Mouse Embryo Fibroblasts Lacking the Tumor Suppressor Menin Show Altered Expression of Extracellular Matrix Protein Genes

Multiple endocrine neoplasia type 1 (MEN1), an autosomal dominant disorder, is characterized by multiple tumors of the parathyroid, anterior pituitary, and enteropancreatic endocrine tissues. At a lower penetrance, foregut carcinoids and adrenal tumors also can occur, as well as hormone nonproducing tumors, such as facial angiofibroma, truncal collagenoma, lipoma, leiomyoma, meningioma, and ependymoma (1). Loss-of-function mutations in the tumor suppressor gene MEN1 are responsible for the syndrome (2). Biallelic somatic mutations in the MEN1 gene are also often found in sporadic tumors of endocrine tissues, such as parathyroid adenoma, gastrinoma, insulinoma, and foregut carcinoid (3). The MEN1 gene encodes a 610-amino acid protein, which is ubiquitously expressed and resides primarily in the nucleus (4). The amino acid sequence of the protein, denoted menin, does not provide any clues as to its function.

Several proteins that bind menin directly or indirectly have been reported. Many of these are nuclear proteins involved in transcriptional regulation, including JunD, nuclear factor-κB, Pem, the COMPASS-like complex, Smad proteins 1, 3, and 5, RunX2, and histone deacetylases (5). The COMPASS-like complexes include the trithorax proteins mixed lineage leukemia 1/2, which possess histone methyl transferase activity. The menin-histone methyl transferase complexes methylate lysine 4 of histone H3 (H3K4), which is generally associated with activation of transcription (6, 7). Although the precise biochemical function of menin in this complex needs to be elucidated, the importance of menin in this complex is apparent from the observation that the loss of menin results in reduced H3K4 methylation and associated reduction in the expression of known mixed lineage leukemia targets Hoxc6 and Hoxc8 (6). Interaction of menin with Smad3 is reported to result in activation of transforming growth factor-β (TGF-β)-induced transcription (8), and the interaction of menin with Smad1, Smad3, Smad5, and Runx2 may participate in the TGF-β–mediated and bone morphogenetic protein-2–mediated pathways of commitment of multipotential mesenchymal stem cells to osteoblast lineage (9-11). Previous studies of gene expression in tumors, cell lines, and embryos have identified nonoverlapping sets of gene(s), whose expression is affected by menin (5, 12, 13). A recent report of chromatin immunoprecipitation-on-chip analysis has identified thousands of menin-binding promoter targets in the human genome (14).

The clinical features of the MEN1 syndrome are well emulated in the mouse MEN1 model. Mice heterozygous for Men1 alleles, on loss of the remaining allele as they age, develop endocrine tumors resembling the human MEN1 condition (15, 16). Homozygous loss of Men1 alleles results in embryonic lethality (E11.5-E13.5), indicating a requirement for menin in early development. Before demise, menin-null embryos are smaller in size than WT, often with body hemorrhages and edema, and a substantial portion of embryos reveal defects in neural tube closure as well as abnormal cardiac and liver development (15, 17).

Lack of cell lines from MEN1 tumors has been an impediment in evaluation of the molecular consequences of the loss of menin. To gain insight into the basic biological function of menin, we generated and characterized multiple mouse embryo fibroblast (MEF) cell lines lacking menin. Expression changes associated with the loss of menin provide insight into the importance of menin in early embryonic development.

The development of a mouse Men1 knockout model provided an opportunity to establish menin-null cell lines and to gain insight into the biological function(s) of menin. Mouse embryos null for both Men1 alleles do not survive beyond E11.5 to E13.5 (15). Therefore, MEF cell lines of wild-type (WT) and nullizygous genotype were established from early-stage (E9.5) embryos obtained by crossing mice heterozygous for the Men1 alleles (Men1 ΔN3-8/+). A total of four WT and six nullizygous (menin-null) cell lines as well as five subclones (N41.1-N41.5) of the menin-null cell line N41 were established. Four WT (W10, W26, W46, and W50) and four menin-null (N12, N17, N40, and N47) cell lines were used for characterization and expression analysis in this study. Both male (W10, W46, N12, and N40) and female (W26, W50, N17, and N47) embryos evenly represented the WT and menin-null MEFs. The presence or absence of Men1 expression, respectively, in WT and menin-null cell lines was shown at both the RNA and protein levels (Fig. 1A and B). Both reverse transcription-PCR (RT-PCR) for the Men1 coding region as well as a Western blot with menin antibodies against a COOH-terminal peptide show the absence of Men1 expression in menin-null MEFs (Fig. 1A and B).

