The Chromosome 10 Monosomy Common in Human Melanomas Results from the Loss of Two Separate Tumor Suppressor Loci¹ Free

Loss of genetic material from cancer cells occurs more frequently than chromosomal amplifications, suggesting a prominent role for the inactivation of tumor suppressor genes during the etiology of solid tumors (1). Although several mechanisms to accomplish this loss have been described (2), the end results of this process can be the deletion of several intact chromosomes and/or subchromosomal regions during tumor development. For example, loss of chromosomes X, Y, 4, 10, 13–15, 18, or 22 and/or chromosome segments 1p22-pter, 3p13-pter, 6q14-qter, 8p, 9p, 11p, and 17p occur repeatedly in several prominent cancers, including carcinomas of the breast, brain, kidney, colon, ovary, lung, and skin (1). A number of these deleted chromosomes or subchromosomal regions that have been identified in tumors have facilitated the demonstration that they contain functionally inactivated tumor suppressor genes; for example, INK4A/p16 in 9p21 (3, 4), RB1 in 13q14 (5), and PTEN³ in 10q23 (6, 7).

In tumors where the recurring genetic alteration entails loss of an entire chromosome (monosomy) or loss followed by chromosomal duplication (uniparental disomy), it is possible that multiple suppressor genes on the same chromosome must be inactivated for a complete effect on tumorigenesis. In these situations, it is possible that tumor suppressor genes on these chromosomes play a more prominent role than oncogenes. Unfortunately, localization of these suppressors by cytogenetic analysis or molecular studies searching for loss have been limited largely due to the small proportion of tumor samples that exhibit defining segmental deletions. In these situations, functional complementation by introducing normal chromosomes and subchromosomal fragments into cancer cells can be used to delineate individual chromosomal regions as the sites of tumor suppressor genes and to functionally demonstrate their additive effects (8, 9, 10).

Loss of an entire copy of chromosome 10 is common in tumors of the brain, lung, ovary, and skin (1). In nonfamilial melanomas, chromosome 10 monosomy occurs in 30–60% of both early- and advanced-stage tumors (11, 12) and is an indicator of a poor clinical prognosis (13). Although segmental deletions are rare, they have been useful in defining one candidate growth-suppressing region on the long arm of chromosome 10; initially cytogenetically localized to 10q (14, 15) and subsequently narrowly defined by LOH to 10q22-qter (16, 17, 18, 19). This region is also a common target for LOH in a wide spectrum of sporadic cancers, including neoplasias of the prostate (20), thyroid (21), kidney (22), endometrium (23), and bone (24). The most noticeable tumor suppressor gene in this region is the dual-specificity phosphatase, PTEN, which is deleted or mutated in a wide variety of advanced tumors, including melanomas (6, 7, 25, 26, 27). We have recently developed a functional approach, termed IVLOH, in which a normal copy of chromosome 10 was transferred into melanoma cells, which then eliminated the responsible growth suppressor gene(s) in culture through chromosomal fragmentation accompanied by loss of a region extending distally from 10q23.1 (28). The region selected for loss had a breakpoint at 10q23, suggesting that PTEN was one candidate melanoma growth suppressor. Subsequent ectopic gene expression studies in this melanoma cell line that lacked PTEN protein confirmed this candidacy (28).

Inactivation of suppressor genes on chromosome 10p has been suggested by cytogenetic analysis, molecular evidence, and functional complementation studies in both prostate cancers (20, 29, 30) and glioblastomas (31, 32). However, no extensive analysis for 10p loss in melanomas has been reported. One comprehensive cytogenetic study found a small minority of melanomas that lost either the short or long arm, or a segment of each, suggesting the possibility of suppressor loci on both arms (33). Collectively, these studies could be interpreted as the tumors’ using chromosomal mechanisms to eliminate tumor suppressor genes located on both arms of chromosome 10.

