Date Published: August 31, 2007
Publisher: Public Library of Science
Author(s): Hilary A Coller, Joshua J Forman, Aster Legesse-Miller, Wayne N Frankel
Abstract: The recent revelation that there are small, noncoding RNAs that regulate the expression of many other genes has led to an exciting, emerging body of literature defining the biological role for these molecules within signaling networks. In a flurry of recent papers, a microRNA polycistron induced by the oncogenic transcription factor c-myc has been found to be involved in an unusually structured network of interactions. This network includes the seemingly paradoxical transcriptional induction and translational inhibition of the same molecule, the E2F1 transcription factor. This microRNA cluster has been implicated in inhibiting proliferation, as well as inhibiting apoptosis, and promoting angiogenesis. Consistent with its seemingly paradoxical functions, the region of the genome in which it is encoded is deleted in some tumors and overexpressed in others. We consider the possibility that members of this polycistronic microRNA cluster help cells to integrate signals from the environment and decide whether a signal should be interpreted as proliferative or apoptotic.
Partial Text: microRNAs are 21–23-nucleotide noncoding RNAs processed from double-stranded hairpin precursors present in a wide range of organisms including worms, plants, flies, and mammals [1,2]. microRNAs are loaded into the RNA-induced silencing complex and subsequently hybridize to complementary sequences in target mRNAs. This results in inhibition of mRNA translation or reduced message stability [3,4]. Microarray analyses suggest that individual microRNAs can regulate hundreds of genes . This finding has raised the interesting possibility that microRNAs can coordinate complex cellular responses. One emerging model of the role of microRNAs is to maintain the robustness of genetic networks by ensuring that genes that ought to be “off” are downregulated not only via decreased transcription but also by translational inhibition (Text Box 1) [6,7]. Recently, however, a microRNA cluster was found to be involved in a complex network structured like a feed forward loop (described further in Text Box 2). This network appears to play a central role in controlling proliferation, apoptosis and tumorigenesis.
Many microRNAs are present within the genome not as an individual microRNA but rather as clusters of multiple microRNAs [8,9]. Usually these clusters contain two or three genes but larger clusters exist, including the miR-17–92 cluster, which contains seven mature microRNAs (miR-17–5p, miR-17–3p, miR-18a, miR-19a, miR-20a, miR-19b, and miR-92–1). These microRNAs are organized in a polycistron; that is, they are all transcribed as a single pri-microRNA, and then subsequently processed to form the individual microRNAs. The entire miR-17–92 cluster is located within the third intron of an open reading frame termed the C13orf25 (Chromosome 13 open reading frame 25) gene located at 13q31–q32. The close proximity of the six clustered microRNAs to each other (all six encompass only ∼800 base pairs of genomic DNA) makes it possible for all of the microRNAs to be transcribed and expressed similarly. In fact, a recent analysis of the expression of microRNAs in hematopoietic cell lines revealed similar expression patterns for all of the microRNAs in this cluster in the samples analyzed .
O’Donnell and colleagues demonstrated that activation of the oncogenic transcription factor c-myc induces the expression of microRNAs within the miR-17–92 cluster . Chromatin immunoprecipitation confirmed that c-myc binds to its recognition sites, E-boxes, upstream of this cluster. Serum stimulation of fibroblasts induced the expression of c-myc and the miR-17–92 cluster with similar kinetics. Two of the microRNAs within the cluster, mir17-5p and miR-20a, downregulated the protein—but not mRNA—abundance of their predicted target, the transcription factor E2F1. Transfection of antisense oligonucleotides that inhibit miR-17–5p or miR-20a resulted in increased E2F1 protein levels without affecting E2F1 transcript abundance. Consistent with this finding, c-myc induction led to a strong increase in E2F1 transcript levels, but only a modest increase in E2F1 protein levels . Since c-myc and E2F1 have been shown to activate each other’s transcription [11–13], this established an unusually structured network in which c-myc activates the transcription of E2F1 while simultaneously inhibiting its translation (Figure 1). The importance of the miR-17–5p/miR-20a binding sites for E2F1 expression were demonstrated with reporter assays [10,14]. The 3′ UTRs of E2F2 and E2F3 were also regulated via miR-17–5p/miR-20a binding sites, but the downregulation of E2F1 protein levels was stronger than for E2F2 or E2F3 .
