Temporally regulated neural crest transcription factors distinguish neuroectodermal tumors of varying malignancy and differentiation

Abstract

Neuroectodermal tumor cells, like neural crest (NC) cells, are pluripotent, proliferative, and migratory. We tested the hypothesis that genetic programs essential to NC development are activated in neuroectodermal tumors. We examined the expression of transcription factors PAX3, PAX7, AP-2alpha, and SOX10 in human embryos and neuroectodermal tumors: neurofibroma, schwannoma, neuroblastoma, malignant nerve sheath tumor, melanoma, medulloblastoma, supratentorial primitive neuroectodermal tumor, and Ewing's sarcoma. We also examined the expression of P0, ERBB3, and STX, targets of SOX10, AP-2alpha, and PAX3, respectively. PAX3, AP-2alpha, and SOX10 were expressed sequentially in human NC development, whereas PAX7 was restricted to mesoderm. Tumors expressed PAX3, AP-2alpha, SOX10, and PAX7 in specific combinations. SOX10 and AP-2alpha were expressed in relatively differentiated neoplasms. The early NC marker, PAX3, and its homologue, PAX7, were detected in poorly differentiated tumors and tumors with malignant potential. Expression of NC transcription factors and target genes correlated. Transcription factors essential to NC development are thus present in neuroectodermal tumors. Correlation of specific NC transcription factors with phenotype, and with expression of specific downstream genes, provides evidence that these transcription factors actively influence gene expression and tumor behavior. These findings suggest that PAX3, PAX7, AP-2alpha, and SOX10 are potential markers of prognosis and targets for therapeutic intervention.

Figure 1

In truncal sections of a human embryo at approximately 8 weeks of gestation, antibodies PAX3, AP-2α, SOX10, or PAX7 specifically labeled NC-derived cells according to their state of differentiation. Slides were labeled by immunocytochemistry using HRP and DAB, with hematoxylin counterstaining. (A) Top panel: Antibodies to PAX3 labeled cells of the dorsal neural tube (NT), dorsal root ganglia (DRG), and migrating NC-derived cells along an intervening trajectory (arrowhead), while also labeling dermatomyotomal cells (arrows, lower panel). (B) Antibodies to AP-2α labeled migratory NC-derived cells as they exit the dorsal NT (black arrowhead), as well as cells of the DRG, the ventral root, as it leaves the NT (white arrowhead), and the embryonic epithelium (arrow). Noto = notochord. (C) Antibodies to SOX10 labeled cells in the DRG, the dorsal and ventral roots (arrowheads), and the para-aortic sympathetic tissue (arrow). Ao = aorta. (D) Upper Panel: Antibodies to PAX7 labeled no cells of the NT or NC at this level, but, like antibodies to PAX3, labeled cells of the dermatomyotome (arrowheads, lower panel). Bar = 100 µm.

Figure 2

NC markers labeled specific terminally differentiated cells, visualized by using immunocytochemistry. (A) SOX10 was expressed by Schwann cells in a peripheral nerve (arrowheads mark examples). (B) SOX10 was expressed by enteric glia (arrowheads mark examples), seen here surrounding enteric neurons (arrows), notable for their large unlabeled nuclei and extensive cytoplasm. (C) Oligodendrocytes in a temporal lobe section expressed SOX10. In this preparation, SOX10 is labeled with Alexa red, GFAP is labeled with Alexa green, and nuclei are counterstained with bisbenzamide. An astrocyte (arrow) strongly expressed GFAP, but did not express SOX10. (D) In a skin biopsy, SOX10 was expressed by melanocytes, identified by their density and position in the basal layer of the epidermis (arrowheads mark examples). (E) Melanocytes labeled with PAX3 antibodies (arrowhead). PAX3-expressing cells were more rare than cells expressing SOX10. (F) All cells within the epidermis expressed AP-2α. Bar = 40 µm.

