Serological Cloning of a Melanocyte rab Guanosine 5′-Triphosphate-binding Protein and a Chromosome Condensation Protein from a Melanoma Complementary DNA Library¹

The antigenic profile of human melanoma has been the object of intense scrutiny, motivated mainly by the desire to find antigens that can serve as suitable targets for vaccine or antibody-based therapies(1, 2, 3, 4). Several general approaches to defining melanoma antigens have been established, and this has led to an impressive list of human melanoma antigens recognized by antibodies or T cells in humans or monoclonal antibodies generated in mice (3, 4). With regard to melanoma antigens recognized by humans, the approach of autologous typing provided the framework for analyzing the specificity of humoral (5) or cytotoxic T cell(6) reactivity against autologous melanoma cells. This led to the molecular characterization of CTL³-recognized peptides by the methodology developed by Boon and van der Bruggen (1) and antibody-recognized antigens according to SEREX methodology by Sahin et al. (7). The analysis of surface and intracellular antigens by mouse monoclonal antibodies represents another approach to defining the antigenic phenotype of melanoma and a wide range of melanoma differentiation antigens, including gp75 (8) and gp100 (9),have been identified this way.

The current list of restricted melanoma antigens recognized by these methods falls into three main categories. The first are melanocyte differentiation antigens, including tyrosinase (10), gp75(11), gp100 (12), and Melan-A/MART-1(13, 14). Of these, tyrosinase, gp75, and gp100 are melanosome-associated transmembrane proteins involved in the melanin synthesis pathway. These three proteins are structurally related, and gp75 and gp100 have been designated as tyrosinase-related proteins 1 and 2, respectively. In contrast, Melan-A is structurally unrelated,and its function is unknown. The second category comprises mutational antigens and a growing list of mutations in melanoma cells giving rise to CTL-recognized epitopes are being defined (15). The third category of melanoma antigens are the CT antigens(16), including MAGE, BAGE, GAGE, SSX, NY-ESO-1, CT7, and CT10 (1, 16). Like melanocyte differentiation antigens,these antigens are highly restricted differentiation antigens, but in the case of CT antigens they are normally expressed only in the germ cells of the testis. In cancer, CT antigens are aberrantly expressed in a wide range of different tumor types, including melanoma. The expression patterns of melanocyte differentiation antigens and CT antigens in melanoma are quite distinct. Melanocyte differentiation antigens are expressed at high frequency, with >80% of all melanomas expressing these antigens, although with substantial intratumor heterogeneity (9, 17). As expected, melanocyte differentiation antigens are generally found in melanomas with more differentiated phenotype and absent in less differentiated variants,such as the desmoplastic and spindle cell variants (18). In contrast, CT antigens are expressed at lower frequencies, in the range of 10–30% for individual CT antigens (15). In the case of MAGE antigens, there is more frequent expression in metastatic melanoma than in primary melanoma (19), suggesting a possible relation to tumor progression.

Regarding immune recognition of these antigens in humans, tyrosinase,gp75, gp100, and Melan-A/MART-1 have been shown to be recognized by CD8⁺ T cells (15). Tyrosinase also elicited a CD4⁺ T cell response (20) and an antibody response (7). In the CT antigen category, NY-ESO-1 is often recognized by both humoral and cellular immunity(21), whereas MAGE, BAGE, and GAGE, initially defined as CD8⁺ CTL antigens, rarely elicit spontaneous antibody responses (22). SSX, CT7, and CT10, identified more recently by SEREX, have not as yet been shown to be CTL targets. Recently, a series of MAGE-3 peptides recognized by CD4⁺ T cells has been defined (23).

SEREX analysis of melanoma by Sahin et al. (7)defined 10 antigens, including MAGE-1, SSX2, and tyrosinase. In subsequent SEREX studies, sera from melanoma patients were screened against a testicular cDNA library (24) and an allogeneic melanoma cell line library (25). The testicular library screening led to the definition of the SSX gene family, and the cell line library screening led to the identification of CT7(25). In the present study, we further extended our study and screened an autologous melanoma cell line library. Forty-three gene products were identified, including a rab GTP-binding protein preferentially expressed in melanocytes and a novel human chromosome condensation protein hCAP-G.

The melanoma cell line MZ19 as well as the autologous melanoma serum were derived from a 50-year old female patient at Krankenhaus Nordwest,Frankfurt, Germany. The cell line, derived from a metastatic lesion,was recognized by CD8⁺ T cells from the same patient. Other melanoma cell lines (SK-MEL series) and breast cancer cell lines were obtained from the repository maintained at the Ludwig Institute for Cancer Research, New York Branch at the Memorial Sloan-Kettering Cancer Center.

