Identification of genes regulated by c-Myc has been an enduring question in the field. c-Myc also regulates the expression of other transcriptional regulators, both activators and repressors, which in turn modulate the expression of their targets.57
To disentangle this complex web, it is critical to understand which genes are the primary or direct targets of c-Myc. Toward this end many have examined the regulatory effects of the conditional MycER fusion protein18
activated by OHT in the presence of cyclohexamide. The rationale was that cyclohexamide, by inhibiting translation, would block the expression of any secondary effectors downstream of c-Myc. In practice, this approach has been disappointing.15
First, cyclohexamide exerts very strong effects on the steady-state levels of numerous mRNAs, and these alone are often of greater magnitude than those elicited by c-Myc. Second, the MycER protein (like c-Myc itself) is unstable, and preventing its resynthesis dampens the regulatory effects, especially with respect to downregulated genes.
Another approach, which we have taken here, is to collect samples for expression profiling in a time course after activation of MycER with OHT. While this approach does not provide direct evidence of direct action, it also does not suffer from the strong perturbations of cellular physiology caused by cyclohexamide. To maximize the resolution of this analysis we collected numerous samples at early time points. The expression profiling was performed in an immortalized but non-tumorigenic rat fibroblast cell line that had been genetically engineered such that c-Myc activity can be regulated with OHT from essentially null to only slightly above physiological.15
The long time course, frequent time points and three biological replicates resulted in a data set with very high resolution and statistical significance, allowing us to define 4,186 differentially regulated genes at 1% FDR. Of these, 1,826 were upregulated and 2,360 were downregulated (). Remarkably, 90.6% of the upregulated genes and 80.3% of the downregulated genes showed an initial response within the first 3 h of c-Myc activation (OHT addition).
The identification of a positive c-Myc target in our experimental regimen would require activation of the c-Myc protein, activation of the target gene promoter and accumulation of the encoded mRNA to a level that could be detected by the microarray. Our data indicate that this process is clearly detectable in 1 h and peaks at 2 h after c-Myc induction, with only relatively few genes responding at the 3 h time point (). In contrast, for a negative target, in addition to repression of transcription, the steady-state levels of the mRNA would have to decay sufficiently to be detectable by the microarray. Accordingly, the kinetics of repression were observed to lag by approximately 1 h, with very few genes being first detected at 1 h, and the peak of detection being maximal and roughly equivalent at 2 and 3 h (). Since the proliferation of the cultures is very slow at these early time points (doubling time >40 h), dilution due to cell division would be minimal, implying that c-Myc repressed genes tend to have unstable mRNA.
For indirect regulation, such as that imposed by a transcription factor regulated by c-Myc, the process would be significantly longer and entail the additional steps of translation and transport of the intermediate factor into the nucleus. It is very unlikely that all this, as well as the ultimate regulation of the target gene, could be achieved in a 3 h window, especially for downregulated genes. In agreement, we found that promoter sequences of genes upregulated in the first 3 h were highly enriched for E-boxes, whereas this enrichment was lost for promoters upregulated at later times (). Based on all these considerations we conclude that genes responding within the first 3 h of c-Myc activation should be highly enriched for direct targets.