Figure 3

SOX10, in combination with AP-2α or PAX3, in benign glial neoplasms, demonstrated by immunocytochemistry. SOX10 was expressed by cells in schwannoma (A) and neurofibroma (B). Schwannoma cells expressed AP-2α (C), whereas neurofibroma cells expressed PAX3 (D). Bar = 40 µm.

Figure 4

SOX10 and AP-2α were expressed in neuroblastoma cells, demonstrated by immunocytochemistry. In ganglioneuroma, antibodies to SOX10 (A) or AP-2α (B) labeled elongated Schwann-like cells. In grade IV neuroblastoma, antibodies to SOX10 (C) or AP-2α (D) labeled round, undifferentiated cells. The labeling was weaker in cells of stage IV neuroblastomas than ganglioneuromas. Bar = 20 mm.

Figure 5

Marker expression in melanoma, MNST, and ES cells. (A–C) Melanoma cells labeled with antibodies to SOX10 (A), AP-2α (B), and PAX3 (C). Most cells express each of the markers. (D–G) MNST cells, labeled with antisera to SOX10 (D), AP-2α (E), PAX3 (F), and PAX7 (G). Cells expressing SOX10 and AP-2α are relatively scattered, whereas PAX3 and PAX7 are more uniformly expressed. (H–K) ES cells labeled with antisera to SOX10 (H), AP-2α (I), PAX3 (J), and PAX7 (K). As in melanoma, markers are expressed by almost all cells. Bar = 40 µm.

Figure 6

PAX3 was expressed by clusters of cells in classic MBL (A) and SPNET (B). This patchy distribution may lead to undercounting of the cases in which PAX3 is detected, as a small sample from a large tumor may not include discrete regions of PAX3 expression. Bar = 40 µm.

Figure 7

Expression by ganglioneuroma cells of genes regulated by SOX10 and AP-2α. P0 (A), a downstream target of SOX10, and ERBB3 (B), a downstream target of AP-2α, visualized by immunocytochemistry. Label was primarily seen in elongated cells resembling Schwann cells. Bar = 20 µm.

Graph 1

SOX10 and AP-2α mRNA, detected by microarray hybridization, were more abundant in ganglioneuromas than in stage IV neuroblastomas. Affymetrix intensity scores are plotted for each of the nine tumors. Lanes 1–3: Scores for ganglioneuromas; lanes 4–9: scores from stage IV neuroblastomas. Abundance of lactate dehydrogenase mRNA, evaluated as a control, did not vary with tumor type.

Graph 2

SOX10 (A) and AP-2α (B) mRNA abundance, measured by real-time RT-PCR with SYBR green detection, was greater in ganglioneuromas than in stage IV neuroblastomas. Each curve represents fluorescence intensity per cycle for an individual tumor. Ganglioneuromas are shown in shades of blue, and neuroblastomas in shades of red. Negative control with no template is shown in gray. All ganglioneuroma curves are to the left of the neuroblastoma curves, indicating greater abundance of specific message. For SOX10, the average crossing point for ganglioneuroma was 18, and for neuroblastoma was 24, indicating a 64-fold difference. For AP-2α, the average crossing points were 25 and 30, indicating a 32-fold difference. In contrast, β-actin mRNA abundance was similar among all tumors and did not vary with tumor type (C).

Graph 3

STX mRNA, measured by real-time RT-PCR with FRET detection, was greater in classic than in desmoplastic MBLs (A). Each curve represents intensity per cycle for an individual tumor. Classic MBLs are shown in shades of red, and desmoplastic tumors in shades of blue. Negative control with no template is shown in gray. All classic MBL curves are to the left of the desmoplastic variant curves, indicating greater expression of STX mRNA. The average crossing point for classic MBLs was 40, and for desmoplastic tumors was 46, indicating an eight-fold difference. In contrast, β-actin mRNA abundance was similar among all tumors and did not vary with tumor type (B). Melting point analysis revealed that the signal from the negative control in panel (B) was due to nonspecific PCR products (not shown).