We used human promoter microarrays that contained 9,303 proximal promoters to identify the direct targets of Myc in two different human cell lines, HeLa cells and primary foreskin fibroblasts. The genomic regions between –750 bp and +250 bp relative to transcriptional start sites (TSS) of well annotated genes were amplified and printed on the microarrays [24]. We carried out independent ChIP reactions from four independently grown HeLa cell cultures and co-hybridized the ChIP samples together with a mock immunoprecipitated sample as the reference. We determined the median ratio of each spot on the microarray from the four independent hybridizations, and also calculated a p-value for significant ratios using a previously described error model [25], [26]. We considered spots with a ratio ≥2.0 and P value ≤0.005 to represent direct Myc target loci. Among the 9,303 promoter spots analyzed, 1351 spots were in this category (Figure 1A). Control hybridizations with differentially labeled mock immunoprecipitated samples did not yield any significantly enriched loci at this threshold (data not shown).

The response to serum stimulation has been used as a model for studying mammalian cell cycle activation and proliferation [27], [28]. These studies have led to the identification of Myc as a critical downstream element of proliferation signaling cascades. As an immediate early response transcription factor, the expression levels of Myc mRNA and protein are rapidly induced by a variety of growth factors and serum stimulation [29], [30]. Based on the fact that expression of Myc is strongly and rapidly induced by serum stimulation, and the fact that Myc is known to be an important regulator of cell proliferation, we hypothesized that Myc might show changes in binding to its target promoters in response to serum stimulation. To determine if this was the case, we measured the occupancy of Myc at target promoters in quiescent foreskin fibroblasts as well as in fibroblasts after serum stimulation. Two independent cultures of quiescent foreskin fibroblasts grown in low serum (0.1% FBS) media for 72 hours, as well as fibroblasts stimulated by serum (10% FBS) for 4 hours were used for ChIP, and hybridizations were carried out as described for HeLa cells above, using a mock immunoprecipitated sample as reference. We averaged the ratios from duplicate microarray hybridizations and used the same threshold as for HeLa cells to define the target promoters occupied by Myc in foreskin fibroblasts. We detected 196 promoters occupied by Myc in quiescent fibroblasts (Figure 1B), whereas 788 promoters were occupied by Myc in response to serum stimulation (Figure 1C). Our result thus reveals differential binding of Myc to its target promoters under different growth conditions in the same cell type. Interestingly, Myc occupied a much larger number of target promoters in HeLa cells although the HeLa cells used in our study were not synchronized. Altogether, we identified 1469 direct downstream targets of Myc from at least one of the conditions we tested.

We compared the Myc targets identified in HeLa cells with the Myc targets in fibroblasts to see if there were any differences in target profiles among cell lines or different growth conditions. There were many common Myc targets among the different conditions we examined. Approximately 86% (676 out of 788) of the Myc targets in serum-stimulated fibroblasts were also targets in HeLa cells (Figure 1D). 86% (169 out of 196) of Myc targets in quiescent fibroblasts were also targets of Myc in serum stimulated fibroblasts, and 95% (185 out of 196) of them were also Myc targets in HeLa cells (Figure 1D). Interestingly, a more continuous analysis of target overlaps based on pair-wise global comparisons of each data set (Figure 2A–C) suggested that the occupancy of target promoters by Myc was largely consistent across cell lines and growth conditions. Thus, the differences between its binding targets in HeLa cells and fibroblasts are likely to reflect lower levels of Myc in fibroblasts, rather than its occupancy of a markedly different set of targets in the different cells or conditions.

Since Myc expression is induced by serum stimulation, the small set of genes (27) whose promoters were occupied by Myc in serum starved fibroblasts and not in serum stimulated fibroblasts is noteworthy. We examined expression changes of this subset of genes upon serum stimulation, using an expression data set published from our lab [31]. We found that this subset of Myc targets included genes that are activated and repressed by serum stimulation, as well as genes that show no significant expression change (Supplementary Figure S1). This behavior is consistent with the previously reported role for Myc as a repressor [14], [15], [16], [17], as well as the finding that Myc binding can occur without a corresponding change in mRNA expression levels [22].

