Two somatic biallelic lesions within and near SMAD4 in a human breast cancer cell line

Loss of chromosome arm 18q is a common event in human pancreatic, colon, and breast cancers and is often interpreted as representing loss of one or more tumor‐suppressor genes. In this article, we describe two novel biallelic deletions at chromosome band 18q21.1 in a recently characterized human breast cancer cell line, HCC‐1428. One lesion deletes a fragment of approximately 300 kb between SMAD4 and DCC that encodes no known genes. The second lesion is an in‐frame SMAD4 deletion (amino acids 49–51) that affects the level of SMAD4 protein but not the SMAD4 message. This change accelerates 26S proteasome–mediated degradation of both endogenous and exogenous mutant SMAD4. Examination of normal DNA from the same patient demonstrated that both lesions are somatic and associated with loss of both normal alleles. These data support the concept that two independent tumor‐suppressor loci exist at chromosome segment 18q21.1, one at SMAD4 and the other potentially at an enhancer of DCC or an unrelated novel gene. © 2005 Wiley‐Liss, Inc.


INTRODUCTION
Breast cancer was responsible for the deaths of an estimated 40,200 women in 2003 just in the United States (Ries et al., 2003). Unfortunately, many of the genetic causes of this devastating disease remain unknown. Cancer geneticists have long labored to discover the tumor suppressors and oncogenes altered in breast carcinoma. One approach to identifying novel tumor suppressors has been to characterize biallelic deletions (also known as homozygous deletions) in the genomes of cancer cell lines and xenografts. This strategy proved indispensable for the positional cloning of tumor-suppressor genes including RB1 (Lee et al., 1987), CDKN2A (Kamb et al., 1994), SMAD4 (Hahn et al., 1996b), and PTEN (Li et al., 1997). With the human genome project now complete, biallelic deletions should prove easier to map and candidate tumor-suppressor genes easier to identify.
This study employed RDA in a screen for biallelic deletions in the genome of the recently isolated breast cancer cell line HCC-1428. RDA uncovered one novel biallelic deletion in the HCC-1428 genome, on the 18q21.1 chromosome segment. We were able to demonstrate that the biallelic deletion does not alter the open-reading frames (ORFs) of the SMAD4 or DCC candidate tumor-suppressor genes, also on chromosome segment 18q21.1. We mapped the extent of the deletion, determined its length as 260-330 kbp, and localized it as being between the SMAD4 and DCC loci. Sequence analysis of the SMAD4 gene in the HCC-1428 cell line identified only an allele with a novel in-frame and protein destabilizing mutation whose predicted effect was to delete amino acids 49-51 of the MH1 domain. The existence of a second biallelic deletion of approximately 300 kilobase pairs (kbp) between the SMAD4 and DCC genes suggests the potential to select for loss of an additional tumor-suppressor locus resident on chromosome segment 18q21.1.

Cell Lines
Human breast carcinoma cell lines MDA-MB-468, MDA-MB-415, MDA-MB-474, MDA-MB-453, MCF7, and T-47D, human colorectal carcinoma cell line SW-480, human glioblastoma cell line U-87, and the African green monkey kidney, SV40-transformed COS-1 cell line were acquired from the American Type Culture Collection (ATCC) (Manassas, VA) and cultured according to the distributor's instructions. Human mammary epithelial cells (HMEC) were purchased from Clonetics (San Diego, CA) and cultured per the distributor's instructions. The Gazdar laboratory isolated breast carcinoma cell lines HCC-1428, HCC-1428BL, HCC-1937, and HCC-1143 they are now available from the ATCC (Gazdar et al., 1998).

Representational Difference Analysis
Representational difference analysis (RDA) was performed essentially as described by Lisitsyn and Wigler (1993). Specifically, the HCC-1428 genome served as the ''driver genome,'' or template for the driver; the HCC-1428 BL genome as the ''tester genome,'' or template for the tester. BglII restriction endonuclease (New England Biolabs, Beverly, MA) was employed to construct the driver and the tester. Our RDA methodology did differ from the original protocols in the ratio of driver to tester that was used for subtraction-hybridization. To set up the first-round subtraction-hybridization reaction, driver and tester were mixed in an 80:1 ratio (40 mg driver:500 ng tester). For the second-round subtraction-hybridization, an 800:1 ratio was employed (40 mg driver:50 ng tester). In the third-round subtraction-hybridization, driver and tester were mixed in a 400,000:1 ratio (40 mg driver:100 pg tester).