Here, we use LOH analysis and functional complementation to localize a melanoma suppressor locus on 10p15.3, centered at D10S249. We also show that a second locus on 10q distal to 10q23.1, containing the PTEN tumor suppressor, also reduces the ability of melanoma cells to form tumors. Finally, we show that the two loci are synergistic in their ability to suppress tumor growth in vivo. These results, therefore, provide an etiological basis for the chromosome 10 monosomy observed during the genesis of sporadic melanomas.

The procedure for DNA extraction from tumor samples and matched normal lymphocytes has been described (18, 34). Fifty-six matched tumor and normal DNA samples were subjected to PCR-based STS genotyping searching for loss of heterozygosity in tumor material. Reaction conditions for each STS marker were as described in the Genome Database⁴ or from a reference cited therein. Criteria required for LOH have been described previously (18, 34).

The melanoma cell line UACC 903, its growth conditions, and the microcell-mediated chromosome transfer technique have been previously reported (35, 36). The rodent donor cell lines 10E-6 and 10ER-4.S2, containing transferable and selectable fragments of human chromosome 10, were created and characterized before transfer, as described previously (37).

The cytogenetic characterization of the 903(10n) microcell hybrids has been reported (28). STS-based markers spaced at ∼10 megabase intervals along the length of chromosome 10 were used to determine the amount of the transferred chromosome present in each of the hybrid cell lines at different periods during cell culture. Fluorescent primers for microsatellite analysis were purchased from Research Genetics (Huntsville, AL) or synthesized from sequence data available in published studies or through the Genome Database. STS-PCR products were separated on an Applied Biosystems automated sequencer. Fluorescence in situ hybridization, with a probe specific for chromosome 10 ∝-satellite sequences (D10Z1), was used to determine the presence of the chromosome 10 centromere. The D10Z1 probe was purchased from Vysis (Downers Grove, IL) and used according to the recommended protocol.

Tumor formation was measured in athymic BALB/c (nu/nu) mice purchased from Simonson Laboratories (Gilroy, CA). Five million cells in 0.2 ml of DMEM containing 5% fetal bovine serum were injected s.c. above both the left and right rib cages of female mice, 4–6 weeks of age. A minimum of three mice in two independent experiments were measured at 7-day intervals for 1–2 months. The dimensions of the developing tumors were determined using calipers, and the sizes were estimated in cubic millimeters. The average tumor size of both experiments is reported together with the SE for each.

In earlier studies, we had used markers specific to 10q to define a region of loss on the long arm extending from 10q22 to qter (18). A comprehensive cytogenetic report has also documented 10q loss and found a minority of tumors (13%) that only lost material on 10p, with one tumor only losing 10p15 (33). Therefore, we sought to determine whether smaller defined regions of LOH could be identified on 10p, specifically at 10p15.

We detected LOH of chromosome 10 in 30% (17 of 56) of independent melanoma tumor samples, and representative examples of the LOH analysis for marker D10S249 on 10p15.3 are shown in Fig. 1,A for patient samples 006 and 116. The 17 tumor samples exhibiting 10p LOH and the results for each marker are listed in Fig. 1 B. Of the 17 tumors exhibiting 10p loss, 6 (35%) were primary tumors and 11 (65%) were metastatic lesions. Two regions of LOH were detected on 10p: the first occurring in 24% (9 of 38) of informative patient samples between markers D10S526 and D10S558 and centered at D10S249 in 10p15.3; and the second, which is lost in 30% (10 of 33) of informative tumors and is centered on marker D10S199 at 10p12. These results suggested two regions on 10p that may contain melanoma tumor suppressor genes.

