Isolation of a Novel Gene, TSP50, by a Hypomethylated DNA Fragment in Human Breast Cancer

INTRODUCTION

Abnormal DNA methylations (hypomethylation or hypermethylation) have been linked to various human diseases including cancers (1, 2, 3, 4, 5, 6, 7, 8) . Because methylated DNA sites are usually close to genes (9, 10, 11, 12, 13) , searching for differentially methylated DNA fragments in cancer could pinpoint genes of interest. Consequently, a modified RDA technique using human breast cancer biopsies as starting material was used to search for differentially methylated DNA fragments. Unlike traditional RDA 3 (14) , to perform modified RDA, the restriction enzyme MspI, which is sensitive to the methylated GC-rich sequence GGC^mCGG (15, 16, 17, 18) , was used as a master enzyme to cleave genomic DNAs for amplicon preparation. MspI is a relatively frequent cutting enzyme (19) , which, when used alone, produces an amplicon of high complexity that can cause unsuccessful subtractive hybridization. Hence, a second restriction enzyme, or “partner enzyme,” has been incorporated into the technique. The amplicons can only be made by PCR from DNA fragments with both ends cut by the MspI enzyme.

As a result of using this modified technique, two DNA fragments, BR50 and BR254, were isolated that detected DNA hypomethylation in breast cancer. Additional studies verified that both fragments also detected hypomethylation in ovarian cancer, and BR254 was amplified in 1 of 10 breast cancer biopsies. On the basis of these findings, we considered both fragments good candidates to search for genes that might be related to various malignancies. This report will focus on presenting the detailed studies related to BR50, which covers its own isolation as a differentially methylated DNA fragment, to its utilization in the discovery of a novel gene, TSP50.

Investigation of the TSP50 gene has found that it encodes a protease-like protein. Northern analysis of multiple human tissue RNA expression panels showed that TSP50 is a tissue-specific gene, which was heavily expressed in human testes. There were almost no visible amounts of TSP50 transcript displayed in the other 15 types of human tissues in the panels. This result indicates that the TSP50 gene holds a special physiological function(s) in human testes. The DNA methylation status of the downstream region of the gene in normal human testes and eight other tissues was also examined. Apparently, DNA methylation silences the TSP50 gene expression in those eight normal tissues, whereas DNA demethylation in human testes could be a key element responsible for gene expression. Furthermore, RT-PCR was performed to examine differential expression in breast cancer and matched normal control tissues. We found that ∼28% of the cancer samples tested expressed the TSP50 gene, whereas the corresponding controls did not. Whether there is a relationship between gene expression and cancer development is presently under investigation.

MATERIALS AND METHODS

DNA from Human Cancer Biopsies.

Dissected human breast and ovarian cancer tissues (tumor and matched normal) were immediately frozen in liquid nitrogen and stored at −70°C. DNAs were isolated from those tissues by the phenol extraction method (20) .

Modified RDA.

