Purpose: Members of the SPAN-X (sperm protein associated with the nucleus mapped to the X chromosome) family of cancer-testis antigens are promising targets for tumor immunotherapy because they are normally expressed exclusively during spermiogenesis on the adluminal side of the blood-testis barrier, an immune privileged compartment.
Experimental Design and Results: This study analyzed the human SPANX genomic organization, as well as SPAN-X mRNA and protein expression in somatic and cancer cells. The SPANX family consists of five genes, one of which is duplicated, all located in a gene cluster at Xq27.1. From the centromere, the arrangement of the five SPANX genes mapped on one contiguous sequence is SPANXB, -C, -A1, -A2, and -D. Reverse transcription-PCR analyses demonstrated expression of SPAN-X mRNA in melanoma and ovarian cell lines, and virtual Northern analysis established SPANX gene expression in numerous cancer cell lines. Immunoblot analysis using polyclonal antisera raised against recombinant SPAN-X confirmed the translation of SPAN-X proteins in melanoma and ovarian tumor cell lines. The immunoreactive proteins migrated between Mr 15,000 and Mr 20,000 similar to those observed in spermatozoa. Immunoperoxidase labeling of melanoma cells and tissue sections demonstrated SPAN-X protein localization in the nucleus, cytoplasm, or both. Ultrastructurally, in melanoma cells with nuclear SPAN-X, the protein was associated with the nuclear envelope, a localization similar to that observed in human spermatids and spermatozoa. Significantly, the incidence of SPAN-X-positive immunostaining was greatest in the more aggressive skin tumors, particularly in distant, nonlymphatic metastatic melanomas.
Conclusions: The data herein suggest that the SPAN-X protein may be a useful target in cancer immunotherapy.
Molecular characterization of human tumor antigens recognized by the cellular or humoral immune systems of cancer patients has generated new prospects for the development of cancer vaccines (1, 2, 3) . The list of human cancer antigens defined to date falls into three major categories: (a) differentiation antigens; (b) mutational antigens; and (c) viral antigens. Cancer-testis (CT) antigens are a distinct class of differentiation antigens that are viewed as attractive candidates for cancer vaccines, because the expression of CT antigens in normal tissues seems to be restricted to the testis. Malignant transformation is associated with activation or derepression of these normally silent genes resulting in CT antigen gene expression in a variable portion of many human tumors. No tumor specimen or cell line screened to date has been found to express all of the defined CT antigens (1, 2, 3) .
CT antigens represent good targets for immunotherapy, because they are widely distributed in a number of tumors and are present primarily in normal testes. The testis is considered an immune privileged site due to the presence of the blood-testis barrier, which inhibits contact between differentiating germ-line and immune cells, and to the apparent lack of HLA class I expression on the surface of germ cells (4) . These considerations suggest that only cancerous cells will be targeted by CTLs during CT antigen immunotherapy. Unfortunately, many of the CT antigens studied to date are first expressed in spermatogonia or spermatocytes located in the basal compartment of the seminiferous epithelium, the nonprotected side of the blood-testis barrier. Thus, targeting of these antigens may reduce or eliminate spermatogenesis and compromise the subsequent fertility of the patient. In contrast, the identification and validation of CT antigens, which lie on the adluminal (protected) side of the blood-testis barrier, may reduce the chances of inducing autoimmune orchitis (5 , 6) and permanent injury to the testes as a result of cancer immunotherapy. Furthermore, polyvalent CT antigen vaccines comprised of a mixture of immunogens represent an important strategy for obtaining adequate therapeutic responses given the variable incidence of CT antigen expression in tumors. For these reasons, it is important to identify and characterize new CT antigens, particularly those that are expressed postmeiotically and are immunogenic in humans.
We have characterized a testis-specific protein designated SPAN-X (sperm protein associated with the nucleus on the X chromosome; Refs.7, 8, 9 ). Ultrastructurally, the proteins are associated with the nuclear envelope of ejaculated spermatozoa. In transfected mammalian CV1 cells the SPAN-Xa protein localized to the nucleus, whereas SPAN-Xb localized to the cytoplasm. SPAN-X mRNAs were expressed exclusively after meiosis in the testis in round and elongating spermatids (7) . The human SPANX gene was mapped to chromosome Xq27.1. The expression of SPAN-X, an X-linked gene product, exclusively in haploid spermatids leads to interesting questions regarding the transcription of sex-linked genes in a precise temporal and spatial sequence during spermiogenesis and transcriptional regulation of this highly regulated gene during tumorigenesis.
