The paired box (PAX) family of genes comprises 9 members, which encode nuclear transcription factors controlling development. PAX proteins are characterized by the presence of a paired domain (PD), a conserved amino acid residue motif with DNA-binding activity. PAX3 plays a critical role in cell proliferation, differentiation and migration during embryonic development of cells in vivo.1 Alternative transcripts of PAX3 have been identified in tissues, including human adult skeletal muscle and mouse embryos. PAX3 produces up to 7 different isoforms in man, which result from alternative splicing.2, 3, 4 An isoform of human PAX3 (PAX3c) was first described by Macina et al.,5 comprising 8 exons and containing a PD, a paired type homeodomain (HD) and a transactivation domain (TA). Barber et al.3 determined that the full-length PAX3 gene (PAX3e) contains 10 exons encoding 510 amino acid residues. Parker et al.4 in our laboratory found the new isoform PAX3g, in which exon 8 and intron 8 are spliced, translation proceeding from exon 7 to 9. PAX3g thus lacks part of the TA encoded by exon 8, but includes exon 9.
Our group has shown different expression patterns of PAX3 isoforms within tumors of the same type or cell line, and also variation among different neural crest tumors or cell lines.4, 6 Functional analysis further indicated that these isoforms have different activities in stably transfected mouse melanocytes in vitro.7 Pritchard et al.8 found that mouse Pax3g (Pax3Δ8) was transcriptionally inactive in a transient transfection assay but could inhibit Pax3d activity in mouse myoblasts, presumably by competing for Pax3 binding sites.
In vitro studies demonstrate that both the PD and HD of PAX3 protein are involved in DNA binding and the TA plays a key role in regulating HD function.9, 10, 11 Potential downstream targets of PAX3/Pax3 have been identified, including Met, a proto-oncogene12; Waardenburg's syndrome type II associated gene, MITF13 and the pigmentation gene, Trp1.14 Other targets such as STX, TIMP3, NF-kB, NCAM, MyoD, Dep-1 and Mbp gene have also been reported.15, 16 Barber et al.17 using a “cyclic amplification and selection of targets” (CASTing) strategy identified further PAX3 targets including Itm2A, Fath, FLT1 and TGFα. The genomic DNA near these genes contained PAX3 binding sites, conferring PAX3-dependent regulation.
The large-scale expression analysis of microarrays enables observation of the broad effects of transcription factors on gene expression and potentially elucidates their roles in tumorigenesis. Alternative isoforms differ in structure as well as in binding and activation of target genes.18 To better understand the biological roles of PAX3 isoforms in melanocytes, it is important to know their downstream target genes and whether they have similar or distinct regulatory effects on target gene expression. In this study, microarrays were used to determine differences in gene expression between the empty vector control cells and PAX3c, e or g isoform transfected melanocytes. Some of the potential target genes discovered on microarrays were validated by semi-quantitative RT-PCR and Western blotting.
Material and methods
Melan a-PAX3c, -PAX3g and -PAX3e stable transfectants and Melan a-pcDNA4 empty vector control transfectants7 were grown in DMEM containing 10% FBS, 1 mM L-glutamine, 100 units/ml penicillin, 100 μg/ml streptomycin and 400 μg/ml zeocin at 37°C and 5% CO2 in air until they were 70–80% confluent.
Total RNA from PAX3c, e and g isoform transfectants and empty vector control cells was isolated using a total RNA isolation system (RNAgents, Promega, Southampton, UK) according to the manufacturer's instructions. The RNA concentration was measured spectrophotometrically, and aliquots were examined for integrity by formaldehyde agarose electrophoresis. RNA samples from 3 clones of each PAX3 isoform transfectants were pooled, and their concentration measured before equal amounts of total RNA were used for microarray analysis.
