Mect1-Maml2 Fusion Oncogene Linked to the Aberrant Activation of Cyclic AMP/CREB Regulated Genes

Malignant salivary gland tumors can arise from a t(11;19) translocation that fuses 42 residues from Mect1/Torc1, a cyclic AMP (cAMP)/cAMP-responsive element binding protein (CREB)–dependent transcriptional coactivator, with 982 residues from Maml2, a NOTCH receptor coactivator. To determine if the Mect1-Maml2 fusion oncogene mediates tumorigenicity by disrupting cAMP/CREB signaling, we have generated in-frame deletions within the CREB-binding domain of Mect1/Torc1 for testing transformation activity and have also developed a doxycycline-regulated Mect1-Maml2 mammalian expression vector for global gene expression profiling. We observed that small deletions within the CREB-binding domain completely abolished transforming activity in RK3E epithelial cells. Further, we have shown that the ectopic induction of Mect1-Maml2 in HeLa cells strongly activated the expression of a group of known cAMP/CREB-regulated genes. In addition, we detected candidate cAMP-responsive element sites within 100 nucleotides of the transcriptional start sites of other genes activated by Mect1-Maml2 expression. In contrast, we did not observe alterations of known Notch-regulated target genes in these expression array profile experiments. We validated the results by reverse transcription-PCR in transfected HeLa, RK3E, and H2009 lung tumor cells and in mucoepidermoid cancer cells that endogenously express the fusion oncopeptide. Whereas overexpression of components of the cAMP pathway has been associated with a subset of human carcinomas, these data provide a direct genetic link between deregulation of cAMP/CREB pathways and epithelial tumorigenesis and suggest future therapeutic strategies for this group of salivary gland tumors.

A major emphasis in modern cancer research has focused on understanding the biological implications of somatic mutational events that occur sequentially during the development of specific malignant tumors. This approach has been successful in providing reagents for molecular diagnosis, in suggesting models for cancer gene pathways, and in identifying candidate targets for new anticancer agents. The discovery and analysis of new and unexpected mechanisms for tumorigenesis have also provided the opportunity to enhance our understanding of growth stimulatory and inhibitory cellular processes that may be important in a wide range of distinct human tumor types. An example of a somatic mutation with an undefined biological mechanism is the recent discovery that a t(11;19) chromosomal rearrangement creates a fusion oncogene product between the Mect1/Torc1 gene on chromosome 19p and the Maml2 gene on chromosome 11q in human mucoepidermoid carcinoma (1, 2). Because pathogenic markers are uncommon in sporadic epithelial carcinomas (3, 4), the recognition that a fusion Mect1-Maml2 gene product underlies the most common type of malignant salivary gland tumor may offer a valuable model for defining general molecular events leading to glandular epithelial carcinogenesis.

Mucoepidermoid cancer can arise from major glands, such as parotid and submandibular, or from minor mucous/serous glands that are scattered throughout the upper aerodigestive tract, including the pulmonary bronchial tree (5). Although mucoepidermoid carcinoma is a rare tumor, with an estimated incidence of 1,000 cases per year in the United States, the incidence of salivary cancers may be increasing slowly (5, 6). In addition, these data do not take into account the possible misdiagnosis of cases of pulmonary mucoepidermoid cancer, as well as the detection of Mect1-Maml2 expression in otherwise unrelated eccrine-like tumors (7). Aside from overexpression of the c-ERBB2 gene product (8, 9) and alterations in H-ras and p53 gene expression (10, 11) in a small subset of tumor samples, no other molecular marker has been associated with these cancers. Therefore, tumor-specific expression of the Mect1-Maml2 fusion gene, which has now been detected in 18 of 24 (75%) primary and derived tumor specimens (1, 2, 12), is the most common genetic event for these cancers.

