Molecular Cloning of a Candidate Tumor Suppressor Gene, DLC1, from Chromosome 3p21.3¹ Free

The short arm of chromosome 3 is thought to include multiple tumor suppressor genes, because one allele of this chromosomal arm often has been lost in carcinomas of various tissues. Toward isolation of the putative tumor suppressor gene(s), we earlier performed detailed deletion mapping of this chromosomal arm using dozens of polymorphic DNA probes and a large number of primary cancer materials. That effort identified candidate regions at 3pter–p25, p22-p21.3, and p21.1-p14 (1, 2, 3). We subsequently found a homozygous deletion at 3p22-p21.3 in a lung cancer cell line (4) and performed large-scale genomic DNA sequencing of this 685-kb region because at least one tumor suppressor gene was likely to be present there (5). All four genes identified in the homozygously deleted region were subsequently excluded as candidates for tumor suppressor functions (6). However, because the homozygous deletion might have exerted a positional effect on expression of genes in the close vicinity, we extended our DNA sequencing further and have been characterizing genomic structures within a 515-kb segment lying distal to the deleted region (7, 8, 9, 10, 11). Here, we report identification of a possible candidate gene, DLC1,³ that showed aberrant splicing patterns in one-third of the carcinomas of esophagus, lung, and kidney we examined. The DLC1 cDNA exerted growth-suppressive activity in vitro.

Fourteen human esophageal carcinoma cell lines [TE series: gifts from Dr. Tetsuro Nishihira, Tohoku University (Miyagi); Ref. 12], six lung cancer cell lines [LC319, a gift from Dr. Takashi Takahashi, Aichi Cancer Center (Aichi); A549, NCI-H23, -H226, -H460, -H522, gifts from Dr. Takao Yamori, Cancer Institute (Tokyo)], and two renal cancer cell lines (RXF631L and ACHN, gifts from Dr. Takao Yamori) were grown in monolayers in RPMI 1640 supplemented with 5–10% fetal bovine serum.

Tumors and corresponding normal tissue samples were obtained from a total of 48 patients with NSCLCs and 10 patients with primary esophageal squamous cell carcinomas, during surgery at the Cancer Institute Hospital (Tokyo) or the Osaka Medical Center for Cancer and Cardiovascular Diseases (Osaka). Of the 48 lung cancers, 35 were adenocarcinomas and 13 were squamous cell carcinomas. Total RNA was extracted from each of the 22 cell lines, from 30 of the frozen paired lung tissues, and from all 10 paired esophageal specimens, using TRIzol Reagent (Life Technologies, Inc.), according to the manufacturer’s protocol. Extraction of DNA from the cancer cell lines and primary tissue samples was carried out as described previously (13).

Five cosmid clones (306, 308, 603, 602, and 594; Fig. 1), each of which contains part of the genomic DNA of chromosome 3p21.3 in YAC936c1, were completely sequenced by means of the shot-gun method detailed previously (6). We analyzed genomic DNA sequences from the target region with an exon-prediction computer program, GRAIL2 (14), and performed an exon-connection experiment (15) by RT-PCR to investigate whether the predicted candidate exons were actually transcribed. The RT-PCR technique was performed as described previously (7). Then we screened human testis cDNA libraries (nearly 1 million plaques) using the exon-connected product as a probe and obtained a full-length cDNA.

Human multiple-tissue blots (Clonetech) were hybridized with a fragment of DLC1 cDNA, labeled by random-oligonucleotide priming. Prehybridization, hybridization, and washing were performed according to the supplier’s recommendations. The blots were autoradiographed and analyzed with a BAS 1000 image analyzer (FUJI).

To achieve c-myc-tagged DLC1, we constructed pcDNA3.1(+)/DLC1SS that contained c-myc epitope sequences (LDEESILKQE) at the COOH-terminal of the DLC1 protein and transfected to COS-7 cells. Transiently transfected COS-7 cells replated on chamber slides were fixed with PBS containing 4% paraformaldehyde, then rendered permeable with PBS containing 0.1% Triton X-100 for 3 min at 4°C. Cells were covered with blocking solution (2% BSA in PBS) for 30 min at room temperature to block nonspecific antibody-binding sites. Then the cells were incubated with a mouse anti-c-myc antibody (diluted 1:800 in blocking solution). Antibodies were stained with a goat anti-mouse secondary antibody conjugated to rhodamine and viewed with an ECLIPSE E800 microscope (Nikon). To confirm the expression of DLC1/c-myc-tagged protein in transfected cells, we also performed Western blotting, in a manner described previously (16).

