The netrin-1 receptors UNC5H are putative tumor suppressors controlling cell death commitment

Transient transfections of 293T or NIH 3T3 were performed as described (2, 11). Similarly, MCF7, LS174T or U373MG cells were transfected by using a Fugene 6 transfection reagent (Roche Molecular Biochemicals). For each cell line, the percentage of transfected cells was determined by measuring after 48 h the number of fluorescent cells in a cell population transfected with a GFP-expressing construct. Rodent UNC5H1–UNC5H2–UNC5H3-expressing constructs have been described (17). Ras-expressing construct was obtained from M. Castellazzi (Institut National de la Santé et de la Recherche Médicale, Lyon, France). pcDNA-p53 and pcDNA-MDM2 were obtained from A. Puisieux (Centre Leon Berard, Lyon, France). UNC5A-B-C probes were obtained by PCR amplification using cDNA from Daudi (for UNC5A and B) or HT29 (for UNC5C) cells as described in the Supporting Text, which is published as supporting information on the PNAS web site, www.pnas.org.

UNC5H gene expression in tumor/normal tissues was monitored by using the Cancer Profiling Array (CLONTECH) following the manufacturer's suggested procedure. UNC5H probes were prepared by using a Megaprime DNA Labeling System (Amersham Pharmacia) and the full-length UNC5H2 cDNA or the UNC5A–UNC5B–UNC5C-specific cDNA (see above) as template. As a control for the samples loading in the array, the ubiquitin probe was also used.

Genomic DNA (50 ng) from matched tumor and corresponding normal tissue was amplified by PCR and analyzed as described (23) by using fluorescently labeled primers for the indicated polymorphic microsatellite markers on chromosomes 5q35, 10q21, or 4q21–23 for, respectively, UNC5A, UNC5B, or UNC5C locus.

Genomic DNA (50 ng) from 30 colorectal tumors was amplified by PCR using the primers described in the Supporting Text. PCR products were then sequenced on a Megabace 1000 DNA Analysis System (Molecular Dynamics and Amersham Pharmacia) sequencer by using forward and reverse primers.

Sense and antisense digoxigenin-labeled RNA probes for UNC5C were prepared by in vitro transcription. In situ hybridization was performed as described (3).

The classical soft agar assay was adapted for transient transfection as described in ref. 24 and the Supporting Text. Biocoat Matrigel invasion chambers (Becton Dickinson) were used to assess the invasiveness of LS174T, U373MG, or Ras-transfected NIH 3T3 cells as described (25).

We first surveyed several human tumor types for loss of UNC5H expression. Based on the analysis of 241 tumors and their corresponding normal tissues from a dot blot array by using a full-length cDNA probe that recognizes UNC5A, UNC5B, and UNC5C mRNA, we observed that UNC5H expression was strongly down-regulated in the majority of the tested tumors (Fig. 1A) whereas such down-regulation was not detected with the ubiquitin control probe (see Fig. 8). Whereas no significant change of UNC5H expression was associated with thyroid or prostate tumors, a reduction of expression by a factor of 2 or more was observed in 88% of ovary tumors, 49% of breast tumors, 48% of uterus tumors, 93% of colorectal tumors, 68% of stomach tumors, 74% of lung tumors, and 81% of kidney tumors (Fig. 1B). Along the same line, we failed to detect UNC5H expression in eight of nine cancer cell lines (Fig. 2). These first set of experiments suggested that the global expression of the three UNC5H members was down-regulated. We next analyzed whether the three UNC5H genes were individually shut down with a specific emphasis for colorectal tumors in which the decrease of expression was the more drastic (i.e., >49% of cancer of the colon and 80% of cancer of the rectum showed at least a 5-fold decrease in UNC5H expression). Using dot blot arrays with specific probes for UNC5A, UNC5B, or UNC5C mRNA, we observed that UNC5A, UNC5B, and UNC5C expression was affected in 48%, 27%, and 77% of the tested colorectal tumors, respectively (not shown and Fig. 3A). Studying further UNC5C expression that appeared strongly reduced in colorectal tumors, we performed in situ hybridization in either normal colorectal tissue or well-differentiated adenocarcinoma. In normal tissue, the staining of UNC5C mRNA was restricted to surface and crypt epithelial cells and was observed mainly in a supranuclear position. Interestingly, consistent with the dot blot array analysis, the expression of UNC5H was reduced in neoplastic epithelial cells (Fig. 3B).