SPAN-Xc, also designated CTp11, was identified previously as a CT gene by comparing mRNA expression between a human parental melanoma cell line and its metastatic variant using mRNA differential display (10) . A recent study demonstrated the expression of SPAN-Xb in myeloma and other hematological cancers (11) . Significantly, high titer antibodies against SPAN-X were present in sera of these patients, but absent in healthy individuals, confirming the immunogenicity of the SPAN-X differentiation antigen. In the present study, we examined the incidence of SPAN-X expression in several tumor types and placed particular emphasis on SPAN-X protein expression in melanomas. We determined that SPAN-X protein, a postmeiotic, germ cell marker, was present in more than half the melanoma tumors examined and that its presence correlated with tumor progression. These findings prompted a detailed study of SPANX gene organization that revealed a multigene family of five members closely linked on Xq27.1. In cooperation with the Human Genome Nomenclature Committee, the nomenclature for the SPANX gene family was established, and the ordering of the SPANX gene cluster was determined.
MATERIALS AND METHODS
Melanoma cell lines were derived from tumor digests obtained from patients at the University of Virginia as described (12 , 13) . The ovarian cancer cell lines CaOV3, ES-2, MDAH2774, OvCar3, SKOV3, and SW626 were obtained from the American Type Culture Collection (Manassas, VA). All of the cell lines were cultured in RPMI 1640 supplemented with 10% fetal bovine serum, penicillin, and streptomycin. Cells were washed once with Dulbecco’s PBS, and the pellets were snap-frozen in liquid nitrogen and stored at −80°C for subsequent protein or RNA extraction.
Reverse Transcription-PCR (RT-PCR) Analysis of Tumor Cell Lines and Clinical Tumor Specimens.
mRNA was isolated with Tri Reagent (Molecular Research Center, Inc., Cincinnati, OH). To detect low abundance SPAN-X transcripts, reverse transcription of total RNA followed by PCR amplification of the SPAN-X cDNA (or control β-actin cDNA) was performed. Briefly, 2 μg of total RNA from each tissue were reverse transcribed in a 20-μl reaction using the Omniscript reverse transcriptase kit (Qiagen, Valencia, CA) according to the protocol provided therein. An RNase inhibitor (10 units) was added to the reaction and oligo(dT)18 (1 μm) was used as the primer. The reactions were incubated at 37°C for 60 min and then at 95°C for 5 min to inactivate the enzyme. Subsequently, PCR was performed on 2 μl of cDNA using the HotStarTaq DNA polymerase and master mix kit (Qiagen). The reaction contained specific primers published previously (10) with modifications. The forward primers were 5′-CTGCCRCWGACATTGAAGAA-3′ (SPAN-XA, C and D) and 5′-CTACTGTAGATCGAAGAA-3′ (SPAN-XB) mixed at a 3:1 molar ratio; the reverse primer was 5′-TCYATGAATTCCTCCTCCTC-3′. Amplification was performed over 30 cycles at 94°C for 1 min, 58°C for 30 s, and 72°C for 30 s.
Western Blot Analysis.
For SDS-PAGE, snap-frozen cell pellets were resuspended in 2× Laemmli sample buffer (7) , vortexed, boiled for 10 min, and centrifuged at 16,000 × g for 15 min. The supernatants were transferred to clean tubes and diluted with an equal volume of distilled water. One dimensional SDS-PAGE was performed on 15% acrylamide gels using 2 × 106 cell equivalents of protein extract per lane and the polypeptides transferred onto nitrocellulose (14) . A 1% SDS extract of sperm (40 μg per lane) was included as a positive control for the SPAN-X protein. All of the samples were obtained with informed consent using forms approved by the University of Virginia Human Investigation Committee. After transfer, the blots were washed with PBS [150 mm NaCl and 10 mm sodium phosphate (pH 7.4)], incubated with blocking buffer (PBS, 0.05% Tween 20, and 5.0% nonfat dry milk), and then with guinea pig anti-SPAN-X antiserum diluted 1:1000 in blocking buffer. SPAN-X antiserum was produced by immunizing guinea pigs with recombinant SPAN-X protein as described previously (8) . Pre-immune serum was used as a negative control for immunostaining. After washing three times in PBS-Tween, blots were incubated with horseradish peroxidase-conjugated donkey antiguinea pig IgG (H+L) F(ab′)2 fragments (Jackson ImmunoResearch, West Grove, PA) diluted 1:2000 in blocking buffer and washed again. Immunoreactive proteins were visualized with TMB membrane peroxide substrate (KPL, Gaithersburg, MD).
Immunoperoxidase Labeling of Tissue Sections.
