Isolation of a Novel Human Gene, APCDD1, as a Direct Target of the β-Catenin/T-Cell Factor 4 Complex with Probable Involvement in Colorectal Carcinogenesis¹

The β-catenin/Tcf⁴ signaling pathway is involved in various morphogenic aspects of development (1, 2, 3). However, recent progress in cancer research has also underscored its importance in the development of tumors arising in the colon, liver, prostate, stomach, brain, endometrium, or elsewhere (4). APC, a tumor suppressor, interacts with β-catenin, axin, conductin, and glycogen synthase kinase 3β and facilitates the degradation of β-catenin via the ubiquitin proteosome system (2). Most sporadic colorectal tumors accumulate β-catenin in the cytoplasm and/or nucleus as a consequence of inactivating mutations in APC, AXIN1, or AXIN2 (conductin) or oncogenic mutations in CTNNB1 (β-catenin); any of these mechanisms can result in constitutive activation of the β-catenin/Tcf transcriptional complex (2). The complex then activates target genes such as c-myc, cyclin D1, matrilysin, c-jun, fra-1, PPARδ, CD44, WISP1, AF-17, ENC-1, Id2, and laminin 5γ2 (5, 6, 7, 8, 9, 10, 11, 12, 13). However, the precise mechanism of tumorigenesis when this pathway is activated in colonic epithelial cells remains unsolved.

In the study reported here, we demonstrate that expression of a novel gene, APCDD1, is directly regulated by the β-catenin/Tcf complex and that its elevated expression promotes proliferation of colonic epithelial cells in vitro and in vivo. These data should lead to a better understanding of colorectal tumorigenesis and thereby facilitate the development of novel strategies for diagnosis, treatment, and prevention of colorectal cancers.

Human colon cancer cell lines SW480 and LoVo and human cervical cancer cell line HeLa were obtained from the American Type Culture Collection (Manassas, VA). All cells were cultured as monolayers in appropriate media [Leibovitz’s L-15 (Life Technologies, Inc., Grand Island, NY) for SW480 cells, Ham’s F-12 nutrient mixture (Life Technologies, Inc.) for LoVo cells, and Eagle’s MEM (Life Technologies, Inc.) for HeLa cells] supplemented with 10% fetal bovine serum (Cansera) and 1% antibiotic/antimycotic solution (Sigma). Cells were maintained at 37°C in an atmosphere of humidified air with 5% CO₂ for LoVo and HeLa cells and without CO₂ for SW480 cells. Cancerous colonic tissues and corresponding noncancerous mucosae were excised from 38 patients during surgery, after informed consent had been obtained.

Fabrication of the cDNA microarray slides and construction of recombinant adenovirus have been described elsewhere (14, 15). We first analyzed expression profiles of 9,216 genes and then extended the analysis to 23,040 genes, using two sets of slides containing duplicated cDNA spots to reduce experimental fluctuation. SW480 cells were infected at a multiplicity of infection of 100 with adenovirus constructs expressing either wild-type APC (Ad-APC) or LacZ (Ad-LacZ), a control gene. Total RNAs were extracted 72 h after infection, and T7-based RNA amplification was carried out using polyadenylated RNA purified from the extracts as described elsewhere (15). Five-μg aliquots of amplified RNA from SW480 cells transfected with Ad-APC or Ad-LacZ were labeled with Cy5-dCTP or Cy3-dCTP (Amersham Pharmacia Biotech), respectively. Hybridization, washing, detection, and analysis of data were carried out as described previously (15).

Human multiple-tissue blots (Clontech, Palo Alto, CA) were hybridized with a ³²P-labeled PCR product of B7323N, a gene on the microarray chosen for investigation. The product was prepared by RT-PCR using primers 5′-CGTGTTCAAAGTGAATCACATG-3′ and 5′-CTAAAACTTCTATCTGCGGATG-3′. Prehybridization, hybridization, and washing were performed according to the supplier’s recommendations. The blots were autoradiographed with intensifying screens at −80°C for 24–72 h.