SSCP analysis was performed to screen tumors for genetic alterations in the DLC1 gene. DNA samples extracted from 35 adenocarcinomas and 13 squamous cell carcinomas of the lung were used as templates. In 31 of the 48 cases, LOH was confirmed at 3p21.3 using either DNA marker D3S685 or F56-CA1, a microsatellite located within the DLC1 gene. Primers to amplify each exon were end-labeled with [³²P]ATP, and the PCR products were analyzed by electrophoresis in 5% acrylamide gels containing 5% glycerol.

Comparative RT-PCR was performed as described elsewhere (17). The cDNAs obtained from all 22 cancer cell lines mentioned above, 30 of the lung cancers, all 10 esophageal cancers, and samples of normal tissue adjacent to each primary tumor served as templates for the PCR in a thermal cycler (Perkin-Elmer). Six primer sets were designed to amplify the entire coding region of DLC1 cDNA, from nucleotides 23 to 5373.

Tumor materials that showed significant reduction of expression of normal DLC1were examined for 5′ CpG island methylation of the DLC1 gene by PCR-based assay. Genomic DNA digested with one of the methylation-sensitive restriction enzymes (AccII, CfoI, HaeII, or HapII) was PCR amplified according to the methods described previously (18). A methylation-insensitive restriction enzyme, MspI, was used as a control enzyme.

Plasmids designed to express DLC1 were constructed by cloning the entire coding region of DLC1 cDNA into the pcDNA3.1(+) vector (Invitrogen), which carries a cytomegalovirus promoter and a gene conferring resistance to neomycin (G418). We constructed two plasmid clones, one that was designed to express a sense transcript [pcDNA3.1(+)/DLC1SS], and the other to express an anti-sense transcript [pcDNA3.1(+)/DLC1AS] that inserted the same cDNA in the opposite direction. In addition, we constructed a plasmid clone designed to express the sense strand of an aberrant transcript, the predicted protein of which was truncated at exon 13 [pcDNA3.1(+)/DLC1T].

The cancer cell lines used for colony-formation assays included human esophageal cancer line TE14, renal cancer lines RXF631L and ACHN, and lung cancer lines LC319 and NCI-H23. Each of these five cell lines was plated in 25-cm² flasks (2 × 10⁵ cells/flask) and transfected at 24 h later with the plasmid [pcDNA3.1(+)/DLC1SS] designed to express the sense strand of the DLC1 cDNA. As controls, we also transfected the pcDNA3.1(+) vector alone, the pcDNA3.1(+)/DLC1AS, or the pcDNA3.1(+)/DLC1T to each of the five cell lines. Five μg of plasmid DNA and 25 μg of TransIT-LT1 (PanVera Corp.) were used for each transfection. Cells were diluted 1:8 after 6 h of transfection and cultured for 14 days in the presence of 400–800 μg/ml of geneticin (G418). The transfection experiments were independently repeated three times.

GRAIL2 computer analysis of DNA sequences in the region covered by cosmid clone 594 (Fig. 1) predicted seven possible exons with “excellent” scores. To examine whether these candidate exons were transcribed in human tissues, we synthesized oligonucleotides corresponding to the possible exon regions and performed exon-connection experiments. We confirmed by RT-PCR that the seven exon candidate segments were parts of the same transcript and temporarily designated the 800-bp PCR product as 594E17. Northern blot analysis using 594E17 as a probe revealed that 6.0- and 8.0-kb transcripts were expressed in all human tissues examined, including lung and kidney; prostate and testis seemed to express this gene more abundantly than other tissues (data not shown). The 6.0-kb transcript of the DLC1 gene was expressed more abundantly (2–3-fold) than the 8.0-kb transcript.

Because the 594E17 PCR product was smaller than either of the transcripts indicated by Northern analysis, we screened a human testis cDNA library and identified cDNA sequences consisting of 5615 nucleotides, including an open reading frame of 5265 nucleotides encoding a 1755-amino acid peptide (Fig. 2). A comparison of genomic sequences with cDNA sequences revealed that the 6.0-kb transcript spanned ∼59 kilobases of genomic DNA and consisted of 37 exons (nucleotide sequences of cDNA and exon-intron boundaries are not shown here but are available from GenBank; accession numbers are AB020522 and AB010443, respectively).