SPAN-X protein localization in tumor cells was examined by immunoperoxidase labeling of paraffin-embedded tumor sections. For fresh specimens, washed cells or tissues were fixed for several hours with 10% neutral-buffered formalin (Sigma). The samples were dehydrated through an ethanol series, embedded in paraffin, sectioned, and mounted onto slides. In addition, multiple archival samples of paraffin-embedded tumor specimens and tissue microarrays were obtained from the Department of Pathology, University of Virginia. Normal testis sections served as positive controls for SPAN-X protein localization. Before use, sections were dewaxed, rehydrated, and treated with 0.25% hydrogen peroxide to block endogenous peroxidase activity. Nonspecific protein binding sites were blocked by incubating the slides in PBS with 5% normal goat serum. Slides were incubated with immune guinea pig anti-SPAN-X antisera diluted in PBS with 1% normal goat serum. SPAN-X antiserum was produced by immunizing guinea pigs with recombinant SPAN-X protein as described previously (8) . The slides were washed and incubated with horseradish peroxidase-conjugated F(ab′)2 fragments of goat antiguinea pig IgG (Jackson ImmunoResearch) in PBS-normal goat serum. Preimmune serum was used as a negative control for immunostaining. Slides were washed with PBS, immunoreactive proteins visualized by staining with the 3,3′-diaminobenzidine peroxidase substrate, and counterstained with H&E. Slides were visualized using an Olympus BH-2 microscope, and digital images were obtained using Image-Pro Plus software (Media Cybernetics, Des Moines, IA).
For postembedding immunolabeling, washed VMM150 cells were fixed on ice with 4% formaldehyde, 0.5% glutaraldehyde, and 1% tannic acid in 0.1 m sodium phosphate buffer (pH 7.4), rinsed in buffer, dehydrated through an ethanol series, and embedded in Lowicryl K4M resin (Electron Microscopy Sciences, Ft. Washington, PA). Thin sections were mounted on nickel grids and immunostained. Primary guinea pig antibodies were used at a dilution of 1:200, and 5-nm gold-conjugated secondary goat antiguinea pig antibodies (Goldmark, Phillipsburg, NJ) were used at a dilution of 1:50. Grids were rinsed with PBS, rinsed with water, stained with uranyl acetate, and carbon coated. Specimens were viewed and photographed with a JEOL 100CX electron microscope. Darkfield images were obtained from the photographs using Image-Pro Plus software (Media Cybernetics).
Analyses of the SPAN-X genes were performed using Basic Local Alignment Search Tool (BLAST) the Human Genome5 and National Center for Biotechnology Information (NCBI) Map Viewer.6 Multiple sequence alignment was performed using the ClustalW program at the European Molecular Biology Laboratory-European Bioinformatics Institute.7 Virtual Northern analysis was performed against SAGE (serial analysis of gene expression) libraries using identified SAGE tags.8 Analysis of the expressed sequence tags databases was performed using nucleotide BLAST and MEGABLAST at NCBI. Single nucleotide polymorphisms were identified from the dbSNP database at NCBI.9
SPANX Is a MultiGene Family.
Genomic analysis demonstrated that the SPANX genes comprise a multigene family on Xq27.1. SPANX genomic sequences include nonredundant human DNA sequences on one contiguous genomic sequence (NT_011786.13) at 138 M to 138.7 M on the X chromosome from human genome build 33 (Fig. 1)⇓ . Within this contiguous sequence, two identical copies of SPANXA, designated SPANXA1 and SPANXA2, were identified in opposing orientations. The SPANXA1 and SPANXA2 genes encode the SPAN-Xa mRNA and protein. One copy each of SPANXB, SPANXC, and SPANXD have also been identified on this contiguous sequence. SPANXD gene expression in the testis has been confirmed recently in our laboratory.10
Each of the five genes, including SPANXB, is represented by two exons with a small intron of 648 bp. The order of the SPANX genes from the centromere is SPANXB, SPANXC, SPANXA1, SPANXA2, and SPANXD. These results demonstrated that SPANX comprises a multigene family consisting of at least four unique members, one of which is duplicated in the human genome, all of which are tightly clustered on Xq27.1.
Other SPAN-X cDNA sequences, designated SPANXE (AJ457795), SPANXF1 (AJ457796), and SPANXF2, have been annotated in the NCBI database. It seems unlikely that these sequences represent additional SPANX loci but rather represent polymorphisms of the loci described above. For example, the SPAN-Xe (AJ457795) sequence maps to the SPANXD locus and exhibits only three nucleotide substitutions compared with the SPAN-Xd mRNA (AF312765). SPAN-Xf1 and -f2 map to the SPANXB locus and exhibit only one nucleotide difference compared with the SPAN-Xb mRNA (AF098307), and this substitution does not change the amino acid sequence. Each of these nucleotide differences has been described in the single nucleotide polymorphism database. Thus, the SPANX gene family consists of five closely linked genes with multiple polymorphisms.