To correlate tumor suppression with the regions of loss we had identified on 10p and 10q (between 10q22 to qter), we measured the tumorigenicity of cells containing varying regions of the transferred chromosome 10. Initially, we compared the tumorigenicity of cells containing the intact transferred chromosome 10 with that of cells retaining only the short arm and a portion of 10q proximal to 10q23.1. Detailed genotypic analyses of these hybrids with markers spaced at ∼10 megabase intervals is shown in Fig. 2,A. Hybrid cell lines 903(10n)29 and 903(10n)36 at passages 4 and 6, respectively, retain the entire transferred chromosome, whereas later passages 16 and 12 only retain markers from 10pter to 10q22.3 (D10S201) due to IVLOH (28). The rates of in vivo tumor growth elicited by these cell lines at both early and late passages are shown in Fig. 3. Cells retaining the intact transferred chromosome showed delays of ∼3 weeks before forming tumors similar in size to the parental cell line. Once cells containing an intact chromosome 10 had formed tumors in nude mice, they were also found to have lost the region distal to 10q23.1 (data not shown). Continued retention of the short arm and region of 10q proximal to 10q23.1 is likely due to the outgrowth of a small undetectable population of cells that had lost this region in culture before injection into nude mice.

Cells that had segregated the region distal to 10q23.1 partially regained their tumorigenic potential, but never fully reverted to the tumorigenicity of the parental cell line. This includes hybrids 903(10n)24, 903(10n)29, 903(10n)36, and 903(10n)37. The average tumor sizes at days 14 and 21 are listed in Table 1. All hybrid cells lacking the region distal to 10q23.1 required at least one additional week to form tumors similar in size to that of the parental cell line. These observations suggest that a second less potent suppressor(s) is present on the retained region of the transferred chromosome, extending from the p-terminus to 10q21 (D10S1422).

Subclones 37.7 and 37.21, which retained only 10p, were isolated from the cell line 903(10n)37 and used to localize suppressor function to 10p. Genotypic analysis of subclones 37.7 and 37.21 (Fig. 2,A) showed that these hybrids retained the centromere and entire short arm of the transferred chromosome. Both formed tumors at rates similar to cells that had lost all material distal to 10q21, thereby localizing the second tumor suppressor region to the short arm of chromosome 10 (Table 1). To address the possibility that variability in clonal tumorigenicity resulted in this behavior, we also examined the tumorigenicity of three randomly selected clones derived from the parental UACC 903 cell line. These clones, designated as 903A, 903B, and 903D in Table 1, each formed tumors at rates significantly faster than cells containing either the 10p or 10q tumor suppressors. Thus, the LOH and functional mapping both converge on the conclusion that the tumor suppressors reside in 10q, between 10q23.1 to qter and on 10p.

To define the region of 10p containing the second melanoma suppressor locus, we used microcell-mediated chromosome transfer to introduce two separate fragments of 10p15, from the rodent donor cell lines 10E-6 and 10ER-4.S2 into the human melanoma cell line UACC 903. Fig. 2,B shows the human chromosome 10 content in each of these donor cell lines before transfer. Donor cell line 10E-6 contains the tip of chromosome 10 from pter to D10S2325, whereas donor line 10ER-4.S2 contains a smaller region from D102477 to D10S591. After fragment transfer, resultant hybrid cell lines were genotyped to confirm the presence of the transferred fragment (see Fig. 2,B). Three hybrids contained the fragment found in the 10E-6 donor, and one contained the fragment found in the 10ER-4.S2 donor. The size of the tumors elicited by the hybrid cell lines containing the 10E-6 fragment are shown in Table 1 at days 14 and 21. The 10E-6 fragment reduced the rate of tumor formation to that observed with the intact short arm; therefore, the 10p suppressor is localized to a region from 10pter to D10S2325 (Fig. 2 B). The 10ER-4.S2 fragment did not reduce tumorigenicity (data not shown), thus, the region in common with fragment 10E-6, between D10S2477 to D10S591, can tentatively be excluded as the suppressor locus; however, this observation is based on only a single microcell hybrid. Correlating the LOH and functional mapping, we can define a common region between 10pter and D10S591, centered at D10S249, as the site of the 10p melanoma tumor suppressor.