Screening Products of Representational Difference
Analysis of the HCC-1428 Genome for Deleted Sequences RDA products were cloned into the pZeroII vector (Invitrogen, Carlsbad, CA) using the BamHI site, electroporated into DH10B E. coli. (Invitrogen), and 800 colonies were selected for sequencing. We performed all sequencing in this study on the ABI Prism 377 DNA sequencer (Applied Biosystems, Foster City, CA). In a PCR screen for deleted fragments, the HCC-1428 BL genome was used as the template for a positive control reaction.

Polymerase Chain Reaction
All polymerase chain reactions (PCR) in this study were performed under the following conditions unless otherwise indicated: 10 mL volume with 16.6 mM (NH) 2 SO 4 , 67 mM Tris-HCl (pH 8.8), 6.7 mM MgCl 2 , 10 mM 2-mercaptoethanol, 0.25 mM dNTPs, 6% DMSO, 350 ng of each primer, and 0.5 U of Platinum Taq (GibcoBRL, Carlsbad, CA). PCR reactions were performed on 10 ng of genomic DNA. All PCR reactions were carried out on a Hybaid Omnigene Thermocycler (Thermo Electron Corporation, Waltham, MA). PCR primers used in this study are available on request.

Bacterial Artificial Chromosome Isolation and End Sequencing
The sequence-tagged site (STS) SHGC-7133 was used to PCR-screen the Human Bacterial Artificial Chromosome Release II Library (Genome Systems, St. Louis, MO). We used a NucleoBond Plasmid Kit (Clontech, Palo Alto, CA) to isolate bacterial artificial chromosomes (BACs) 494J9 and 471M4. SP6 and T7 primers were used to sequence BAC ends.

Sequence-Tagged Site Mapping
PCR conditions and primers for sequencetagged sites used in this study are available on request as are STS coordinates on the NCBI chromosome 18 contig (accession no. NT_010966.13, October 17, 2003, version). PCR products of novel STSs probe B, 84-2, 494J9-T7, rpS8-like EST, 800-20K, and 494J9-SP6 were sequenced directly in order to confirm their identities. The control PCR fragment was from chromosome 2, NCBI accession number AC012305.7, coordinates 78,323-78,502.

Single-Nucleotide Polymorphism Analysis
Candidate single-nucleotide polymorphisms (SNPs) in the deleted region were identified with the NCBI dbSNP database and PCR-amplified from the HCC-1428 BL genome. Sequence analysis of candidate SNP PCR products demonstrated that rs2615548 was present in two variants in the HCC-1428 BL genome. An ambiguous nucleotide (peaks at both G and T), referred to as N, was detected at the predicted rs2615548 SNP position. To confirm that both the G and T alleles were amplified from the HCC-1428 BL genome, this PCR product was cloned into the pZeroII vector, and individual clones were sequenced. The G and T alleles were detected with equal frequency. Results were confirmed with a second rs2615548 SNP PCR amplification from the HCC-1428 BL genome and subsequent product sequencing. The control PCR fragment also was from chromosome 2, NCBI accession #AC012305.7, coordinates 78,323-78,502.

Southern Blot Analysis
For Southern blot analysis, restriction-digested DNA was electrophoresed through a 0.8% agarose gel and transferred to Zeta-Probe GT membranes (Bio-Rad, Hercules, CA). Probes were labeled with 32 P dCTP by the random hexamer method (Feinberg and Vogelstein, 1983). Probe hybridization to membrane was carried out overnight at 688C in 10 mL of ExpressHyb (Clontech); membranes were then washed for 15 min at 68C in 0.3Â SSC/0.1% SDS, and the results were visualized with the Storm 840 phosphoimage system (Molecular Dynamics, Sunnyvale, CA). Probe A was derived from the NCBI chromosome 18 contig, accession #NT_010966.13, coordinates 30,787,107-30,786,164. Probe B also was derived from the NT_010966.13 contig, coordinates 30,682,681,194. The cytogenetic locations of probes A and B relative to BID-CHR18 are indicated in Figure 1A. Both probe A and probe B were synthesized by PCR and sequenced to confirm their identities.