The proteins encoded by the SPANX genes are presented in Fig. 2⇓ . The SPANXA1, SPANXA2, SPANXC, and SPANXD genes encode highly similar proteins of 97 amino acids. SPANXB encodes the SPAN-Xb protein that is 103 amino acids in length. The SPAN-Xb protein sequence differs from SPAN-Xa, SPAN-Xc, and SPAN-Xd at the NH2 and COOH termini, and contains a six amino acid insertion. No homologous sequences or motifs, with the exception of consensus nuclear localization signals described previously (8) , have been identified to date for the SPAN-X proteins.
SPAN-X mRNA Is Expressed in a Variety of Tumor Cell Lines.
To examine the expression of the SPANX gene in tumors, virtual Northern analysis was performed by searching the SAGE and expressed sequence tag databases with SPAN-X cDNA sequences. The SAGE database uses serial analysis of gene expression to quantify transcript levels in both malignant and normal human tissues. Using the SAGE tag sequence, 5′-GAAATAATGG-3′, present in SPAN-Xa, SPAN-Xc, and SPAN-Xd, a variety of cancer cell lines expressing SPANX, were identified (Table 1)⇓ . These cell lines include: (a) GBM H1110, a glioblastoma multiforme primary tumor derived from a 51-year-old male; (b) three treatment groups of MCF7, derived from a mammary gland carcinoma; (c) DCIS-3 and DCIS-5, derived from two patients with mammary gland ductal in situ high grade carcinomas; (d) DCIS-6 from a vascular endothelium breast carcinoma; (e) a mammary gland carcinoma metastasis, designated B2; (f) ES2–1, an ovarian clear cell carcinoma cell line; (g) a prostate carcinoma cell line PC3 AS2; (h) two brain ependymomas R582 and R1023; and (i) a normal first trimester placental cell line. No SAGE tags were identified in the SPAN-Xb sequence by BLAST analysis of expressed sequence tag databases; SPANX gene expression, including sequences corresponding to all four of the SPANX gene transcripts, was identified in several libraries from normal testes and germ cells but not from any other normal somatic tissues. These data are consistent with the testis-specific expression of SPAN-X observed previously for somatic tissues by Northern and dot blot analyses (9) . Expressed sequence tags homologous to SPAN-X were also identified in prostate tumors (15) , chronic myelogenous leukemia (accession no. BF212177), and amelanocytic melanoma (accession no. BM554791). These data indicate that, whereas in normal tissue SPANX is expressed exclusively in the testis, SPANX is expressed in a number of histologically distinct tumors, a characteristic shared with other CT antigens.
SPAN-X mRNA expression in melanoma and ovarian cancer cell lines was also examined using RT-PCR (Fig. 3)⇓ . A 297-bp band at the expected size for SPAN-X was observed in normal human testis, the melanoma cell lines VMM150 and VMM12, and the ovarian cancer cell lines MDAH2774, SKOV3, SW626, OVCar3, and ES-2. No SPAN-X transcripts were observed in normal ovary, the ovarian cancer cell line CaOV3, or the melanoma cell lines VMM5 and VMM18. These observations indicate that the expression of SPAN-X in tumors is variable, as has been observed for other CT antigens.
SPAN-X Protein Is Translated in Tumor Cell Lines.
To examine the distribution and molecular weight of SPAN-X protein in tumor cells, SDS extracts were prepared from four melanoma cell lines, the ES2 ovarian tumor cell line (American Type Culture Collection #CRL 1978), and ejaculated human spermatozoa. The proteins were separated by reducing one-dimensional SDS-PAGE and immunoblotted as described in “Materials and Methods.” In extracts from VMM150 and ES2 cell lines, postimmune sera reacted with bands migrating between Mr 15,000 and Mr 20,000 similar to that observed in human spermatozoa (Fig. 4)⇓ . Immunoreactive bands with slightly higher apparent molecular weights were also detected in VMM150. Protein extracts from VMM5, VMM12, and VMM18 did not show SPAN-X immunoreactivity in the expected Mr 15,000–20,000 molecular weight range. However, a high molecular weight band of Mr ∼90,000 was present in all of the cell lines examined but not in spermatozoa. Interestingly, this band was consistently detected in cultured human cell lines, but not in extracts of any normal or cancerous primary human tissue samples examined to date (data not shown). This may represent a cross-reactive protein that shares a SPAN-X epitope or epitopes or a nonspecific reactivity induced in guinea pigs after administration of Complete Freund’s Adjuvant. No immunoreactive bands were observed in preimmune serum. These data are consistent with the expression of SPAN-X mRNA in VMM150 and ES2 observed by Northern blot and SAGE analyses, respectively. Furthermore, these results confirm translation of SPAN-X immunoreactive proteins in the melanoma cell line VMM150 and the ovarian cancer cell line ES2.
SPAN-X Protein Localizes to the Nuclear Envelope in a Melanoma Cell Line and in Spermatids.