When loss of an entire chromosome is a common and frequent event in carcinogenesis, the localization of causative genes becomes extremely difficult because mapping is forced to rely on only those few tumors with partial chromosomal deletions. This is further complicated because suppressors having weaker effects could easily be masked by stronger ones making their localization impossible until the dominant suppressor is eliminated by deletion or mutation (38). Under these circumstances, it is possible that tumor suppressor genes on these chromosomes could play a more prominent role than oncogenes. Chromosome 10 is an example of a chromosome commonly lost in multiple tumor types, including melanomas; therefore, we have used this as a model system to demonstrate the value of functional complementation mapping to provide a plausible explanation for monosomy observed during tumorigenesis. We localize separate suppressor regions on chromosome 10, assign varying tumorigenic phenotypes to particular suppressor regions, and demonstrate the existence of multiple functional melanoma suppressor genes on chromosome 10. This latter point provides a rational explanation for the selection of monosomic chromosomes during the development of melanoma and possibly other tumor types.

The genetic content of the cell line to be used for functional complementation is critical to the success of functional mapping. The melanoma cell line UACC 903 is ideal for chromosome 10 complementation because during its evolution it has lost one copy of chromosome 10 and duplicated the remaining one resulting in uniparental disomy (28). With two identical chromosomal copies, any mutations or deletions would be duplicated and fit the two-hit-model for tumor suppressor genes (2, 39). Therefore, the reintroduction of a wild-type chromosome 10 should restore suppressor gene activity, disrupting the homeostasis of the cell line and cause either an in vitro or in vivo cellular response to restore the equilibrium. In fact, these cells do fragment the transferred normal chromosome in culture to eliminate trans-acting suppressor elements in a region distal to 10q23.1 (28). In addition to an in vitro effect, this region also dramatically reduced the ability of these cells to form tumors in nude mice. Tumors that developed at a slowed rate also lost the region distal to 10q23.1. Retention of 10p and the region of 10q proximal to 10q23.1 in these tumors is likely due to the outgrowth of a small undetectable population of cells that had lost this region in culture before injection in nude mice. Therefore, these observations define the 10q23.1 to q-terminus as the site of a potent tumor suppressor that delayed tumor formation by 3 weeks.

A prominent tumor suppressor found in this functionally defined region of 10q is the PTEN phosphatase gene at 10q23.3. Although UACC 903 cells transcribe a mutant PTEN mRNA, they do not produce functional PTEN protein because both copies contain an identical T to G truncating point mutation at codon 76 (28). The reintroduction of a wild-type chromosome 10 does restore functional PTEN in these cells, as well as any other yet unidentified mutated or deleted melanoma suppressor genes that occur in this region. PTEN can suppress the ability of glioblastoma cell lines to form tumors in nude mice (40, 41) and similar results are observed in melanoma cells after the retroviral-mediated introduction of PTEN into the UACC 903 cell line.⁵ Missense and truncating mutations, as well as intragenic deletions affecting the expression of PTEN, have been reported for primary melanoma tumors and cell lines (25, 26, 27). These observations make PTEN a likely melanoma tumor suppressor involved in the development of possibly 30–40% of sporadic human malignant melanomas. It is likely that other suppressors may reside in this region, and their locations could be delineated functionally using subchromosomal fragments to map areas of suppression (10). This would be important to the melanoma patient because loss of 10q is associated with a poor clinical outcome (13).

In addition to the suppressor locus on 10q, we have found areas of loss on 10p that are likely sites of tumor suppressor gene loss. Using subchromosomal complementation, we demonstrated that one of these regions at 10p15.3, surrounding marker D10S249, can delay tumor formation by at least 1 week. Although this suppressor has a less dramatic effect on melanoma cells than the 10q site, it is clearly a second functionally distinct melanoma tumor suppressor locus. This area has also been implicated as a site of suppressor gene loss in both prostate cancers (20, 29, 30) and glioblastomas (31, 32).

Because functional melanoma tumor suppressor loci are present on both arms of chromosome 10, it seems that loss of the entire chromosome is favored over multiple independent mutations or deletions. Thus, this study provides both biological and molecular evidence supporting the presence of multiple melanoma tumor suppressors on chromosome 10 and provides a plausible functional explanation for the observed loss of the entire chromosome 10 during melanoma tumorigenesis.

We thank Hal Hoffman and John Weger for assistance with genotyping of cell line samples and Edith K. Podewski and Susanne Mommert for assistance with LOH screening of tumors.