Northern Blot Analysis
Total RNA was isolated from the indicated cell lines according to the cesium chloride/guanidium isothiocyanate method (Sambrook et al., 1989). RNA was resolved on a 1Â MOPS gel by electrophoresis and transferred onto a NYTran membrane (Schleicher and Schuell BioScience, GmbH, Dassel, Germany). DNA probes were labeled with 32 P dCTP by the random hexamer priming method. Blots were hybridized at 428C in UltraHyb (Ambion, Austin, TX). The membrane was washed in 0.1Â SSC/0.1% SDS at 608C and exposed to film at À708C using intensifying screens. After SMAD4 hybridization and exposure, the membrane was stripped by boiling for 30 sec in 0.1Â SSC/0.1% SDS before being hybridized with the beta actin probe. The SMAD4 probe was constructed by EcoRI and PstI restriction digest of the SMAD4-FLAG plasmid and isolation of the 300-bp fragment at the 3 0 end of the SMAD4 ORF. The beta actin probe was described previously (Wu et al., 1999).

Reverse Transcription Polymerase Chain Reaction
cDNA was synthesized from 5 mg of the whole RNA template with random primers and Super-Script II RNase H À Reverse Transcriptase (GibcoBRL) according to the manufacturer's instructions. Human breast RNA was acquired from Invitrogen, human fetal brain RNA from Stratagene (La Jolla, CA). Professor V. Band, of the Radiation Oncology Department of New England Medical Center, provided RNA from the human mammary epithelial cell cultures 70N and 76N and the transformed breast cell cultures 70E6 and 76E6 (Wazer et al., 1995). PCR primers hed1R and hed3F amplify a cDNA product spanning exons 2-4 of the DCC transcript. Cyclophilin A was chosen as a housekeeper control, as a recent report indicated it is present at low copy number in both normal and cancerous breast tissue (Tricarico et al., 2002). Primers CYCA-F1 and CYCA-R1 amplify a cDNA product that includes exons 1-5. The cyclophilin A and DCC reverse transcription polymerase chain reaction (RT-PCR) products were sequenced to confirm their identities.

Pulse Chase Analysis
COS-1 cells at 50% confluence were transfected with either the SMAD4-FLAG or the D(49-51)-SMAD4-FLAG plasmids. For transfection, 5 mg of plasmid and 60 mL of lipofectamine (Invitrogen) per 75 cm 2 flask were used; each 75-cm 2 flask of cells represented a time point in the pulse-chase experiment. Forty-eight hours posttransfection, the cells were placed in L-methionine-and L-cystinefree DMEM and 5% dialyzed fetal bovine serum with 35 S-labeled L-methionine and L-cystine at 180 mCi/mL for 2 hr at 378C/5% CO 2 (Amersham Biosciences, Piscataway, NJ). Chase with cold L-methionine and L-cystine was carried out for the specified periods. Sequential immunoprecipitation with M2-FLAG antibody agarose beads was performed twice before the labeled protein was resolved on a 4%-20% Tris-glycine gel according to standard protocols (Struhl, 2001). After drying, the gels were placed between 6-mm-thick single layers of Standard Mylar(Somar International Inc., Sparks, NV), exposed for 5-7 days to the LE storage phosphoimage screen (Amersham Biosciences), and the images processed with the Storm 840 system.

Fluorescence In Situ Hybridization
Fluorescence in situ hybridization (FISH) was performed by standard methods on metaphase cells prepared from the HCC-1428 cell line. A Spectrum Orange-labeled CEP18 centromeric probe and a SpectrumGreen-labeled chromosome 18 paint probe (both obtained from Vysis, Downers Grove, IL) was used in dual-color FISH. Fluorescence signals were captured after a DAPI counterstain on an Applied Imaging Cytovision Imaging system attached to a Nikon Eclipse 600 microscope.
To determine whether this novel deletion represented the somatic loss of two alleles, sequence analysis of candidate SNPs was performed. Sequence analysis of the HCC-1428 BL (paired normal) PCR products of candidate SNP rs2615548 revealed two alleles, G and T. Candidate SNP rs2615548 did not amplify from the HCC-1428 genome (Fig. 1C). The deleted locus was named biallelic deletion of chromosome 18, BID-CHR18.