One to two μg of DNA (tester) isolated from human breast cancer biopsies and their matched normal DNA (driver) were cleaved with MspI (20 units/μl; Boehringer Mannheim) and MseI (20 units/μl; New England Biolabs, Inc.) in a 50-μl reaction for 3 h. To prepare tester and driver master amplicons, the MspI- and MseI-digested tester and driver genomic DNAs were ligated to 1.5 μg of MSA24-mer and 0.75 μg of MSA12-mer (Table 1) ⇓ ; these were the first pair of oligonucleotide linkers that only recognize the ends generated by MspI. The procedures for amplicon preparation were performed as described (14) . The DNA amplicons were then purified by phenol, phenol/chloroform extraction. To remove the first set of linkers from the driver amplicon, 80 μg of driver amplicon DNA were digested with the MspI enzyme (10 units/μl). To change the tester master amplicon DNA linkers, 5 μg of tester master amplicon were digested with MspI (20 units/μl) and ligated to 0.6 μg of MSB24-mer and 0.3 μg of MSB12-mer (Table 1) ⇓ ; these were the second set of oligonucleotide linkers. Subtractive hybridization was performed as described (14) . The first round of difference products (DP1) were amplified as described (14) . To prepare the second round of subtractive hybridization, 3 μg of DP1 were digested with the restriction endonuclease MspI (20 units/μl). To put a new set of linkers on DP1, 0.1 μg of DP1 was mixed with 0.6 μg of MSC24-mer and 0.3 μg of MSC12-mer (Table 1) ⇓ . Another round of subtractive hybridization/PCR amplification was repeated. The second round of difference products (DP2) usually contained several individual DNA fragments when electrophoresed on a 2% agarose gel. The individual DNA fragments were purified by DNA gel extraction kit (Qiagen, Inc.) and subcloned into pUC118 vector, which was linearized by the restriction endonuclease AccI and transformed into Escherichia coli (DH5α). Twelve cloned inserts were chosen to be amplified, from which different-sized probes were selected for master amplicon Southern blot. The candidate probes were then further tested by human genomic DNA Southern blot.

Amplicon DNA Southern Blot.

The first round of positive probe screening was performed with amplicon DNA Southern blots. Non-Radiation Southern Blot and Detection kits (Genius) were purchased from Boehringer Mannheim. Probe labeling and detection followed the instructions of the manufacturer. Two to three μg of tester and driver amplicon DNA were electrophoresed on a 2% agarose gel and blotted to positively charged nylon membranes (Boehringer Mannheim). For prehybridization, the membranes were placed at 68°C for 2–4 h in solutions containing 6× SSC, 5× Denhardt’s solution, 0.5% SDS, 0.1 m EDTA, and 50 μg/ml of salmon sperm DNA. Under the same conditions, the probes were added and hybridized to the membranes overnight. The membranes were then rinsed three times with 2× SSC, 1× Blot wash (12 mm Na₂HPO₄, 8 mm NaH₂PO₄, 1.4 mm Na₄P₂O₇, and 0.5% SDS) at 68°C and further washed three times (30 min each) with the same buffer at 68°C. Next, the membranes were equilibrated in buffer A (100 mm Tris·HCl, 150 mm, pH 7.5) and transferred into buffer B (2% block reagent in buffer A), which was incubated at room temperature for 1 h. The membranes were then washed 2 times for 15 min with buffer A and equilibrated in buffer C (100 mm Tris·HCl, 100 mm NaCl, and 10 mm MgCl₂). Before the membranes were exposed on Kodak X-OMAT film for 1 h, they were rinsed in lumi-P530 for 1 min and kept in a plastic sheet protector.

Genomic DNA Southern Blot.

Genomic DNAs were digested with a desired restriction enzyme (20 units/μl) and electrophoresed on 1.5% agarose gels, which were then transferred to Hybond N⁻ membranes (Amersham). These membranes were exposed to UV light to immobilize the DNA. Probes for the Southern blot were labeled with High Prime DNA labeling kits (Boehringer Mannheim) following the instructions of the manufacturer. The procedure for hybridization and blot wash were the same as in the Amplicon DNA Southern blot section.

Northern Analysis.

Two Human Multiple Tissue Northern blot panels, MTN and MTN II, were purchased from Clontech, Inc. The MTN blot contains ∼2 μg of poly(A)⁺ RNA per lane from eight different human tissues (heart, brain, placenta, lung, liver, skeletal muscle, kidney, and pancreas). The MTN II blot contains the same amounts of mRNA from an additional eight different human tissues (spleen, thymus, prostate, testes, ovary, small intestine, colon, and peripheral blood leukocyte). The labeling and detection of the probes were the same as above.

DNA Sequence and Chromosome Assignment.