Mouse known gene SGC oligo set arrays (version 2) on glass slides were obtained from the HGMP Resource Centre, funded by the MRC. These arrays utilize a commercial oligonucleotide library designed by Compugen and synthesized by Sigma-Genosys, designed to detect a major sample (7,524) of known mouse genes. Each oligo is represented in duplicate on every array (http://www.hgmp.mrc.ac.uk/research/microarray). cDNA fluorescently labeled with either Cy3 or Cy5 (Amersham Pharmacia Biotech, Buckinghamshire, UK) was synthesized from 50 μg total RNA by oligo(dT)15 (Promega) primed polymerization using Superscript II reverse transcriptase (InVitrogen, Paisley, UK) as described by Khan et al.19 The same reference RNA from Melan a-pcDNA4 empty vector transfected cells was used as a control in each hybridization. Control cDNA was labeled with one of the 2 fluorochromes, Cy3 or Cy5, and the test cDNA (from Melan a-PAX3c, -PAX3g, or -PAX3e stable transfectants) with the other. The labeled cDNA probes were purified by column chromatography (AutoSeqG-50, Amersham Pharmacia Biotech). The microarray slides were soaked in prehybridization buffer for 1 hr at 42°C. A pair comprising labeled control and one test cDNA were hybridized to the same microarray at 50°C overnight. The arrays were washed twice with 2× SSC for 5 min, twice with 0.1× SSC/0.1% SDS for 5 min, and twice with 0.1× SSC for 5 min, followed by ddH2O for 1 min and isopropanol for 1 min to help minimize background.
Imaging and image analysis
The fluorescent probes bound to the arrayed oligo DNAs were detected using a GenePix® 4100 scanner (Axon Instruments, USA). Image analysis was performed using GenePix Pro 5.0 software (Axon Instruments, USA). Each image is the ratio of pixel intensity at 635 nm to that at 532 nm, reflecting the ratio of test to control. The mean of the ratios for housekeeping genes was taken and adjusted to 1.0 for normalization of the expression levels of other genes. Comparing the expression levels of genes in the isoform transfectants with those in control cells, a 2-fold up-/downregulation was chosen as the cut-off point for the data; 1.5- to 2-fold up-/downregulation (ratios 1.5–2 or 0.67–0.5) were regarded as borderline and used only for comparison with a value of over 2-fold in another isoform transfectant; 1.0- to 1.5-fold up-/downregulation (ratios 1.0–1.5 or 1.0–0.67) were regarded as no change. Genes with changed expression were classified into different groups according to their functions, as determined by a search through the Medline database. Each microarray comparison of paired cDNAs was carried out in duplicate, and the results had to have the same trend on both arrays, in order to be included in the list of changed genes.
RNA (1 μg) from PAX3c, e, g isoform transfectants and empty vector controls was reverse transcribed into cDNA using a single-strand cDNA synthesis kit (Promega) according to the manufacturer's instructions. cDNA from individual transfectants (1 μl) was amplified with different primers (InVitrogen) to examine Dhh, Ffg17, Kitl, MyoD, Met, Msx1, Rac1 and Muc18 gene expression by PCR (Table I). Briefly, PCR was carried out for 30 cycles (94°C for 1 min, 58°C for 1 min and 72°C for 1.5 min) with a final incubation at 72°C for 10 min. This was the linear region of the graph in which cycle number proportional to amount of DNA produced. The example amplifications of Ffg17, GAPDH at 20, 25, 30, 35, 40 cycles are shown in Figure 1. The housekeeping gene (GAPDH) was amplified 30 cycles in parallel as a loading control and mouse cDNA synthesized from commercial RNA pooled from 5 tissues including brain, heart, liver, kidney and spleen (Clontech, Oxford, UK) was a positive control. The negative control contained water in lieu of cDNA. The PCR products were electrophoresed on a 1% agarose 1× TBE gel and stained with ethidium bromide. The relative band intensities were quantified using Scion Imaging Software (http://www.scioncorp.com) and normalized to GAPDH in the same cDNA sample.
Table I. Primer Sequence and Expected Fragment size in RT-PCR (F: Forward Primer; R: Reverse Primer)
Predicted fragment size (bp)
Western blot analysis
Cell lysates containing 30 μg of total protein from PAX3c, e.g., isoform transfectants or empty vector controls were separated using 12% SDS/PAGE electrophoresis and transferred to a nitrocellulose membrane (Amersham Pharmacia Biotech). The membrane was incubated for 1–2 hr at room temperature in Tween-PBS buffer supplemented with 5% (w/v) dry milk to block nonspecific binding sites. After 6 hr incubation with primary antibody (Table II), the membrane was washed with Tween-PBS buffer and hybridized with secondary antibody. Rabbit anti-actin antibody was used as a loading control. Horseradish peroxidase-conjugated secondary antibodies, including mouse anti-rabbit IgG (1:2,000 dilution), rabbit anti-mouse IgG (1:1,000 dilution) and rabbit anti-goat IgG (1:2,000 dilution), were from DAKO, Denmark. Immunoreactive proteins were visualized by the enhanced chemiluminescence system ECL or ECLplus (Amersham Pharmacia Biotech). Signal intensity was obtained by film scanning on a laser densitometer. The relative protein band densities were normalized to α-actin density from the same sample.