Among the genes that showed high signal intensity ratios of Cy3:Cy5 on the microarray, we focused on one with in-house identification number B7323N because its ratio was as high as 4.2. A subsequent homology search in public databases identified a genomic sequence with GenBank accession number NT_019631.4, as well as more than 100 human expressed sequence tags that contained B7323N sequences. Because the assembled cDNA sequence was shorter than the transcript indicated by Northern blotting, we performed 5′ rapid amplification of cDNA ends using the Marathon cDNA amplification kit (Clontech) according to the supplier’s recommendations. The cDNA template was synthesized from human colon mRNA (Clontech) with a gene-specific reverse primer (5′-GCTCGTCTGACCGATAGATGATCC-3′) and the AP1 primer supplied in the kit. Nucleotide sequences were determined with an ABI PRISM 3700 DNA sequencer (Applied Biosystems) according to the manufacturer’s instructions.

Total RNA was extracted from cultured cells and clinical tissues using Trizol reagent (Life Technologies, Inc.) according to the manufacturer’s protocol. Extracted RNA was treated with DNase I (Boehringer Mannheim, Mannheim, Germany) and reverse transcribed for single-stranded cDNAs using oligo(dT)_12–18 primer with Superscript II reverse transcriptase (Life Technologies, Inc.). We prepared appropriate dilutions of each single-stranded cDNA for subsequent PCR amplification by monitoring the glyceraldehyde-3-phosphate dehydrogenase gene (GAPDH) as a quantitative control. Primer sequences were 5′-ACAACAGCCTCAAGATCATCAG-3′ and 5′-GGTCCACCACTGACACGTTG-3′ for GAPDH and 5′-GGATCATCTATCGGTCAGACG-3′ and 5′-TGGGTCACATCCTGCTGGATG -3′ for APCDD1. All reactions involved initial denaturation at 94°C for 2 min followed by 18 (for GAPDH) or 25 cycles (for APCDD1) of 94°C for 30 s, 57°C for 30 s, and 72°C for 45 s on a GeneAmp PCR system 9700 (PE Applied Biosystems, Foster City, CA).

Western blotting with mouse anti-β-catenin antibody (Transduction Laboratories, Lexington, KY) was performed as described elsewhere (14).

An initiation site for transcription of APCDD1 was determined by comparing the cDNA sequence of APCDD1 with the human genomic sequence. Two potential Tcf4 binding sites with the consensus sequence CTTTGWW were identified in the 5′-flanking region between −910 and −904 (TBM1) and between −466 and −460 (TBM2). A wild-type reporter plasmid containing both TBM1 and TBM2 (P1) and deletion constructs of the plasmid (P2 and P3) were prepared by cloning the fragments encompassing regions between −971 and +26 (P1), −700 and +26 (P2), and −151 and +26 (P3) into an appropriate site of pGL3-Basic vector (Promega, Madison, WI). Site-directed mutagenesis was performed using the QuikChange site-directed mutagenesis kit (Stratagene, La Jolla, CA) for the two fragments (P1 and P2) that contained the putative Tcf/LEF-binding motifs. Plasmids expressing an activated form of β-catenin (mut β-catenin) and wild-type and dominant-negative forms of Tcf4 (wtTcf4 and dnTcf4) were prepared as described previously (11). Two μg of each reporter plasmid and 1.5 μg of each expression construct were cotransfected with 0.5 μg of pRL-TK plasmid (Promega) into HeLa cells plated on 6-cm dishes, using FuGENE 6 (Boehringer Mannheim) to normalize the efficiency of transfection. Reporter assays were carried out using a dual-luciferase reporter assay system (Promega) according to the supplier’s recommendations.

We performed EMSA experiments using extracts from intact nuclei of SW480 cells as described previously (14). Two double-stranded 16-nucleotide DNA probes were prepared by annealing FF (5′-GCTTTGATTGTGGTGA-3′) and RR (5′-TCACCACAATCAAAGC-3′) for APCDD1-TBM1 and FF2 (5′-CCCCTTTGAACACCTT-3′) and RR2 (5′-AAGGTGTTCAAAGGGG-3′) for APCDD1-TBM2.