The pUC118 plasmid containing the candidate DNA fragment was sequenced using the Ampli-Cycle sequencing kit (Perkin-Elmer), under conditions described by the manufacturer. Chromosome assignment for the candidate DNA fragment was determined by genomic Southern blot of the HindIII digested monochromosomal human/rodent somatic cell hybrid mapping panel #2 (NIGMS Human Genetic Mutant Cell Repository) while it was used as a probe. Fine chromosome mapping was performed with GeneBridge 4 Radiation Hybrid Panel (Research Genetics, Inc.) by PCR amplification (21) .

Human Genomic DNA Library Screening.

A human placenta genomic phage library, EMBL3 SP6/T7 (Clontech, Inc.), was used for cloning a longer genomic fragment containing the candidate fragment. Phage infection procedure was based on the instructions supplied by the manufacturer. Plaques (2 × 10⁶) were evenly distributed on 20 plates (150 × 15 mm), then transferred onto Hybond N⁻ membranes. The treatment of the membranes, preparation of the probe, and the blot wash were the same as that described in the Genomic DNA Southern blot section. The phage DNA, with human DNA insert, was purified by the Lambda TRAP Plus kit (Clontech, Inc.) following the instructions of the manufacturer. The individual insert was released by restriction enzyme SstI cleavage from the phage DNA arms, then subcloned into pUC118 plasmid.

Human cDNA Library Screening.

A human testes λgt11 cDNA library (Human Testis cDNA Library; Clontech, Inc.) was used to obtain an intact gene following the instructions of the manufacturer. Plaques (2 × 10⁵) evenly distributed on six plates (150 × 15 mm) were transferred onto Hybond N⁻ membranes. Southern analysis was performed as before. The phage DNA, with human cDNA insert, was purified by the λ Quick! Spin kit (BIO 101, Inc.) following the instructions of the manufacturer. The individual insert was released by restriction enzyme EcoRI cleavage from the phage DNA arms, then subcloned into pUC118 plasmid.

RT-PCR.

Total RNAs were isolated from paired breast cancer and normal tissues by RNA isolation kit, RNA STAT-60 (TEL-TEST, Inc.). The first-strand cDNA was synthesized by SuperScript Preamplification system kit (Life Technologies, Inc.). Oligomers E (ACCAGAGCGTCCAGTGTGTCC, sense) and F (TGGGACTTGATGATCTGAACC, antisense) were used to synthesize the TSP50 gene. The predicted size was 699 bp. β-actin was used as an internal control, the sense and antisense primers of which were 5′ -GACGACATGGAGAAGATCTGG-3′ and 5′ -TGTAGAGGTAGTCAGTCAGG-3′. The predicted size for β -actin was 335 bp. The PCR reaction mixture was comprised of cDNA derived from 125 ng of RNA, 10 pmol of sense and antisense primers from both TSP50 and β-actin, 200 μm of four deoxynucleotide triphosphate, and 0.125 unit of Taq DNA polymerase with reaction buffer (Perkin-Elmer) in a final volume of 25 μl. Thirty-eight cycles of PCR were carried out. Each cycle of PCR included 30 s of denaturation at 95°C, 60 s of annealing at 60°C, and 60 s of extension at 72°C. The PCR products were separated on a 2% agarose gel.

RESULTS

Isolation of Hypomethylated Sequences from Human Breast Cancer Biopsies.

The DNAs isolated from three paired human breast cancer biopsies (tester) and surrounding normal tissues (driver) were cleaved with the MspI and MseI enzymes. The tester and driver amplicons with both ends cleaved by MspI were selectively prepared by PCR amplification (see “Materials and Methods”). After two rounds of DNA hybridization/subtraction and PCR amplification, individual fragments (DP2) were isolated from two breast cancer patients (see “Materials and Methods”). The DP2 fragments were subcloned into the pUC118 vector, and the inserts were amplified by PCR. Twelve different-sized inserts were selected from each modified RDA and used as probes for the master amplicon Southern blot. Two probes, BR50 and BR254, isolated from two patients, were identified as candidate probes for additional study. In this report, we focus on presenting the work that has been done by probe BR50.

