PMEPA1, a Transforming Growth Factor-β-induced Marker of Terminal Colonocyte Differentiation Whose Expression Is Maintained in Primary and Metastatic Colon Cancer¹

TGF-β³ is a cytokine that has been shown to have a role in inducing many different biological effects, including differentiation, migration, and adhesion (reviewed in Ref. 1). In normal epithelial cells, TGF-β commonly demonstrates an antiproliferative effect, including cell cycle arrest and apoptosis (reviewed in Ref. 1), whereas epithelial cancers commonly demonstrate resistance to TGF-β (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14). In particular, TGF-β functions as a tumor suppressor in the colon where it induces apoptosis of normal colon epithelial cells. Mutations of the TGF-β receptor and downstream Smad signaling elements have been shown to commonly disable TGF-β signaling (6, 15, 16, 17, 18, 19, 20, 21, 22).

To define potential effectors of TGF-β-mediated tumor suppression, we looked for target genes induced by TGF-β treatment of VACO 330, a nontransformed TGF-β-sensitive cell line derived from a human colonic adenomatous polyp. One such TGF-β-induced gene was PMEPA1, located on chromosome 20q13 (23). As PMEPA1 expression has paradoxically been described to be induced by androgen in the LNCaP prostate cancer cell line and increased in renal and other cancers (23, 24), we decided to characterize both the nature of the PMEPA1 transcript in the human colon and cellular origin of PMEPA1 expression in normal and potentially malignant colon epithelium. We defined a novel splice variant of PMEPA1, named Variant A, which lacks the putative transmembrane domain encoded by the originally published sequence (23). Consistent with TGF-β being an important regulator of PMEPA1 expression in a colon cell line, we localized PMEPA1 expression to the terminally differentiated luminal compartment of the colonic crypts. However, both primary and metastatic colon cancers demonstrated strong retention of PMEPA1 expression in transformed epithelial cells. PMEPA1 appears to typify a class of TGF-β target genes that are markers of terminally differentiated colonocytes but mediate functions other than tumor suppression. PMEPA1 also demonstrates that even late colon cancers retain a strong capacity to execute many steps of the normal colonocyte terminal differentiation program.

VACO cell lines were established as described (25) and maintained in MEM supplemented with 2% fetal bovine serum, 2 mm l-glutamine, 50 μg/ml gentamycin, 1 μg/ml hydrocortisone, 2 μg/ml human transferrin, 5 nm sodium selenite, and 10 μg/ml bovine insulin (MEM2+ medium). VACO 400/RII and VACO 400/neo were maintained in culture with MEM2+ medium supplemented with G418 at 300 μg/ml (26). FET was a generous gift from M. G. Brattain and grown as described (27). TGF-β1 was added to cultures where indicated at 10 ng/ml.

All normal colon, primary colon tumor, and liver met tissues were obtained from the archives of University Hospitals (Cleveland, OH).

We designed two custom expression-monitoring DNA microarrays using Affymetrix GeneChip® technology (28) that contained essentially all expressed human genes in the public domain at the time of design. Briefly, we selected the sequences for inclusion on the arrays using genes predicted from the available human genome sequences and sequences derived from the expressed mRNA and EST databases in GenBank (29). Consensus sequences representing human expressed sequences were generated using the Clustering and Alignment Tool software (DoubleTwist, Oakland, CA) using the mRNAs (nt) and EST (dbest) databases in GenBank. Prediction of the expressed genome from the human genome sequence was done using ab initio exon prediction (30).

The arrays were hybridized with labeled cRNA derived from 10 μg of total RNA using standard protocols (31). The intensity data from the arrays were analyzed using a statistically based analysis methodology, which allows for estimating expression levels and providing confidence intervals for these estimates. This method uses a Gamma distribution model of the intensity data for normalization to control for the systematic variation attributable to nonbiological factors, such as array-to-array variability, and attributable to variation in sample quality. For each probe set, a single measure or average intensity was calculated using Tukey’s trimean of the intensity of the constituent probes (32).