Recent clues to understand how Mect1-Maml2 expression may be linked to human tumorigenesis were provided by data on the function of the normal Mect1/Torc1 and Maml2 genes. For example, the Mect1/Torc1 gene was shown to be a coactivator of cyclic AMP (cAMP)/cAMP-responsive element binding protein (CREB) signaling in two independent screens using large-scale arrayed cDNA methodology (13, 14). In contrast, the Maml2 gene is related to the Drosophila gene, Mastermind, and to the mammalian Mastermind-like gene, Maml1, and was shown to be an essential coactivator for NOTCH receptor transcriptional activation and signaling (1, 15). To determine if ectopic expression of the Mect1-Maml2 product alters the pattern of transcription of downstream target genes and to determine the effect of the fusion oncogene on specific transcripts normally regulated by either cAMP/CREB or NOTCH signal transduction pathways, we have developed a doxycycline-regulated mammalian expression vector for the Mect1-Maml2 transcript. Using global gene expression profiling, we have shown that Mect1-Maml2 strongly activates the expression of a set of genes that are largely regulated by cAMP/CREB signaling. These data, therefore, support a model system for Mect1-Maml2 tumorigenesis that involves the aberrant activation of downstream cAMP/CREB signaling genes. Whereas overexpression of components of the cAMP pathway has been associated with a subset of human carcinomas (16), this functional analysis of the recurrent t(11;19) translocation now provides a direct genetic link between deregulation of cAMP/CREB pathways and epithelial tumorigenesis, suggesting future therapeutic strategies for this group of salivary gland tumors.

Materials. GM6001, AG 1478, and H89 were purchased from Calbiochem (San Diego, CA), and gefitinib was from obtained from AstraZeneca (Wilmington, DE).

Description of cells and plasmids. To generate the pTRE-Mect1-Maml2 plasmid (pTRE-M-M2), a 3.5 kb fragment containing the Mect11-Maml2 full-length open reading frame was cloned in-frame into the SalI-NotI site of the pTRE-Myc plasmid (BD Biosciences-Clontech, Palo Alto, CA; ref. 17). HeLa cells stably expressing a tetracycline (Tet)-on plasmid, which contains the reverse Tet repressor fused to the carboxyl-terminal portion of the herpes simplex virus VP16 activation domain, were obtained from Clontech (Palo Alto, CA). H2009 and H360 (human lung carcinoma) and H292 and H3118 (human mucoepidermoid) tumor cell lines were propagated as previously described (18). RK3E cells were obtained from the American Type Culture Collection (Manassas, VA) and propagated in DMEM supplemented with 10% serum. Transfection of RK3E cells for foci formation was done as previously described (1).

Generation of stable transfected cell lines. HeLa Tet-on cells stably expressing the Tet-on plasmid were grown in DMEM containing 10% tetracycline-free fetal bovine serum (FBS) and 100 μg/mL G418 to maintain expression of the plasmid. These cells were plated in 100 mm plates and transfected with 10 μg of pTRE-M-M2 and 1 μg of the pTK-Hyg selection plasmid using LipofectAMINE reagent according to the manufacturer-recommended protocol (Invitrogen, Carlsbad, CA). Cells were washed after 3 hours and then 36 hours after transfection the media was supplemented with G418 and hygromycin. Media was changed every 3 days. After 3-week incubation in 37°C/6% CO₂, antibiotic-resistant, single-cell clones were transferred to separate 12-well plates for clonal expansion. Stable clones were tested for RNA and protein expression of M-M2 in both the absence and presence of 0.1 to 0.25 μg/mL of doxycycline.

Paired oligonucleotide target sequences for reverse transcription-PCR reactions. Mect1-Maml2 forward: 5′-ATG GCG ACT TCG AAC AAT CCG CGG AA-3′, reverse: 5′-CCA TTG GGT CGC TTG CTG TTG GCA GGA G-3′; matrix metalloproteinase 10 (MMP10) forward: 5′-AGT CTG CTC TGC CTA TCC TCT GAG-3′, reverse: 5′-CTT CAT ACA GCC TGG AGA ATG TGA G-3′; NR4A2 forward: 5′-GCT GTT GGG ATG GTC AAA GAA G-3′, reverse: 5′-TCG CCT GGA ACC TGG AAT AGT C-3′; GPI-anchored metastasis (C4.4A) sense: 5′-TCC CCG AAC AAG ATG AAG ACA G-3′, reverse: 5′-CAG GCA AGG ACA CAG TCA CAT TAG-3′; NR4A3 forward: 5′-TTC CCC TCC AGG TTC CAG TTA TGC-3′, reverse: 5′-TGG TGG TGG TGA TGG TGA TGG TAG-3′; amphiregulin forward: 5′-TGG TGC TGT CGC TCT TGA TAC TCG-3′ reverse: 5′-TCA CTT TCC GTC TTG TTT TGG G-3′.