BID-CHR18 Spans 260-330 Kilobase Pairs and Is
Between Candidate Tumor-Suppressor Genes SMAD4 and DCC BLAST searches of the NCBI HTGS and NR databases mapped BID-CHR18 to NCBI accession number AC027216.6, at chromosome band 18q21.1. PCR radiation hybrid mapping of the 84-2 fragment with the Stanford G3 Human/Hamster RH Panel also localized the deletion to 18q21.1, within 5 cM, or about 100 kbp, of STS SHGC-7133. A physical map of the deletion was constructed with publicly available STSs and the NCBI HTGS and NR databases. The centromeric end of the deletion was determined to within 40 kbp. SHGC-105608, at AC027216.6, coordinates 109,631-109,933, did not PCR-amplify from the HCC-1428 genome, whereas stSG28163, at AC027216.6, coordinates 150,505-150,627, did amplify from the HCC-1428 genome (Fig. 1A).
To map the telomeric extent of the deletion, we performed a PCR screen of a human BAC library for SHGC-7133 isolated BAC 494J9. STS 494J9 SP6, present in the HCC-1428 genome, maps to NCBI accession number AC105032.4, coordinates 22,666-22,965. Overlap of AC105032.4 with NCBI accession number AC080051.6 and a novel STS we generated from the latter, 800-20K, localized the telomeric extent of BID-CHR18 to 27 kbp between 800-20 K and 494J9SP6 (Fig. 1A). Ultimately, sequence information from NCBI accession numbers AC027216.6, AC022701.6, and AC080051.6, as well as from the UCSC human genome server, July 2003 freeze, allowed us to conclude that BID-CHR18 is 260-330 kbp in size and between candidate tumor suppressors SMAD4 and DCC (Fig. 1A).
We performed a PCR screen of STSs SHGC-7133 and 84-2 on a panel of 40 breast cancer cell line genomes and 40 breast cancer xenograft genomes. Both SHGC-7133 and the 84-2 fragment amplified in all 80 genomes (results not shown).

Bioinformatic Analysis of BID-CHR18 Revealed No Known Genes
BID-CHR18 is unremarkable, save for its paucity of known coding sequence. No known genes or spliced expressed sequence tags (ESTs) are contained in BID-CHR18 as assessed by BLAST searches and analysis with the UCSC human genome browser, April 2003 freeze. BID-CHR18 has no known microRNA genes. We performed Paracel BLAST (Paracel Inc., Pasadena, CA) of BID-CHR18 (using accession nos. AC027216.6, AC022701.6, and AC080051.6) versus 1,138 candidate chromosome 18 exons found by trap analysis (Chen et al., 2003). This approach identified three putative exons, NCBI accession nos. BH608849, BH608723, and BH608687, within BID-CHR18.
However, repeated exon connection RT-PCR reactions on human fetal brain, fetal testis, and breast cDNA templates failed to amplify the spliced transcripts containing these sequences.
BID-CHR18 contains an unexceptional number of human-mouse orthologous regions. BID-CHR18 (April 2003 UCSC freeze) was compared to the corresponding mouse region (February 2003 UCSC freeze) with the VISTA 2.0 browser. This approach revealed 51 regions of 50 bp or more that had more than 90% nucleotide identity between human and mouse. All human-mouse hits were checked for synteny and best possible matches. The similar regions covered a combined 4,998 bp, or 1.5% of the maximum possible (*330 kbp) BID-CHR18. This is consistent with the findings of the Mouse Genome Sequencing Consortium, which concluded that roughly 2.3% of the human genome is conserved in the mouse in small segments ( 50 bp) that are not coding sequences, 5 0 UTRs, 3 0 UTRs, or repeats (Waterston et al., 2002). There is synteny of the SMAD4-(BID-CHR18)-DCC arrangement between humans and mice. However, this also was expected because 90% of the human genome shows synteny with the mouse genome (Waterston et al., 2002).