Several MEF cell lines from WT (W10, W46, and W50) and menin null (N12, N17, N40, N47, and N41.4) were analyzed by spectral karyotyping (SKY) to identify any discernible chromosomal changes associated with the loss of menin. All the cell lines were found to be aneuploid, primarily tetraploid, although the cell lines were heterogeneous in chromosome number. Translocations were also seen in both genotypes; however, there was no consistent variation that could be associated with the homozygous loss of menin (data not shown).

We tested whether MEFs lacking menin would be tumorigenic by evaluating their ability to form colonies in soft agar. Five menin-null cell lines (N12, N17, N40, N47, and N41.4) along with the four WT (W10, W26, W46, and W50) were tested. None were capable of forming colonies in soft agar with the exception of N41.4 (a subclone of N41), which formed a few smaller colonies (data not shown). Thus, these menin-null cell lines in general did not show tumorigenic characteristics in culture.

Previous studies have shown that menin interacts with Smad3 and enhances TGF-β–induced transcriptional activation (8). We evaluated whether the menin-null cell lines lacking menin would respond poorly to TGF-β by using a Smad3-mediated, TGF-β–responsive luciferase reporter assay. The plasminogen activator inhibitor-1 gene promoter contains Smad3-binding sites. A luciferase reporter gene construct driven by this promoter (3TP-Lux) is expected to be activated by TGF-β. This construct was transfected into five menin-null and four WT MEF cell lines and the luciferase activity was quantitated with and without TGF-β treatment. The ratio of luciferase activity in the presence and absence of TGF-β (Fig. 1C) indicates that WT cells generally respond better than the menin-null cell lines (P < 0.05). Transient expression of menin in menin-null MEFs increases the luciferase activity (Fig. 1D), suggesting that the deficiency may be due to the absence of menin. This supports the prior observation that reduction in menin results in reduced response to TGF-β. Thus, the menin-null MEFs are suitable to explore the TGF-β–induced menin-mediated expression changes.

As a nuclear protein that interacts with many other factors that modify transcription, menin presumably affects gene expression of many targets. Establishment of the spontaneously immortalized MEF cell lines lacking menin allowed for a thorough investigation of transcriptional changes. The spontaneously immortalized cell lines were aneuploid, so we chose multiple cell lines (four or more for WT and menin null each) that would reduce the effect of clonal variations. RNA isolated from four WT and four menin-null MEF cell lines was labeled with Cy5 and Cy3 fluorochrome and hybridized to spotted arrays consisting of 21,120 cDNA clones representing 16,000 unique mouse transcripts. A total of 32 pairwise hybridizations were carried out between four WT and four menin-null cell lines, all with a dye swap protocol. In addition, two hybridizations with pooled WT cell lines against pooled menin-null cell lines, and three self-self hybridizations, were done for comparison purposes. On average, we obtained 15,082 (of 21,120) positive probes when we required both Cy3 and Cy5 channel intensity (background subtracted signals) to be >200. All the data from the expression arrays are deposited at the National Center for Biotechnology Information⁵

By requiring 2-fold or greater changes in the expression ratio (median ratio over all 32 hybridizations), and a Benjamini-Hochberg adjusted P value (t test) of <0.01, we obtained in menin-null MEFs 66 and 70 underexpressed and overexpressed genes, respectively. By eliminating redundant and probes with no annotation, we further reduced the unique probes to 54 and 56 for underexpression and overexpression (Figs. 2 and 3). As illustrated in these two figures, the majority of these genes also showed high expression ratio in pooled WT versus menin-null hybridizations (35 of 54 genes in Fig. 2 have expression ratios >2-fold in both pooled experiments, and first 20 genes are the same except Matn2). The expression level in three self-self control experiments showed no significant changes (expression ratio, 1.10 ± 0.16).

We then analyzed the Gene Ontology function enrichment by using the EASE program⁶

We confirmed the changes in expression of several transcripts independently by Northern blot, Western blot, and quantitative PCR. The two genes that showed the maximum change in menin-null cell lines were the underexpression of fibulin 2 (Fbln2) and overexpression of heat shock protein b1 (Hspb1/Hsp25; Figs. 2 and 3). Northern blot results were consistent with the array data for Fbln2 and Hspb1 (Fig. 4A and B), although there were some clonal variations in expression. Reduced expression in menin-null cells of another extracellular matrix protein gene periostin (Postn) was also verified by Northern and Western blot analysis (Fig. 4A and C). We also looked at the increased expression in menin-null MEFs of Cd24a protein, a cell surface marker, by fluorescence-activated cell sorting analysis of cells stained with antibodies for Cd24a (Fig. 4D). The results indicate that the menin-null cell lines generally express higher levels of Cd24a compared with the WT cell lines, although the W10 cell line turned out to be an exception, as was apparent from Fig. 3; quantitative PCR results also show unusual high Cd24a expression in W10 (see Fig. 5B).