The upregulation of target gene expression is mediated by the binding of Myc-Max complexes to E-boxes. These mechanisms have been extensively studied and are relatively well understood. In contrast, although c-Myc represses an equivalent (in our study an even greater) number of genes, our knowledge of these processes is still quite incomplete. One mechanism involves the binding of c-Myc to other, positively acting, transcription factors and interfering with their activities, typically not at the DNA-binding level but by antagonizing the interaction of the transcription factors with their co-activators.58
Whether E-boxes are required for repression has been controversial.36
Our data show that all E-box PWMs found in the TRANSFAC database are significantly negatively enriched (under-represented) in the promoters of c-Myc repressed genes at all time points (). This is especially the case for the early responders, which are most likely to be the direct targets, arguing against a link between E-boxes and repression. This does not rule out, however, that interaction with non-canonical E-boxes may stimulate the repressive activity of the c-Myc complexes.36
The most well studied transcription factor that is the target of c-Myc interference is Miz1 (encoded by the Zbtb17
Other well documented factors include Sp1 and NF-Y. Given the wide range of protein-protein interactions that c-Myc engages in,59
it is widely believed that additional transcription factors targeted by c-Myc remain to be identified.60
We therefore took advantage of our database of genes likely to be directly repressed by c-Myc and performed a bioinformatic analysis of their promoters for enriched transcription factor binding sites (). Our approach was validated by clearly identifying Miz1 and NF-Y (the PWM of Sp1 was not significantly enriched above background in any promoter set because of its very short and degenerate sequence). The novel factors identified by our analysis are one (or more) Ets family members, Mzf1 (myeloid zinc finger 1), Rreb1 (Ras responsive element binding protein 1), Klf4 (Kruppel-like factor 4) and Irf1 and Irf2 (interferon regulatory factors 1 and 2). It is interesting to note that Mzf1, Rreb1 and Klf4, like Miz1 and Sp1, are zinc finger transcription factors.
Another mechanism of c-Myc repression is mediated by small noncoding RNAs known as microRNAs (miRNAs). A single miRNA can destabilize and/or inhibit the translation of multiple mRNAs, and recent work has shown that c-Myc dramatically affects miRNA expression.61
miRNAs are typically synthesized by RNA Pol II as larger polycistronic precursors from which the mature miRNAs are processed. Several miRNA precursor genes, such as the miR-17-92 cluster, have been shown to contain E-boxes in their promoters and to be directly activated by c-Myc61,62
The miRNAs processed from these precursors have been shown to promote a variety of c-Myc functions, such as stimulation of proliferation, survival and metabolism, by targeting the mRNAs that restrain these pathways.61
For example, miR-17, miR-20a, miR-93 and miR-106b have been shown to target the mRNA of the cyclin dependent kinase inhibitor p21Cip1
(encoded by the Cdkn1a
gene), a known target of c-Myc repression.61,63
Hence, Cdkn1a can be downregulated by c-Myc at two levels: transcriptionally by interfering with Sp1 activation of Cdkn1a
and posttranscriptionally by inducing the expression of miRNAs that destabilize the Cdkn1a mRNA.63
The direct upregulation by c-Myc of miRNAs that subsequently target mRNAs for degradation is consistent with the rapid response of c-Myc repressed genes that we have observed. In the majority of reported studies of miRNA regulation, c-Myc was overexpressed in the context of cancer promotion.61,64
We have previously shown that a significant number of genes affected by c-Myc overexpression are not regulated during its normal physiological range of expression.14,15
The relative proportions of repressed genes in our data set that are regulated transcriptionally and posttranscriptionally thus remains to be determined.
As is the case with protein-coding genes, c-Myc also represses a large fraction of its miRNA targets.65
Initial reports indicate that these effects are likely to be mediated by both transcriptional and posttranscriptional mechanisms.66
The identification of transcription factors that are targets of c-Myc interference will thus remain a topic of considerable importance for understanding the repression of both coding and non-coding genes. In this context, it will be of interest to identify miRNA-targeted mRNAs in our data set so that the search for PWMs can be limited to transcriptionally repressed promoters, as well as to search for the PWMs we already identified in the promoters of miRNA transcripts.
The c-Myc transcriptome data set presented here is the largest reported to date (4,186 differentially regulated genes at 1% FDR), and the analysis of the gene expression patterns fits well with the known biological functions of c-Myc. To allow data-mining by the community we have not only deposited the primary data in the GEO database (accession GSE20550), but provide additional tools on our website (www.brown.edu/Project/Myc/
). Finally, we have uncovered several novel aspects of c-Myc activity, such as downregulation of mitosis and the JAK-STAT pathway, which are discussed more fully in the Results.