BID-CHR18 Is in the 5 0 Extension Intergenic Region of DCC
The BID-CHR18 deletion does not extend into the DCC ORF of the HCC-1428 genome (Fig. 1A). At *300 kbp, the apparently gene-barren BID-CHR18 constitutes part of a large intergenic region upstream of DCC. The 350 kbp between the first exon of DCC and BID-CHR18 lack any known genes or spliced ESTs, meaning that DCC has a 5 0 extension intergenic region of at least 700 kbp (distance from the 5 0 end to the nearest upstream gene). In contrast, a recent analysis of 5 0 extension intergenic regions found medians of 46 and 18 kbp for chromosomes 21 and 22, respectively (Chen et al., 2002).
RT-PCR analysis detected DCC transcripts in human fetal brain and adult breast cDNAs; sequence analysis of these PCR products confirmed their identity. The DCC cDNA did not amplify from the HCC-1428 cDNA, nor did it amplify from nontransformed, cultured breast epithelial cells HMEC, 70N, and 76N cDNAs, possibly because of dedifferentiation of breast cells in culture (Fig. 2F). Western blot analysis of the DCC protein failed to detect it in the HCC-1428 or other breast cancer cell lines or cultures (results not shown).
The HCC-1428 18q21.1 Band Also Contains a Biallelic Deletion at SMAD4 Northern blot analysis demonstrated that the SMAD4 message was not disrupted in the HCC-1428 cell line (Fig. 2D). However, sequence analysis of SMAD4 in the HCC-1428 genome revealed only an allele with a 9-bp deletion of exon 1 ablating amino acids 49, 50, and 51 (Fig. 2C). Sequence analysis of the entire ORF, 5 0 UTR, and 3 0 UTR of the HCC-1428 SMAD4 gene uncovered no additional mutations. The mutation was confirmed in the SMAD4 transcript of the HCC-1428 cell line. The normal SMAD4 exon 1 sequence did not amplify from the HCC-1428 genome, indicating that both normal copies of SMAD4 exon 1 had been lost from the HCC-1428 genome (Fig. 2B). In addition, Western blot analysis demonstrated that endogenous mutant SMAD4 protein was almost undetectable in the HCC-1428 proteome (Fig. 2E). This SMAD4 mutation was named, in line with the guidelines of the HUGO Mutation Database Initiative in genomic DNA, cDNA, and protein contexts, g.144_152del, c.144_152del, and p.Glu49_Lys51del, respectively (den Dunnen and Antonarakis, 2000), referred to henceforth as D(49-51)-SMAD4.

D(49-51) Mutation Enhanced SMAD4 Protein Degradation
Western blot analysis demonstrated that endogenous D(49-51)-SMAD4 protein was almost undetectable in the HCC-1428 proteome (Fig. 2E). To explore the effect of D(49-51) on the steady-state level of SMAD4 protein, we transfected normal SMAD4-FLAG and D(49-51)-SMAD4-FLAG constructs into COS-1 cells. Immunoprecipitation was needed to detect the mutant SMAD4-FLAG protein, which was found to have a much lower steady-state level than did normal SMAD4-FLAG (Fig. 3A). To determine whether D(49-51) decreased SMAD4 protein synthesis or stability, we transfected normal SMAD4-FLAG and D(49-51)-SMAD4-FLAG plasmids into COS-1 cells and performed pulse chase analysis. A 2-hr pulse labels similar amounts of normal and mutant protein. In this assay, labeled SMAD4-FLAG was stable for 18 hr, but D(49-51)-SMAD4-FLAG had a half-life of less than 4 hr (Fig. 3B).

Proteasome Inhibition Increased the Level of Exogenous and Endogenous D(49-51)-SMAD4 in HCC-1428 Cells
To characterize the mechanism of the enhanced degradation of D(49-51), COS-1 cells transfected with normal SMAD4-FLAG and D(49-51)-SMAD4-FLAG were incubated with MG-132, a 26S proteasome inhibitor. A concentration of 50 mM was needed to observe even modest stabilization at 4 hr (Fig. 3A). At this concentration, MG-132 is toxic to COS-1 cells, leading to a dramatic decrease in normal SMAD4-FLAG after more than 4 hr (not shown). However, 24-hr treatment of HCC-1428 cells with the proteasome inhibitors MG-132 (10 mM) or PS-341 (100 nM) caused a dramatic increase in the level of endogenous D(49-51)-SMAD4 protein (Fig. 3C).
Chromosome Paint Analysis Demonstrated an Unbalanced Translocation of Chromosome Arm 18q in HCC-1428 Cells FISH, using orange chromosome 18 centromeric and green chromosome 18 paint probes, was performed on HCC-1428 to determine the structure of chromosome 18. The analysis of HCC-1428 cells identified one intact chromosome 18 in all cells and an unbalanced chromosome 18 translocation, with the proximal half of 18q detected, making a net loss of one copy of distal 18q21 material (Fig. 4). Therefore, only one allele of chromosome 18 is present in the SMAD4-BID-CHR18-DCC interval.