The entire coding region of human APCDD1 was amplified by RT-PCR using primers 5′-GCGGAATTCAGGGCCCAGAGCAGGACTG-3′ and 5′-TAGCTCGAGCTAAAACTTCTATCTGCGGATGT-3′ and cloned into expression vector pcDNA 3.1(+) and pcDNA3.1(−) (Invitrogen, Carlsbad, CA) under the control of the cytomegalovirus promoter/enhancer. Plasmids expressing APCDD1 (pcDNA-APCDD1), its complementary strand (pcDNA-antisense), or the empty vector (pcDNA) were transfected into colon cancer LoVo cells for colony formation assay. After 2 weeks of incubation with an appropriate concentration of Geneticin, cells were fixed with 100% methanol and stained with Giemsa solution.

Cells that abundantly expressed exogenous APCDD1 (LoVo-APCDD1-S7 and LoVo-APCDD1-S10 cells) and control cells transfected with empty vector (LoVo-vector-V3 and LoVo-vector-V4 cells) were selected in medium containing 500 μg/ml Geneticin. LoVo-APCDD1 and LoVo-vector cells were seeded onto 6-cm dishes, and viable cells were counted by the trypan blue dye exclusion method in triplicate from day 0 to 8.

LoVo-APCDD1-S7, Lovo-APCDD1-S10, LoVo-vector-V3, or LoVo-vector-V4 cells (5 × 10⁶) were injected s.c. into the posterior mid-dorsum of three BALB/cAnN Crj-nu/nu mice. Tumors were measured every 7 days for 9 weeks, and the volumes were estimated by the following formula: V = π/6 × a² × b, where a is the short axis, and b is the long axis.

SW480 cells plated onto 10-cm dishes (2 × 10⁵ cells/dish) were transfected with synthetic S-oligonucleotides corresponding to APCDD1 using Lipofectin reagent (Life Technologies, Inc.) and maintained in media containing 10% fetal bovine serum for 7 days. The cells were then fixed with 100% methanol and stained with Giemsa solution. Sequences of the S-oligonucleotides were as follows: sense (S2), 5′-ATGTCCTGGCCGCGCC-3′; antisense (AS2), 5′-GGCGCGGCCAGGACAT-3′; reverse (R2), 5′-TACAGGACCGGCGCGG-3′; and scrambled (Sc2), 5′-ATCTGGTCCGGCGCGG-3′. Additionally, a MTT assay was performed in triplicate as described elsewhere (16).

The data were subjected to ANOVA and Scheffé’s F test.

To identify downstream genes in the β-catenin/Tcf pathway, we analyzed expression profiles of SW480 cells in which a large amount of β-catenin accumulates in the nucleus and cytoplasm. The initial analysis using in-house microarray slides containing 9,216 genes identified a total of 84 candidates including cyclin D1, AF17, and MT1-MMP (10). We extended the analysis using sets of in-house microarray slides containing an additional 13,824 genes. Comparison of the expression profiles between SW480 cells infected with either Ad-APC or Ad-LacZ identified a number of genes whose expression had been altered by APC. Because transduction of APC reduces the level of β-catenin in the nucleus and subsequently represses signaling mediated by the β-catenin/Tcf complex (17, 18, 19), we focused on genes that appeared to be down-regulated by APC. Among them, we identified a gene whose expression level had decreased approximately 4-fold in response to Ad-APC as compared with Ad-LacZ. This expressed sequence, designated as in-house accession number B7323N, corresponded to Hs.355947, an expressed sequence tag in a UniGene cluster in the National Center for Biotechnology Information. Subsequent semiquantitative RT-PCR experiments corroborated its reduced expression after transduction of APC (Fig. 1,A). Treating the cells with Ad-Axin, an adenovirus expressing wild-type AXIN1, which also down-regulates the activity of the β-catenin/Tcf4 complex, also led to decreased expression of B7323N (Fig. 1 A).