To examine the localization of SPAN-X protein in tumor cells, two melanoma cell lines, VMM150 and VMM5, and the metastatic melanoma tumor from which the VMM150 cell line was derived were fixed in paraformaldehyde and embedded in paraffin for immunoperoxidase staining with anti-recSPAN-X antiserum. Human testis sections served as a positive control for SPAN-X protein localization (Fig. 5A)⇓ . A brown reaction product from the 3,3′-diaminobenzidine substrate indicated a positive localization. SPAN-X protein was localized to the nuclear envelope of round spermatids (Fig. 5A⇓ , red arrows) and to the posterior head, corresponding with the redundant nuclear envelope in elongating spermatids (Fig. 5A⇓ , purple arrowhead). This localization was consistent with that observed previously in ejaculated spermatozoa (7 , 8) . The melanoma cell line VMM5 demonstrated no expression of SPAN-X mRNA or protein as demonstrated by Northern and Western blot analyses, respectively, and served as an additional negative control. The VMM5 cells demonstrated only background staining of the cytoplasm possibly due to the high molecular weight cross-reactive protein seen by Western analysis (Fig. 5B)⇓ . In contrast, the VMM150 cells demonstrated strong SPAN-X immunoreactivity at the nuclear periphery, within the nucleoplasm close to the nuclear envelope, and slight reactivity of the surrounding cytoplasm (Fig. 5C and D)⇓ . In some VMM150 cells, SPAN-X reactivity was seen in invaginations of the nuclear envelope (Fig. 5D⇓ , green arrowhead). Not all of the VMM150 cells demonstrated SPAN-X immunostaining (Fig. 5C and D⇓ , blue arrows).
Immunohistochemical localization of SPAN-X was also performed on paraffin-embedded sections obtained from the tumor sample that gave rise to the VMM150 cell line (Fig. 6)⇓ . As in the VMM150 cell line (Fig. 6A)⇓ , SPAN-X was localized to the interior and periphery of the nucleus of cells in this tumor, a distant metastatic melanoma (Fig. 6B)⇓ . Preimmune serum demonstrated only slight background staining in all of the tissues and cells examined (Fig. 6A and B)⇓ . Similar localization was seen in formaldehyde-fixed, air-dried cells immunolabeled with SPAN-X antiserum and fluorescent conjugated secondary antibodies (data not shown). This immunolocalization in the metastatic melanoma tumor is consistent with SPAN-X mRNA and protein expression in the derived melanoma cell line VMM150 as observed by Northern and Western blot analyses, respectively. Furthermore, localization of the SPAN-X protein to the nuclear envelope of the patient tumor specimen correlates with the observations in the VMM150 melanoma cell line and the localization observed previously in spermatids, spermatozoa, and SPAN-Xa transfected CV1 cells (7 , 8) .
To examine the subcellular localization of SPAN-X protein in melanoma cells by electron microscopy, postembedding immunolabeling was performed on VMM150 and VMM5 melanoma cells using anti-SPAN-X antibodies and gold-labeled secondary antibodies. Gold particles representing SPAN-X protein were observed overlying the nuclear envelope of VMM150 cells (Fig. 7A and B)⇓ . However, the association of gold particles with any specific structural elements of the nuclear envelope such as the nuclear lamina or pore complexes could not be resolved using this method of fixation and staining. Only background labeling was observed in VMM5 cells (data not shown). The localization of SPAN-X to the nuclear envelope of VMM150 cells and the absence of reactivity in VMM5 cells is consistent with the data on SPAN-X protein and mRNA expression in these two cells lines. Cells incubated with preimmune serum showed only background labeling (data not shown).
SPAN-X Localization in Skin Tumors Correlates with Tumor Aggressiveness.
Immunohistochemical localization of SPAN-X protein was performed on paraffin-embedded tissues from a number of skin tumor patients. Table 2⇓ summarizes the results of the immunohistochemical localization of SPAN-X on sections of 21 skin tumors of varying degrees of aggressiveness. Of these 21 tumors, 10 demonstrated positive immunoreactivity for SPAN-X (48%). Localization of SPAN-X was observed in the nucleus (60%), cytoplasm (20%), or both (20%). Significantly, immunostaining for SPAN-X protein correlated with increased aggressiveness of skin tumors, particularly in distant, nonlymphatic metastatic melanomas. Fig. 8A⇓ shows a metastatic melanoma with positive SPAN-X staining in the cytoplasm. Fig. 8B⇓ shows a metastatic melanoma demonstrating heavy staining of the nucleus. Fig. 8C and D⇓ are skin metastatic melanomas showing SPAN-X staining of the cytoplasm and nucleus, respectively. A distant metastatic melanoma of the parotid gland demonstrates nuclear localization of SPAN-X (Fig. 8E)⇓ . Fig. 8F⇓ shows a metastatic melanoma with no SPAN-X staining. Slides incubated with preimmune serum showed no positive staining (data not shown). The nuclear localization of SPAN-X (Fig. 8B, D, and E)⇓ is consistent with localization of SPAN-Xa and SPAN-Xc in transfected mammalian cells (7 , 10) , whereas the cytoplasmic staining (Fig. 8A and C)⇓ is consistent with the transfection of SPAN-Xb (7) . The differential localization of SPAN-X in tumor cells may be a result of differential expression of the multiple SPANX genes in individual tumors and may prove useful in diagnosing and/or treating tumors.