Analysis using the FASTA program (19) indicated that the M_r 166,000 protein sequence showed no significant homology to any known proteins in the public database. However, the PROSITE database (20) identified three possible cyclic AMP-dependent protein kinase phosphorylation sites, 24 protein kinase C phosphorylation sites, 27 casein kinase II phosphorylation sites, 12 N-glycosylation sites, and 15 N-myristoylation sites. Because this gene was isolated from the commonly deleted region at 3p21.3 defined by the LOH study of lung cancers and its gene product was able to suppress growth of some cancer cells (see below), we designated the novel gene DLC1.

To investigate the cellular localization of DLC1 protein in mammalian cells, we transfected COS-7 cells with pcDNA3.1(+)/DLC1SS, a plasmid that contained c-myc epitope sequences (LDEESILKQE) at the COOH-terminal of DLC1 protein. We confirmed expression of DLC1 in transfected cells by immunoblotting. After transient expression of the DLC1SS/c-myc protein in COS-7 cells, the proteins were extracted and separated by SDS-PAGE. The M_r 166,000 DLC1SS/c-myc protein was detected by Western blot analysis using anti-c-myc antibodies (Fig. 3,a). Using the same antibodies, we detected the DLC1SS/c-myc protein mainly in cytoplasm, when the COS-7 cells were transfected with the DLC1 expression plasmid (Fig. 3 b).

Because the DLC1 gene is located within a region that is deleted in many lung cancers, we considered that mutated forms might contribute to the etiology of lung carcinoma. To investigate this possibility, we performed SSCP analyses involving all coding exons in 48 NSCLCs, including 31 that had shown LOH at the DLC1 locus on 3p21.3. Apart from a few probable polymorphisms, we detected no genetic alterations that should cause dysfunction of the gene product.

To look for abnormalities in DLC1 transcripts from the cancer materials that showed frequent LOH at 3p21.3, we reversely transcribed the mRNAs from the 22 cancer cell lines (14 esophageal, 6 lung, and 2 renal) as well as from tumors and corresponding normal tissues of the lung, esophagus, and kidney. We amplified the entire coding region of DLC1 cDNA in each case using six primer sets and sequenced the PCR products. In addition to the cDNA corresponding to the 6.0-kb transcript, we detected six alternatively spliced transcripts in normal lung, esophagus, and kidney tissues (Fig. 1 D). Among them, three (DLC1-N2, -L1, and -L2) included an additional exon corresponding to part of intron 12 or 13 or all of intron 13; these transcripts had involved aberrant splice acceptor/donor sites. The largest transcript, DLC1-L1 contained an additional exon between exons 13 and 14, and its open reading frame was disrupted within this additional exon. The 8.0-kb transcript detected on Northern blots seemed to correspond to the DLC1-L1 by the following criteria: (a) about 2–3-fold difference in expression level between normal transcript (DLC1-N1) and DLC1-L1 observed by RT-PCR was compatible with that between 6.0-kb and 8.0-kb transcript; and (b) no cDNA clone corresponding to the larger transcript except for DLC1-L1 has been obtained by screening human cDNA libraries. The remaining three alternative transcripts (DLC1-S1, -S2, and -S3) lacked exons 11, 13, or 36, respectively. The proteins predicted by all of the aberrant transcripts except DLC1-N2 would be truncated close to the alternatively spliced sites.

We subsequently examined these various transcripts in cancer cell lines and found that the patterns varied; some lines apparently expressed no normal transcript, and some showed a significant reduction or total absence of any DLC1 expression. Esophageal cancer-derived TE2, TE7, and TE13 expressed no DLC1 at all; TE15 lacked a normal product. TE1 and TE14 showed significant reduction of expression of normal DLC1 transcript in comparison to the other types of transcript. Renal cancer-derived RXF631L did not express the DLC1 gene in any form, and ACHN showed only aberrant products (Fig. 4 a).

We also examined DLC1 transcripts in 30 of the primary NSCLCs in our panel and their adjacent normal lung tissues and found abnormal expression patterns in 11 of the tumor samples. Of those, eight produced no detectable DLC1 transcripts. In the remaining three cases, no normal transcript was detected, but aberrant transcripts were (Fig. 4 b). Furthermore, 3 of the 10 primary esophageal carcinomas examined lacked normal products (data not shown). However, analysis of genomic sequences with SSCP and direct DNA sequencing of lost exons in the cell lines or primary tumors that exhibited abnormal patterns of DLC1 transcripts revealed no alterations in the gene itself.