Ten μg of total RNA were size fractionated on a 1% formaldehyde agarose gel and transferred to Nytran SuPerCharge (Schleicher & Schuell, Keene, NH). The membrane was prehybridized using Rapid-hyb Buffer (Amersham Biosciences, Corp., Piscataway, NJ) for 30 min at 65°C. The PMEPA1 probe was labeled with ³²P-dATP using the random-primed, DNA-labeling kit Strip-EZ DNA (Ambion, Inc., Austin, TX). The labeled probe was added to the membrane in Rapid-hyb buffer for 1.5 h at 65°C. The membrane was washed in 0.1% SDS/2 × SSC 2 × 15 min at room temperature, then 0.1% SDS/0.1 × SSC 2 × 15 min at 65°C. The membrane was then exposed to a phosphor screen overnight and then analyzed using a STORM optical scanner (Molecular Dynamics, Sunnyvale, CA).

The PMEPA1 probe AA451748 was PCR amplified using the forward primer 5′-GGTGAAAAGGCAGAACACTCCGC-3′ and reverse primer 5′-GGCAAAACCACTACCTGGAACTCG-3′. The PMEPA1 probe T17185 was PCR amplified using the forward primer 5′-TGTGAGCGTAAATAGAGGTGACCAG-3′ and reverse primer 5′-CAAAGGAGAGCAGTTCCCACG-3′.

FET was treated with 0.5 μg/ml cycloheximide for 3 h, and then TGF-β was added at 10 ng/ml, whereas cycloheximide treatment continued. Cells were harvested after 18 h of TGF-β treatment, and total RNA was isolated as described (16).

Total RNA was isolated from normal colon epithelium by extraction with guanidine isothiocyanate as described (16). The FirstChoice RLM-RACE Kit (Ambion) was used with SuperTaq Plus (Ambion) as described by the manufacturer’s protocol. The outer gene-specific primer used was 5′-CTTCTGAGGACAGGGCATCTTCTCTC-3′, and the inner gene-specific primer used was 5′-GCTTGTAGTGGCTCAGCAGGCAC-3′.

Total RNA from normal colon epithelium and TGF-β-treated and nontreated VACO 330, VACO 394, VACO 425, and FET were prepared by extraction with guanidine isothiocyanate as described (16). Total RNA from LNCaP treated with 1 μm dihydrotestosterone (Sigma-Aldrich, St. Louis, MO) was a generous gift from D. Danielpour (Case Western Reserve University, Cleveland, OH). DNase-treated RNA was reverse transcribed with random hexamer primers using the Thermoscript RT-PCR System (Invitrogen, Carlsbad, CA). Full-length PMEPA1 was then PCR amplified with Advantage-GC cDNA polymerase mix (Clontech, Palo Alto, CA) using the forward primer 5′-GGTCGTCCTCCTTGGGTTCG-3′ and reverse primer 5′-AAGCATCCACTCTGCAGGG-3′. The PCR conditions for the amplification of all PMEPA1 fragments listed in this manuscript were 1 cycle of 94°C for 3 min; 40 cycles of 94°C for 30 s, 55°C for 30 s, and 68°C for 1.5 min; and 1 cycle of 68°C for 4 min. The product was size fractionated on a 1% agarose gel, and the full-length product was isolated, purified using Qiaquick Gel Extraction Kit (Qiagen, Valencia, CA), cloned into pCR2.1-TOPO (Invitrogen), and sequenced.

FLAG-tagged PMEPA1 Variant A was made by PCR amplifying PMEPA1 as described above with the forward primer 5′-AAAAAAGCTTGGATCCATGATGGTGATGGTGGTGGTGATCACGTGC-3′ and reverse primer 5′-AAAAGCGGCCGCCGAGAGGGTGTCCTTTCTGTTTATCCTTC-3′. The full-length product was purified, digested with HindIII and NotI, which were incorporated into the forward and reverse primers, respectively (underlined sequences), and inserted into the HindIII and NotI sites of FLAG/pcDNA3.1/V5/His-TOPO. FLAG/pcDNA3.1/V5/His-TOPO was made by annealing the oligos 5′-GGCCGCGACTACAAGGACGACGATGACAAGTGAA-3′ and 5′-CCGGTTCACTTGTCATCGTCGTCCTTGTAGTCGC-3′ that encode FLAG and inserting this fragment into the NotI and AgeI sites of pcDNA3.1/V5/His-TOPO (Invitrogen). The underlined sequences encode NotI and AgeI sites in the first and second oligonucleotides listed, respectively.