DISCUSSION
Frequent loss of specific chromosome arms often is interpreted as representing loss of canonical, genetic two-hit tumor-suppressor genes. Potential candidate tumor-suppressor genes for human chromosome arm 18q include SMAD4 and DCC. However, neither SMAD4 nor DCC has a frequency of mutation of both alleles approaching that of the observed LOH and CGH in breast cancer. The rate of loss of both SMAD4 alleles, as judged by immunohistochemistry, is 2% in sporadic breast cancer (Xie et al., 2002). Only one case of mutation of both SMAD4 alleles has been reported previously in breast cancer . A two-hit lesion of DCC has never been documented in breast cancer.
The relevance of DCC to tumor suppression is a topic of much controversy. DCC was first identified as a candidate tumor-suppressor gene by virtue of its position in a biallelic deletion in a colon tumor genome (Fearon et al., 1990). However, character-ization of biallelic deletions of the DCC locus in cancer demonstrated that these lesions were always intergenic, never limited to the DCC gene itself. Indeed, some reported DCC deletions include the BID-CHR18 locus (Hilgers et al., Figure 2. The HCC-1428 SMAD4 locus contains a 9-bp deletion mutation in exon 1. (A) Schematic diagram of SMAD4 gene [above, exon/intron structure; below, position of D(49-51) mutation relative to domains of SMAD4; MH1, mad homology 1 domain, responsible for sequence-specific DNA binding; MH2, mad homology 2 domain, involved in heteromerization and transactivation (Miyaki and Kuroki, 2003); amino and carboxy terminal residues indicated as 1 and 552, respectively]. (B) PCR of SMAD4 exon 1 from HCC-1428 genome amplifies only a truncated product. (C) Sequence analysis reveals a 9-bp deletion from exon 1 of SMAD4 of HCC-1428 (nucleotides deleted from HCC-1428 BL genome in box, deleted amino acids in boldface).