Multiple tissue Northern blotting with B7323N cDNA as a probe showed that a 2.6-kb transcript was expressed abundantly in human heart, pancreas, prostate, and ovary and moderately in lung, liver, kidney, spleen, thymus, colon, and peripheral lymphocytes (Fig. 1 B).

To obtain the most 5′ portion of its cDNA sequence, we performed 5′ rapid amplification of cDNA ends. The full-length cDNA sequence contained 2,607 nucleotides encoding a putative 514-amino acid protein with a predicted molecular mass of 58.8 kDa (GenBank accession number, XM_097240). Although motif searches using the computer programs SMART⁵ and PSORT II Prediction⁶ did not identify any known domains, the predicted amino acid sequence had some degree of homology with the product of the endo-1,4-β-xylanase of Thermobacillus xylanilyticus (31% identity in amino acids). However, the region showing homology did not contain the fragment responsible for enzymatic effects (20). Therefore, we termed this gene APCDD1 (adenomatous polyposis coli down-regulated 1). Additional BLAST searches identified a human genomic sequence (GenBank accession number NT_019631.8) assigned to chromosomal band 18p11. Comparison of our cDNA with the genomic sequence allowed us to determine that the APCDD1 gene consisted of five exons and covered a genomic region of approximately 40 kb (data not shown).

Because accumulation of β-catenin is a frequent feature of colorectal tumors, we examined expression of APCDD1 in colon cancer tissues and corresponding noncancerous mucosae using semiquantitative RT-PCR and detected its increased expression in 18 of the 27 (67%) tumors examined (Fig. 1 C).

To test whether the β-catenin/Tcf4 complex regulates transcription of APCDD1, we transfected into HeLa cells the reporter plasmid P1, which contained two putative Tcf/LEF-binding motifs (TBM1 and 2), with/without an activated form of mutant β-catenin and wild-type Tcf4 (Fig. 2,A). The reporter activity of plasmid P1 was significantly enhanced by introduction of the activated form of β-catenin and wild-type Tcf4 (Fig. 2,B). The enhancement of activity was reduced when P1 was cotransfected with the dominant-negative form of Tcf4. To determine the element(s) responsible for promoter activity, we compared each of the deletion mutants of P1. The activity of P1 was significantly higher than that of either P2 or P3; the activity of P2, which contained only TBM2, was significantly higher than that of P3 (Fig. 2 B). These data suggested that a region encompassing −971 to −151 may associate with the β-catenin/Tcf4 complex in promoting transcription of APCDD1.

Because the region in question contained two possible Tcf/LEF-binding motifs, we hypothesized that one or both of these motifs might be responsible for transcriptional activation. To investigate that possibility, we constructed reporter plasmids P1Mutant (P1M) and P2Mutant (P2M), in which the candidate Tcf/LEF-binding motif was changed to CTTTGGC; the β-catenin/Tcf4 complex is unable to bind to this sequence (Fig. 2,A). Reporter assays using these five plasmids revealed that the P1M and P2M fragments containing mutated motifs were less able to activate transcription of APCDD1, with luciferase activities equivalent to those of the P2 or P3 fragments, respectively (Fig. 2 B). These results implied that both putative Tcf/LEF-binding motifs are involved in transcriptional activation of APCDD1.

To examine whether the β-catenin/Tcf4 complex associates directly with TBM1 and TBM2, an EMSA was carried out using oligonucleotides designed to encompass each sequence (APCDD1-TBM1 and APCDD1-TBM2). A shift was observed in the bands corresponding to β-catenin/Tcf4 bound to either APCDD1-TBM1 or APCDD1-TBM2 after the addition of anti-β-catenin antibody, but not P53 antibody (control; Fig. 2 C). As expected, this binding was abrogated by the addition of wild-type unlabeled oligonucleotides, but not by mutant unlabeled oligonucleotides.