FLAG-tagged Variant B PMEPA1 was made by PCR amplifying PMEPA1 with the forward primer 5′-AAAAAAGCTTGGATCCATGGCGGAGCTGGAGTTTGTTC-3′ and reverse primer 5′-AAAAGCGGCCGCCGAGAGGGTGTCCTTTCTGTTTATCCTTC-3′. The full-length product was purified, digested with HindIII and NotI, which were incorporated into the forward and reverse primers, respectively (underlined sequences), and inserted into the HindIII and NotI sites of FLAG/pcDNA3.1/V5/His-TOPO.

VACO 400/RII was seeded at 8 × 10⁴/chamber in Nunc Lab-Tech chambered slides (Fisher Scientific, Pittsburgh, PA) the day before transfection. The cells were transfected with Effectene (Qiagen) using the manufacturer’s protocol. Where indicated, TGF-β was added at 10 ng/ml 3 h after transfection and continued for the time indicated.

Forty-eight h after transfection, cells were washed three times in PBS and fixed with methanol for 10 min at −20°C. Cells were washed three times in PBS, and anti-FLAG M2 antibody (Sigma-Aldrich) was added at 10 μg/ml in 10% goat serum/PBS for 1 h at room temperature. Cells were washed three times in PBS. Alexa-Fluor 488 goat antimouse IgG (Molecular Probes, Inc., Eugene, OR) was added to the cells at 1:400 diluted in 10% goat serum/PBS for 1 h at room temperature. Cells were washed five times in PBS and mounted in Vectashield mounting medium with 4′,6-diamidino-2-phenylindole (Vector Laboratories, Burlingame, CA).

Formalin-fixed, paraffin-embedded sections were deparaffinized in xylene and hydrated in a graded ethanol series. Dako Genpoint for biotinylated nucleic acid probes (DakoCytomation, Carpinteria, CA) was used according to the manufacturer’s protocol for the in situ hybridization. The biotinylated PMEPA1 probe was added to the hybridization solution at a final concentration of 1 × 10⁻⁷ g/ml and incubated with the tissues overnight at 42°C. Washes were performed at 50°C. After development with DAB, the tissues were counterstained with hematoxylin (Sigma-Aldrich).

The PMEPA1 probe used for in situ hybridization was PCR amplified using the forward primer 5′-GCAGCCATCTGGAGCAAAGAG-3′ and reverse primer 5′-TTGCCTTCAAGACACAGCTCAAC-3′. The amplified fragment was inserted into pCRII-TOPO (Invitrogen) and in vitro translated using the Biotin RNA labeling mix (Roche Applied Science).

To identify potential effectors of TGF-β-mediated suppression of colon cancer, we used GeneChip® expression microarrays to identify TGF-β-induced genes in VACO 330, a nontransformed TGF-β-sensitive cell line derived from a human adenomatous colon polyp (25). TGF-β induces apoptosis in VACO 330 cells with a long latency of ∼120 h. To define potential effectors of TGF-β-negative regulation of VACO 330, we compared at 24, 48, and 72 h the expression profiles of TGF-β-treated VACO 330 versus the expression profiles of nontreated VACO 330 or VACO 330 treated with EGF, a potent stimulator of VACO 330 cell growth. PMEPA1 was identified as a TGF-β target gene, because it was noted to show steadily increasing expression during the TGF-β time course (23). PMEPA1 expression was increased ∼6.5-fold after 72 h of TGF-β treatment compared with control VACO 330 or VACO 330 treated with EGF (Fig. 1).