SPANX Genes Comprise a MultiGene Family.
From the centromere, the arrangement of the five tightly clustered SPANX genes, on genome build 33, is SPANXB, SPANXC, SPANXA1, SPANXA2, and SPANXD between 138 M and 138.7 M on Xq27.1. The existence of multiple SPANX genes is consistent with other CT antigens that are present in the human genome as multigene families. For example, >30 melanoma antigen (MAGE) gene loci have been identified including 13 MAGEA genes, 4 MAGEB genes, and several MAGE-like hypothetical genes for which evidence of transcription has not yet been found. The majority of the MAGE genes map to human chromosome X on both the long and short arms, and the cluster of SPANX genes is in close proximity to the loci of a number of MAGE genes and LDOC1 (leucine zipper, down-regulated in cancer 1), also known as BCUR1 (breast cancer, up-regulated 1). Interestingly, expression of these closely linked genes has been implicated in tumorigenesis. For example, SPANX, MAGEC3 (HCA2), MAGEC1, MAGEA10, and MAGEE1 are members of gene families that encode CT antigens and are found in a number of melanomas and other tumors (16, 17, 18, 19, 20, 21) . LDOC1 is located adjacent to the SPANXC locus and was identified by screening for genes expressed in cancer cells using RNA differential display (22) . The high number of cancer-related genes in this region of the X chromosome suggests a high susceptibility to dysregulation during tumor progression.
The mapping of SPANX to the q27 region of the X chromosome additionally emphasizes the potential importance of SPAN-X in tumorigenesis. Rapley et al. (23) mapped the locus for familial testicular germ cell tumors, designated TGCT1, to Xq27. Testicular germ-cell tumors (TGCTs) affect 1 in 500 men and are the most common cancer in males ages 15–40 in Western Europe. The known risk factors for TGCT include undescended testis and a family history of the disorder, demonstrating a greater risk in brothers consistent with X-linked inheritance. Linkage analysis of the set of families compatible with X linkage provided evidence for a TGCT predisposition locus at Xq27. In addition to TGCT1, a prostate susceptibility gene, symbolized HPCX, has also been mapped to Xq27-q28 and is thought to account for 16% of hereditary prostate cancers (24 , 25) . Although other genes may be embedded within these cancer susceptibility loci, SPANX is one of the few genes identified in this region to date. Furthermore, the unique distribution of SPAN-X in 50% of human spermatozoa, its germ cell-specific expression, and its identification in human tumors also implicate the SPANX genes in the etiology of TGCT and/or prostate cancer.
The mechanism(s) of tumor expression of these testis-specific genes is not well understood. The lack of coordinate expression of different CT antigen family members rules out a general activation of silent X-linked genes in cancer and indicates a selective gene-specific regional activation. This could be due to aberrant promoter/enhancer binding factors by cis- or trans- genes, or to specific modification in DNA methylation or acetylation patterns. During spermatogenesis, the germ cell genome becomes more methylated, particularly at CpG dinucleotides (26) . Demethylation of testis-specific gene promoters occurs before the activation of transcription during spermatogenesis (27 , 28) . For MAGE, both methylation and DNA binding factors appear to be involved in aberrant expression of this gene family (29, 30, 31, 32, 33, 34) . The expression of MAGE in tumor cells seems to be a result of genome-wide demethylation that occurs frequently during tumor progression (31 , 35 , 36) . In addition, the demethylating agent 5-aza-2′-deoxycytidine can induce the expression of the MAGE genes in MAGE-negative cells suggesting that gene activation results from the demethylation of the promoter region (29, 30, 31, 32, 33, 34) . These data suggest that promoter methylation is a shared mechanism involved in regulating the expression of CT antigens in metastatic melanomas. Furthermore, the clinical use of 5-aza-2′-deoxycytidine may provide a new chemoimmunotherapeutic strategy for melanoma patients. Using the computer algorithm Grail CpG Islands and NCBI Map Viewer on the genomic contig NT_011786.13, no CpG islands were identified <65,000 bp from any of the six SPANX coding sequences. Thus, it seems unlikely that methylation/demethylation of the SPANX promoter is involved directly in SPANX gene expression during spermatogenesis and/or tumor progression.
SPAN-X Protein Is Expressed in a Large Number of Melanomas.