We also evaluated expression of 14 different transcripts by quantitative PCR. RNA from individual cell lines as well as pooled WT and menin-null RNA samples (using equivalent amounts of RNA from each of the four cell lines) were evaluated. Expression in pooled WT and menin-null cells, as well as representation of the same data in terms of relative changes between WT and menin null, is shown in Fig. 5A: Cspg2, Igfbp3, Wnt5a, Pbx3, Sh3kbp3, Sgk, and Ccnb2 represented those that were lower in menin null and Btg2, Peg3, Cp, Vil2, and Clu represented the ones higher in menin nulls. The fold difference observed for most of these transcripts from the array data was consistent with that observed by quantitative PCR. Quantitative PCR for Mox2, Cd24a, Cspg2, and Btg2 in eight individual MEF cell lines is also shown (Fig. 5B) and was generally consistent with the array data.

We treated a WT (W10) and a menin-null (N47) MEF cell line with varying amounts (0, 2.5, 5, and 10 ng/mL) of TGF-β for varying lengths of time (12, 24, and 48 h), and expression of the three extracellular matrix proteins, Postn, Cspg2, and Fbln2, was evaluated (Fig. 6). RT-PCR, quantitative RT-PCR, and Western blot analysis indicate that Postn was induced by TGF-β in WT MEF but not in the null MEF (Fig. 6A). Response of Cspg2 (Fig. 6B) expression was similar to Postn, whereas that of Fbln2 (Fig. 6C) was not much different between W10 and N47 cell lines, and an increase in N47 was seen at 12 h after treatment. Therefore, the expression changes observed in Postn and Cspg2 in the WT and menin-null MEFs may reflect the TGF-β–induced, menin-mediated expression changes.

Two other studies have looked at the expression changes in mouse in WT and menin-null conditions in embryos or fibroblasts using Affymetrix arrays. Four genes (Hoxc6, Cyt19, Hoxc8, and Asb4) with reduced expression (1.29- to 1.56-fold) in menin-null embryos were identified (6). None of these appeared in our list, perhaps because of differences between whole embryos and MEFs. Menin-null MEFs, immortalized with retrovirus expressing human papillomaviral E6 and E7, and infected with retrovirus expressing menin, led to the identification of 25 transcripts each with >2-fold down (18) and up (12). Of these, Sdfr2 and Lgals3 were also found to be in our list of underexpressed and overexpressed transcripts, respectively, in menin-null MEFs. Differences in the transcripts printed on different array platforms, the methods in immortalization of MEFs may contribute toward an incomplete overlap of the transcripts observed. Our posting of the expression analysis on the entire set of 21,000 transcripts provides an opportunity for other investigators to compare their results with these data.⁷

The menin-null cell lines were found to respond poorly to TGF-β. This is consistent with earlier observations that blocking of menin function by antisense oligos results in blocking of TGF-β–mediated activation of a Smad3-responsive promoter (8). As discussed below, some of the known TGF-β/Smad3–induced target genes (19) in dermal fibroblasts were underexpressed in menin-null MEFs that included Igfbp3 and Nid1 and members of the collagen family (Col1a2, Col3a1, Col5a2, and Col6a1). In addition, Fbln2, Postn, and Cspg2 were also underexpressed in menin-null MEFs, are required for (or markers of) heart development, and are induced by TGF-β. Unlike in a menin-null MEF, expression of Cspg2 and Postn increased in a WT MEF on TGF-β treatment (Fig. 6), supporting that menin might play a role in TGF-β–mediated expression changes of some of the extracellular matrix genes.