Probe BR50 was selected from the DP2 isolated from breast cancer patient no. 14’s biopsy by the modified RDA technique (Fig. 1a) ⇓ because it hybridized a band of much greater intensity in the tester amplicon than that in the driver amplicon (Fig. 1b) ⇓ . To confirm the differences observed in the tester and driver amplicons, a Southern analysis was performed on patient no. 14’s tumor and matched normal genomic DNAs. Six μg of each DNA were cleaved with the MspI enzyme and hybridized by probe BR50. The results showed that in the tumor DNA, the probe hybridized a lower band, ∼1 kbp long, of much greater intensity than an upper band, which was ∼2 kbp in length. In the normal DNA, just the opposite occurred (Fig. 2) ⇓ . Because the sizes of the upper and lower bands in the tumor and normal control DNAs were the same, the only reasonable explanation causing uneven hybridization intensities is DNA hypomethylation in the tumor cells. To examine whether the event also existed in other breast cancer patients, paired tumor and normal DNAs isolated from additional breast cancer biopsies were cleaved with MspI and subjected to Southern analysis. The results showed that of 10 samples tested, 4 had similar hybridization patterns to those of patient no. 14 (Fig. 2) ⇓ . It is notable that in the normal DNAs of patient nos. 3, 4, and 5, more than one upper band was evident. We believe this can be attributed to partial DNA demethylation, instead of incomplete enzymatic digestion. This is because the Southern membrane was reblotted by a control probe, which was a background probe isolated along with probe BR50, and only a single hybridized band was displayed in each lane (Fig. 2) ⇓ .

BR50 Also Detected Altered Methylation Patterns in Ovarian Cancer.

To examine whether the hypomethylation event also occurred in human ovarian cancer, paired DNA samples isolated from eight ovarian cancer patients were analyzed by probe BR50. The DNAs were cleaved with MspI and hybridized with probe BR50 in a Southern blot experiment. The results demonstrated that of eight patients tested, four displayed similar hybridization patterns to those observed in the breast cancer samples. In the tumor DNAs, the lower band was heavily hybridized by the probe, whereas in the normal control DNAs, a lower and upper band were hybridized. The completion of DNA digestion was confirmed by the same control probe used before, which only hybridized a single band in each lane (Fig. 3) ⇓ . Thus, BR50 also detected altered DNA methylation in ovarian cancer.

Sequencing and Chromosome Assignment of Probe BR50.

DNA sequencing found that BR50 contains 1005 bps, with a CG content of 58% (Fig. 4a) ⇓ . The DNA fragment has a GGCCGG sequence on one end and CCGG sequence on the other end. Because MspI is sensitive to GGC^mCGG and not sensitive to CC^mGG sequences, it is conceivable that the GGCCGG sequence was methylated in normal breast tissue DNA while being hypomethylated in tumor DNA. This prediction probably holds true because all of the candidate probes, including probe BR254 and subsequent probes isolated from breast cancer biopsies, have one end terminating with a GGCCGG sequence and the other with a CCGG sequence (data not shown). A homologue search of the NIH GenBank discovered that BR50 is not homologous to any existing sequences. 4 The chromosome assignment determined by monochromosomal human/rodent somatic cell hybrid mapping panel #2 established its location on chromosome 3 (data not shown). The fine chromosome position of BR50 was analyzed by PCR amplification using GeneBridge 4 Radiation Hybrid Panel as templates. The top and bottom strand primers for BR50 are 5′-ACCAGATGGAGGCAGTTGAC-3′ and 5′-AAGTGGGTGCT-CTTTCTGTG-3′, respectively. The result obtained from radiation hybrid mapping suggests that BR50 is placed 4.29 cR below the adjacent STS, AFMB362WB9, which is 179.84 cR from the top of the chromosome 3 linkage group. This result confirmed that BR50 is proximally located on 3p12–14.