Table II. Antibodies used in Western Blotting to Examine the Protein Expression of Genes Selected from Microarrays and Expected Antigen Size
Expected protein size (kDa)
Observed protein size (kDa)
Santa Cruz Biotechnology
Santa Cruz Biotechnology
Santa Cruz Biotechnology
Santa Cruz Biotechnology
Santa Cruz Biotechnologyc
Microarray screen for PAX3c, PAX3e and PAX3g isoform downstream target genes
Representative examples of arrays hybridized to samples are shown in Figure 2. Not all of the genes that are represented on the microarray are expressed in the cell lines examined. This is normal in any microarray experiments.20 Pixel intensities for unexpressed or poorly expressed genes were close to the background intensity. Red spots indicate the expression of gene transcripts in PAX3 isoform transfectants cDNA (test samples), that are upregulated when compared with those in the control sample. Green spots represent downregulated gene transcripts in the test samples compared with those in the control sample. Yellow spots indicate no difference in gene transcription between test and control. The arrays confirmed the expected PAX3 upregulation in PAX3c, PAX3e and PAX3g transfectants, for which the normalized data were 1.8-, 1.7-, 1.7-fold respectively. This compares very well with the RT-PCR data on the original transfected cell lines where the relative PAX3 isoform expression was similar in PAX3c, e and g transfectants (our unpublished data).
The expression of 109 genes was significantly altered among the 3 PAX3 isoform transfectants in comparison with vector control cells, when a 2-fold change of expression was used as threshold. By this criterion, 39 genes were upregulated and 20 downregulated in PAX3c transfectants; 29 genes upregulated and 8 downregulated in PAX3e transfectants; 19 genes upregulated and 16 downregulated in PAX3g transfectants. These genes were classified into different groups according to their major functions, such as differentiation, proliferation, migration or apoptosis (Table III). In addition, if the response to 1 PAX3 isoform was greater than a 2-fold change, then the responses to the other isoforms were included for comparison. For example, Met oncogene expression was upregulated 10-fold in PAX3c transfectants, 1.85-fold in PAX3e but not in PAX3g transfectants (1.00-fold). If an expression ratio over 1.5-fold was regarded as changed expression, among these 109 genes, 58 were upregulated, 26 downregulated and 25 unchanged in PAX3c transfectants; 56 genes upregulated, 18 downregulated and 35 unchanged in PAX3e transfectants; 50 genes upregulated, 21 downregulated and 38 unchanged in PAX3g transfectants (Fig. 3a). Only 11 of these 109 genes were found to be similarly affected by all 3 PAX3 isoforms (asterisks in Table III). Another 31 genes displayed a similar transcription pattern in PAX3c and PAX3e transfectants, while 11 genes showed a similar transcription pattern in PAX3c and PAX3g transfectants or 11 in PAX3e and PAX3g transfectants (Fig. 3b). It appears that PAX3c and PAX3e transfectants have more altered genes in common and the 2 isoforms might have similar regulatory effects.
Table III. ALTERATION OF GENE EXPRESSION IN PAX3 ISOFORM TRANSFECTANTS COMPARED WITH EMPTY VECTOR CONTROL CELLS, ASSESSED BY OLIGO MICROARRAY ANALYSIS
Main functions of the 109 genes listed above are differentiation (D), proliferation (P), migration (M), adhesion (Ad), apoptosis (Ap) or angiogenesis (An).
