Identification of c-Myc Responsive Genes Using Rat cDNA Microarray

EST 2 Clones and Cell Lines.

Our rat microarray contained 4400 EST cDNAs clones selected from the Rat Gene Index of the Institute for Genomic Research. A small set of redundant clones (5%) were included for quality control. Selected clones were from 52,000 3′ ESTs generated from 12 different libraries constructed from 10 distinct tissues and two PC12 cell lines. All clones are available through the American Type Culture Collection. The clones used for our microarray were sequence-verified by generating both 5′ and 3′ ESTs. The EST sequences have been deposited into the GenBank database dbEST. 3

The somatic c-myc null cells (Ho15.19) and its parental c-myc wild-type Rat1 cells (TGR-1) were from John Sedivy (Brown University, Providence, RI) and cultured as described (4) . The log-phase cells were harvested at 60% confluence. For serum starvation condition, the cells were grown to confluence and then cultured without serum for 48 h. The cell lines overexpressing c-Myc were generated by cotransfection of pBabe-puro vector and CM19, which contains genomic c-myc gene under the control of a long terminal repeat promoter, into Ho15.19 or TGR-1 cells.

cDNA Microarray and Data Analysis.

We fabricated our rat cDNA microarrays essentially as described previously (6) . The plasmids of EST clones were purified using the QiaPrep Turbo kit (Qiagen). The cDNA inserts were amplified by PCR for 30 cycles with M13 froward and reverse primers. The PCR products were verified by gel electrophoresis, purified, and arrayed robotically onto polylysine-coated glass microscope slides with a GeneMachine arrayer (San Carlos, CA). The microarrays were then postprocessed to denature and immobilize the DNA (6) .

The cDNA probes were made from total RNAs with cy5- or cy3-dUTP (Amersham Pharmacia) and Superscript II polymerase (Life Technologies, Inc.) as described (6) . Microarrays were hybridized with probes, washed, and then scanned using an Axon scanner (Foster City, CA), with the sample intensities for cy5 and cy3 collected separately. The normalization of the sample intensities and calculation of the calibrated fluorescence ratios after background subtraction were as described (7 , 8) .

Northern Hybridization.

The cDNA probes were made by random primer labeling. Fifteen μg of total RNA were separated on a formaldehyde 1.2% agarose gel, transferred to a Hybond-N nylon membrane (Amersham), and then immobilized by UV irradiation. The hybridization was carried out using ExpressHyb solution (Clontech), and bands were quantitated using a PhosphorImager (Molecular Dynamics).

Some Genes Are Differentially Expressed between c-myc Wild-Type and c-myc Null Cells.

Because c-Myc is actively expressed in growing cells, we first compared gene expression profiles between c-myc null cells (Null) and its parental c-myc wild-type cells (WT) during log-phase growth, as diagrammed in Fig. 1 ⇓ . Using the criteria of at least 2-fold difference in expression in two separate experiments, we found 319 reproducible gene expression differences, representing 283 distinct genes, between WT cells and Null cells during log-phase growth. One hundred eighty-eight of these genes were up-regulated, and 95 were down-regulated by 2.3–15-fold in WT cells. That 7.3% (319 of 4400) of the gene elements on the microarray were differentially regulated is highly significant, because the false-positive rate of our system is only 0.2% (8 of 4400) at an expression threshold of 2-fold difference.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Experimental scheme and summary of results. Some of the known genes were represented by more than one EST clone.

Ectopically Expressed c-Myc Can Recover the Gene Expression Profiles in the c-myc Null Cells.

To validate that the differential gene expression we found truly depends on the expression of c-Myc, we ectopically expressed c-Myc in the c-myc null cells (Null^+MYC) and compared the gene expression between Null^+MYC cells and Null cells (Fig. 1) ⇓ . We found that these c-myc-reconstituted cells behaved like their wild-type counterparts. As shown in Fig. 2 ⇓ , the gene expression profile of Null^+MYC cells resembled that of the WT cells, because they shared 198 differentially expressed genes (represented by 227 EST clones on the microarray) when each was compared with Null cells at log phase (Figs. 1 ⇓ and 2 ⇓ , A and B). The correlation between the Null^+MYC/Null comparison and the WT/Null comparison (r = 0.854) is very similar to that between two repeated WT/Null comparisons (r = 0.846; Fig. 2, C and D ⇓ ). By comparing log-phase Null^+MYC cells and Null cells, we were also able to eliminate those gene expression differences attributable to potential clonal variations between log-phase WT and Null cells. Therefore, these 198 overlapping genes, which were differentially expressed between log-phase WT and Null cells, and also between Null^+MYC and Null cells, represent a stringent list of c-Myc-responsive genes in log-phase growth. Although all genes listed are responsive to the presence of c-Myc, we acknowledge that only some of these genes will be direct targets of c-Myc, and others will be indirect targets regulated by other c-Myc-responsive gene products. Our current experimental approach cannot distinguish between the two mechanisms of induction but identifies the universe of RNAs whose levels are correlated with the expression of c-Myc.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Comparisons between WT and Null cells or between Null^+MYC and Null cells during log-phase growth. A, comparison of WT and Null cells at log phase. A section of the microarray is shown. Total RNA from WT cells was labeled with cy5-dUTP; total RNA from Null cells was labeled with cy3-dUTP. B, comparison of Null^+MYC and Null cells at log phase. The same section of the microarray as in A is shown. Total RNA from Null^+MYC cells was labeled with cy5-dUTP; total RNA from Null cells was labeled with cy3-dUTP. C, scatter plot of the calibrated ratios (log scale) in the two independent comparisons between WT cells and Null cells. D, scatter plot of the calibrated ratios (log scale) in the comparisons between WT cells and Null cells (vertical axis) and between Null^+MYC cells and Null cells (horizontal axis).