Search for the Coding Regions by Probe BR50.

On the basis of our preliminary analysis, BR50 was an interesting probe to be used for an adjacent gene search because it was differentially methylated in some breast and ovarian cancer samples tested, although the sample size was small. BR50 is a genomic fragment, and consequently the chances that it encodes a polypeptide are slim because a majority of human genomic DNA sequences are noncoding sequences. As a result, we decided that the first step in searching for a gene should be to isolate a longer DNA piece from a human genomic DNA library, with the hope that it may contain exon(s). It is well known that DNA methylation sites can be near genes (18) ; therefore, we decided to screen a human placenta genomic phage library, EMBL3 SP6/T7 (Clontech, Inc.), where the insertion sizes were relatively small (9–28 kbp). For this purpose, 2 × 10⁶ plaques derived from the library were screened by probe BR50. As a result, a 17-kbp length clone was isolated. To obtain the sequence information, the DNA clone was released from the phage DNA arms by restriction enzyme SstI cleavage, which generated eight DNA fragments. All eight fragments were subcloned into pUC118 plasmids. The fragments smaller than 1 kbp were completely sequenced, whereas the fragments larger than 1 kbp were partially sequenced from both the plasmid and insert junctions. A homologue search of the NIH GenBank revealed two exons, BR50-44 and BR50-45, which contained 112 and 132 nucleotides, respectively. Both exons encoded polypeptides that were ∼50% identifiable to several mammalian proteases, such as serine proteases and tryptases. The BR50-45 sequence was found 142 bp upstream of the BR50 sequence.

Because we did not know the exact positions of the two exons in the 17-kbp fragment, it was possible that other exon(s) might lay between them. To gain a longer coding sequence, we designed four oligomers (A, B, C, and D), based on the sequence information of both exons, to perform PCR. Oligomer A (5′-CCTGGATGGTCAGCGTG-3′) and B (5′-CTGGGAGGCAATGATGGT-3′), which were on the complimentary strand, were based on the sequence information of BR-44; and C (5′-CTGGAGAGCCCTTGGTCT-3′) and D (5′-CAGTGTTGGTAGGAGGAG-3′), which were on the complimentary strand, were based on the sequence information of BR-45. A strategy using four different combinations of oligomer pairs was used to perform PCR by using the Human Universal cDNA Library Panel (Clontech, Inc.). A PCR product, which was about 700 bp in length, was generated from one oligomer combination (A/D, 5′-CCTGGATGGTCAGCGTG-3′/5′ -CAGTGTTGGTAGGAGGAG-3′). This PCR product was directly sequenced. Combining the sequence information of the PCR product and the two exons, we obtained a cDNA fragment that contained 974 bps. The DNA homologue search of the NIH GenBank revealed again that it coded for a protease-like protein, and the overall identity was ∼40%.

The Candidate Gene Is Highly and Specifically Expressed in Human Testes.

To obtain a full-length cDNA, it is critical to use the right cDNA library where the gene of interest is expressed. Thus, two Human Multiple Tissue Northern blot panels, MTN and MTN II, containing 16 different tissue mRNAs (Clontech, Inc.), were used to test the expression of the candidate gene by using the 700-bp cDNA PCR product as a probe. The results showed that there were no visible transcripts of this gene in the eight mRNAs (heart, brain, placenta, lung, liver, skeletal muscle, kidney, and pancreas) included in the MTN panel (data not shown). In the MTN II panel, a 1.7-kbp band was heavily hybridized by the probe exclusively in the testes mRNA as compared with the control probe, which was the human rab6 gene (Fig. 5) ⇓ . These results suggested that the gene that we were searching for is a tissue-specific gene. We have named the gene TSP50. At this moment, the biological function(s) of the gene in human testes remains unknown.

Isolation of the Full-length TSP50 Gene from a Human Testes cDNA Library.