Total RNA was extracted from the MZ19 melanoma cell line by conventional CsCl-guanidine thiocyanate gradient method. A cDNA library was constructed in a λ-ZAP express vector, using a commercial cDNA library kit (Stratagene).

The unamplified cDNA expression library was screened with the autologous serum at 1:200 dilution. The screening procedure was as described previously (7, 26).

The reactive clones were purified and in vivo excised to pBK-CMV plasmid forms (Stratagene). Plasmid DNA was prepared and sequenced (DNA Sequencing Service, Cornell University, Ithaca, NY). DNA and amino acid sequences were analyzed against GenBank and EST databases using the BLAST program. Genes identical with entries in the GenBank were classified as known genes, whereas those that shared sequence identity only with ESTs and those that have no identity in both GenBank and EST databases were designated as unknown genes.

To evaluate the mRNA expression pattern of the cloned cDNA in normal and malignant tissues, gene-specific oligonucleotide primers were designed to amplify cDNA segments 300–600 bp long. RT-PCR was performed by using 30 amplifications at an annealing temperature of 60°C, and the products were analyzed by agarose gel electrophoresis.

Northern blot analysis was performed using commercial poly(A) (2μg/lane) Human Multiple Tissue Northern (MTN) Blot I and II(Clontech), or total RNA (20 μg/lane) derived from different melanoma cell lines. ³²P-labeled PCR probes 300–600 bp long were used, and the blots were hybridized and washed following the manufacturer’s protocol (ExpressHyb kit; Clontech).

A cDNA library of 7 × 10⁵primary clones was prepared from melanoma cell line MZ19 and screened with autologous serum at 1:200 dilution. Sixty-four immunoreactive clones were purified and DNA sequenced. Sequence comparison against GenBank and EST databases showed that these 64 clones were derived from 43 distinct genes, designated NY-MEL-1 through NY-MEL-43. Among these, 29 genes showed sequences identical with or highly homologous to known GenBank entries. These genes are referred to as “known genes” and are summarized in Table 1. The remaining 14 genes showed no homologous sequences in the GenBank but shared sequences with ESTs derived from various tissues. These 14 genes, referred to as “unknown genes,” are summarized in Table 2.

Of 29 previously known genes, 22 genes (NY-MEL-8 to NY-MEL-29) are broadly expressed in human tissues, based on the finding that cDNAs derived from various tissues could be found in the GenBank and EST databanks. Similarly, 10 of 14 unknown genes have ESTs isolated from a number of normal tissues, presumably also reflecting ubiquitous mRNA expression in adult tissues. Such universal expression, however, was not immediately evident for seven known genes and four unknown genes, and the expression of these genes were evaluated by RT-PCR analysis of total RNA from five representative normal tissues (brain, colon, kidney, testis, and liver), using the isolated clone as the positive control. Results showed universal expression at similar levels with four of four unknown genes(NY-MEL-30 to NY-MEL-33) and five of seven known genes (NY-MEL-2, and NY-MEL-4 to NY-MEL-7; see Table 1). With two genes, NY-MEL-1and NY-MEL-3, a pattern of differential tissue expression was seen. NY-MEL-1 (homologous to rat rab-related GTP-binding protein) showed a strong RT-PCR signal in MZ19 cells,weaker signals in testis and kidney, and was negative in brain, colon,and liver. NY-MEL-3 (homologous to Xenopuschromosome-associated peptide G) showed strong RT-PCR signals in MZ19 and testis, with positive but weaker signals seen in brain, colon,liver, and kidney (data not shown).

mRNA expression of NY-MEL-1 was evaluated in a larger panel of normal tissues by Northern blot analysis. Northern blotting with 2μg of poly(A) RNA and a 5′ NY-MEL-1 probe showed no visible signal in any of the normal tissues tested, including heart, brain,placenta, lung, liver, skeletal muscle, kidney, pancreas, spleen,thymus, prostate, testis, ovary, small intestine, colon, and peripheral blood leukocyte (Fig. 1,a). Control hybridization with actin probe confirmed the quality and quantity of the poly(A) RNA (Fig. 1,b). Expression in cells of melanocytic lineage was then examined, using total RNA from cultured melanocytes and eight melanoma cell lines (20μg each, equivalent to <1 μg of poly(A) RNA). Results showed a single strong mRNA species migrating at ∼1.6 kb in cultured melanocytes and variable but weaker signals in melanoma cell lines SK-MEL-12, -14, -26, -28, and -37. No detectable NY-MEL-1 mRNA was found in SK-MEL-10, -24, or -MZ19 after overnight exposure (Fig. 1 c). A weak signal was visualized in MZ19 after 1 week of exposure, indicating a lower level expression then in SK-MEL-10 and SK-MEL-24. This predominant expression in melanocyte and melanoma cells but not in other normal tissues indicates that NY-MEL-1 belongs to the category of melanocyte differentiation antigens.

