|Home | About | Journals | Submit | Contact Us | Français|
Micro (mi)RNAs are emerging as important regulators of cellular differentiation, their importance underscored by the fact that they are often dysregulated during carcinogenesis. Two evolutionary conserved families, let-7 and miR-200, regulate key differentiation processes during development. Loss of let-7 in cancer results in reverse embryogenesis and dedifferentiation, and miR-200 has been identified as a powerful regulator of epithelialto-mesenchymal transition (EMT). Recent findings have connected let-7 with stem cell maintenance and point at a connection between EMT and stem cell formation. A part of tumor progression can be viewed as a continuum of progressive dedifferentiation (EMT) with a cell at the endpoint that has stem cell-like properties. I propose that steps of this process are driven by specific changes in the expression of let-7 and miR-200 family members.
Micro (mi)RNAs are small, 19−22 nucleotide (nt) long, non-coding RNAs that inhibit gene expression at the posttranscriptional level. They are first transcribed as parts of longer molecules, up to several kilobases in length (pri-miRNA), that are processed in the nucleus into hairpin RNAs of 70−100 nt by the double-stranded RNA-specific ribonuclease, Drosha.1 The hairpin pre-miRNAs are then transported to the cytoplasm by exportin 5 where they undergo final processing by a second, double-strand specific ribonuclease, known as Dicer. In animals, single-stranded miRNAs are incorporated into RNA induced silencing complexes (RISC) that bind primarily to specific messenger RNA (mRNA) at specific sequence motifs within the 3' untranslated region (3'UTR) of the transcript, which are significantly, although not completely, complementary to the miRNA. The mRNA/miRNA duplex then inhibits translation either through a (mRNA 5') cap-dependent mechanism affecting initiation2,3 or through increased degradation of the mRNA.4 Given the frequency with which miRNA target motifs are conserved within 3'UTRs, it is estimated that 20 to 30% of all human genes are targets of miRNAs, and that for each miRNA hundreds of genes exist that carry conserved sequence motifs within the 3'UTR.5-7
A strong link between miRNA dysregulation and human cancer has been established. A comparison of miRNA expression in normal and tumor tissues demonstrated global changes in miRNA expression in various human malignancies.8 In addition, mapping of 186 human miRNA genes has revealed that they are frequently located at fragile sites and other cancer-associated chromosomal regions.9 Consequently miRNAs have been demonstrated to act either as oncogenes (e.g., miR-155, miR-17−5p, miR-21)10,11 or tumor suppressors (e.g., miR-34, miR-15a, miR-16−1, let-7).12-22
Specific miRNAs have been shown to regulate mammalian cellular differentiation as well as developmental patterning and morphogenesis in a tissue specific fashion.23-29 In addition to these specific regulators, there exist classes of miRNAs that have universal functions which are dependent not on the tissue, per se, but rather on the differentiation state of the tissue. Two of the largest miRNA families, let-7 and miR-200 seem to have such activities.
The ubiquitously expressed let-7/miR-98 family was one of the first mammalian miRNAs to be identified.13-16,30,31 The let-7 family is comprised of 12 family members (let-7-a1, a2, a3, b, c, d, e, f1, f2, g, i and miR-98) located on 8 different chromosomes.32 These 12 family members represent 9 distinct let-7 sequences with identical seed sequences (the 5' sequence of the miRNA responsible for initating binding) and, very likely, overlapping sets of targets. Let-7 is expressed late in mammalian embryonic development and plays an evolutionarily conserved role from Caenorhabditis elegans to Drosophila to mammals.31,33-35 The let-7 targets that have been identified include cell cycle regulators such as CDC25A and CDK6,36 promoters of growth including RAS and c-myc16,37,38 and a number of early embryonic genes including HMGA2, Mlin-41 and IMP-1.39-44
The miR-200 family is comprised of 5 members (miR-200a, b, c, miR-141 and miR-429). They are located within two clusters on separate chromosomes. Interestingly the 5 members can be subdivided into two subgroups according to their seed sequences. MiR-200a and miR-141 comprise group I and miR-200b, c and miR-429 comprise group II. Although target prediction algorithms predicted little overlap in the targets of these groups, experimental approaches suggested that their sets of targets are highly overlapping.45 The most prominent targets of the miR-200 family are two E box binding transcription factors, ZEB1 (also known as TCF8 and δEF1) and ZEB2 (also known as ZFXH1B and SMAD interacting protein 1 (SIP1)).45-48 Both are key regulators within a complex network of transcriptional repressors that regulate the expression of E-cadherin and a number of master regulators of epithelial polarity.49-52 Consistent with their function, the miR-200 family was recently identified as both a marker and a powerful regulator of the epithelial-to-mesenchymal transition (EMT).45,48,53,54 For simplicity I will refer to these families of miRNAs as either let-7 or miR-200 unless specific activities of individual miRNA species are being discussed.