Global gene expression profiling using oligonucleotide microarrays. Double-stranded cDNA and biotin-labeled cRNA were generated from 6 μg of total RNA (Qiagen RNeasy Mini Kit) using protocols recommended by the manufacturer (Affymetrix, Santa Clara, CA). The oligonucleotide microarray platform used was the GeneChip Human Genome U133A that has 18,400 transcripts and variants, including 14,500 known genes, which are represented by 22,000 oligonucleotide probe sets (Affymetrix GeneChip). Prehybridization of the probe arrays was done for 10 minutes in hybridization buffer (1× buffer: 100 mmol/L MES, 1 mol/L Na⁺, 20 mmol/L EDTA, 0.01% Tween 20 final concentration) at 65°C. The hybridizations were done at 65°C with 60 rpm rotation for 16 hours in 1× hybridization supplemented with 0.06 μg/μL fragmented cRNA (15 μg in total); 50 pmol/L control oligonucleotide B2; 1.5, 5, 25, and 100 pmol/L of 20× eukaryotic hybridization controls (bioB, bioC, bioD, cre); and 0.1 mg/mL herring sperm DNA. The probe arrays were sequentially washed with nonstringent (6× saline-sodium phosphate-EDTA, 0.01% Tween 20 final concentration) and stringent buffers (100 mmol/L MES, 0.1 mol/L Na⁺, 0.01% Tween 20 final concentration). The arrays were then sequentially stained with 2 mg/mL acetylated bovine serum albumin (BSA) and 10 μg/mL streptavidin phycoerythrin in 1× MES buffer and acetylated BSA and normal goat IgG in 1× MES.

Microarray image acquisition and data analysis. The probe arrays were scanned in an Affymetrix GCOS argon-ion Scanner at 488 nm. Statistical algorithms that consider both perfect and mismatch probes were used to create expression analysis files and the program normalized the data using a global scaling method based on a trimmed mean. Statistical analyses were done using the BRB ArrayTools Version 3.2.0 (BRB-Array Tools Users Guide, version 3.2; R. Simon and A. Peng Lam, Biometric Research Branch Technical Report 7, National Cancer Institute)¹

. We did class comparison analyses to determine which genes were significantly different in expression values following 24, 48, and 72 hours of doxycycline induction in independent HeLa cell clones that stably express the Mect1-Maml2 fusion protein (expressor clones) compared with independent HeLa cell clones that retained antibiotic resistance but expressed undetectable Mect1-Maml2 protein by reverse transcription PCR (RT-PCR) and protein immunoblot analyses (nonexpressor clones). In addition, class comparison analysis was done on gene expression obtained from the individual Mect1-Maml2 expressor clones induced in the presence of doxycycline compared with the same HeLa cell clones grown in the absence of doxycycline. For class comparisons, random variance t test was used (19). Genes were considered statistically significant if their P value was <0.001, which controlled for the number of false discoveries. That is, if N is the number of genes that passed the filters, then the expected number of false discoveries would be 0.001 N. If M is the number of genes significant at the 0.001 level, then the proportion of false discoveries would be 0.001 × N / M. We also did a global test of whether the expression profiles differed between the classes by permuting the labels of which arrays corresponded to which classes. For each permutation, the P values were recomputed and the number of genes significant at the 0.001 level was noted. The proportion of the permutations that gave at least as many significant genes as with the actual data was the significance level of the global test (20). Clustering of genes and samples were done using the BRB clustering tools by centering the genes, using a one minus correlation, using an average linkage, and by selecting a subset of genes that were significantly different in the class comparisons.

Cell proliferation assay. The H2009, H292, H3118, and H360 cells were plated at 1,000 cells/well in 96-well plate in RPMI 1640 supplemented with 10% (v/v) FBS overnight, and agents were added on the next day. The cells were subsequently incubated for 4 days and the cell viability was assessed by WST-1 assay according to the instruction of the manufacturer (Roche Applied Science, Indianapolis, IN).

