Phenotypic differences between two subtypes of SH-SY5Y
We observed clear differences in the RA-induced neuronal phenotype of the two SH-SY5Y subclones, SH-SY5Y-A and SH-SY5Y-E, which were obtained from two different bioresource centres (ATCC and ECACC, respectively). After 5 days of RA treatment, ECACC (SH-SY5Y-E) cells were slightly larger in size (Fig. 1A) and contained a significant amount of neuroblastic (N-type) cells and a small fraction of epithelial (S-type) cells (Fig. 1B). The appearance of S-type cells was suppressed by RA-mediated differentiation on laminin, which promotes neurite outgrowth of SH-SY5Y cells [15]. Full differentiation of RA-treated cells required sequential treatment with BDNF (Fig. 1C), thus supporting previous findings [10]. Conversely, ATCC (SH-SY5Y-A) cells easily differentiated into a neuronal phenotype with extended neurites and formed a neural network after 5 days of RA treatment (Fig. 1E), again supporting previous observations [16]. The neuronal phenotype of SH-SY5Y-A cells was further enhanced by sequential treatment with RA and BDNF in serum-free medium for 3 days (Fig. 1F). The original SK-N-SH human neuroblastoma cell line comprises two different cell types: a large fraction of epithelial Schwann cells (S-type cells); and a small number of neuroblast cells (N-type cells) [17]. No neural network is formed by the larger fraction of SK-N-SH cells (> 80%) in the presence of RA (Fig. 1H). Instead, these cells elongate and gradually undergo cell death with sequential BDNF treatment under serum-free conditions (Fig. 1I).
Identification of expressed genes by using microarray analysis
In order to identify the genes required for progression of neuronal differentiation in RA-treated neuroblastoma cells, we first performed a microarray analysis of global gene expression profiles for the two subtype clones of SH-SY5Y and SK-N-SH cells treated with RA for 5 days. We also included a perturbation experiment using a potent PI3K inhibitor, LY294002, to impair RA-induced differentiation of SH-SY5Y cells (data not shown) [6, 7]. For full differentiation of RA-treated SH-SY5Y-E cells, gene expression profiles were also analysed following sequential BDNF treatment of cells for an additional 3 days [10].
We selected 3690 genes (5119 probes) for comparative analysis of gene expression profiles. These genes were extracted by combining 3100 genes (3628 probes) derived from a sum of the top 1000 probes exhibiting significant expression patterns by ANOVA [18, 19] under the 5 culture conditions described above, with 1059 genes (1950 probes) identified as being more significant (Fig. 2).
Classification of genes regulated by the PI3K signalling pathway
We first compared the differential gene expression profiles of SH-SY5Y-A cells and SK-N-SH cells, because there is a clear phenotypic difference between these cell lines under RA-treated conditions (Fig. 1). Two-factor ANOVA has previously been used for the statistical analysis of normalised data in order to determine differences in gene expression between cell lines and time points [20]. As summarised in Figure 2, we identified a gene cluster containing 2517 genes with significantly different expression profiles (p < 0.001) between SH-SY5Y-A cells and SK-N-SH cells treated with RA for 5 days. We also identified 513 genes with expression levels that were significantly down- or up-regulated when PI3K was inhibited in RA-treated SH-SY5Y-A cells. Interestingly, 448 of these genes (87.3%) were common in the product sets of gene clusters selected on the basis of differences in expression pattern between SH-SY5Y-A and SK-N-SH. Moreover, expression behaviours of the selected genes were almost identical to those in RA-treated SK-N-SH cells. By removing genes with contradictory profiles or low expression levels, we finally identified gene clusters A and B, comprising 386 genes, and a third "other" cluster with 33 genes regulated by the PI3K signalling pathway in RA-treated SH-SY5Y-A cells.
Figures 3 and 4 show a heat map of expression profiles for all genes in the clusters. We calculated relative signal values (0.000–1.000) of gene expression levels in each cluster under all culture conditions, as summarised in Additional File 1. Cluster A comprises 316 genes with down-regulated expression in RA-treated SH-SY5Y-A cells cultured with LY294002 or extremely low-level expression in RA-treated SK-N-SH cells. Cluster B includes 71 genes with up-regulated expression in RA-treated SH-SY5Y-A cells cultured with PI3K inhibitor or high-level expression in SK-N-SH cells (Figs. 2, 3 and 4 ). Half of the genes in cluster A showed average signal values < 0.1 in RA-treated SK-N-SH cells, too low for quantitative comparison (Additional File 1). To address the effect of a PI3K inhibitor on gene expression in the two clusters, we compared the average signal values of gene expression levels greater than 0.179, the average signal value in the clusters, between RA-treated SK-N-SH cells in the presence and absence of LY294002. In cluster A, the average signal values of 32 of 77 (42%) gene expression levels did not dip below 75% of the average values when RA-treated SK-N-SH cells were exposed to LY294002. Similarly, the average signal values in 35 of 65 (54%) gene expression levels in cluster B did not exceed 125% of the average values by the addition of a PI3K inhibitor (Additional file 1). These findings are consistent with a substantial defect in transcriptional regulation mediated by the PI3K signalling pathway in RA-treated SK-N-SH cells.