Because c-myc and E2F1, E2F2, and E2F3 can activate one another’s transcription [11–13], one might imagine a cell in a runaway positive feedback loop in which c-myc induces the E2F transcription factors, which induce each other and, in turn, c-myc. This would result in excessively high levels of proliferative transcriptional regulators. High levels of c-myc have been clearly shown to have a tumor-promoting effect . Just 2-fold differences in c-myc expression can affect cell size in flies or cell number in mice [17–20]. Indeed, dysregulated expression of c-myc is one of the most common abnormalities in human malignancy . Thus, maintaining c-myc levels within a tight range can be considered critical for the prevention of cancer and accordingly, c-myc levels are tightly controlled at the level of transcription [16,22]. One hypothesized role for the miR-17–92 cluster is to further the goal of carefully minimizing noise in c-myc levels. By ensuring that E2F1 protein levels do not rise precipitously in response to c-myc activation, miR-17–92 may act as a brake on this possible positive feedback loop, thus helping to ensure tightly controlled expression of c-myc and E2F1 proteins [10,23]. From this perspective, the miR-17–92 cluster might act as a tumor suppressor. Consistent with this hypothesis, loss of heterozygosity or deletion of the 13q31–32 region is observed in multiple types of tumors, including squamous cell carcinoma of the larynx , retinoblastoma , hepatocellular carcinoma , breast cancer , and nasopharyngeal carcinoma [28,29]. Copy number loss was observed frequently for the microRNAs of the miR-17–92 cluster in breast cancers (21.9%), ovarian cancers (16.5%), and melanomas (20.0%) . In addition, miR-17–5p in particular is expressed at low levels in multiple breast tumor cell lines and has antiproliferative effects on breast cancer cells .
Paradoxically, while the miR-17–92 cluster has a putative role as an inhibitor of proliferation described above, the observed functional effect of overexpressing the microRNAs within the miR-17–92 cluster is not to inhibit proliferation but rather to induce proliferation and/or inhibit apoptosis. As an example of the role of microRNAs in the miR-17–92 cluster in inhibiting apoptosis, introduction of miR-17–92 in conjunction with c-myc overexpression resulted in B-cell lymphomas characterized by an absence of the high levels of apoptosis normally associated with c-myc-induced lymphomas . In in vitro studies, Sylvestre and colleagues demonstrated that overexpression of miR-20a decreased doxorubicin-induced cell death in a prostate cancer cell line, while inhibition of miR-20a with antisense oligonucleotides increased cell death after doxorubicin . Matsubara and colleagues discovered that antisense oligonucleotides directed against miR-17–5p, miR-20a, or both result in increased apoptosis as measured by TUNEL assays in cancer cell lines overexpressing the miR-17–92 cluster . Little effect was observed when either miR-17–5p or miR-20a was introduced into cell lines that do not overexpress miR-17–92, or when antisense oligonucleotides against either miR-18a or miR-19a were introduced. These findings demonstrate that the microRNAs within this cluster have distinct functional effects, with at least miR-17–5p and miR-20a specifically inhibiting apoptosis.
A physiological role for apoptosis inhibition mediated by miR-17–92 has been suggested in spermatocytes . In individual pachytene spermatocytes within a normal testis, high levels of E2F1 message but low levels of protein were associated on a cell-by-cell basis with high miR-17–92 expression. This has been suggested to reflect the particular needs of spermatocytes, cells in which meiotic recombination induces extensive crossing over of sister chromatids and multiple double-strand breaks, which would be expected to result in apoptosis . Inhibition of E2F1 by miR-17–92 could be important for preventing apoptosis in these cells during meiotic recombination. These findings suggest that the miR-17–92 cluster may be important for inhibiting apoptosis under conditions in which it would be detrimental to the organism. In addition, this same mechanism for apoptosis inhibition may be associated with tumorigenesis, as described further below.