BCL2/adenovirus E1B 19 kDa-interacting protein 1, NIP3
Fas-associating protein with death domain
Fas activated serine/threonine FAST kinase
Growth differentiation factor 15
Heat shock protein, 25 kDa
Myeloid cell leukemia sequence 1
Phospholipid scramblase 2
Tissue inhibitor of metalloproteinase 3
Glial cell line derived neurotrophic factor family receptor α1
Angiogenin related protein 2
Interferon β, fibroblast
Interferon-stimulated protein (15 kDa)
An, P, M
Pigment epithelium-derived factor
Tissue inhibitor of metalloproteinase 2
An, D, P
Analysis of genes altered in response to PAX3c, PAX3e or PAX3g expression revealed 3 situations: some genes were induced/repressed by all 3 PAX3 isoforms; some induced/repressed by 1 or 2 isoforms but not the other; some induced by 1 isoform, but repressed by the other two. These results indicate that different PAX3 isoforms are involved in regulation of cellular activities through direct or indirect modulation of different downstream target genes.
Confirmation of differential expression by RT-PCR and Western blotting
Semi-quantitative RT-PCR was carried out on 8 selected differentially expressed genes to confirm the expression pattern observed in microarrays. Eight genes, for which similar results were obtained by microarray and RT-PCR analysis (Fig. 4), were analyzed further for their protein expression by Western blotting (Fig. 5). The relative up-/downregulation of these genes in PAX3c, e or g isoform transfectants is shown in Figure 6.
The RT-PCR results demonstrate that Dhh was strongly upregulated in PAX3c and PAX3e transfectants, but only slightly in PAX3g transfectants (Fig. 4a). The pattern of Dhh mRNA expression detected in PAX3c and PAX3e transfectants was repeated at the protein level (Fig. 5a). These results confirmed the microarray results, but Dhh repression in PAX3g transfectants in the microarray and Western blot analyses was not observed in RT-PCR.
Fgf17 was increased in PAX3c, PAX3e and PAX3g transfectants at the mRNA level (Fig. 4b). Upregulation of Fgf17 in PAX3c and PAX3e transfectants in the RT-PCR assay confirmed the microarray result, but Fgf17 was unchanged in PAX3g transfectants (<1.5-fold on microarrays). Western blotting analysis confirmed the RT-PCR results; the expression of Fgf17 protein was increased in all 3 PAX3 isoform transfectants (Fig. 5b).
Kitl mRNA was increased in PAX3c, PAX3e and PAX3g transfectants (Fig. 4c), consistent with the microarray data. Western blot analysis revealed similar results to those of RT-PCR and microarrays. Kitl protein was increased in PAX3c, PAX3e and PAX3g transfectants (Fig. 5c).
RT-PCR analysis demonstrated that Met transcription was upregulated in PAX3c and PAX3e transfectants. No change occurred in PAX3g transfectants (Fig. 4d), thus confirming the microarray results. Met protein was increased in PAX3c transfectants, but was not altered in PAX3g transfectants (Fig. 5c). This finding is consistent with the RT-PCR results. However, the slight upregulation of Met mRNA in PAX3e transfectants in the microarray and RT-PCR assays was not repeated at the protein level.
RT-PCR analysis demonstrated that Msx1 transcription was upregulated in PAX3e and PAX3g but not in PAX3c transfectants (Fig. 4e), consistent with the microarray data. Msx1 protein was upregulated in PAX3e and PAX3g transfectants but not in PAX3c transfectants (Fig. 5e), confirming the RT-PCR results, although the changes were not as obvious.
Muc18 transcription as demonstrated by RT-PCR was upregulated in PAX3c transfectants, but no change occurred in PAX3e or PAX3g transfectants (Fig. 4f), confirming the microarray results for PAX3c and PAX3e, but not PAX3g transfectants. Muc18 protein expression was upregulated in PAX3c transfectants, but no change was observed in the other 2 isoform transfectants (Fig. 5f), consistent with the RT-PCR results.
MyoD1 was significantly upregulated in PAX3c and PAX3e, but unchanged in PAX3g transfectants (Fig. 4g), consistent with the microarray data. MyoD1 protein expression was upregulated in PAX3c transfectants, but unchanged in PAX3e or PAX3g transfectants (Fig. 5g). These data confirmed the RT-PCR results, except that upregulation of MyoD1 transcription in PAX3e transfectants was not obvious at the protein level.
Rac1 was slightly downregulated in PAX3c and PAX3e transfectants, but upregulated in PAX3g transfectants (Fig. 4h), confirming the microarray results. Rac1 protein expression in PAX3c, PAX3e and PAX3g transfectants (Fig. 5h) was consistent with the microarray and RT-PCR results.