Our results confirmed a high incidence of SPAN-X protein expression in melanoma tumors, verified the testis specificity of SPAN-X, and demonstrated a pattern of postmeiotic gene expression in the adluminal compartment of the blood-testis barrier. The presence of SPAN-X protein in 48% of skin tumors and 53% of melanomas suggests that SPAN-X may prove to be a useful marker for staging melanoma tumor progression and that studies of additional cohorts of tumor samples should be conducted. Detection of SPAN-X protein in over half of the melanomas tested in the present study is generally comparable with the nested RT-PCR results of Zendman et al. (10) , who identified SPAN-X mRNA in 70% of melanoma tumors, but not in normal skin. In a recent study using RT-PCR analysis of 52 melanoma samples in three tumor stages (37) , SPAN-X was identified predominantly in primary melanomas: 45% in primary tumors, 27% in local metastases, and 10% in distant metastases. In the same study, the CT antigen NY-ESO-1 was found in the majority of local metastasis (45%) and distant metastatic (50%) melanomas but in only 10% of primary tumors. It is not yet clear whether the difference in the incidence of mRNA expression in the two cited studies and the incidence of protein localization in the present study is indicative of translational regulation of SPAN-X. The possibility that extremely sensitive and nonquantitative nested RT-PCR approaches may simply overestimate the incidence of protein expression for CT antigens such as SPAN-X must also be considered, because these assays allow the efficient detection of transcripts that may be present only at very low copy numbers in a small subset of cells within a specimen. Larger controlled studies, including studies that correlate mRNA and protein detection with serological data, are indicated to additionally refine the incidences of SPAN-X gene expression and protein localization in melanomas and other tumor types.
The nuclear localization of SPAN-X in melanomas is consistent with localization of SPAN-Xa (7) and SPAN-Xc (10) found in mammalian cells transfected with these constructs, whereas the cytoplasmic localization is consistent with transfection results for SPAN-Xb (7) . These transfection experiments indicated that the nuclear localization signals found in the SPAN-X protein sequences were effective in trafficking the SPAN-X a and c proteins to the nucleus. However, the SPAN-Xb protein apparently contains a variation that inhibits nuclear targeting in somatic cells. The open reading frame of SPAN-Xb contains a consensus N-linked glycosylation site within the six amino acid insert that is not present in the other SPAN-X protein sequences. Glycosylation of this asparagine in the transfected cells expressing SPAN-Xb would account for the increased molecular weight of the SPAN-Xb protein observed by immunoblotting and may relate to its localization in the cytoplasm (7) . Thus, localization of the SPAN-X protein to the cytoplasm in some melanomas may be the result of SPAN-Xb expression and glycosylation. Similarly, the differential localization of SPAN-X to both the nucleus and cytoplasm in tumor cells may be a result of differential expression of the multiple SPANX genes in individual tumors and may prove useful in diagnosing and/or treating tumors.
The SPAN-X protein is associated with the nuclear envelope in 80% of melanomas positive for SPAN-X staining. This is consistent with the localization of SPAN-X to the redundant nuclear envelope in spermatozoa. Unlike the nuclear envelope of the anterior part of the sperm head, which is tightly apposed to the nuclear chromatin and lacks nuclear pores, the redundant nuclear membranes, embedded within the cytoplasmic droplet, have a wider apposition between their lamellae and possess numerous, regularly arranged nuclear pore complexes. Although the SPAN-X peptide sequences do not contain FG-repeat nucleoporin domains (38) , the SPAN-X protein from spermatozoa is highly insoluble suggesting that this protein is a structural protein or associated with structural elements of the nuclear envelope (7) . The localization of SPAN-X to the redundant nuclear envelope and its structural characteristics suggests that SPAN-X may be a component of a large macromolecular structural complex such as the nuclear pore. Additional studies are under way to determine whether SPAN-X is associated with the nuclear pore complexes in spermatids, spermatozoa, and cancer cells expressing SPAN-X. Although the predominantly nuclear localization of SPAN-X in tumors and tumor cell lines suggests that SPAN-X nuclear epitopes may be inaccessible to antibody or T cells invoked by a therapeutic vaccine, the results of the recent study by Wang et al. (11) demonstrating high-affinity antibodies to SPAN-X in cancer patients supports the hypothesis that SPAN-X peptides are presented at the surface of tumor cells in the context of MHC or that SPAN-X protein is released during necrosis. These hypotheses now require verification.
SPAN-X Protein Is on the Protected Side of the Blood-Testis Barrier.