To confirm the findings of the GeneChip® array, TGF-β induction of PMEPA1 was evaluated by Northern blot analysis of VACO 330, as well as three additional nontumorigenic TGF-β responsive colon cancer cell lines: FET, VACO 394, and VACO 425. As shown in Fig. 2 a, in each of VACO 330, FET, VACO 394, and VACO 425, 48 h of TGF-β treatment resulted in substantial induction of PMEPA1 transcript.

As shown in Fig. 2,a, TGF-β induced two distinct PMEPA1 transcripts of 2.7 and 5 kb. PMEPA1 transcripts of these two sizes had in other tissues been attributed to alternative use of the two alternative polyadenylation signals present in the PMEPA1 3′ UTR (Fig. 5,a; Refs. 23 and 24). Northern blot analysis using probe AA451748 from the proximal 3′ UTR detected two transcripts (Fig. 2,a), and probe T17185 from the distal 3′ UTR detected only the larger transcript (Fig. 2 b), consistent with these two colon transcripts arising from alternative splicing in the 3′ UTR (23, 24).

To determine the kinetics of PMEPA1 induction by TGF-β, a time course was performed in the TGF-β-sensitive nontumorigenic FET colon cancer cell line (Fig. 3). Increased PMEPA1 expression in response to exogenous TGF-β was detected as early as 2 h after TGF-β treatment (Fig. 3,a) and continued to increase steadily over 72 h (Fig. 3 b).

To determine whether PMEPA1 is a direct target of TGF-β signaling, cycloheximide was used to block protein synthesis in FET treated with TGF-β. FET was treated with cycloheximide for 3 h; then TGF-β was added for 18 h, at which point RNA was extracted for Northern blot analysis. Under these conditions, incorporation of ³⁵S-labeled methionine into protein was undetectable (data not shown). As shown in Fig. 4, treatment with cycloheximide itself demonstrated a small increase in PMEPA1 transcript, which was clearly further increased by treatment with TGF-β. Thus, PMEPA1 appears to be a direct target of TGF-β whose induction does not require synthesis of any intermediate protein mediators. TGF-β induction of PMEPA1 could result either from activation of a PMEPA1 transcription factor or inactivation of a PMEPA1 transcriptional repressor.

Recognition of PMEPA1 as a target of TGF-β signaling in the colon was potentially paradoxical, because PMEPA1 had also been noted to be induced by androgen, a positive growth factor in a prostate cell line, and increased on Northern blot analyses of several human cancers (23, 24). Accordingly, we next defined the PMEPA1 transcript as expressed in colon epithelium. Cloning of the full-length transcript of PMEPA1 from normal colonic epithelium was done first by 5′ rapid amplification of cDNA ends, then by RT-PCR. Two alternatively spliced transcripts of PMEPA1 were expressed in normal colon, named Variants A and B. Variant A is a novel splice variant, and the sequence of Variant B corresponded to the published sequence of Xu et al. (Ref. 23; Fig. 5, a and b). Variant A differs from Variant B by using a splice donor site 7 nt upstream of the originally defined end of exon 1. This causes Variant A to splice out the first PMEPA1 ATG and forces Variant A to initiate translation from a position corresponding to the second ATG of Variant B. Variant A thus encodes a shorter form of the PMEPA1 protein of 237 amino acids long, which specifically is lacking part of the putative NH₂-terminal transmembrane region encoded by Variant B.

To determine which variant was induced with TGF-β, the relative proportion of PMEPA1 Variants A and B was determined by sequencing the PMEPA1 transcripts from TGF-β-treated cell lines, normal colon, and androgen-treated LNCaP. Table 1 shows that Variant B accounts for 100% of PMEPA1 transcript in androgen-treated LNCaP prostate cancer cells, for 89% of PMEPA1 transcript in normal colon, and for 73–92% of basal PMEPA1 transcript expression in four TGF-β-sensitive colon cell lines. Moreover, Variant B remains the predominant PMEPA1 transcript after substantial induction by TGF-β in the four TGF-β responsive colon cell lines, with Variant B indeed accounting for 64–100% of the TGF-β-induced PMEPA1 transcripts (Table 1). Thus, both TGF-β in cell lines of the colon and dihydrotestosterone in LNCaP predominantly target increased expression of PMEPA1 Variant B.