We further categorised genes into two subgroups of neural functions and other functions (Figs. 3 and 4), on the basis of Gene Ontology (GO) annotation [21, 22] and a comprehensive search of the literature [22–24]. Genes with neural functions were defined when significant GO term(s) and/or literal description(s) implicating involvement in neural events were obtained by the gene annotation. We also defined genes with other functions when neither the GO term nor the literal description was related to neural events. GO annotation in two clusters yielded the identification of 114 genes with GO terms related to neural events (Additional File 2). Cluster A genes included 158 genes in the neural function subgroup and 158 genes in the other functions subgroup. Cluster B genes were similarly classified, with 11 genes in the neural function subgroup and 60 genes in the other functions subgroup (Fig. 3 and 4). Neural functions were further divided into 9 functions, as summarised in Additional File 2.
Differences in PI3K-mediated transcriptional regulation between two subtypes of SH-SY5Y
To investigate whether the PI3K signalling pathway is sufficiently activated in RA-treated SH-SY5Y-E cells, we further focused on the transcriptional regulation of the 386 genes included in these clusters. Differences in expression profiles of RA-treated SH-SY5Y-E cells with and without LY294002 were compared to differences in expression profiles of RA-treated SH-SY5Y-A cells with and without LY294002 (Fig. 3 and 4). Neurite outgrowth induced in RA-treated SH-SY5Y-E cells was also inhibited by LY294002, thus suggesting that the PI3K signalling pathway is functional in SH-SY5Y-E cells. However, most of the gene expression in the clusters was regulated in a PI3K-independent manner, and half of the neural genes were expressed at extremely low levels (Figs. 3, 4 and 5A), whereas transcriptional inhibition by LY294002 exclusively emerges after 2–3 days in the presence of RA (Fig. 5A, 5B; bottom panels). These results suggest that the PI3K signalling pathway is partially disrupted in RA-treated SH-SY5Y-E cells, with this disruption leading to reduced differentiation. These results also suggest that supplementation with an additional TRKB signalling pathway is required for full differentiation of SH-SY5Y-E cells after down-regulation of the gene cluster required for neural events.
Human TRKB is alternatively spliced into at least 3 variants: TRKB-FL; TRKB-T1; and TRKB-T-Shc [25]. TRKB-FL is a tyrosine kinase-containing variant, whereas the intracellular tyrosine kinase domain is truncated in the other proteins [26, 27]. Two typical gene expression profiles were observed by microarray analysis in accordance with different transcriptional regulation on the alternative promoters (Fig. 5B; bottom panels) [28]. In SH-SY5Y-A cells, TRKB-FL and TRKB-T-Shc variants showed continuously induced transcription over 5 days, whereas gene expression of TRKB-T1 was initially induced but plateaued 1 day after RA treatment. In SH-SY5Y-E cells, expression of all TRKB variants, particularly TRKB-T1, was abruptly induced 3 days after RA treatment and peaked after 5 days in the presence of RA (Fig. 5B; bottom panels), thus supporting previous results [29]. These TRKB variants were also transcriptionally regulated by the PI3K signalling pathway (Fig. 5B; bottom panels), indicating clear cross-talk between the TRKB and PI3K signalling pathways. Conversely, induction of TRKB variants was not observed in SK-N-SH cells at any time, providing a possible explanation for the defect in the neuronal phenotype of these cells (Fig. 5B, bottom panels).
Transcriptional compensation by an additional TRKB-mediated signalling pathway in SH-SY5Y-E
Additional BDNF treatment of SH-SY5Y cells stimulates tyrosine phosphorylation of TRKB [9], promoting neurite outgrowth in serum-free medium [8, 10]. Supporting these findings, additional exposure of RA-treated SH-SY5Y-E cells to BDNF resulted in full differentiation and induced 89 genes that were down-regulated or maintained at low expression levels with RA treatment only (Figs. 5B, 6). These genes were further sorted into two gene pools: 54 genes from cluster A (shown in blue in Fig. 3; Fig. 6, left panel) and 35 genes regulated in a PI3K-independent manner (Fig. 6, right panel). As expected, both of these pools contain many genes required for neural events (Fig. 6; Additional File 2), indicating that these genes are critical for TRKB-mediated differentiation of SH-SY5Y-E cells. These results also demonstrate that genes down-regulated either by an impaired PI3K signalling pathway or by another abrogated signalling pathway are transcriptionally compensated for via an additional TRKB-mediated signalling pathway, leading to full differentiation (Additional File 3). However, a significant number of genes down-regulated or maintained at low expression levels in RA-treated SH-SY5Y-E cells did not exceed the threshold (Fig. 6 legend) following additional BDNF treatment (Fig. 3; Additional File 1), thus suggesting that the TRKB-dependent signalling pathway only partially compensates for an impaired PI3K signalling pathway. The TRKB variants TRKB-FL and TRKB-T-Shc were induced and plateaued following sequential treatment of SH-SY5Y-E cells with RA and BDNF. Conversely, TRKB-T1 expression, which was strongly induced by RA treatment of SH-SY5Y-E cells, was abruptly down-regulated after a shift to serum-free medium with BDNF (Fig. 5B, bottom panels), which supports the notion that the dominant-negative function of tyrosine kinase-deficient receptors [30] is totally diminished by induction of functional TRKB-FL. Further studies may be required to identify the components and transcription factors involved in the TRKB and PI3K signalling pathways.
In conclusion, we identified gene clusters that are transcriptionally controlled by two different signalling pathways mediated by PI3K and TRKB during the differentiation of two subtypes of SH-SY5Y cells. These expression profiling data may prove useful in further elucidating the molecular mechanisms regulating the promoter activities of genes required for neuronal differentiation. These promoter activities are mediated by an upstream signal transduction-transcriptional network including PI3K and/or TRKB.