Altered HOX and WNT7A expression in human lung cancer

Abstract

HOX genes encode transcription factors that control patterning and cell fates. Alterations in HOX expression have been clearly implicated in leukemia, but their role in most other malignant diseases remains unknown. By using degenerate reverse transcription-PCR and subsequent real-time quantitative assays, we examined HOX expression in lung cancer cell lines, direct tumor-control pairs, and bronchial epithelial cultures. As in leukemia, genes of the HOX9 paralogous group and HOXA10 were frequently overexpressed. For HOXB9, we confirmed that elevated RNA was associated with protein overexpression. In some cases, marked HOX overexpression was associated with elevated FGF10 and FGF17. During development, the WNT pathway affects cell fate, polarity, and proliferation, and WNT7a has been implicated in the maintenance of HOX expression. In contrast to normal lung and mortal short-term bronchial epithelial cultures, WNT7a was frequently reduced or absent in lung cancers. In immortalized bronchial epithelial cells, WNT7a was lost concomitantly with HOXA1, and a statistically significant correlation between the expression of both genes was observed in lung cancer cell lines. Furthermore, we identified a homozygous deletion of beta-catenin in the mesothelioma, NCI-H28, associated with reduced WNT7a and the lowest overall cell line expression of HOXA1, HOXA7, HOXA9, and HOXA10, whereas HOXB9 levels were unaffected. Of note, both WNT7a and beta-catenin are encoded on chromosome 3p, which undergoes frequent loss of heterozygosity in these tumors. Our results suggest that alterations in regulatory circuits involving HOX, WNT, and possibly fibroblast growth factor pathways occur frequently in lung cancer.

Figure 1

RT-PCR amplification of HOX and FGF loci. (A) Degenerate RT-PCR products from the homeodomain (arrow) were amplified from nonmalignant (“normal”) lung, NHBE, two passages of T-antigen-immortalized bronchial epithelial cells (TR5214), and lung cancer cell lines. The leukemia line, MV4;11, is shown for comparison. (B) RT-PCR for FGF10 and FGF17 in lung tumor lines, a colorectal carcinoma, NCI-H630, and indicated controls.

Figure 2

(A) (Upper Left) Quantitative RT-PCR products for G3PDH, HOXA1, HOXA9, HOXA10, HOXB9, and WNT7a from NHBE cells and TR5214, passage 83. (Upper Right) Examples of quantitative RT-PCR products from tumor lines for HOXB9 and HOXA9. NT, no template. (Lower) Similar products shown for WNT7a. (B) Western blot for HOXB9. (Left) The indicated cell lines and direct tumor pair, as described in the text, were analyzed for the 30-kDa HOXB9 protein. Tubulin (55 kDa) served as a loading control. (Right) Identical Western blots contained (lane 1) 3-fold and (lane 2) 1-fold amounts of GST-HOXB9 fusion protein and (lane 3) tumor NCI-H513. No signal was observed for the endogenous HOXB9 (arrow) when the primary antibody was preincubated with GST-HOXB9-bound beads (blocked). (C) Southern blot demonstrating the specific absence of the β-catenin gene in EcoRI-digested genomic DNA from NCI-H28. A control probe (D3S30) is shown for comparison. Immunofluorescence confirmed loss of β-catenin protein in NCI-H28. Nuclei were visualized with 4′,6-diamidino-2-phenylindole (DAPI).

Figure 3

Correlation plots for HOX gene expression in cell lines. (A) ΔCt values for HOXA10 (x axis) are plotted against the sum of ΔCt values for all HOX loci analyzed (y axis). The trend line is shown along with the correlation coefficient (Pearson r = 0.77; P = 0.0001). (B) Similar analysis for HOXA1 (x axis) vs. WNT7a (y axis). In this case, the Spearman correlation coefficient is r = 0.56 with P = 0.0038. Note that the trend line is skewed by the presence of one outlying data point.