Mect1/Torc1 domain is required for transformation in RK3E cells. Insight into the functional consequences of the Mect1-Maml2 fusion transcript can be inferred from recent data suggesting that the chimeric gene product can potentially disrupt both cAMP/CREB and NOTCH signal transduction pathways. For example, a cDNA (called Torc1) was isolated from high-throughput reporter assays designed to identify transcriptional cAMP/CREB coactivators and was found to be identical to the Mect1 gene on chromosome 19p13 (13, 14, 21). Importantly, the minimal region required for CREB binding was localized to the amino terminal 42-amino-acid exon 1, which is the only domain of Mect1/Torc1 that is included in the oncogenic fusion peptide (13, 21). In contrast, the Maml2 gene, localized to chromosome 11q21, is a homologue of the Drosophila NOTCH coactivator, Mastermind, which has been shown to be required for normal NOTCH signaling in Drosophila and mammalian cells (1, 15). Mect1-Maml2 was subsequently shown to suppress the basal NOTCH/Mastermind activation of a promoter containing tandem CSL/NOTCH binding sites, which suggested that the fusion protein may be exerting a dominant-negative effect (1). Surprisingly, Mect1-Maml2 also activated the Hes1 promoter, a NOTCH receptor target gene, in the absence of NOTCH ligand and NOTCH/CSL binding sites. This observation initially suggested an alternate hypothesis for Mect1-Maml2 that involved the constitutive deregulation of NOTCH signaling (1). Recent data, however, has revealed that the Hes1 promoter contains a cryptic cAMP/CREB response DNA binding element (13, 22), suggesting the hypothesis that the biological effect of Mect1-Maml2 may be the aberrant activation of cAMP signaling through activation of selected CREB target genes.

To examine the importance of the amino-terminal Mect1/Torc1 region in Mect1-Maml2 tumorigenesis, we generated two different in-frame deletions within this 42-residue domain of the chimeric oncogene (Fig. 1). We had previously reported the ability of wild-type Mect1-Maml2 to efficiently generate foci formation in RK3E rat epithelial cells at 3 weeks, whereas full-length Maml2 or the parental vector alone had no transforming activity (1). We observed that both deletions (20 or 14 residue deletions) within the CREB-binding domain of Mect1/Torc1 exon 1 resulted in complete loss of foci formation. These data support the hypothesis that the CREB-binding domain within Mect1/Torc1 exon 1 plays an essential role in the ability of Mect1-Maml2 to transform epithelial cells and rejects an alternate hypothesis that proposed that the role of the t(11;19) chromosomal translocation was to simply truncate and deregulate Maml2 activity under control of the Mect1/Torc1 promoter. Whereas a small, in-frame deletion (residues 77-118) within the Maml2 sequence had no effect on transformation ability, serving as a negative control, we also observed that a larger deletion (residues 703-838) within the conserved carboxyl-terminal Maml2 transactivation domain showed loss of foci formation, consistent with the model that Mect1-Maml2 functions to aberrantly transactivate selected target genes.

Global gene expression profiling using doxycycline-inducible Mect1-Maml2. To further study possible mechanisms underlying Mect1-Maml2 tumorigenesis, we have generated a doxycycline-inducible Mect1-Maml2 expression vector in HeLa cells to study the pattern of mRNA expression following induction of ectopic Mect1-Maml2 protein. We transfected the pTRE-M-M2 plasmid containing the full-length Mect1-Maml2 open reading frame and isolated five independent clones following selection with G418 and hygromycin. Three of these clones expressed high or moderate levels of the Mect1-Maml2 protein following doxycycline expression (expressor clones 1-3), whereas two other clones expressed undetectable (or extremely low) levels of Mect1-Maml2 protein despite doxycycline exposure (nonexpressors clones 4-6; Fig. 2). We selected these independent paired cell clones for further analysis because the expressor and nonexpressor clones could be maintained continuously under identical antibiotic (G418, hygromycin, and doxycycline) incubation conditions during the harvesting of RNA. We generated total RNA from independent frozen stocks of these transfected clones after exposure to doxycycline for 24, 48, or 72 hours to use as labeled probes in oligonucleotide microarray analyses. We planned to first generate a set of candidate genes that were differentially regulated in doxycycline-treated Mect1-Maml2 expressor versus doxycycline-treated, nonexpressor cell clones. This set of genes could then be independently validated for differential expression in the individual Mect1-Maml2 expressor clones that were incubated in the presence versus the absence of doxycycline.