To disclose a potential role of APCDD1 in colorectal tumorigenesis, we prepared plasmids expressing APCDD1 (pcDNA-APCDD1) and a complementary strand of APCDD1 (pcDNA-antisense) and carried out a colony formation assay in LoVo cells. LoVo cells were chosen because they exhibited the lowest expression of APCDD1 among the five colon cancer cell lines examined (Fig. 3,A). pcDNA-APCDD1 produced markedly more colonies than did control plasmids (pcDNA-antisense or pcDNA; data not shown). This result was confirmed by three independent experiments. To determine its effect on cell growth in vitro, we established LoVo cells expressing exogenous APCDD1 (Fig. 3,B, LoVo-APCDD1-S7 and LoVo-APCDD1-S10) and compared their growth with control cells transfected with mock vector (Fig. 3,B, LoVo-vector-V3 and LoVo-vector-V4). As shown in Fig. 3 C, LoVo-APCDD1 cells grew at a markedly increased rate compared with control cells.

To investigate the role of APCDD1 in vivo, we s.c. transplanted either LoVo-APCDD1 cells or LoVo-vector cells into BALB/cAnN Crj-nu/nu mice. All 12 transplantations of either LoVo-APCDD1 (S7 and S10) or LoVo-vector (V3 and V4) cells formed tumors in the nude mice. Although there was a tendency for LoVo-APCDD1 tumors to grow larger and faster compared with those of LoVo-vector, the difference was not statistically significant (Fig. 3 D).

To assess the growth-promoting role of APCDD1, we synthesized five pairs of control and antisense S-oligonucleotides corresponding to APCDD1 and transfected them into SW480 cells, which had shown the highest expression of APCDD1 (Fig. 3,A). Among the pairs tested, APCDD1-AS2 significantly suppressed the expression of APCDD1 compared with control sense S-oligonucleotides (APCDD1-S2; Fig. 4,A). Six days after transfection, introduction of APCDD1-AS2 clearly suppressed focus formation compared with APCDD1-S2, suggesting that suppression of APCDD1 reduced growth and/or survival of transfected cells (Fig. 4,B). To con-firm gene-specific growth suppression by APCDD1-AS2, we synthesized additional control oligonucleotides (reverse S-oligonucleotides APCDD1-R2 and scrambled S-oligonucleotides APCDD1-Sc2) and carried out a MTT assay. The results confirmed the decrease in cell survival in response to APCDD1-AS2 compared with controls (Fig. 4 C).

We have shown here that expression of APCDD1 correlates well with the transcriptional activity of the β-catenin/Tcf4 complex because its expression is clearly affected by the amount of β-catenin, and the dominant-negative form of Tcf4 is able to suppress its promoter activity. Moreover, we have demonstrated that two Tcf-binding motifs in the 5′-flanking region of APCDD1 are responsible for promoting transcription of this gene and that these motifs do associate directly with the β-catenin/Tcf4 complex. The data support our view that APCDD1 is a direct target of the β-catenin/Tcf4 complex.

The great majority of colon cancers manifest aberrant accumulation of β-catenin resulting from mutations in APC, AXIN2, or CTNNB1 (19, 21). Accumulated β-catenin associates with and activates Tcf/LEF transcription factors. Numerous downstream genes in the β-catenin/Tcf/LEF pathway have been identified thus far. These include genes regulating proliferation, such as c-myc, cyclin D1, and AF17 (5, 10); genes associated with cellular differentiation, such as ENC1 and Id2 (11, 12); and genes involved in cell adhesion and/or migration, such as MMP7, CD44, and laminin 5γ2 (5, 13). The data presented here strongly suggest that APCDD1 plays a role in proliferation because it promoted growth of cancer cells in vitro and in vivo. Because we have not analyzed its role in differentiation, adhesion, or migration, it would be of great interest to investigate whether APCDD1 might also have a role in those functions. Additional studies will be necessary to reveal how APCDD1 promotes cellular proliferation.