To analyze NY-MEL-1 mRNA expression in a larger panel of melanoma and other tumor samples, RT-PCR assay was performed and compared with the Northern blot data. Of the eight melanoma cell lines tested, seven were positive by RT-PCR, including SK-MEL-12, -14, -26, -28, and -37(Northern blot positive), SK-MEL-10 (Northern blot negative), and SK-MEL-36 (not tested by Northern blot; Fig. 2 a). Two cultured melanocyte preparations were both strongly positive by RT-PCR. SK-MEL-24, the only melanoma line negative by RT-PCR, was also negative by Northern blot. Six breast cancer cell lines, 734B, MDA-MB-444, HBL 100, BT20, ZR71-1, and MDA-MB-231, were also RT-PCR negative. In accordance with the greater sensitivity of RT-PCR, several Northern blot-negative tissues showed weak to moderate RT-PCR signals, including testis, kidney, uterus, prostate, and pancreas. Adrenal gland, a tissue source not included in our Northern blot panel, showed a strong RT-PCR signal comparable with that of cultured melanocytes. This product was confirmed to be NY-MEL-1 by nested PCR with internal primers and by sequencing of the RT-PCR product.

A panel of tumor tissues was then examined by RT-PCR. Of 19 melanoma tumor specimens, 17 (90%) were positive, with moderate to strong signals seen in 12 (63%) cases. Nonmelanocytic tumors tested included 4 colon cancer, 4 lung cancer, 4 breast cancer, and 4 squamous carcinomas. All 16 specimens were negative, except for 1 lung cancer(lung ca 4; Fig. 2).

DNA sequencing of NY-MEL-1, represented by cDNA clone MZ19–32a, revealed a full-length cDNA clone of 1407 bp, including 47 bp of 5′-untranslated region, 724 bp of 3′-untranslated region, and an open reading frame of 636 bp, encoding a polypeptide of 211 amino acids, a predicted molecular mass of 23.714 kDa (Fig. 3, GenBank number AF235022). Protein motif analysis showed that the five highly conserved GTP-binding domains that are typical for the ras superfamily (ras, rab, rho, ran, arf) of small GTP-binding proteins(26) are present in this sequence, as highlighted in Fig. 3.

Sequence comparison with the GenBank database confirmed that this gene is highly homologous to the rat rab-related GTP-binding protein,sharing 81.5% nucleotide identity and 97% amino acid identity. Rab GTP-binding proteins have been shown to play a crucial role in the intracellular processes of membrane trafficking and vesicular fusion and targeting (28). More than 40 rab GTP-binding proteins have been isolated from mammalian cells (28), at least 28 from human cells. Among human rabs, NY-MEL-1 is closest to rab32 at both DNA and protein levels (GenBank NM006834),sharing 75% amino acid sequence identity, 88% homology including conservative changes. On the basis of these findings, NY-MEL-1 was designated as rab38 (International System for Gene Nomenclature gene symbol: RAB38). Fig. 4 shows the amino acid sequence comparison of rab38, its rat homologue gene, and the human rab32.

Despite its overall strong homology to rab proteins, rab38 is unique in its COOH terminus. Rab proteins typically contain two cysteines in the carboxyl end, with a motif of -CC, -CXC, or-CCXX. This allows posttranslational lipid modification,most likely in the form of geranylgeranyl prenylation of both cysteine residues (29). Such hydrophobic modification enables the rab protein to interact with lipid membranes and is functionally crucial. Rab38, however, ends with -CSGCAKS, a sequence closer to ras than to rab (30). This primary structure would dictate a different pattern of lipid modification, likely farnesyl prenylation at the carboxycysteine residue and a palmitoylation at the upstream cysteine, as has been shown in ras protein (30).

In addition to NY-MEL-1, the only other SEREX-defined gene in this study that showed a differential expression pattern in normal tissues is NY-MEL-3, represented by clone MZ19–59b. This clone was completely sequenced, revealing a full-length cDNA of 3198 bp, including 68 bp of 5′-untranslated region, 82 bp of 3′-untranslated region, and an open reading frame of 3048 bp, encoding a polypeptide of 1015 amino acids, predicted molecular mass 114.3 kDa (submitted to GenBank, accession number AF235023).