In the testis, the MAGE CT antigens are expressed in spermatogonia and primary spermatocytes (39) . MAGE-A1 and MAGE-A3 have been localized to the cytoplasm, but MAGE-A11 is present in the nucleus (40, 41, 42) . MAGE proteins have limited homology to the tumor suppressor protein p53 in a region important for the growth of p53 inhibitory properties (2) , suggesting that MAGE may function in the control of tumor proliferation, although no experimental evidence for such activity has yet been found. The CT antigen NY-ESO-1 is also expressed in spermatogonia and localized to the cytoplasm (43) . The CT antigen SCP-1 is a synaptonemal complex protein that is expressed during meiotic prophase in spermatocytes and is thought to function in the alignment of homologous chromosomes during meiosis, the promotion of cross-over events, and chromosome segregation (44) . Except for these examples, the expression pattern, localization, and function of other CT antigens in the testis remain unknown despite their growing number and importance in tumor immunology, highlighting the value of generating anti-CT antigen antibodies to characterize their localizations and function in normal testicular cells and tumors.
Unlike the CT antigens noted above, SPAN-X is expressed in postmeiotic spermatids. This is interesting, as few X-linked genes exhibit postmeiotic expression (45) . Thus, the expression of SPAN-X mRNA and protein in haploid spermatids places SPAN-X on the protected, adluminal side of the blood testis barrier (46 , 47) . Because spermatozoa appear in males only after puberty, when the immune system has long been established, the development of active tolerance to male germ cell autoantigens is of considerable importance. Immune privilege in the testis is established as a result of both a physical barrier and an immunosuppressive microenvironment. The structural barrier, composed largely of tight junctional complexes between adjacent Sertoli cells, separates the seminiferous epithelium into two cellular compartments, basal and adluminal. From an immunological perspective, this barrier has two effects: (a) it prevents circulating antibodies from entering the seminiferous tubules beyond the Sertoli cell tight junctions; and (b) it isolates the meiotic and postmeiotic spermatogenic cells from the immune system of a body in a compartment where differentiation antigens are expressed and may be recognized as “foreign.” The lack of HLA gene expression in the testis may also be necessary to create a protecting environment to prevent T-cell-mediated autoimmune reactions (4 , 48) .
Thus, it may be predicted that immunization of cancer patients with the SPAN-X protein, which is expressed in normal testis after puberty and is protected from the immune system by the blood-testis barrier, should produce an immune response to this differentiation antigen in both male and female patients, whereas being unlikely to induce autoimmune orchitis in males. The lack of SPAN-X expression in adult somatic tissues predicts that this protein will be immunogenic in humans, a prediction verified by the recent demonstration that SPAN-Xb was aberrantly expressed and elicited immune responses in patients with myeloma and hematological cancers (11) . These considerations support the possible inclusion of SPAN-X proteins in immunotherapeutic vaccines. Indeed, SPAN-X appears to be the first CT antigen restricted to the adluminal compartment that has shown immunogenicity in humans. This finding indicates that the identification of additional postmeiotic CT antigens may be important to the development of safe polyvalent CT antigen vaccines, which will address the variability of CT antigen expression in tumors, whereas simultaneously mitigating pathological consequences such as immune orchitis.
Note Added in Proof
An article by Furlong et al., Journal of the National Cancer Institute, 1999, indicates that the SW626 shown in Figure 3⇓ is a cell line of colon origin, rather than ovary as previously described. A recent article by Zendman et al., Gene, 2003, has recently described in part the genomic organization of the SPANX genes.
Grant support: U54HD29099 and D43TW/HD00654 (J.C.H.) from the Fogarty International Center, a pilot project grant funded through the National Cancer Institute Cancer Center, Support Grant to the University of Virginia Cancer Center (M. A. C.), and a grant from the Cancer Antigen Discovery Collaborative of the Cancer Research Institute and the Ludwig Institute for Cancer Research (M. A. C.).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Requests for reprints: John C. Herr, Department of Cell Biology, Center for Research in Contraceptive and Reproductive Health, University of Virginia, P. O. Box 800732, 1300 Jefferson Park Avenue, Charlottesville, Virginia 22908. Phone: (434) 924-2007; Fax: (804) 982-3912; E-mail:
↵5 Internet address: http://www.ncbi.nlm.nih.gov/genome/seq/page.cgi?F = HsBlast.html&&ORG = Hs.
↵6 Internet address: http://www.ncbi.nlm.nih.gov/mapview/.
↵7 Internet address: http://www.ebi.ac.uk/clustalw/.
↵8 Internet address: http://www.ncbi.nlm.nih.gov/SAGE/.
↵9 Internet address: http://www.ncbi.nlm.nih.gov/SNP/.
↵10 V. A. Westbrook, P. D. Schoppee, A. B. Diekman, C. J. Flickinger, M. A. Coppola, J. C. Herr. The SPAN-X family of genes are differentially expressed during spermiogenesis and SPAN-X segregates in human and chimpanzee spermatids to non-acrosomal domains during nuclear envelope morphogenesis, manuscript in preparation.
- Received April 21, 2003.
- Revision received August 11, 2003.
- Accepted August 14, 2003.