Of these 198 c-Myc-responsive genes, 124 genes were up-regulated by c-Myc, which represent 96 known genes and 28 unnamed ESTs. There were 74 genes down-regulated by c-Myc, representing 46 known genes and 28 unknown genes. The known genes are listed in Tables 1 ⇓ and 2 ⇓ . 4 Among the c-Myc-responsive genes we observed, cyclin D1, Cdc2, LDH-A, and MHC-Iα have been reported previously as c-Myc targets (2) . Cyclin D1 has been reported to be either up-regulated or down-regulated by c-Myc in different studies (9 , 10) . In our cDNA microarray study, cyclin D1 was consistently down-regulated by c-Myc in WT cells compared with Null cells, and this was confirmed by Northern hybridization (Fig. 3) ⇓ . The Myc-mediated decrease of cyclin D1 expression was also observed when Null cells were compared with Null^+MYC cells or WT cells were compared with wild-type cells overexpressing c-Myc (WT^+MYC; Fig. 3 ⇓ ). Although it has been reported that the expression of LDH-A is not changed between c-myc wild-type cells and c-myc null cells (3) , we found that LDH-A is consistently induced in WT cells or Null^+MYC cells when they were compared with Null cells in our microarray analysis. This result was also confirmed by Northern hybridizations (Fig. 3) ⇓ . In addition to cyclin D1 and LDH-A, we chose 10 other genes from Tables 1 ⇓ and 2 ⇓ and examined their expressions by Northern hybridizations in Null, Null^+MYC, WT, and WT^+MYC cells. Except for the two genes whose expression was not detectable by Northern, all others were differentially expressed, in concordance with our microarray results (Fig. 3) ⇓ . In a limited number of instances, reported c-Myc target genes did not appear to be differentially expressed between WT and Null cells in our microarray analysis (e.g., RCC1, ODC, and Cad). Several factors might explain these discrepancies:

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Confirmation of differentially expressed genes in microarray analysis by Northern hybridizations. (−/−), Null cells; (+/+), WT cells; (−/−)+Myc, Null^+MYC cells; (+/+)+Myc, WT^+MYC cells. The cDNA probes were: i, c-myc; ii, CPD (carboxypeptidase D); iii, Cyclin D1; iv, GCN1; v, PGAM (peptidylglycine α-amidating monooxygenase); vi, rp-S12 (ribosomal protein S12); vii, rp-L6 (ribosomal protein L6); viii, NPM (nucleophosmin or nucleolar phosphoprotein B23); ix, α-NAC (nascent polypeptide associated complex); x, LDH-A; xi, 28S rRNA (inverted picture of agarose gel stained with ethidium bromide solution); xii, actin.

(a) The criteria we set forth for defining c-Myc-responsive genes were very stringent. A greater than 2-fold difference in gene expression was required in repeated comparisons of WT and Null cells and also in the comparison between Null^+MYC cells and Null cells. These stringent inclusion criteria may exclude potentially real differences. For example, the gene expression differences for Cad in the three experiments were 2.81, 1.57, and 2.36 (ratios). Because the difference in one of the experiment is <2-fold cutoff (1.57), Cad was not considered positive.

(b) Some of the previously identified Myc target genes were found in different cellular systems and may not be regulated in the same manner as in the rat fibroblast cells.

(c) All of the previously reported c-Myc target genes were found in c-Myc overexpression systems. Cells may respond differently to overexpressed c-Myc than to physiologically expressed c-Myc. In fact, some of these discrepancies have been reported previously in other systems (3) .

c-Myc Regulates the Expression of Genes Involved in Protein Synthesis and Metabolism.

c-Myc has been considered an important regulator of the cell cycle machinery. We observed, however, that only a few cell cycle-associated genes were regulated by c-Myc. c-Myc up-regulates CDC2, PCNA, CKS2, AIM1, and DNA polymerase α but paradoxically down-regulates cyclin D1. Unexpectedly, however, we found that a majority of the c-Myc target genes (61 of 96 c-Myc-induced genes with known functions) are those involved in macromolecular synthesis, protein turnover, and metabolism (e.g., leucyl-tRNA synthetase, glycerate dehydrogenase, enolase, and lactate dehydrogenase, ubiquitin, E2-EPF ubiquitin-carrier protein, and heat shock proteins). Interestingly, the largest category of c-Myc-induced genes were those involved in protein synthesis, including 30 of the 43 distinct ribosomal protein genes present on our microarray. In addition to the 43 ribosomal protein genes, there are 20 other genes represented on our microarray involved in protein synthesis. Taken together, 57% of genes involved in protein synthesis (36 of 63) were identified as overexpressed, despite of the fact that these protein synthesis genes represented only 1.4% of the cDNAs on the microarray. It is, therefore, very unlikely that the induction of ribosomal/protein synthesis-associated genes by c-Myc is attributable to their overrepresentation on our microarray.