A sequence homology search showed NY-MEL-3 to be the human counterpart of Xenopus laevis chromosome-associated polypeptide group G (XCAP-G; GenBank accession number AF111423), a chromosome condensation protein which is part of the 13S condensin complex formed during the early stage of mitosis. NY-MEL-3shares 28% nucleotide identity and 59% amino acid identity with XCAP-G, and we have therefore designated this gene hCAP-G.

Expression of hCAP-G in normal and tumor cells was evaluated by RT-PCR and by Northern blot analysis. Northern blotting of normal tissue RNAs showed the strongest expression in testis, a weak signal in thymus, and no detectable signal in other normal tissues (Fig. 5). Melanoma cell line RNAs (20 μg of total RNA) showed variable signals of weak to moderate intensity. RT-PCR was performed for normal tissues, melanoma cell lines, and melanoma tumor samples. All RNA tested were positive, with highly variable levels of RT-PCR products in these tissues and cell lines. The highest level of expression was seen in testis, as also observed in Northern blots.

To compare the genes isolated in this NY-MEL series with those isolated from other SEREX studies, the sequences of NY-MEL-1 through NY-MEL-43 were analyzed against entries in the SEREX databank (http://www.licr.org/SEREX.html). Results showed that 8 of the 43 genes have previously been identified in the SEREX analysis of a range of tumor types, including renal cancer, colon cancer, gastric cancer, etc. (Table 3). NY-MEL-14 shares 92% homology to a HERV-K long terminal repeat (GenBank Z21852) derived from an endogenous retroviral present in multiple copies in the human genome (31). HERV-K has been identified in a previous SEREX screening of renal cancer by Sahin et al. (32).

Rab proteins, the largest branch of the ras superfamily of GTP-binding proteins, were first identified in yeast in 1983(33). Search for the homologous genes in mammalian cells initially identified its counterpart in rat brain, hence the name rab(ras-like in rat brain). More than 40 rab genes have since been identified in mammals, including at least 28 human genes found in the public databanks. These genes share structural features with ras and other GTPases in having the 5 highly conserved regions necessary for GTP binding and hydrolysis. In addition, all rab proteins typically contain 2 cysteine residues in the carboxyl end,usually in the format of -CXC, -CC, or -CCXY. Isoprenylation of both cysteine residues by geranylgeranyltransferase renders the rab proteins hydrophobic, providing them the ability of reversible membrane association (34). Given this capacity,Rab protein functions by cycling between a GDP-bound cytosolic form and a GTP-bound membrane form, and various rab proteins have been shown to play a crucial role in the docking/fusion of transport vesicles or organelles with their target/acceptor membranes. Subcellular localization studies have indeed shown that with the exception of lysosomes, all of the organelles involved in biosynthetic/secretory and endocytic pathways contain at least one rab protein on their cytoplasmic surface (28).

Although most of the rab proteins identified are ubiquitously expressed in many tissues, some of them have been found to be cell type- or tissue-specific. For example, rab17 is detected only in epithelial cells (35), rab3a is expressed in neurons and neuroendocrine cells (36), and rab3d is mainly in adipocytes (36). It is likely that the cell-specific expression of these rab proteins is critical for cells to perform unique functions related to vesicle targeting and docking. In this regard, melanocytes might be suspected to have one or more unique rabs,because they contain a highly cell-specific organelle, the melanosome. Melanosomes are organelles that contain tyrosinase and other melanin synthesis-related proteins. As melanocytes mature, the melanosomes gradually move from the cytoplasm of the melanocytes toward the dendritic processes in the periphery. Morphological and biochemical changes occur during this process, defining melanosomal developmental stages I to IV. Mature melanosomes can then be transferred from the melanocytes to the adjacent keratinocytes as a result of active phagocytosis of the melanocytic dendrites by the keratinocytes(37). Considering this highly polarized movement of melanosomes and the eventual fusion and transfer to keratinocytes, it seems likely that unique rab proteins are involved in this process. This idea led to the isolation of an array of 17 rabs from melanocytes by a PCR-based strategy (38). Sequence analysis revealed a few new rab genes, including rab22b and rab30. However, none of the genes isolated in that study showed a melanocyte-specific expression pattern.

By screening the MZ19 melanoma cell line cDNA expression library with autologous antibody, we have identified NY-MEL-1 in the present study, encoding a new member of the rab family protein. The closest human rab gene to NY-MEL-1is rab32 which shows strong nucleotide and amino acid homology within and beyond the GTP-binding domains. For this reason, we designated the NY-MEL-1 gene rab38 following the recommendation of the International System for Gene Nomenclature.