2000)
. No convincing somatic missense, nonsense, frameshift, or in-frame deletion mutations of the DCC ORF have been reported in the literature (Cho et al., 1994;Miyake et al., 1994). Inherited mutation of DCC also has been excluded as predisposing to hereditary colon cancer (Peltomaki et al., 1991). Finally, transgenic mice heterozygous for a DCC knockout allele are not predisposed to colon cancer (Fazeli et al., 1997). Nonetheless, immunohistochemical studies indicated that DCC protein is absent in a subset of a variety of cancers including breast cancer (Koren et al., 2003). The presence of DCC protein has been shown to be a positive prognostic factor in stage II and stage III colorectal carcinoma (Shibata et al., 1996), and overexpression of Netrin-1, a DCC ligand, stimulates the formation of adenocarcinoma in Apc mutant mice (Mazelin et al., 2004).
In contrast to DCC, SMAD4 has been established as a canonical two-hit tumor-suppressor gene. SMAD4 has been shown to be biallelically deleted from a variety of pancreatic carcinoma cell line and xenograft genomes. Many of these deletions are intragenic, eliminating only the SMAD4 gene (Hahn et al., 1996b). Many somatic missense, nonsense, and frameshift mutations have been reported to occur in the SMAD4 ORF in pancreatic and colon carcinoma without concurrent amplification of normal the SMAD4 ORF sequence (Iacobuzio-Donahue et al., 2004). Inherited mutated alleles of SMAD4 are responsible for 50% of juvenile polyposis syndrome cases (Howe et al., 1998). SMAD4 protein has not been detected in polyps that develop in carriers of SMAD4 mutations, indicating loss of the normal SMAD4 allele (Woodford-Richens et al., 2001). Mice heterozygous for a SMAD4 knockout allele are predisposed to gastric polyposis, with loss of the normal SMAD4 allele in subsequent carcinomas . The SMAD4 protein probably acts as a tumor suppressor through its signal transduction role in the TGFB pathway (Massague et al., 2000).
The novel deletions described in this report are true two-hit lesions. The normal paired genome possesses two separate alleles at each deleted locus; both were lost en route to oncogenesis. Analysis of deletion maps from previous reports of chromosome arm 18q deletion in human cancer found complete (presumably biallelic) loss of BID-CHR18 from other cancer cell lines and xenograft genomes. Specifically, studies of SMAD4 (Hahn et al., 1996a;Hilgers et al., 2000) and SMAD4 and DCC (Thiagalingam et al., 1996) reported BID-CHR18 deletion from genomes of cancer cell lines and xenografts, although with the concomitant loss of the respective candidate tumor-suppressor genes. It therefore is possible that the phenotypic consequences of these large deletions are in part a result of the genetic information in BID-CHR18. Other studies have identified homozygous deletions centromeric to the SMAD4-BID-CHR18-DCC interval (Thiagalingam et al., 1996;Takei et al., 1998;Wilentz et al., 2000). Some of these deletions reduced the SMAD4 message by affecting a noncoding exon of SMAD4.
The putative relationship between DCC expression in human breast epithelial cells with the BID-CHR18 locus is difficult to assess. Other investigations have shown the DCC protein to be breast in epithelia lining ducts and acini in vivo by immunohistochemical techniques (Koren et al., 2003). In our study, the DCC transcript was detected in fetal brain and normal breast tissue, but not in three separate normal breast epithelial cell cultures or in HCC-1482 or other breast cancer cell lines (results not shown). DCC transcription appears to be down-regulated in response to cell culture, independent of malignant transformation. This phenomenon has been observed by Thompson et al. (1993), who used Northern analysis to detect DCC transcript in breast cancer tumors and MCF-7 xenograft cells but did not detect DCC transcript in MCF-7 cells in culture or in the breast cancer cell lines MDA-MB-231 and T-47D (Thompson et al., 1993). Therefore, we cannot rule out that deletion of BID-CHR18 may have affected the expression of DCC in the context of the primary tumor, perhaps via elimination of a DCC enhancer sequence. The definitive gene content of BID-CHR18 is difficult to assess with current resources, and it does not contain any currently known genes.
The 9-bp D(49-51) SMAD4 mutation represents an additional two-hit lesion of the HCC-1428 genome. No other in-frame 3-amino-aciddeletion mutations have been reported in SMAD4 in sporadic cancer or juvenile polyposis (JP), although one JP kindred has a 9-nucleotide deletion that eliminates amino acids 64-66 and adds a threonine (Woodford-Richens et al., 2000). The D(49-51) SMAD4 mutation is in the MH1 (Mad homology 1 domain) DNA-binding domain. Based on a comparison with the crystal structure of fellow SMAD protein family member SMAD3, amino acids 49-51 are not directly involved in DNA contact; however, SMAD4 missense mutations in the vicinity reduce DNA binding in vitro (Moren et al., 2000).
Notwithstanding the DNA-binding impairment conferred by MH1 cancer-specific mutations, Moren et al. (2003) and others have convincingly demonstrated that such lesions, specifically, L43S, G65V, R100T, and P130S, all inactivate the SMAD4 protein by enhancement of polyubiquitination and 26S proteasome-mediated degradation (Xu and Attisano, 2000;Moren et al., 2003). Our own pulse chase analysis of exogenous D(49-51) SMAD4-FLAG protein and partial pharmacologic rescue of endogenous D(49-51) SMAD4 by inhibition of the 26S proteasome confirmed that this mechanism is responsible for low levels of mutant protein. Other recent studies have demonstrated that the SMAD4 MH1 domain cancer mutants R100T, G65V, and L43S have increased susceptibility to ubiquitination and proteolysis mediated by E3 ubiquitin ligase complex SCF skp2 . Additional investigations are needed to confirm that D(49-51) SMAD4 also is targeted by the SCF skp2 complex (Liang et al., 2004).
Partial pharmacologic rescue of endogenous D(49-51)-SMAD4 by inhibition of the 26S proteasome confirm that increased degradation was responsible for the low level of SMAD4 protein, not loss of any information from 18q21.1, the deleted region, or elsewhere. This finding raises the possibility that TGFB signaling attenuated by SMAD4 mutation could be rescued in vitro or in vivo by pharmacological inhibition of the 26S proteasome.