To investigate the possibility of hypermethylation-based inactivation of the DLC1 gene, we examined 20 cancer materials that showed significant reduction of expression of normal DLC1 transcript for 5′ CpG island methylation of the DLC1 gene by PCR-based assay. However, we detected no 5′ CpG island methylation in the tumor samples examined (data not shown).

We performed colony-formation assays to investigate whether the DLC1 gene can act as a growth suppressor in transfected cells. Among the five cell lines tested, the number of geneticin-resistant colonies was significantly reduced in dishes containing TE14 (18.3% in an average of three independent experiments), RXF631L (19.6%), ACHN (40.6%), or LC319 (36.2%) cells that had been transfected with the sense-strand of cDNA corresponding to the normal transcript, in comparison to cells transfected with the mock vector or with plasmids designed to express anti-strand cDNA or a sense-strand cDNA encoding a protein truncated at exon 13 (Fig. 5). However, no difference in the colony numbers was observed when we transfected these plasmids to NCI-H23. NCI-H23 expressed the functional transcript of the DLC1 gene relatively abundantly (Table 1).

The DLC1 cDNA encodes a novel M_r 166,000 protein. Northern blot analysis detected 6.0- and 8.0-kb transcripts in several normal human tissues, including lung and kidney, although prostate and testis seemed to express this gene more abundantly than other tissues. The minor 8.0-kb transcript detected seemed to correspond to the largest transcript, DLC1-L1, which contained an additional exon between exons 13 and 14, and its open reading frame was disrupted within this additional exon (Fig. 1). Because the predicted amino acid sequence of DLC1 has no significant homology to known proteins or domains, we were unable to speculate on the function of this protein. However, the PROSITE database identified a total of 54 putative phosphorylation sites including 27 casein kinase (CSNK) II phosphorylation sites, 12 putative N-glycosylation sites, and 15 putative N-myristoylation sites. CSNK II is a ubiquitous, highly conserved enzyme consisting of subunits α, α-prime, and β. A ubiquitous, messenger-independent serine/threonine kinase, CSNK II is localized in both the cytoplasm and the nucleus and functions as a protease (21, 22, 23). The kinase domain of human CSNK I delta, a serine/threonine-specific protein kinase, may function in DNA metabolism through excision and recombinational repair (24). Using immunostaining, we determined that DLC1 is localized only in cytoplasm. Therefore, the DLC1 protein may act as a downstream gene in the serine/threonine kinase pathway.

Lung cancers exhibit multiple genetic lesions including mutations activating the dominant oncogenes in the myc and ras families, as well as those inactivating the tumor suppressor genes. With respect to tumor suppressor genes in NSCLCs, LOH studies suggested possible involvement of multiple tumor suppressor genes on chromosomal arms 1p, 1q, 3p, 5q21, 9p, 9q, 13q, 16q, and 17p that include the loci containing the p53, RB, or p16 genes (13). At present, several different regions of chromosome 3p are considered to contain tumor suppressor genes: 3p12, 3p14, 3p21.3, and 3p25. One candidate, the FHIT gene, was isolated from 3p14.2 (25) and another, the von Hippel-Lindau (VHL) disease gene, was isolated from 3p25 (26). To our knowledge, several genes that have reduced expression but no mutations in small cell lung cancers or other tumors were identified on chromosome 3p21. Except for BAP1(27), most of them (28, 29, 30, 31, 32, 33), such as PTPG, UBEIL, semaphorin family [III/F, IV, A(V)], UNP, and ACY1, were not analyzed by detailed LOH studies or re-introducing gene products into malignant cells. We previously observed LOH on 3p21.3 in 22 of 27 (81%) squamous cell carcinomas and 34 of 86 (40%) adenocarcinomas of the lung (13). We have now identified and characterized a novel transcriptional unit, DLC1, lying within the commonly deleted region at 3p21.3 defined by that LOH study of lung cancers (3, 4). The frequent appearance of aberrant transcripts of DLC1 in lung, esophageal, and renal cancers and the fact that introduction of the cDNA corresponding the normal transcript of DLC1 into several cancer cell lines caused significant suppression of growth suggest that aberrant patterns of DLC1 transcription may play important roles in carcinogenesis of those tissues.