Searching for the mechanisms that may explain such down-regulation, we analyzed whether the three human UNC5H genes may be subjected to LOH. In a panel of 31 colorectal tumors versus corresponding normal tissue, we performed allelic loss analysis by PCR amplification of polymorphic microsatellite markers that neighbor UNC5A, UNC5B, or UNC5C locus. Fig. 4 shows that an important allelic loss was observed within the UNC5A, UNC5B, or UNC5C locus (Fig. 4 A and B). UNC5A and UNC5C genes, whose expression was more frequently reduced, were found to display 52% and 39% of LOH, respectively, whereas 26% of LOH occurred within the UNC5B locus.

We next checked whether epigenetic processes like DNA methylation may also account for the reduction/loss of UNC5H expression. The possible role of promoter methylation was then assessed in different colorectal cell lines. Fig. 4C shows that a treatment with methylation inhibitor 5-aza-2′-deoxycytidine (5aza2dC) (26) only led to a modest increase of UNC5H expression in HT-29 cells, hence suggesting that promoter methylation may slightly affect UNC5H expression. Putative CpG islets can be found in both UNC5A, UNC5B, or UNC5C promoter but the in vivo methylation of these islets has yet to be determined. Interestingly, if 5aza2dC had only a slight effect on UNC5H expression, treatment with trichostatin A, a specific inhibitor of the histone deacetylases that are known to interact with methylated CpGs and to participate to stable transcriptional repression (27), led to a robust increase of UNC5H expression (Fig. 4C). Taken together, these data suggest that histone deacetylase activity mediates part of the transcriptional repression of UNC5H genes even though ongoing work would have to determine whether this regulation is specific for tumors. Interestingly, UNC5B promoter has been recently presented as a direct target of p53 (20), the prototype tumor suppressor and transcription factor whose expression is regulated by histone deacetylase (28). In that sense, as also described by Tanikawa et al. (20), we observed that p53 overexpression led to an increased level of UNC5B (not shown and see Fig. 6A Inset). Thus, the loss of p53 activity in numerous tumors might also account for part of the loss/reduction of UNC5H expression. Taken together, these results suggest that UNC5H expression may be down-regulated in tumors by at least three mechanisms including allelic loss, promoter silencing, and the absence of p53-dependent transactivation.

The loss/reduction of UNC5H expression suggested that UNC5H genes might be putative tumor suppressors. Very often, such tumor suppressors are subjected to mutation in cancers. We then analyzed whether UNC5H genes were targets for mutations. We focused our study on UNC5C because it appears to be strongly affected in colorectal cancer. We then performed in a panel of 30 colorectal tumors PCR amplification and sequencing of the 13 accessible exons covering the complete coding sequence of UNC5C except from amino-acids 200–370 whose genomic region is not available. In 40% of samples analyzed, we failed to amplify at least one of the exons, hence strongly suggesting homozygous deletions in UNC5C gene in these tumors. However, we cannot completely exclude a repeated PCR amplification problem, even though this problem would not be related to (i) the quality of the genomic DNA (because other exons from the same genomic DNA were perfectly amplified) and (ii) the quality of the primers (because genomic DNA from other tumors were easily PCR-amplified with the same primers). More importantly, three different mutations were found in five tumors. All mutations were missense mutations. The first mutation converted a proline (amino acid 57) into a tyrosine. The second mutation converted an alanine (amino acid 628) into a glutamic acid. The third mutation, found in three different tumors, converted an alanine (amino acid 841) into a tyrosine. If the impact of the different mutations in UNC5C activity is to be evaluated, it is interesting to note that the A628E mutation is present in the ZU5 domain of UNC5C and that the frequent (i.e., 10% of the tumors tested) A841Y mutation occurs in the death domain of UNC5C, a domain that was shown to be crucial for UNC5H proapoptotic activity (17). The three mutations detected were not, however, observed only in the tumor but also in the adjacent normal tissue. This finding would argue that the mutations detected are probably not the targeted mechanism for inactivating UNC5H in cancer. However, these mutations may result in a loss/reduction of UNC5H activity that taken together with the decreased UNC5H expression described above can act as a selective advantage for tumor development.