The properties, regulation and functional roles of PAX3 isoforms have not been fully elucidated. PAX3c consists of 8 exons, containing complete versions of the DNA binding domains and a TA; PAX3e is a full-length isoform with 10 exons; PAX3g has an incomplete TA with exon 8 spliced out. These 3 isoforms are expressed differently in melanomas and their cell lines: PAX3c is predominantly expressed, while PAX3e and g are expressed at low levels.4 They have different effects on melanocyte proliferation, migration and transformation in vitro.7
Microarray analysis identified 109 genes that were up- or downregulated >2-fold in PAX3 isoform transfectants when compared with controls. A comparison of PAX3c, PAX3e and PAX3g isoform transfectants revealed that only 11 genes were similarly up- or downregulated by all 3 isoforms. How the differences in gene regulation are related to the molecular structure of the PAX3 isoforms is unknown, but the varying isoform structure leads to differences in binding and activation of target genes.18 The unique sequences of the carboxyl termini of PAX3c, e and g isoforms may be essential for such differences. A second, alternative splicing in PAX3 at the intron 2 and exon 3 junction results in inclusion or exclusion of a glutamine (Q) residue in the links between N- and C-terminal paired box subdomains. Both Q+ and Q− forms are expressed with similar abundance in developing mice, and whether they significantly influence the recognition of target genes is unknown.21
There was not complete overlap between the genes identified in this study and previous studies that had identified PAX3 target genes. MyoD, Met, Ret, Timp3 and Muc18, which had been detected previously,15 were identified in this study also. Others such as Dhh, Msx1, Kitl, Rac1 and the markers of melanoma progression, RhoC and Timp3, were identified only in the present study.22, 23 Genes identified in other studies such as Mitf, Trp1 and Stx13, 14, 15 were absent in this study. Several possible reasons exist. First, if the changed expression of a gene was <2-fold, it was excluded from the list. For example Mitf upregulation was 1.9-, 1.2- and 1.4-fold in PAX3c, PAX3e and PAX3g transfectants respectively. Secondly, the array used here represented only part of the murine genome and some PAX3 targets such as Trp1, Six1 and Myf4 were not on the microarrays. Third, different cell types produce different PAX3 isoforms4 and the altered expression of some gene targets may be cell-type specific. BMP4 was repressed by both PAX3 and PAX3/FKHR in SaOS-2 osteosarcoma cells, while in RD rhabdomyosarcoma cells, it was induced by PAX3/FKHR, but not PAX3.24
Microarray analysis is just a preliminary step in studying gene expression. Three independent assays revealed that mRNA expression detected by microarray and RT-PCR was not always reflected by expression of its corresponding protein. For example, MyoD and Met mRNAs were upregulated in PAX3e transfectants, but there was no significant alteration at the protein level. Such discrepancies involve post-transcriptional and post-translational mechanisms affecting mRNA or protein stability. The 8 genes on which the present study focused are, like others, not monofunctional (Table III). However, for the sake of simplicity, they have been considered in 3 groups representing cell proliferation (Fgf17, Kitl and Rac1), cell differentiation (Dhh, Msx1 and MyoD) or cell migration and adhesion (Met and Muc18).
FGFs control cell proliferation, migration, differentiation and homeostasis.25 FGF17 is expressed in the human central nervous system, and its overexpression in transfected NIH-3T3 cells has transforming and tumorigenic effects.26 In the present study, Fgf17 was upregulated in PAX3c, e and g transfectants, but these 3 cell lines did not have the same proliferation and transformation characteristics.7 Fgf17 may not influence melanocyte proliferation and transformation directly or other factors may be involved such as Kit ligand (Kitl), a glycoprotein growth factor required for melanocyte migration and proliferation.27 The survival effect of Kit signaling in melanocytes is mediated by MAPK, which may upregulate Mitf, another putative target gene of Pax3.28 Microarray analyses demonstrated Kit upregulation in PAX3c and PAX3e transfectants and the expression of Kitl was upregulated in PAX3c, PAX3e and PAX3g transfectants at both mRNA and protein levels. Thus, the proliferation of PAX3 isoform transfected melanocytes may be determined partly by Kitl/Kit signaling. Rac is required for G1 to S progression in 3T3 fibroblasts and regulates cell morphology, proliferation and survival.29 Activated Rac1 inhibits apoptosis triggered by anticancer drugs in melanoma cells.30Rac1 mRNA and protein expression were downregulated in PAX3c and Pax3e transfectants, but upregulated in PAX3g transfectants, indicating that it is regulated differently by PAX3c, e and g isoforms in melanocytes in vitro.