TSP50 is highly expressed in human testes; to search for the full-length TSP50 gene, a human testes cDNA library (Human Testis cDNA Library; Clontech. Inc.) was screened by the 700-bp TSP50 cDNA sequence. A cDNA clone containing the probe sequence was isolated. Sequence analysis suggests that this fragment encodes a protein with 385 animo acids (Fig. 4b) ⇓ . There is a stop codon located at the 117th bp upstream of the first initial translation site, and there is a 125-bp untranslated region before a polyadenosine sequence. These results imply that a full-length gene has been obtained. It is also notable that the BR50 probe sequence is located at the 3′ end of the gene and is only 17 bp downstream from the polyadenosine adding signaling site. The exons BR50-44 and BR50-45 encode animo acids from 103 to 157 and 308 to 385, respectively (Fig. 4. b.) ⇓ . The 3′- untranslated region before the polyadenosine site is also included in the sequence of BR50-45.

DNA Methylation Status of the TSP50 Gene in Human Testes and Other Normal Tissues.

Our studies have proven that TSP50 is a tissue-specific gene, and the methylation patterns in its 3′ region were altered in some breast and ovarian cancers. It is also known that many tissue specific genes are methylated, and this methylation may regulate their expression (22, 23, 24) . To explore the possible relationship between TSP50 gene expression and DNA methylation in different normal human tissues, Southern analysis was performed. The normal tissues tested included the testes, where TSP50 was expressed, and bladder, blood, breast, colon, lung, kidney, placenta, and ovary samples, where TSP50 was apparently not expressed. To perform the Southern analysis, BR50 was used as a probe. DNAs isolated from the nine tissues were digested by MspI and HpaII, which is an isoschizomer of the MspI enzyme and the most popular enzyme used to study DNA methylation patterns (25) . HpaII digestion showed that in the testes DNA, two bands, probably released from each allele by enzyme cleavage, were hybridized by the probe. However, in the DNAs of other tissues, the corresponding bands were either not hybridized or hybridized to a much smaller degree (Fig. 6a) ⇓ . For MspI cleavage, both bands were released in different tissues to various extents (Fig. 6b) ⇓ . Both blots used a genomic fragment that did not detect differential DNA methylation as a control to determine complete enzymatic digestion (Fig. 6) ⇓ . These results demonstrated that the TSP50 gene was differentially methylated in various human tissues. In general, DNA demethylation in the testes is correlated with high levels of gene expression. Conversely, DNA methylation is correlated with gene silencing in the bladder, blood, breast, colon, lung, kidney, placenta, and ovary tissues.

TSP50 Was Differentially Expressed in Some Breast Cancer Tissues.

Preliminary results demonstrated that TSP50 was differentially methylated in 40% of the breast cancer tissues tested. This suggested that it could also be differentially expressed in cancer. To test this possibility, RT-PCR was carried out to determine TSP50 expression levels in 18 paired breast cancer tissues. Our findings showed that TSP50 PCR products were generated in five tumor tissues, whereas in their normal controls, they were not visible relative to the control gene, β-actin (Fig. 7) ⇓ . Products generated from the five patients were gel purified and sequenced. DNA sequence analysis confirmed that the PCR products synthesized were the TSP50 gene (data not shown). Therefore, among the 18 paired samples tested, 28% of the tissues expressed the TSP50 gene. At this moment, we cannot answer the question of whether activation of the TSP50 gene in cancer is a consequence or a causal factor of neoplastic growth. To find the truth, it will be necessary to perform in vitro cellular transformation and in vivo tumor induction assays.

DISCUSSION

A modified RDA technique was used to study genetic alterations by using breast cancer biopsies as a working model system. As a result, two hypomethylated genomic DNA fragments were successfully isolated. The extensive study of one of the two fragments, BR50, is the subject of this report.