Menin-null embryos display defects in neural tube, heart, skeletal, and liver development (15, 17). It is apparent from transcript profiling of embryonic fibroblasts that many genes affected by the loss of menin are important for extracellular matrix/space and cell adhesion. Among these genes, Fbln2 and Postn display the highest degree of down-regulation between menin-null and WT fibroblasts. Both Fbln2 and Postn are prominently expressed in the endocardial cushion tissue within the developing heart tube during embryonic days 10 to 13 (20, 21). In addition, both genes are expressed during skeletal development (22). Another extracellular matrix protein whose expression was similarly down-regulated in the menin-null fibroblasts is versican (Cspg2), a large proteoglycan that is critical for endocardial cushion tissue formation during cardiac development (23). The endocardial cushion tissue, which ultimately divides the straight heart tube into four chambers, is formed by the process of epithelial-mesenchymal transformation in response to signaling of the TGF-β and bone morphogenetic protein growth factor families (24). We suggest that the reduced expression of Fbln2, Postn, and Cspg2 in the absence of menin could contribute to the defects in heart morphogenesis of menin-null embryos.

The embryos were dissected and treated with trypsin (5 min at 37°C) to dissociate them into a single-cell suspension. The cell suspensions were plated onto gelatin-coated six-well plates with DMEM + 15% fetal bovine serum. The cells were allowed to immortalize and transferred to larger plates as needed. Genotypes were established by PCR of amnion DNA from embryo using primers as described earlier (25). Expression of Men1 was analyzed by RT-PCR, and reverse-transcribed RNA was used with primers in exons 2 and 6: TGCCCGATTCACCGCTCAGA (691F) and GCGCCATGGGGTATCTTTCC (1269R).

Equal amounts of whole-cell lysates (10-20 μg) were separated on a 12% Tris-glycine gels and transferred to nitrocellulose membranes. Menin antibodies were elicited in rabbits using the COOH-terminal peptide (4). Rabbit anti-Hsp25 (Hspb1; Stressgen Biotechnologies Corp.), monoclonal p84 (GeneTex, Inc.), and goat anti-rabbit and anti-mouse secondary (IgG-horseradish peroxidase) antibodies (KPL, Inc.) were from commercial sources. The membranes were visualized using SuperSignal West Pico Chemiluminescent substrate (Pierce Biotechnology, Inc.).

For determination of secreted proteins, 100 μL of serum-free culture medium were precipitated with 900 μL of 100% ethanol for 1 h at −70°C, run on 4% to 12% NuPAGE Bis-Tris gels (Invitrogen Corp.), and subjected to Western blotting using the ECL Plus detection reagents (GE Healthcare). Primary antibodies used were polyclonal antibodies against mouse OSF-2/Postn (R&D Systems).

WT (W10, W46, and W50) and menin-null (N14, N46, N47, and N41.4) cell lines were analyzed for chromosomal changes using spectral karyotyping as described earlier (26).

Soft agar colony-forming assays were done in duplicate as described previously (27). Cells were trypsinized, and 5,000 to 6,000 cells were resuspended in 3 mL of molten 0.33% agar and poured over a 5-mL 0.66% solidified layer of agar in 6-cm plates. The agar solutions were made in complete DMEM. Agar plates were incubated at 37°C in a 5% CO₂, humidified incubator and observed after 2 weeks for colony formation.

WT and menin-null MEF cell lines were plated in six-well plates at a density of 1 × 10⁵ per well. One microgram of the reporter construct (3TP-Lux, provided by the late Dr. Anita Roberts, National Cancer Institute, Bethesda, MD) was transfected using LipofectAMINE 2000 (Invitrogen). For evaluation of the effect of menin, 0.5 and 1 μg of either Men1 expression construct (pCMV-sport-Men1) or the vector (pCMV-sport) alone were cotransfected with 3TP-Lux construct. After 24 h, the medium was replaced with low-serum (0.2% fetal bovine serum) medium to reduce the amount of TGF-β that may be present in the fetal bovine serum. Ten hours later, 2.5 ng/mL TGF-β (R&D Systems) was added to the medium, and the next day, the cells were lysed. The reporter constructs were cotransfected with 1 ng of Renilla luciferase construct as an internal control. Luciferase levels were measured using the Berthold Centro LB 960 luminometer and reagents from the Dual-Luciferase Assay kit (Promega Corp.). The enzyme activity was normalized relative to Renilla luciferase activity.