Ectopic Mect1-Maml2 induces cAMP/CREB-regulated genes. A total of 164 genes were differentially expressed at P = 0.001 between Mect1-Maml2 expressor and nonexpressor HeLa clones of which we selected the first 100 genes with the lowest parametric P value (Fig. 3). Of the differentially expressed genes, 95% showed increased expression following Mect1-Maml2 induction consistent with a model for target gene activation by the chimeric oncogene. Strikingly, no gene was repressed >2-fold in this data set. Only five genes in this data set were induced ≥8-fold by ectopic Mect1-Maml2 protein expression (Table 1). These included cytosolic phosphoenolpyruvate carboxykinase (PEPCK1/PCK1), 171-fold induction; amphiregulin (AREG), 27-fold; GPI-anchored metastasis-associated gene (C4.4A), 17-fold; MMP10, 14-fold; and nuclear receptor subfamily 4, group A, member 3 (NOR1, NR4A3), 8-fold. Importantly, three of these five genes, PEPCK1/PCK1, AREG, and NR4A3/NOR1, are previously published cAMP/CREB target genes (23, 24). In addition, inspection of the promoter sequences of the human GPI-anchored metastasis gene (C4.4A) revealed a variant cAMP responsive element (TGACG) at −82 relative to the transcriptional start site and alignment of the promoter region from the mouse C4.4A orthologue also showed the same cAMP-responsive element (CRE) element sequence at −81 from the mouse transcriptional start site. Many CREB-inducible genes are regulated through variant or half-palindromic sites (25) and a recent genome-wide analysis showed that 70% of CREB-binding loci were within 1 kb of these half-palindromic CRE sites (22), suggesting that the conservation of this sequence within the murine and human GPI-anchored gene is functionally relevant. In addition, Mect1/Torc1 was recently noted to activate the interleukin-8 (IL-8) promoter through a cryptic CRE-like variant site at −69 relative to the IL-8 start site (14).

Fourteen additional genes were induced ≥4-fold among this data set (Table 1). These genes include human chorionic gonadotropin (HCG), IL-6, nuclear receptor subfamily 4, group A, member 2 (NR4A2/Nurr1), and c-MAF, which are also previously known CREB-regulated genes (22, 23, 26, 27). Moreover, inspection of the cystathionase (CTH) and keratin 17 promoters shows the (TGACG) CRE sites at nucleotide position −150 from the CTH transcriptional start site and at −50 and −220 from the keratin 17 start site. These latter sites, however, were not conserved in the promoter region of the corresponding murine homologues and, therefore, are of uncertain significance. Several other previously reported CREB-inducible genes were also detected at lower levels of gene induction, including crystallin α-β (3.8-fold induction; ref. 14), c-fos (3.0-fold; ref. 28), IL-8 (2.2 fold; ref. 14), and dual-specificity phosphatase-1 (1.8-fold induction; refs. 23, 29). We also recognize that certain gene activations may be events that are indirectly linked to CREB activation. For example, IL-6 is known to activate several genes, including MMP10/stromelysin 2 (30) and ID2 (31), which we detected in our data set. Whereas we limited our search for potential CRE sites to within 500 bp of the transcriptional start site, another group recently published an algorithm for a genome-wide analysis of candidate CRE sites that scored for positional characteristics, clustering of sites, and mammalian conservation (32). This analysis, which included 5,000 bp upstream of the transcriptional start site and 300 bp of exon 1 sequences, identified several more genes from our list of Mect1-Maml2–induced transcripts (Table 1). Of interest, we did not detect induction or repression of Notch target genes, such as Hes gene family members, nor did we detect repression of genes >2-fold in the data set list of the 100 highest scoring genes. These observations, therefore, further support the model that the Mect1-Maml2 product targets CREB-binding promoters using its amino-terminal Mect1/Torc1 domain to aberrantly activate certain cAMP-responsive genes during tumorigenesis.

We have taken two different approaches to confirm the global gene expression data obtained from the comparison of the different Mect1-Maml2–expressing versus Mect1-Maml2–nonexpressing cell clones. First, we harvested RNA from each of the three individual Mect1-Maml2 expressor HeLa clones in the presence or absence of doxycycline and pooled the gene expression pattern using the same 18,400 oligonucleotide array. Of 80 genes that were identified at the P < 0.001, 10 genes were induced ≥4-fold. These genes were as follows: PEPCK1/PCK1, P = 2.2 × 10⁻⁵, 20-fold induction; AREG, 1 × 10⁻⁵, 14-fold induction; MMP10, 4 × 10⁻⁶, 12-fold induction; GPI-anchored associated, 1 × 10⁻⁷, 8-fold induction; CTH, 2.2 × 10⁻⁶, 6-fold induction; NR4A2, 1 × 10⁻⁷, 5.4-fold induction; HCG, 7 × 10⁻⁴, 5-fold induction; crystallin α-β, 1 × 10⁻⁴, 5-fold induction; NR4A3, 1 × 10⁻³, 4-fold induction; and cytokeratin 2, 4.6 × 10⁻⁵, 4-fold induction. These observations confirmed the identification of the same set of activated genes that were detected in the earlier comparison of Mect1-Maml2 expressor and nonexpressor clones.