β-Catenin is a key mediator of the canonical Wnt/wingless pathway. Only two negative feedback regulators, namely, AXIN2 and naked cuticle (22, 23), have been identified, implying that this pathway is tightly regulated. Therefore, expression of APCDD1 is likely to be regulated precisely in adult as well as fetal tissues. Although we detected abundant expression of APCDD1 in adult heart, pancreas, prostate, and ovary, its expression during embryogenesis remains to be clarified. In normal mucosae, β-catenin is predominantly expressed in epithelial cells at the basal zone of crypts (24), whereas its aberrant expression is frequently observed in the invasion front of cancer tissues (25). We are now preparing specific antibody to APCDD1 to examine the location of this protein in normal mucosae and cancer tissues as well as to reveal spatio-temporal expression of APCDD1 in fetal tissues.

Although APCDD1 has some degree of homology with the product of the endo-1,4-β-xylanase gene of T. xylanilyticus, the homologous region does not include the domain responsible for enzymatic effects in the bacterial cells (20). Therefore, its role may be very different from that of endo-1,4-β-xylanase. Analysis of motifs using the SMART computer program predicted the presence of a signal peptide between codons 1 and 24 and a transmembrane segment between codons 493 and 512, indicating that APCDD1 may be a secreted protein. We note that WISP1, a gene regulated by Wnt-1, encodes a secreted protein (8, 9). Hence, some of the other genes regulated by the β-catenin/Tcf complex might be secreted and affect surrounding cells by modulating microenvironmental factors that could include components of the extracellular matrix, interactions between epithelium and stroma, immunoreactions, and/or angiogenesis. Alternatively, APCDD1 might increase proliferation of epithelial cells in an autocrine manner. Because its expression was greatly enhanced in more than half of the colon cancers we examined, it would be intriguing to test whether APCDD1 is secreted or not. If it is, APCDD1 could be a diagnostic marker for colonic tumors.

Our results raise a possible scenario in which activated APCDD1 may lead to neoplasms by constitutively accelerating proliferation of colonic epithelial cells. Because activation of the β-catenin/Tcf4 complex is a major and early step of tumorigenesis in the colon, blocking this pathway should be a rational therapeutic strategy. In fact, suppression of Tcf4 activity by transducing either the 20-amino acid repeat domain of APC or the wild-type AXIN1 gene can induce apoptosis in colon cancer and hepatocellular carcinoma cells (14, 26). A recent investigation showed that suppression of β-catenin expression by specific antisense oligonucleotides inhibited neoplastic growth of colon cancer cells in vivo and in vitro (27). However, β-catenin is expressed ubiquitously in epithelial cells and anchors cadherins by interacting with their cytoplasmic domains and α-catenin. Accordingly, suppression of β-catenin could have adverse effects on normal epithelial cells.

Given the crucial role of β-catenin in cell adhesion, a better therapeutic intervention might be to target the activity of the β-catenin/Tcf4 complex through inhibition of β-catenin-Tcf4 or the β-catenin/Tcf4-DNA association. Possible agents for such targeting would include vitamin D and retinoic acid, which activate their specific receptors; those receptors (vitamin D receptor and retinoic acid receptor, respectively) interact with β-catenin and decrease β-catenin/Tcf-mediated transcription (28, 29). Alternatively, because Tcf4 plays an important role in maintenance of epithelial stem cells in the intestine (30), pinpointing downstream gene(s) essential for growth and/or survival of cancer cells but not for noncancerous stem cells should be an ideal therapeutic strategy. APCDD1 may be a logical target for such an approach because down-regulated expression of APCDD1 by antisense oligonucleotides suppressed the growth of colon cancer cells in our experiments.

In conclusion, our discovery that APCDD1 is a downstream target of the β-catenin/Tcf4 complex brings new insight concerning colorectal tumorigenesis. Although further investigation of its function is necessary, APCDD1 may serve in the future as a novel diagnostic marker as well as a therapeutic and preventive target for colon tumors.

We thank Yumi Nakajima, Yuka Yamane-Tanaka, and Dr. Natini Jinawath for technical assistance.