It has been reported that aberrant DNA methylations occur constantly in human tumors (9, 10, 11, 12 , 24 , 26) . DNA hypomethylations could activate oncogenes, whereas DNA hypermethylation could inactivate recessive oncogenes. Both events could result in neoplastic growth (27, 28, 29, 30, 31, 32, 33, 34, 35) . The correlation between aberrant DNA methylations and malignancies suggests that differentially methylated fragments in tumors isolated by a modified RDA technique could be a valuable tool in the search for genes that might be related to cancer development. BR50 was considered to hold such value because it not only detected DNA hypomethylations in the original breast cancer tissues from which it was isolated but also detected DNA hypomethylations in other breast and ovarian cancer samples.

Our first step in processing the gene search was to screen a human genomic phage library. A 17-kbp DNA fragment was isolated, and sequence analysis suggested that this fragment contained at least two exons that were homologous to mammalian proteases. The sequencing information of the exons led to the discovery of a 974-bp gene fragment from a human cDNA library panel by PCR amplification. To obtain a full-length gene, a Northern evaluation on 16 different types of human RNAs was performed. The results demonstrated that the target gene was specifically expressed in human testes tissue. This information secured the isolation of an intact gene, TSP50, by screening a human testes cDNA library. The sequence analysis revealed that the TSP50 gene encodes a protein that shares ∼40% identity with mammalian proteases, such as human tryptase or mouse serine protease. This would suggest that the product of the TSP50 gene is a protease. However, at this point, we do not know the physiological function(s) of this protease. One may assume, though, that it could be a component in the human reproductive pathway due to it being solely expressed in the testes.

It is common knowledge that the expression of many tissue-specific genes is regulated by DNA methylations, which usually modify the promoter, or sometimes, 3′ regions (3 , 4 , 22, 23, 24 , 36) . Our preliminary results, although only obtained from analyzing the DNA methylation status of the 3′ flanking region of the gene, have proven that TSP50 is one of those tissue-specific genes. It will be interesting to discover whether the promoter region of the gene is also methylated when the corresponding sequence information is available. The HpaII and MspI methylation-sensitive Southern analysis of the 3′ region of the TSP50 gene demonstrated that, in HpaII-digested DNAs, probe BR50 hybridized two bands in the testes tissue but none in the other samples. The lower band, which was the same size as the probe, represented the unmethylated DNA pattern, whereas the upper band obviously contained the internal HpaII recognition site(s), which remained methylated. In MspI-digested DNAs, the upper band was dominant in most tissues, whereas in the testes, the lower band was dominant. These results suggest that the GGCCGG end of BR50 was methylated in other tissues but not in the testes. The DNA methylation patterns observed in both blots are probably allelic orientated. It seems that DNA hypomethylation was accompanied by the expression of the gene in the testes, and conversely, DNA hypermethylation was accompanied by the silencing of the gene in other tissues. The correlation between DNA methylation and gene expression provided additional proof that DNA methylation could be an important mechanism in governing the expression of the genes in various differentiated human cells (12 , 37 , 38) . In addition, the differential expression of the TSP50 gene has been tested in 18 paired breast cancer biopsies. Our findings have shown that this gene was activated in five cancer samples. In the near future, more samples from different types of cancer will be examined, and the possibility that the TSP50 gene product might be one of the factors that stimulate human cancer will be further explored.

Recently, by using the same technique, DNA fragments that represent DNA amplifications, deletions, and rearrangements were also obtained (data not shown). Hopefully, this technique will lead to the discovery of additional novel genes that may be related to cancer development. On the basis of our experience, the process of isolating the TSP50 gene was made considerably easier by the modified technology, where the MspI enzyme was used as the master enzyme. The ability of MspI to recognize GC-rich sequences and its sensitivity to DNA methylation (17 , 18) apparently accelerated our gene search. Furthermore, the double enzyme cleavage strategy provides another unique and efficient feature for this technique because ∼40% of a human genome can theoretically be analyzed by a single master enzyme when it is combined with a different partner enzyme.