A comparison with Myc target genes identified in previously published large-scale experiments [14], [22], [32], [33] also confirmed that there were many targets in common between our study and previous studies, even though each of these studies used different cell lines and different experimental platforms (see Supplementary Table S1). These overlaps do not however reflect some kind of broad ChIP-chip artifact. As a negative control for our pair-wise comparisons, we used data from an E2F4 ChIP-chip experiment. As expected, the Myc occupancy and E2F4 occupancy datasets showed much weaker correlations than the different Myc datasets (Figure 2D).

The 1469 target genes of Myc in HeLa cells and foreskin fibroblasts we report here form one of the largest sets of direct Myc targets identified in a single experiment. 1104 of our 1469 targets were not identified in other recent large-scale experiments [14], [22], [34] (also listed in Supplementary Table S1). This is likely in part because many of these studies used different cell types. For example, Li et al used a smaller promoter microarray to identify direct Myc targets in Burkitt lymphoma cells [33]. However, even though the overlap between the two target sets was modest (46%), it was statistically highly significant (p<10⁻⁵⁰). Based on our results, and extrapolating from the coverage of promoters on this array, we estimate that the number of direct target genes of Myc in human cells totals approximately 16% of all human genes.

The direct Myc target genes we identified fell into a variety of functional categories that suggested a role for Myc in diverse cellular processes. We used DAVID [35] and FatiGO [36] to quantitate the enrichment of different functional groups of genes among its targets and confirm the involvement of Myc in many processes identified in previous large scale experiments, such as DNA, RNA, and protein metabolism, transport, apoptosis, cell cycle, signal transduction, and cellular proliferation (Supplementary Table S2). KEGG pathway analysis [35] revealed that the Myc direct target genes we identified by ChIP-chip are additionally involved in many previously known and unknown pathways (Table 1). Ribosomal genes comprised one of the most prominent previously known sets of genes that Myc regulates directly, but our analysis identified that it also directly regulates genes involved in oxidative phosphorylation, aminoacyl-tRNA biosynthesis, and glutamate metabolism.

It is known that Myc activates nuclear-encoded mitochondrial gene expression [18], and that many direct targets of Myc are involved in mitochondrial function [14], [32], [33]. Myc directly regulates TFAM, a transcription factor important for mitochondrial transcription and mitochondrial DNA replication [21], as well as PGC-1β, a regulator of mitochondrial biogenesis [37]. However, the full extent of the relationship between the direct role of Myc acting at its target promoters and mitochondrial function has not been systematically addressed. In addition to these previously reported Myc targets that are involved in mitochondrial biogenesis, it is possible that additional targets in this process are also directly regulated by Myc by its binding to their promoters, indicative of a concerted and multi-pronged regulation of mitochondrial biogenesis by Myc. Indeed, our data and analysis revealed that many functions and pathways that Myc target genes are involved in are closely related to general mitochondrial functions, such as metabolism, apoptosis, oxidative phosphorylation, and glutamate metabolism. It has been suggested that the proliferation and cell cycle progression caused by Myc is in part due to its activation of mitochondrial biogenesis and oxidative phosphorylation [21], [38], [39], which are critical for cell proliferation. However, only a few genes in these pathways have been identified as direct targets of Myc [39]. We found that Myc occupies the promoters of genes such as POLG, POLG2, and NRF1, which have important roles in mitochondrial biogenesis. The promoter of TFBM1, a mitochondrial transcription specificity factor, showed a positive red/green ratio on the promoter microarray, indicative of occupancy by Myc, although it did not meet the p-value threshold for defining targets (R/G >2, p = 0.0283). NRF1, one of the direct Myc targets we identified, is known to regulate nuclear encoded mitochondrial genes including cytochrome c [39], [40]. Our result thus suggests possible co-regulation of mitochondrial biogenesis by a feed-forward network loop involving Myc and NRF1. It has been previously shown that the apoptotic function of Myc is mediated by NRF1 target genes, and that the balance of mitochondrial genes regulated by Myc and NRF1 is important in cell viability [39]. Our finding suggests that Myc could be an upstream regulator of NRF1 and both of these transcription factors are thus important regulators of mitochondrial biogenesis and diverse mitochondrial functions.

In order to systematically quantitate the extent of the involvement of Myc in mitochondrial biogenesis, we searched MitoP2, an integrated database of mitochondrial proteins [41] which lists 785 RefSeq annotated genes, of which promoters for 337 were represented on our promoter arrays. 107 of these 337 mitochondrial genes showed a red/green ratio greater than 2 and are listed in Supplementary Table S3. The red/green ratios of the set of all mitochondrial gene promoters tended to be higher than the overall red/green ratio of all promoters on the array, reflecting higher occupancy of the former set of promoters by Myc (Figure 3A).