The group of genes involved in cell differentiation include Dhh, Msx1 and MyoD. Dhh is a functional homologue of Shh activity during epidermal stem cell homeostasis and skin tumor formation; activation of an ectopic Dhh pathway regulates epidermal basal cell proliferation.31Dhh expression was upregulated in PAX3c and PAX3e transfectants at both mRNA and protein levels, but the protein was downregulated in PAX3g transfectants. Further studies are required to ascertain whether PAX3 isoforms directly regulate Dhh.
The muscle segment homeobox 1 (Msx1) can repress gene transcription.32Msx1 and Msx2 are expressed in many tissues including the dorsal neural tube and neural crest, where Pax3 is also present.16 In chicken embryos, Msx1 and Pax3 were co-expressed in migrating limb muscle precursors and their proteins can form a complex in vitro, mediated by the HD of Msx1 and the PD of Pax3, thereby inhibiting DNA binding by Pax3.33Msx1 was upregulated at both mRNA and protein levels in PAX3e and PAX3g, but not in PAX3c transfectants. Since the Pax3-Msx1 complex inhibits DNA binding by Pax3, this might explain why expression of putative target genes of Pax3 in muscle, viz Met and MyoD, was not upregulated in PAX3e and PAX3g transfectants but was in PAX3c transfectants.
MyoD is a myogenic regulatory factor. Ectopic Pax3 expression can activate MyoD, although indirectly.34MyoD expression was upregulated in PAX3c transfectants, but unchanged in PAX3e and g transfectants. Msx1 is able to repress myogenesis, and by binding to the MyoD enhancer, it may directly inhibit MyoD transcription.35
Met expression in primary cutaneous malignant melanomas is significantly associated with the vertical growth phase, thick tumors, high mitotic index, lymphatic and vascular invasion.36 Met expression is significantly higher in metastatic melanoma and in several melanoma cell lines, compared with normal human melanocytes.37Met mRNA and protein were upregulated in PAX3c transfectants, but unchanged in PAX3e and PAX3g transfectants. These results agree with the observation that luciferase activity driven by a c-Met promoter was induced by the transient transfection of PAX3c, but not PAX3e or g (our unpublished data). Like Met, MUC18, a cell surface molecule, is a marker of melanoma progression, its expression being correlated with tumor thickness and metastasis.38 Muc18 is a downstream target of Pax3 in a medulloblastoma cell line.15 It is not detected on normal melanocytes, but occurs on primary melanoma cells and to a greater extent on metastatic melanoma cells.39Muc18 expression was upregulated in PAX3c transfectants at the mRNA and protein level, but unchanged in the other isoform transfectants. Therefore, PAX3c may be involved in melanocyte migration at least partly by upregulating Muc18 and Met.
In conclusion, the identification of genes that are differentially expressed in melanocytes by transfection of different PAX3 isoforms provides clues as to the biological actions of these isoforms. A significant number of the genes upregulated encode proteins that influence cell proliferation, such as Fgf17 and Kitl. A number of genes encoding proteins with roles in cellular differentiation were changed also, suggesting that PAX3 isoform expressing melanocytes may have altered differentiation. The list also included genes with roles in cell migration, apoptosis and angiogenesis. Thus, PAX3 isoforms may have many important biological functions in melanocytes. A key question arising from gene expression profiling is whether any particular gene is important for cell signal transduction and which level of its up- or downregulation is a critical “break point” for melanocyte transformation. Therefore, further work is required to elucidate the roles that specific downstream targets of PAX3 isoforms play in melanocytes, in vitro and in vivo. Since PAX3 expression is a characteristic of tumors derived from the neural crest (melanoma and neuroblastoma) or from the dorsal dermomyotome (rhabdomyosarcoma), further studies are needed into the roles of its isoforms in the different lines of differentiation represented by these tumors.