The proportion of nonfunctional RNA transcripts (36%) in lung cancer materials was less than the frequency of LOH at 3p21.3. We selected 48 tumor samples from many NSCLCs, which had been analyzed for LOH and considered to be less contaminated with normal cells than the others; however, in some of these cases, the normal transcripts derived from admixed normal cells could still have been detected in the RT-PCR amplification. Hence, the frequency of loss of the normal DLC1 transcripts in noncultured tumor samples is likely to be underestimated.

Tumor suppressor genes can be inactivated by genetic or epigenetic changes. Genetic changes may consist of: (a) mutations in regulatory regions that cause elimination or suppression of mRNA expression; (b) deletions of part of or an entire gene; and (c) missense, nonsense, frameshift, or splice-site mutations resulting in absence of a functional protein. The causes and mechanisms involved in epigenetic changes, such as abnormal methylation, deregulation of imprinting, or aberrant splicing, are still not well understood. However, dysfunction of a gene occasioned by epigenetic changes is often observed in cancer cells. The abnormal transcription patterns of DLC1 in the tumors examined here can be divided into two categories: (a) absence of normal transcript but presence of transcripts lacking one exon (exon 11 or 13) or containing an additional exon corresponding to part or all of intron 13, both mechanisms that disrupt the intact open reading frame; and (b) complete absence of DLC1 transcripts of any kind. In this study, we detected alteration of DLC1 expression in 20 of our primary cancer materials; among them, 15 were classified as category (b) and five belonged to category (a), although no genetic alterations were detected in 5′ noncoding region, exons, or flanking introns, and no hypermethylation of the 5′ CpG island was detected in any of these cases. The mechanism that caused the loss of transcript(s) or aberrant splicing detected in our experiments remains unclear. However, we suggest several possible explanations for category (b) alterations: (i) mutations may have occurred within intronic sequences or in the 3′ noncoding region of DLC1; or (ii) genetic or epigenetic mechanisms not yet identified might have inactivated this gene.

Most of the tumor suppressor genes described to date are inactivated by global loss of a region containing one allele (detected by LOH) and subtle mutation in the other. Genes that are commonly inactivated by such “two-hit” events have been termed class I tumor suppressors (34). Other genes, such as the protease inhibitor maspin, have been proposed as tumor suppressors of class II on the basis of low expression in tumors combined with a decrease in tumorigenicity when the normal cDNA is transfected into cancer cell lines. On the basis of these criteria, DLC1belongs to class II.

Some candidate tumor suppressors were identified from many chromosomal regions; however, it is especially difficult to determine which of several genes within a minimal region of LOH or homozygous deletion may function as a tumor suppressor gene, because most of them did not fulfill the criteria of classical tumor suppressor. Haber and Harlow (35) drew attention to an increasing vagueness in the use of the term “tumor suppressor gene.” They suggested that the simplest, most inclusive, and cleanest genetic definition would be “genes that sustain loss-of-function mutations in the development of cancer.” At the simplest level, mutations associated with cancer can be split informatively into gain-of-function and loss-of-function mutations. Additional studies might be required to confirm whether DLC1 meets this simpler definition.

p27^Kip is a candidate tumor suppressor protein, because it blocks cell proliferation and abnormally low levels of the p27 protein are frequently found in human carcinomas. However, only rare instances of homozygous inactivating mutations of the p27 gene have been found in human tumors. Recently, Fero et al. (36) showed that both p27 nullizygous and p27 heterozygous mice were predisposed to tumors in multiple tissues when challenged with γ-irradiation or a chemical carcinogen. Molecular analyses of tumors in the p27 heterozygous mice show that the remaining wild-type allele is neither mutated nor silenced. Hence, p27 is haplo-insufficient for tumor suppression. The assumption that null mutations in tumor suppressor genes are recessive excludes those genes that exhibit haplo-insufficiency. Further proof that DLC1 is a tumor suppressor gene might need to demonstrate that the abolition of this gene expression contributes to a tumorigenic phenotype in vitro and in vivo.

We thank Drs. T. Nishihira, T. Takahashi, and T. Yamori for esophageal, lung, and renal cancer cells.

Note Added in Proof

The total genomic sequence of the 1200-kb region depicted in Fig. 1 is now available on the Internet (www-alis.tokyo.jst.go.jp/HGS/team_GK/3p21.3/map.html), and the 59-kb genomic sequence of DLC1 will appear in the DDBJ, EMBL, and GenBank databases with accession number AB010443.