The relationship between Myc binding and target gene expression is not a simple one [42]. However, ectopic expression of Myc is known to transcriptionally activate many nuclear genes involved in mitochondrial biogenesis [21]. To determine if this was also the case for the binding targets we identified by ChIP-chip, we examined the level of mitochondrial gene expression in microarray data. In HeLa cells containing high levels of Myc, the expression levels of mitochondrial genes tended to be higher than the background (Figure 3B), which is similar to the pattern of Myc occupancy of its the target promoters (Figure 3A). We independently verified the occupancy by Myc of the promoters of many of these mitochondrial function genes, as well as several other targets identified by our core promoters using quantitative real-time PCR (Figure 3C). A gene-by-gene comparison of Myc binding and target gene expression levels also showed that most of the nuclear mitochondrial genes whose promoters were occupied by Myc were transcriptionally active (Figure 3D). Taken together, our data strongly support the notion that Myc directly binds to the promoters of many nuclear encoded mitochondrial genes and positively regulates the transcription of these genes.

Myc regulates its downstream target genes by binding to cis-acting DNA motifs called E-boxes. The canonical E-box sequences CACGTG and CATGTG [23] as well as non-canonical E-boxes, such as CATGCG, CACGCG, CACGAG, and CAACGTG have been identified as cis-acting elements bound by Myc [43]. Although several E-boxes and E-box derivatives have been identified, the E-box sequence preference of Myc for binding to its chromosomal sites in vivo remains largely unknown. To elucidate the relationship between Myc binding and each of the E-box derivatives occurring in proximal promoters in vivo, we analyzed our ChIP-chip data set.

We first examined the distribution of the previously defined E-box sequences in all human core promoters, considering the regions from 3,000 bp upstream to 1,000 bp downstream of the transcription start sites of 18,193 well-annotated human genes in RefSeq. Three of the six E-box sequences we analyzed–CACGTG, CATGCG and CACGCG–showed a relative enrichment in the promoter region between 800 bp upstream and 200 bp downstream of transcription start site (TSS), whereas in contrast, the sequence CATGTG showed a relative depletion in the same region (Figure 4A). The sequences CACGAG and CAACGTG did not show any significant bias within the 4 kb core promoter region that we analyzed. When we considered only the promoter region between 800 bp upstream and 200 bp downstream, about 55% of all promoters of the RefSeq genes contained at least one of the six E-box elements (Figure 4B). However, each E-box sequence was found in less than 20% of all human promoters.

Approximately 64% of promoters that were occupied by Myc in our ChIP-chip analysis contained at least one of the six E-box sequences. Thus, 36% of the promoters occupied by Myc in vivo lacked any known E-box sequence. We then tested the relationship between the presence of each E-box sequence and Myc occupancy at the promoters. We found that there was a strong correlation between the presence of the canonical E-box sequence CACGTG and Myc occupancy of a promoter. The E-box sequence CACGCG showed lower correlation, while CACGAG, CATGCG, CATGTG, and CAACGTG did not show any significant correlation with Myc binding (Figure 4C). Another postulated binding sequence, CACATG (the reverse complement of CATGTG) also showed no enrichment in Myc target promoters in our data (Supplementary Figure S2), although CACATG/CATGTG have been previously shown to be enriched among Myc binding sites [22]. However, the previous study also examined intronic and intergenic sites which were not included in our promoter arrays, and this could potentially reflect differences in the occurrence of Myc motifs in different types of target sites. Overall, our data show that Myc has a preference to bind only some of the previously defined E-box sequences in vivo. Interestingly although the E-box CATGCG is enriched at proximal promoter regions (Figure 4A) it did not show a significant correlation with Myc occupancy in vivo.

To obtain a more detailed understanding of in vivo Myc usage of its binding sites, we examined the relationship between Myc occupancy and the presence of each E-box sequence within the subset of promoters that were most strongly or most weakly occupied by Myc in vivo. More than half the promoters that were most strongly occupied by Myc (the top 2%) contained the CACGTG sequence (Figure 5A). The promoters most strongly occupied by Myc showed a increased frequency of all the previously defined E-box sequences except, interestingly, CATGTG. We observed the same trend when we examined the top or bottom 10% of promoters occupied by Myc (Figure 5B). Our data thus suggest that Myc has a marked preference to utilize promoters in vivo that contain any of the known E-box sequences except CATGTG. However, E-box sequences CACGTG and CACGCG showed the strongest correlations with Myc occupancy (Figure 4C).