The absence in PMEPA1 Variant A of the putative transmembrane region of Variant B suggested that these two different PMEPA1 proteins would likely localize to different cellular compartments. To investigate this hypothesis, COOH-terminal FLAG-tagged constructs for PMEPA1 Variants A and B were transfected into the colon cancer cell line VACO 400/RII. VACO 400/RII was derived as described previously by transfection of a wild-type TGF-β RII transgene into the TGF-β RII mutant colon cancer cell line VACO 400 (33). The confocal image presented in Fig. 6 demonstrates that Variants A and B both displayed cytoplasmic localization but in distinctly different patterns. Variant A localized in punctate clusters concentrated around the nucleus (Fig. 6,a), whereas Variant B was distributed in a diffusely filamentous pattern (Fig. 6,c). TGF-β treatment of these transfectants did not induce redistribution of the staining patterns for either PMEPA1 Variants A or B (Fig. 6, b and d).

As PMEPA1 expression in colon cell culture served as a marker of TGF-β activity, it was of interest to determine the localization of PMEPA1-expressing cells in normal colonic epithelium in vivo. Therefore, in situ hybridization was performed on normal colon tissue using a probe derived from the 3′ UTR of PMEPA1. This region was used for hybridization because of its divergence in sequence from C18orf1, a protein whose coding region shares homology to PMEPA1 (34). As shown in Fig. 7, a and c, PMEPA1 expression localized to the terminally differentiated epithelium at the luminal surface of the bowel and at the tops of the colonic crypts. In addition, some occasional stromal staining was observed. Figs. 7, b and d, the negative controls, demonstrated absence of staining in the colon tissue using a PMEPA1 sense hybridization probe. PMEPA1 expression thus appears to be a novel marker of terminally differentiated colonocytes at the upper reaches of the colonic crypt, a region that is distinct from the proliferative zone that occupies approximately the lower third of the colonic crypt. As this upper crypt region has also been suggested to be the region of in vivo expression of TGF-β (35), this observation is consistent with PMEPA1 expression in normal colonocytes serving as an in vivo marker for TGF-β activity.

The observation that PMEPA1 expression marked terminally differentiated colonocytes contrasted with previous observations by Northern blot analysis of elevated PMEPA1 expression in renal and other human cancers (24). We therefore characterized PMEPA1 expression in colon adenomas and cancers using GeneChip® expression microarrays (Fig. 8). This analysis demonstrated that compared with normal colon epithelium, PMEPA1 expression was significantly increased in colon adenomas and cancers (P < 0.05 for each comparison), with a 2.11-fold increase expression level in metastatic colon cancer. PMEPA1 expression in colon adenomas was intermediate between normal colon and metastatic colon cancer (P < 0.05). Northern blot analysis of six paired normal colon epithelium and matched primary tumors of the colon also demonstrate that PMEPA1 was indeed expressed in colon cancer tissue at levels elevated above that of matched normal tissue (Fig. 9).

To determine the source of PMEPA1 expression in primary tumors of colon tissues, PMEPA1 in situ hybridization was used. Fig. 7,e shows a representative example demonstrating that PMEPA1 expression in colon cancer indeed arises from the malignant colon epithelial cell. Fig. 7,f shows the absence of staining of the same colon cancer by a control PMEPA1 sense probe. A semiquantitative analysis of the PMEPA1 in situ hybridizations of eight such paired normal and colon cancer samples confirmed that PMEPA1 expression in colon cancer arises from the colon epithelial cell and that PMEPA1 expression in colon cancer cells is similar in magnitude to that seen in normal colonocytes’ upper portions of the normal colonic crypt (Table 2). PMEPA1 expression in colon cancer is not a marker of either tumor stage or differentiation.