At the RNA level, rab38 is preferentially expressed in melanocytes and in their malignant counterpart, melanomas. When compared with other human rab gene sequences, rab38 has a unique COOH terminus, -CSGCAKS, different from the -CC, -CXC, or -CCXX (X being any amino acid) motifs seen in other rab proteins (30). This sequence, in contrast, is similar to the -CAAX (A being an aliphatic residue, and X being M, S, Q, C, or A) motif seen in the ras subfamily proteins, which also contain an upstream cysteine residue as does rab38. It is known that posttranslational lipid modifications occur at both cysteine residues at the carboxy ends of the ras and rab proteins, allowing their hydrophobic interactions with the lipid membranes. However, the two cysteines in the rab proteins are geranylgeranylated, whereas the two cysteines in the ras protein are farnesylated (carboxycysteine) and palmitoylated (upstream cysteine),respectively. Whether these differences in lipid modification lead to specific interaction of rab38 with proteins related to melanosome trafficking is an idea worth exploring. Subcellular localization studies using anti-rab38 antibodies would be important to determine whether rab38 is indeed melanosome-associated. Two of our findings,however, suggest that rab38 may have a broader function besides its role in melanocytes. One is that the rat homologue gene has been isolated from lung alveolar cells, and the other is that rab38 mRNA was detected at low levels in other normal tissues, and at substantial levels in the adrenal gland. Whether this indicates a leaky expression of rab38 in many cell types or specific expression by a small subset of cells in these organs remains to be investigated.

Another novel human gene that we isolated by SEREX in the present study is the human counterpart of the Xenopuschromosome-associated polypeptide G (XCAP-G) gene, designated hCAP-G. This gene belongs to a family of SMC (structural maintenance of chromosomes) genes, which have been shown to be required for proper condensation and segregation of mitotic chromosomes(39, 40). By analyzing protein extracts of Xenopus oocytes, it has been shown that the chromosome condensation complex that forms during mitosis consists of two main fractions, the 13S condensin and the 8S condensin. The 13S condensin contains five main proteins, the XCAP-C, XCAP-E, XCAP-D2, XCAP-G, and XCAP-H. Together, these five proteins constitute the most abundant protein components besides histones and are considered central players in the mitotic protein assembly (41). By RT-PCR and Northern blot analysis of different normal and tumor tissues, we found that hCAP-G mRNA is highly expressed in normal testis, weaker (but detectable by Northern blot) in normal thymus, and at variable levels in different tumor specimens and tumor cell lines. Because hCAP-G presumably functions in mitosis and is cell cycle related, it might be expected that its expression shows a relationship with cell replication rate: highest in testis because of spermatogenesis; high in thymus as a lymphoid organ with constant cell turnover; and variable in tumor cells. Other genes involved in mitosis have been shown to be proliferation markers with prognostic significance in the evaluation of human malignancy. For example, high level expression of mitosin, a nuclear phosphoprotein involved in cell division (42), has been shown to be an independently significant predictor of recurrence in breast cancer without lymph node metastasis (“node-negative”breast cancer) (43). Another proliferative marker, Ki-67,has also been documented as clinically useful (44). In this regard, it would be important to compare the tumor expression of hCAP-G to the expression of these existing markers by immunohistochemical analysis, and antibodies to hCAP-G are being prepared.

The cloning of two biologically important genes, rab38 and hCAP-G, by SEREX further attests to the power of this approach to identify novel human genes and gene products with potential relevance to cancer. SEREX-defined antigens can be categorized into several groups, including differentiation antigens, mutational antigens, amplified/overexpressed antigens, viral antigens, CT antigens, and splice-variant antigens (16). The immunogenicity of these antigens have been ascribed to tumor-specific expression (e.g., mutational antigens), restricted expression (e.g., differentiation antigens), or altered expression (e.g., overexpressed antigens). In this context,rab38 is clearly a new differentiation antigen, mainly expressed in the melanocyte lineage, and hCAP-G would represent an overexpressed antigen in cancer.

To date, >1000 antigens have been isolated by SEREX analysis. The challenge of SEREX analysis is to define the complete repertoire of cancer gene products that elicit an immune response in humans (cancer immunome) and to distinguish those antigens that have a direct relevance to cancer etiology or cancer progression from those that represent general autoimmunogenic cellular components. Of the 43 genes isolated in the current study, 8 had been identified in previous SEREX analysis. This ∼20% overlapping in sequences shows the progress that has been made in defining the cancer immunome, but also the work left to be done in completing the task.