Because ribosomal genes are known to be up-regulated by cell growth, we sought to minimize the general effects of growth by comparing WT and Null cells under serum-starved conditions. In this state, c-Myc is also expressed at low levels in WT cells, as confirmed by Northern hybridization (Fig. 3) ⇓ . Twenty-nine genes differentially expressed between WT and Null cells during serum starvation overlapped with the c-Myc-responsive genes found in log phase (Tables 1 ⇓ and 2) ⇓ . Among these 29 genes, 9 of 19 c-Myc up-regulated genes were ribosomal protein genes. Supporting these findings, recent studies have identified other genes involved in protein synthesis not represented in our arrays that appear to be regulated by c-Myc: eIF-2a, eIF4E, eIF5A, eIF4G, MrDb, nucleolin, isoleucine-tRNA synthetase, and ribosomal protein S11 (2 , 11) . Thus, an important function of c-Myc is in activating pathways of protein synthesis and metabolism in addition to cell cycle.

This conclusion is in concordance with results from recent genetic studies of c-myc in Drosophila[ d-myc; (5)] . The phenotype of the three d-myc mutants is remarkably similar to a group of Drosophila mutants named Minutes (5) . All molecularly characterized Minute mutants involve ribosomal protein genes (5) . The only reported candidate d-Myc target gene pit, a Drosophila homologue of MrDb in mammalian cells, is also involved in ribosome assembly and protein synthesis (12) . In addition to the evidence in Drosophila, the constitutively expressed c-myc transgene in mice and induced c-Myc expression in human B cells have also been found recently to increase protein synthesis (13 , 14) . These studies of c-Myc in Drosophila, mouse, and human would predict that the major c-Myc target genes involve protein synthesis and metabolism. Our findings support this prediction.

Overexpressed c-Myc Regulates Additional Set of Genes Compared with Physiologically Expressed c-Myc.

Because c-Myc overexpression is very common in human and animal cancers, we ectopically expressed c-Myc in WT cells (WT^+MYC) and compared the gene expression between WT^+MYC cells and WT cells during log-phase growth (Table 3) ⇓ . At log phase, WT cells had a sustained c-Myc expression, and the c-Myc expression in WT^+MYC cells was ∼4.5-fold higher than that in WT cells (Fig. 3 ⇓ ; Table 3 ⇓ ). We found that 32 known genes and 19 unknown genes were differentially expressed between WT^+MYC cells and WT cells during log-phase growth. Forty-seven of these 51 genes uncovered here were not found in the comparison between log-phase WT cells and Null cells and may represent candidate genes responsive mainly to high levels of c-Myc expression. Among them, Gadd45 and thrombospondin have been reported previously as target genes of overexpressed c-Myc (2) . It is reasonable to hypothesize that the overexpressed c-Myc may turn on or off the expression of some genes not normally regulated by physiological levels of c-Myc either through direct effects on promoter occupancy or through the differential induction of other transcriptional activators.

Recently, using oligonucleotide microarray analysis, Coller et al. (11) found 36 genes regulated by ectopically overexpressed c-Myc in density arrested or serum-starved human fibroblasts. Of these 36 c-Myc-regulated genes, 28 were not present on our rat cDNA microarray. Thus, direct comparisons between array results cannot be made. Nevertheless, of the other 8 genes present on both microarrays, tropomyosin, collagen, and HSP60 were also found in our study. Moreover, Coller et al. (11) also presented evidence for induction of some genes involved in protein synthesis and metabolism, although they did not show a predominance of ribosomal and protein synthesis gene induction as described herein. We suspect that the use of a primarily overexpression system involving a myc-ER construct, a different array system, and different growth conditions (arresting versus log-phase) may be the major reasons for the discrepancy.

That c-Myc up-regulates genes involved in protein synthesis may explain its role in carcinogenesis. In yeast, the ribosome assembly is precisely adjusted to the physiological demands of the cell, and the coordinated expression of ribosomal proteins is primarily regulated at the transcriptional level (15) . In human, several ribosomal proteins (L5, L21, L27a, L28, S5, S9, S10, and S29) have been shown to be overexpressed in colorectal carcinoma (16) . The increased expression of ribosomal protein S27 has been found in a wide variety of actively proliferating cells and tumor tissues, and its expression correlated with the degree of aggressiveness in prostate and colon malignancy (17 , 18) . Our results and the recent findings of others suggest that c-Myc may potentiate the sensitivity of cells to other mitogenic signals by preparing the protein synthesis machinery to meet the demand of cell proliferation.