To further confirm that these transcripts identified by gene array profiling are bona fide Mect1-Maml2 targets, we harvested RNA from expressor clones for direct confirmation by RT-PCR analysis using paired oligonucleotide primers (Fig. 4A). We tested a fresh source of RNA from a separate stock of one of the Mect1-Maml2 expressor clones, which had been treated with doxycycline for 24, 48, or 72 hours, to determine if we could detect induction of the PCK1, AREG, GPI-anchored metastasis associated, and MMP10 candidate genes. RT-PCR analysis showed maximal gene induction at 48 and 72 hours postexposure to doxycycline induction. We also observed no gene induction when the cells were incubated for the same time period in the absence of doxycycline, nor when RNA from a nonexpressor Mect1-Maml2 clone was tested under identical conditions (Fig. 4A). To quantitate the induction over time for seven cAMP-regulated genes, we pooled the combined oligonucleotide array hybridization data from all three Mect1-Maml2 expressor clones at 24, 48, and 72 hours. We observed a consistent increase in gene induction between 48 and 72 hours for each of the induced genes: PEPCK (1.41-fold increase from 48 to 72 hours), MMP10 (1.50-fold increase), GPI-anchored/C4.4 (1.34-fold increase), NR4A2 (1.08-fold increase), amphiregulin (1.44-fold increase), NR4A3 (1.21-fold increase), and IL-6 (1.18-fold increase). These observations validate the experimental design and data obtained from our global gene profiling and have shown the identification of a group of cAMP/CREB-regulated genes that are aberrantly activated by the oncogenic Mect1-Maml2 product.

To test the effect of Mect1-Maml2 expression in a different cell lineage, we first obtained RNA from the immortalized RK3E rat epithelial cell line or from RK3E foci that were generated and expanded following transfection with a Mect1-Maml2 plasmid as previously described (1). Similar to our data with HeLa cells, we observed induction of NR4A3, GPI-anchored/C4.4, and MMP10 in the transfected RK3E cells (Fig. 4B). We also transfected either Mect1-Maml2, full-length Maml2, Mect1/Torc1, or the control parental constitutive cytomegalovirus promoter plasmid into the H2009 lung adenocarcinoma cell line and harvested RNA at 72 hours. We observed induction of the CREB-regulated genes PCK1 and NR4A3 expression following transfection of Mect1-Maml2, but not with any of the other vectors, confirming the induction of these CREB target genes by exogenous Mect1-Maml2 expression in a different cell type (Fig. 4C). The inability of Mect1/Torc1 to induce expression of these target genes in the transfected H2009 cells may reflect a more potent transcriptional activation domain in the Mect1-Maml2 fusion, which was derived from the carboxyl-terminal region of the Maml2 product (1). We also tested RNA from two different mucoepidermoid cancer cell lines, H292 and H3118, that we had previously shown to contain a t(11;19) translocation with endogenous expression of Mect1-Maml2 transcripts (1). We detected expression of CREB-regulated genes by RT-PCR in both tumor cell lines (Fig. 4D), suggesting that Mect-Maml2 may be aberrantly activating a set of cAMP inducible genes in vivo.