To determine whether the most strongly preferred Myc sites, CACGTG and CACGCG, were both also used under distinct physiological conditions that might affect Myc binding, we separately analyzed ChIP-chip data from quiescent and serum-stimulated foreskin fibroblasts. Promoters containing the highly preferred CACGTG sequence were preferentially occupied by Myc in both conditions in foreskin fibroblasts, but promoters with the CACGCG sequence did not show significant correlation with Myc binding (Figure 6A and B). We also did not observe a strong correlation between any of the other E-box sequences and Myc binding in serum-starved fibroblasts, or in serum-stimulated fibroblasts. As a control, we examined the occupancy of E2F4 at these same promoters in quiescent fibroblasts. Despite the fact that there are several common targets for both Myc and E2F4, we did not observe a significant correlation between the presence of the E-box sequences and E2F4 binding, which implies that the E-box sequence is specific to Myc, but not to E2F4 (Figure 6C).

Taken together, our data indicate that the oncoprotein Myc strongly prefers to use promoters containing the E-box sequence CACGTG in vivo, although most of the other E-box sequences except CATGTG appeared to be utilized in vivo to varying extents. The discrepancy between our observations and previous findings might be due to the fact that most of the previous experiments were performed in vitro, and also did not consider the number of actual E-box sequences present in the promoter. We observed an increase in Myc binding to promoters that contain each of the previously defined E-box sequences, except CATGTG. However, the contribution of most E-box sequences to Myc binding in vivo is low except for the E-box sequence CACGTG.

Human HeLa cells and human fibroblasts derived from foreskin (ATCC CRL 2091) were grown to about 60% confluency in 15 cm plates in DMEM containing glucose (1 g/liter), antibiotics, and 10% fetal bovine serum (FBS, Hyclone). Exponentially growing HeLa cells were used for ChIP. For the serum starvation experiment, fibroblasts were washed once with phosphate-buffered saline (PBS) and twice with DMEM medium lacking FBS. Low-serum DMEM medium (containing 0.1% FBS) was added to the plates and incubated for 72 hours. For serum activation, cells in low-serum medium were washed twice with PBS, and DMEM medium containing 10% FBS was added to the cells and incubated for 4 hours. For crosslinking, 37% formaldehyde was directly added to the media, and ChIP procedures were subsequently performed using an anti-Myc antibody (sc-764x, Santa Cruz).

HeLa cells and fibroblasts were cross-linked by the addition of formaldehyde (1% final concentration) directly to tissue culture plates and incubated for 7 min at room temperature. Cross-linking was terminated by adding glycine to a final concentration of 125 mM. Cells were washed with cold PBS containing PMSF, scraped off the plates, collected by centrifugation, and washed again with PBS. After centrifugation, the pellet was resuspended in SDS lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-Cl pH 8.1, plus protease inhibitors) and incubated at room temperature for 20 min. Cells were sonicated on ice, and fragmented DNA was visualized on an agarose gel (average size 1 kb). The sample was centrifuged at 12000 rpm at 4°C for 10 min, and ChIP dilution buffer (1% Triton X-100, 2 mM EDTA, 20 mM Tris-Cl pH 8.1, 150 mM NaCl, plus protease inhibitors) was added to the collected supernatant. The sample was precleared with protein A-agarose beads (previously washed with ChIP dilution buffer) at 4°C for 1 hour. Precleared chromatin was incubated with an antibody against Myc (sc-764x, Santa Cruz) at 4°C overnight. For the mock IP controls, the antibody was not added. Prewashed protein A-agarose beads were added and protein-DNA complexes were recovered after a 2 hour incubation at 4°C. Immunoprecipitated complexes were successively washed twice with low salt wash buffer (0.1% sodium deoxycholate, 1% Triton X-100, 1 mM EDTA, 50 mM HEPES pH 7.5, 150 mM NaCl), once with high-salt wash buffer (0.1% sodium deoxycholate, 1% Triton X-100, 1 mM EDTA, 50 mM HEPES pH 7.5, 500 mM NaCl), once with LiCl wash buffer (250 mM LiCl, 0.5% NP-40, 0.5% sodium deoxycholate, 1 mM EDTA, 10 mM Tris-Cl pH 8.1), and twice with TE buffer (10 mM Tris-Cl pH 7.5, 1 mM EDTA). SDS elution buffer was added and incubated at 65°C for 30 min to recover protein-DNA complexes. Crosslinks were reversed by incubating at 65 °C overnight. The sample was treated with RNase A and Proteinase K, extracted with phenol/chloroform, and precipitated. The pellet was resuspended in 25 µl of water.