To further characterize the expression of PMEPA1 by colon cancer in vivo, we examined PMEPA1 expression in a group of colon cancer hepatic metastases. In previous studies, our group had demonstrated functional resistance to TGF-β in 9 of 10 cell lines derived from hepatic metastases and indeed had demonstrated TGF-β RII mutations accounting for this resistance in some of these cases (33).⁴ However, PMEPA1 in situ hybridization demonstrated PMEPA1 was highly expressed by malignant epithelial cells in 8 of 10 colon cancer liver metastases (two representative cases shown in Fig. 10) and was not expressed in any of the matched normal liver tissues (data not shown). Thus, as suggested by GeneChip® analysis, PMEPA1 expression is maintained even in metastatic colon cancers in what must be a TGF-β-independent manner.

To determine whether PMEPA1 transcripts in colon cancer were wild type, PMEPA1 was RT-PCR amplified and sequenced from 10 colon cancer cell lines derived from hepatic metastases and from 7 cell lines derived from colon cancer primary tumors. In each case, an RT-PCR-amplified PMEPA1 transcript was obtained and proved wild type. Although no PMEPA1 mutations were found, three PMEPA1 polymorphisms were revealed: (a) a G→C change at position 378 (changing glutamic acid to aspartic acid); (b) a silent A→G change at position 435; and (c) a silent A→G change at position 453 (Fig. 5 b).

The growth suppressor properties of TGF-β in colon epithelial cells are mediated at least in part by a TGF-β-induced growth arrest in the late G₁ phase of the cell cycle (36). A convenient surrogate for this arrest is the effect of TGF-β in suppressing E2F-mediated transcriptional activity as reflected by TGF-β squelching of a reporter construct derived from the E2F promoter that contains six consensus E2F-binding sites (37). To determine whether PMEPA1 played a role in the TGF-β-mediated G₁ arrest, PMEPA1 Variants A and B were each cotransfected with the E2F reporter into FET colon cancer cells. Although TGF-β treatment squelches E2F reporter activity by 5-fold, expression of the PMEPA1 variants with the E2F reporter was without inhibitory effect (data not shown). Moreover, expression of PMEPA1 Variants A and B showed no ability to rescue the E2F reporter from the TGF-β-mediated squelching. We thus conclude that PMEPA1 Variants A and B subserve the TGF-β-induced responses unrelated to TGF-β control of cell growth.

In summary, we have demonstrated that PMEPA1 is a direct target of induction by the TGF-β pathway in nontransformed colon epithelium and is a marker in vivo of terminally differentiated epithelial cells in the upper reaches of the colonic crypt. Paradoxically, we find that PMEPA1 expression is well maintained and actually modestly increased in both primary and metastatic colon cancer. Thus, transformed colon cells continue to recapitulate at least some patterns of gene expression that in the normal colonic epithelium are markers of terminally differentiated cells. PMEPA1 was initially described as an androgen-induced gene well expressed in normal and malignant prostate tissue (23). Subsequently, increased PMEPA1 expression on Northern blot analysis was noted in human renal cell and other malignancies. We have described here that the PMEPA1 locus encodes two alternative forms of the protein, which exhibit differing subcellular localizations. We find that the Variant B of PMEPA1 predominates in normal colon, colon cancer cell lines, TGF-β-stimulated colon cancer cell lines, a prostate cancer cell line, and an androgen-treated prostate cancer cell line and that this transcript is wild type in all colon cancers examined. The seeming paradox that PMEPA1 expression in the normal colon is a marker of terminal differentiation, but that PMEPA1 expression is well maintained in colon and other cancers, is in part resolved by our data that PMEPA1 itself does not appear to be directly involved in growth regulation. The actual function of PMEPA1 must, for the moment, remain speculative. However, PMEPA1 will be useful as a marker of a differentiated subpopulation within the colonic crypt, as an example of retention of differentiation markers in transformed colon cancer cells, and as an example of bifurcation of the TGF-β pathway between genes involved in tumor suppression and those that subserve other functions.

We thank Joan Massague for the kind gift of the E2F-lux reporter construct, David Danielpour for the RNA from androgen-treated LNCaP, and Rachel Kassai for assistance in propagating Vaco330 cells.