P30 using this type of mating. All animal use was performed under the approved protocol BCH05082N (RMG) of the University at Buffalo Institutional Animal Care and Use Committee in the UB Laboratory Animal Facility, an facility licensed by the Association for the Assessment and Accreditation of Laboratory Animal Care.

Total RNA was extracted from the entire brains of Nfia^-/- and Nfia^+/+ mice using TRIzol (Invitrogen, Karlsruhe, Germany), in accordance with the manufacturer's instructions. Further purification of RNA was performed using the RNeasy Mini or Midi Kits (Qiagen, Hilden, Germany). Total RNA concentration was determined using a spectrophotometer at 260 nm and 280 nm wavelength. To check RNA integrity, total RNA was separated in a 1% agarose gel containing formaldehyde, and the intensity ratio of 28S and 18S ribosomal RNA bands was assessed after ethidium bromide staining.

Procedures for cDNA synthesis, labeling, and hybridization were carried out in accordance with the manufacturer's protocol (Affymetrix Inc., Santa Clara, CA, USA). All experiments were performed using Affymetrix mouse genome Genechip U74A version 2. Briefly, 15 μg total RNA was used for first-strand cDNA synthesis with a high-performance liquid chromatography purified T₇-(dT)₂₄ primer. Synthesis of biotin-labeled cRNA was carried out using the ENZO RNA transcript labeling kit (Affymetrix). For hybridization, 15 μg fragmented cRNA were incubated with the chip in 200 μl hybridization solution in Hybridization Oven 640 (Affymetrix) at 45°C for 16 hours. Genechips were washed and stained with streptavidin-phycoerythrin using the microfluidic workstation™ (Affymetrix) and scanned with a laser scanner (Agilent Technologies, Böblingen, Germany).

Quality controls were performed using Affymetrix Microarray Suite 5.0 (MAS 5.0) software. Data and statistical analysis and data visualization were performed with GeneSpring™ (Agilent Technologies). Expression values were extracted from the CEL file, and imported to GeneSpring using robust multi-array average for further analysis 26. Two criteria were used to select for significant changes in gene expression. First, probe sets with less than 1.2-fold change on comparing Nfia^-/- with Nfia^+/+ (control) samples were removed in order to reduce the number of genes undergoing statistical analysis. Second, genes were then tested statistically for significant changes in all three wild-type or NFI-A deficient samples, using Student's t-test and multiple corrections with a false discovery rate of 5%. Finally, genes were grouped based on their biologic function using Affymetrix NetAffx and the National Center for Biotechnology Information's Locuslink.

All microarrays used in this study exhibit a comparable distribution of signal intensities (Figures 9 and 10), fulfilling the requirements for performing gene expression comparisons. Microarray quality control parameters are summarized in Additional data file 4.

Related microarray data are deposited at Gene Expression Omnibus under the following GenBank accession numbers: GSM103419, GSM103420, GSM103421, GSM103422, GSM103423, GSM103424, GSM103425, GSM103426, GSM103427, GSM103428, GSM103429, and GSM103430.

The reverse transcription was performed using Superscript II (Invitrogen), in accordance with the manufacturer's protocol. Briefly, 5 μg total RNA was reverse transcribed with random hexamers. First-strand reverse transcribed cDNA was then diluted 1:10 in water before use for real-time PCR. Primers (Metabion, Martinsried, Germany), as listed in Table 6, were used together with the qPCR Core Kit (Eurogentec, Köln, Germany) in an ABI 7900 HT system (Applied Biosystems, Darmstadt, Germany), in accordance with the manufacturer's instructions. Real-time PCR data analysis was performed using the comparative C_T method (Applied Biosystems), with hypoxanthine guanine phosphoribosyl transferase (Hprt) as an endogenous reference. Statistical significance of expression differences detected by qRT-PCR was examined using a two-tailed, two-sample equal variance t-test.

For each given Affymetrix probe set identity (ID), the corresponding Ensembl gene IDs were retrieved from Ensembl 44. Promoter sequences were then downloaded from Ensembl (release 34, October 2005). One species was defined as reference species (Mus musculus), and orthologous genes from Homo sapiens and Rattus norvegicus were extracted. For each gene, a sequence corresponding to 2 kb upstream of the known transcription start site of each transcript was taken. In case of genes yielding more than one transcript, the window was enlarged to give a stretch of sequence ranging from 2 kb upstream the most 5' starting transcript to the 5' end of the most downstream starting transcript. No repeat masking was involved. Each set of orthologous sequences was screened for the presence of NFI motifs given a threshold for the score of 80%. The NFI motif is represented by a positional weight matrix. All possible permutations of detected sites from all species under study were aligned using ClustalW. The 'percentage identity' value, as calculated by ClustalW, was taken as motif conservation. Sets with a conservation better than 80% were reported.

In total, 395 Affymetrix probe sets resulted in 714 Ensembl gene IDs. The motif finding process detected 172 conserved motifs corresponding to 83 Affymetrix probe names, 88 Ensembl gene IDs, 79 unique gene symbols, and 128 motifs (Additional data file 3).