c-myc has been reported to promote neovascularization via upregulation of pro-angiogenic VEGF  and downregulation of anti-angiogenic thrombospondin-1 (tsp-1) [40,41]. In addition, Dews and colleagues recently discovered that the miR-17–92 microRNA cluster also plays a role in c-myc–induced angiogenesis in a ras–myc tumorigenesis model. When engrafted into mice, cells overexpressing ras form small tumors while overexpression of both c-myc and ras results in larger tumors with much more robust neovascularization . In cell culture, miR-17–92 levels were elevated in the presence of overexpressed c-myc, and levels of predicted targets tsp-1 and connective tissue growth factor (CTGF) declined. Transfection of antisense oligoribonucleotides revealed that miR-18 is especially important for CTGF regulation and miR-19 is important for tsp-1 regulation. Inhibition of miR-17, miR-20a, or miR-92 had no effect on the levels of angiogenesis target genes. Further, cells overexpressing both ras and miR-17–92 had lower levels of CTGF and formed tumors that were larger and more vascular than tumors formed by cells expressing only ras.
While deletion of 13q31–32 has been observed in some tumors as described above [24–29], paradoxically, amplification at 13q31–32 is also frequently observed in multiple tumor types, including liposarcoma , diffuse large B-cell lymphomas , and colon carcinomas . Fine mapping of the 13q31–32 region in diffuse large B-cell lymphomas revealed that C13orf25 is elevated in lymphoma cell lines and patients [32,46]. Overexpression of miR-17–92 may be particularly tumor-promoting when c-myc is also activated. Amplification of miR-17–92 in the B-cell lymphoma tumor type is consistent with a role for miR-17–92 in conjunction with c-myc, because human B-cell lymphomas are often characterized by high c-myc expression . Elevated expression levels of miR-17–92 members has also been associated with lung cell tumors, especially those with c-myc amplification. miR-17–92 overexpression was detected at the RNA level and in some cases at the DNA level as well, suggesting amplification events . In yet another study, a human mantle cell lymphoma was shown to contain genomic amplification of both c-myc and miR-17–92 .
We conclude by hypothesizing that the unusual structure of the c-myc–miR-17–92–E2F network may help the cell to integrate external signals to make a cell fate decision. c-myc activation can result in proliferation, and indeed, c-myc levels are often markedly elevated in tumors. However, in different cellular contexts, c-myc can be a potent inducer of apoptosis [49–55]. In a mouse model with inducible c-myc, activation of c-myc in pancreatic β cells induced uniform β cell proliferation, but also overwhelming apoptosis, thus counteracting the oncogenic potential of c-myc . When c-myc was activated in conjunction with overexpression of Bcl-xL, which suppresses c-myc–induced apoptosis, then c-myc triggered rapid, uniform, and reversible progression to tumorigenesis. c-myc/Bcl-xL-induced β cell tumors also contained an extensive network of blood vessels that regressed after c-myc was switched off.
In summary, a series of elegant recent papers has illuminated a fascinating and unexpected network of interactions involving the c-myc and E2F transcription factors, and the members of a microRNA cluster. This network may be organized in the format of an incoherent feed forward loop, in that c-myc induces E2F1 transcription while repressing E2F1 translation. The microRNAs within this cluster may act as a brake on proliferation, inhibiting apoptosis and promoting angiogenesis. We look forward to further research that will clarify the roles of c-myc, different E2F transcription factors, and other regulators in controlling expression of the miR-17–92 cluster. In particular, experiments addressing whether c-myc and the E2Fs act synergistically or independently to control miR-17–92 expression would help to define its potential role as a signal integrator. It will also be interesting to determine whether miR-17–92 plays a role in tumorigenesis mediated by other genetic mechanisms. For instance, in chronic myeloid leukemias, expression of the miR-17–92 cluster was downregulated by RNAi directed against the pro-oncogenic fusion protein bcr-abl . Other microRNAs have been implicated in tumorigenesis, either as oncogenes or tumor suppressors [58–60]. The mechanisms by which these microRNAs affect tumorigenesis, in particular, whether they affect the same or different molecules and employ the same or different molecular circuitry, would shed light on the key elements in the transition to tumorigenesis. Recent analysis of gene expression patterns between microRNAs and their targets suggest that networks of this type, in which the expression of the microRNA and its targets are positively correlated, are common in human and mouse, especially in neural tissues . This makes it of particular importance to discover the potential advantages conferred by this seemingly paradoxical network structure.