LM-PCR was performed as described [26] with minor modifications. LM-PCR consisted of three steps: blunting of ChIP DNA, ligation of linkers to ChIP DNA, and PCR. Approximately 200 ng of immunoprecipitated DNA was diluted to a final volume of 100 µl with water. A mixture of 11 µl 10X T4 DNA polymerase buffer, 0.5 µl BSA (10 mg/ml), 0.5 µl dNTP mix (25 mM each), and 0.2 µl T4 DNA polymerase (3 U/µl) was added, mixed by pipetting, and incubated at 12°C for 20 min. After the incubation, 11.5 µl 3M sodium acetate, and 0.5 µl glycogen (20 mg/ml) were added. The sample was extracted with 120 µl of phenol/chloroform/isoamyl alcohol (25:24:1) in a 0.5 ml Phase-Lock Gel tube (Eppendorf) and centrifuged for 5 min at maximum speed in a microcentrifuge. The upper phase was transferred to a new 1.5-ml Eppendorf tube and precipitated with 230 µl ice-cold ethanol (100%). After centrifugation, the pellet was washed with 70% ethanol and air dried briefly. The pellet was resuspended in 25 µl water and placed on ice. Ligase mix containing 8 µl water, 10 µl 5X ligase buffer, 6.7 µl annealed linkers, and 0.5 µl T4 DNA ligase was added and incubated overnight at 16°C. Linkers were prepared as follows: 300 µl each of oligo Long (40 µM stock, sequence: 5′-GCGGTGACCCGGGAGATCTGAATTC), and oligo Short (40 µM stock, sequence: 5′-GAATTCAGATC) were mixed. 50-µl aliquots of oligo mixture were placed at 95°C for 5 min, 70°C for 5 min and placed at room temperature for 20 min. Aliquots were then transferred to 4°C and allowed to stand overnight. The annealed linkers were stored at −20°C.

To precipitate the linker-ligated DNA, 6 µl of 3 M sodium acetate (pH 5.2) and 130 µl of cold ethanol were added to the tube and centrifuged for 15 min at 4°C. The pellet was washed with 70% ethanol and resuspended in 25 µl water. A mixture of 4 µl 10X PCR buffer, 6.75 µl water, 2 µl low TTP mix (5 mM each dATP, dCTP, dGTP, and 2 mM dTTP), 1 µl Cy3-dUTP or Cy5-dUTP, and 1.25 µl of oligo Long (40 µM), was added to the sample DNA and transferred to PCR tubes. PCR was performed as follows: first, 2 min at 55°C, followed by addition of a mixture of 8 µl water,1 µl 10X PCR buffer, and 1 µl Taq polymerase (5 U/µl ), 5 min at 72°C, 2 min at 95°C, 35 cycles of 30 sec at 95°C, 30 sec at 55°C, and 1 min at 72°C. After PCR amplification, 5 µl of sample was loaded on a 1.5% agarose gel to verify the size of amplified ChIP DNA. The average size of amplified ChIP DNA was approximately 400 bp. The remaining PCR product was purified using a Qiaquick PCR purification kit (Qiagen), and eluted in 13 µl of water.

Human core promoter microarrays were manufactured as described [44]. Amplified and fluorescently labeled ChIP and reference DNA samples were simultaneously hybridized to the microarrays and ChIP enrichment of target loci was calculated based on significant ratios of ChIP signal to reference signal at each spot. Independent biological replicate experiments were combined by averaging the red/green ratio of each spot on microarrays. The confidence of the ratio measurement was calculated as a p-value using a previously described error model [25], [26], [45]. Microarray elements whose red/green ratio was ≥2 and p-value was ≤0.005 were considered to be direct targets of Myc.