Inhibition of signaling pathways in mucoepidermoid cancer. The presence of CRE or variant, half-palindromic CRE sites in multiple target genes (22, 25) has suggested that CREB functions to regulate energy homeostasis, growth factor signaling, cell survival, and cell-to-cell communications through the coordinated expression of distinct genes. The mechanisms that allow the selective activation of distinct target genes in different tissues, however, remains undefined and may involve chromatin modifications (33) or the activity of accessory cofactors that promote either CRE occupancy or promoter activation. Our global gene profiling data suggests that Mect1-Maml2 expression can potently induce a small group of CREB-inducible target genes that might serve as candidate effectors of tumorigenesis in cancers that have acquired the t(11;19) chromosomal rearrangement. For example, NOR1/NR4A3 is a known oncogene that is tightly linked with the development of chondrosarcoma (34), whereas AREG, MMP10, GPI-anchored metastasis associated gene, IL-6, and PCK1 are associated with cell proliferation, invasion, or metastasis (30, 35–39). We were interested in the CREB-inducible epidermal growth factor receptor ligand, amphiregulin, (a) because we had determined that it was induced >20-fold by Mect1-Maml2 expression, (b) because small molecule, competitive inhibitors for this pathway were available for in vitro testing, and (c) because activation of EGFR signaling and expression of amphiregulin had been previously reported in H292 cells that express endogenous Mect1-Maml2 (1, 40, 41). To test the sensitivity of the mucoepidermoid cancer cell lines H292 and H3118, as well as the nonmucoepidermoid lung cancer lines H2009 and H360, we incubated the cells with a variety of different small molecules that can inhibit either EGFR (amphiregulin) or PKA (cAMP-dependent kinase) pathways (Fig. 5). We observed dose-dependent decreases in proliferation of H292 and H3118 cells with two different inhibitors of EGFR, AG1478 or gefitinib. At the highest dose tested (1 μmol/L), AG1478 inhibited proliferation by 78% in H292 cells. In the case of gefitinib, sensitivity in lung cancer cell lines has been defined as an IC₅₀ <1 μmol/L (42), and we observed that gefitinib inhibited proliferation of H292 cells by 69% at 0.1 μmol/L and 78% at 1 μmol/L. The PKA inhibitor, H89, inhibited proliferation by 78% (Fig. 5A,, top left). H3118 cells, however, were less responsive to EGFR or PKA inhibition than H292 cells, but were still inhibited by 40% to 50% with these agents (Fig. 5A,, top right). Because responsiveness to low doses of gefitinib can be associated with certain acquired EGFR mutations (43, 44), we documented that these cell lines showed a wild-type sequence in exons 18 to 21 (data not shown). In contrast, we found that two tumor cell lines that lack expression of Mect1-Maml2 were resistant to EGFR inhibition; however, the H360 cells were still sensitive to H89 treatment (Fig. 5A , bottom). In addition, a recent study showed that activation of MMPs contributes to activation of EGFR in H292 cells (45). Because we also detected induction of MMP10 in our gene array experiments, we tested whether the proliferation of H292 or H3118 cells is dependent on MMP activity. We observed that treatment of cells with 1 μmol/L GM6001, a pan-MMP inhibitor, resulted in growth inhibition in both H292 and H3118 cells by 45% to 50% (data not shown). These data, whereas preliminary, suggest that the targeting of either global cAMP/CREB signaling through PKA inhibition or CRE decoy strategies (16) or by the targeting of accessible, downstream Mect1-Maml2–inducible gene pathways may offer a new approach for the systemic treatment of these malignant salivary gland tumors.

In summary, evidence that Mect1-Maml2 is an important oncogene that underlies the development of malignant salivary gland tumors includes the observations that (a) the t(11;19) rearrangement has been detected as the sole cytogenetic abnormality in several primary tumor samples (12, 46, 47), (b) Mect1-Maml2 can efficiently transform rat epithelial cells in vitro whereas the reciprocal Maml2-Mect1 transcript was not present in primary or derived tumor samples (1), and (c) Mect1-Maml2 is expressed in 75% of all mucoepidermoid cancers tested to date (1, 2, 12). The present study shows that the CREB-binding amino-terminal domain of Mect1/Torc1 is required for transforming activity. Further, we have shown that ectopic expression of the chimeric Mect1-Maml2 transcript induces the activation of genes that are either known cAMP/CREB targets or genes that contain CRE sequences near their transcriptional start sites and may be previously unrecognized CREB-regulated targets. These observations support the model where Mect1-Maml2 binds to CREB-regulated promoters to deregulate a subset of cAMP-inducible genes during tumorigenesis, thus providing a direct genetic link between an oncogenic chromosomal translocation and cAMP/CREB activation. Ultimately, these data may provide clues for preclinical opportunities in treating this class of malignant salivary gland tumors.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Richard Simon for his help with the statistical analysis of the data and review of the manuscript.