The very frequent loss/reduction of UNC5H expression in the majority of the tested tumors suggest that the presence of UNC5H is a negative constraint for tumor development. We then analyzed whether UNC5H might actually inhibit in vitro the classical hallmarks of malignant transformation that are the anchorage-independent growth and the ability to invade (25). Colorectal LS174T, Ras-transformed NIH 3T3, or large T-transformed 293 cells were then transiently transfected with UNC5H-expressing constructs and allowed to grow in soft agar or invade through a reconstituted 3D basement membrane gel (Matrigel). Expression of one of the three UNC5H receptors in both cell lines (Fig. 5A and data not shown) was sufficient to drastically inhibit growth in soft agar and invasion in Matrigel (Fig. 5 and not shown). Interestingly, when netrin-1 was added, both growth in soft agar and invasion were restored (Fig. 5 and not shown). To further investigate the role of UNC5H/netrin-1 in the regulation of these malignant properties, we sought to determine whether endogenous UNC5H, as opposed to ectopically expressed UNC5H, might also regulate anchorage-independent growth and invasive ability. We took advantage of the recent observation that p53 transactivates the UNC5B gene (20). Thus, colorectal LS174T or glioblastoma U373MG cells were transfected with p53 and analyzed for growth in soft agar and invasion in Matrigel. As described, p53 expression led to enhanced UNC5B expression in these cell lines (ref. 20 and Fig. 6A Inset). In these conditions, we detected a reduction of growth in soft agar (Fig. 6A) and a reduction of invasion in Matrigel (Fig. 6B). Remarkably, netrin-1 addition was sufficient to block this reduction (Fig. 6), hence demonstrating that, in these experimental conditions, p53-induced inhibition of these hallmarks of cell transformation is mediated via UNC5B (Fig. 6). Moreover, these data support the fact that the endogenous expression of UNC5B, provoked by p53, is sufficient to inhibit in vitro cell transformation. Along the same line, when the p53 transactivity was inhibited in MCF7 cells by MDM2 overexpression, an increase of growth in soft agar was observed, such an increase being inhibited by UNC5H expression (Fig. 6). Taken together, these data support the role of the pair netrin-1/UNC5H as regulators of cell transformation.

We finally analyzed whether the regulation of apoptosis by the pair netrin-1/UNC5H may be involved in the inhibition of anchorage-independent growth and invasive ability. We observed that a treatment with the general and potent caspase inhibitor zVAD-fmk completely suppress both UNC5H proapoptotic activity (17) and the UNC5H-mediated inhibition of anchorage-independent growth or invasion (Fig. 7 and not shown). Moreover, we described that UNC5H-mediated apoptosis requires the presence of the UNC5H death domain (17). Interestingly a mutant form of UNC5H2 deleted of its death domain was not able to suppress the hallmarks of malignant transformation (Fig. 7). Moreover, expression of UNC5H2 induced a marked increase of trypan blue staining in nascent soft agar cell colonies (Fig. 7C), strongly suggesting that the transformation-suppressing properties of UNC5H are related to the proapoptotic activity of UNC5H.