The cDNA microarrays consisted of 21,120 cDNA clones that included 15,000 from embryonic cDNA libraries (28) representing 16,219 unique mouse UniGene clusters. cDNA microarrays were fabricated in the National Human Genome Research Institute microarray core facility, and one print batch of 100 slides was reserved for this study. MEF cell lines were grown to 70% confluence and harvested, and total RNA was extracted using Trizol reagent (Invitrogen) and RNeasy Maxi kit (Qiagen, Inc.). RNA was reverse transcribed, labeled by incorporation of Cy5 or Cy3 fluorochrome, and hybridized to the slides using standard protocols. For each menin-null MEF cell line (N12, N17, N40, and N47), eight two-channel hybridizations were done against all four WT MEF cell lines (W10, W26, W46, and W50) with standard fluorescent dye swapping protocol, obtaining a total of 32 expression profiles. For further comparison, two pooled RNAs (menin-null and WT pool) were cohybridized with dye swapping (two arrays), and three self-self hybridizations (menin-null pool/menin-null pool, WT pool/WT pool, and W50/W50) were done for quality control purpose. A total of 37 arrays were obtained. The hybridized microarray slides were scanned in the Agilent DNA Microarray scanner (model G2565AA; Agilent Technologies) and images were analyzed as described previously (29). The expression ratios for all the experiments are available online.⁸

Given the fact that all hybridizations were done with menin-null and WT pairs (ratios were flipped for dye swap hybridization), we chose single-group t test for significant differential expression with all 32 hybridizations in one group. The t test was done using Avadis 4.0 (Strand Genomics), with asymptotic P value and Benjamini-Hochberg correction for multiple tests.

Total RNA (20 μg) was fractionated on 1.2% agarose-formaldehyde gel, transferred to nylon membrane, cross-linked using UV, and hybridized using ExpressHyb (Clontech) with ³²P-labeled cDNA probes. The full-length cDNA inserts were isolated by NotI and MluI digestion of the plasmids: IMAGE clone IDs 3490759, 4457222, and 3498180 had cDNA inserts for Fbln2, Postn, and Hspb1/Hsp25, with insert sizes of 3,969, 3,109, and 862 bp, respectively.

Trypsinized cells (3 × 10⁶) were washed with PBS twice and suspended in 150 μL bovine serum albumin (1% in PBS) solution. Aliquots (50 μL, 1 million cells each) were treated for 1 h with 2 μL FITC-Cd24a or FITC-isotype antibodies (eBioscience), and the third aliquot was used as untreated control. Cells were washed and then suspended in 400 μL of bovine serum albumin solution. A FACSCalibur and CellQuest software (Becton Dickinson) was used to collect and analyze the data.

Quantitative PCR was carried out using the Taqman assay. cDNA was prepared using the SuperScript double-stranded cDNA synthesis kit (Invitrogen) and an oligo(dT) primer. cDNA template and Taqman probes (50 ng) were used for ABI PRISM 7700 (Applied Biosystems). Taqman probes were purchased from Applied Biosystems: Btg2 (Mm00476162_m1), wnt5a (Mm00437347_m1), Pbx3 (Mm00479413_m1), Vil2 (Mm00447761_m1), Peg3 (Mm00493299_m1), Men1 (Mm00484963_m1), Sh3kbp1 (Mm00451715_m1), Ccnb2 (Mm00432351_m1), Cp (Mm00432654_m1), Sgk (Mm00441380_m1), Cspg2 (Mm00490179_m1), Igfbp3 (Mm00515156_m1), Cd24a (Mm00782538_sH), Mox2 (Mm00487740_m1), Clu (Mm00437347_m1), Postn (Mm01289595_m1), and Fbln2 (Mm01188733_m1). β2-microglobulin (Mm00437762_m1) or Gapdh (Mm99999915_g1) was used as internal control. The relative mRNA expression level was calculated using the comparative expression level 2^−ΔΔCT method (CT is the threshold cycle; ref. 30). The experiments were repeated in triplicates, and the mean value with SD is reported.

WT (W10) and menin-null (N47) MEF cell lines were plated and replaced with low-serum (0.2% fetal bovine serum) medium after 24 h, and 12 h later, varying amounts of TGF-β were added to the medium (0, 2.5, 5.0, and 10.0 ng/mL), and cells were harvested 12, 24, and 48 h after treatment. Isolation of total RNA and reverse transcription were as described above. RT-PCR was done using primers for Postn (GCTCCTGTAAGAACTGG and CCCTTCCATTCTCATATA), Cspg2 (ACCAGCAGGTACACTCTGAAC and ACACATCATAGGTTTCCTGGGGAC), Fbln2 (GAACTTCTCGGATGCTGAGG and CAACTGGCCAGGGTGTTACT), and Gapdh (CAGTGGCAAAGTGGAGATTG and AATTTGCCGTGAGTGGAGTC).

We thank Amalia Dutra for help with spectral karyotyping, Martha Kirby for flow cytometry, Sreenath Taduru and Vyomesh Patel for guidance with mouse embryos, and ChenWei Wang for analysis of microarray data.