Embryonic stem (ES) cells have been heralded as the great hope for cures for many degenerative diseases in the future. ES cells have an enormous potential to differentiate into many different tissues (reviewed in ref. 55). This pluripotent property of ES cells is subject to regulation by the homeobox transcription factors, Oct4 and Nanog, which are essential regulators of early development and ES identity.56-59 Disruption of either of these factors causes loss of pluripotency.56,59,60 It has been suggested that Oct4 initiates the pluripotency state whereas Nanog maintains it.61 The HMG-box transcription factor, Sox2, heterodimerizes with Oct4 in ES cells and regulates expression of Oct4.62 Oct4, Sox2 and Nanog are part of an autoregulatory loop. Each binds to its own promoter and to the promoter of the other two genes.63 Autoregulatory loops such as this typically maintain stability of gene expression.64 In ES cells Oct4, Sox2 and Nanog act in concert to maintain the pluripotent state.62 They are often called master regulators of cell state.65 These three transcription factors co-occupy promoters of hundreds of genes, activating some genes and silencing others, in a coordinated fashion.63,66 Among genes that are activated are transcription factors, signaling molecules and chromatin-modifying enzymes, that together promote ES cell self-renewal.63,66 Among the silenced genes are genes that determine lineage commitment. A number of groups have recently reported derivation of induced pluripotent stem cells (iPS) from human somatic cells. Cells were reprogrammed from either differentiated progeny of stem cells,67 fetal and neonatal fibroblasts68 or adult fibroblasts.69 Reprogramming was accomplished using a combination of Oct4, Sox2, c-myc and Klf-4, or a combination of Oct4, Sox2, Nanog and Lin28, omitting the potentially oncogenic gene, c-myc, from the cocktail.70 Oct4 and Sox2 were determined to be essential for the reprogramming process.57,62,71
miRNAs have been shown to regulate embryonic development72 and to play a role in regulating stem cells. Evidence for this activity comes from experiments demonstrating that ES cells that were deficient in miRNA processing enzymes exhibited defects in their capacity for differentiation and self renewal.73-76 In addition, Dicer deficiency is embryonic lethal, and Dicer deficient embryos exhibit greatly reduced expression of Oct4 suggesting a stem cell defect.77 Furthermore, cells from mice deficient for expression of the Drosha cofactor, DGCR8, proliferate and form colonies, but cannot differentiate.76 These data suggest that upregulation of miRNAs is required for various differentiation processes. Little is known with respect to mechanisms by which miRNA function in controlling the developmental potential of ES cells. A number of ES cell/early development specific miRNAs have been described,78-81 as have miRNAs that are expressed in mouse and human ES cells.80,82 However, it is still largely unknown how ES cell-specific transcription factors and miRNAs work together. The three stem factors (Oct4, Sox2 and Nanog) were recently found to occupy the promoters of many transcription factors and of 14 miRNAs.63 A more direct link between miRNAs and these genes was recently suggested by the report that miR-134, miR-296 and miR-470 target Oct4, Nanog and Sox2 within their open reading frames.83 Because miRNAs are presumed to preferentially target the 3'UTR of genes rather than the coding region, this points to an unusual level of regulation. Clearly, the function of these miRNAs in maintaining embryonic cells in vivo needs to be established. Recently miR-302 was demonstrated to convert human skin cancer cells into a pluripotent ES-like state.84 miR-302 is comprised of a cluster of 8 related miRNAs all of which are regulated by the binding of Oct4 and Sox2 to the miR-302 promoter.85 It appears, therefore, that a substantial portion of the activity of stem cell fate-specific transcription factors can be mimicked by miRNAs, which points at the power of miRNAs in maintaining cells and differentiated tissues.
A small number of cells within a tumor have properties that resemble those of stem cells. These cells have the ability to self-renew, to reproducibly form the same tumor phenotype, and to undergo multipotent differentiation into nontumorigenic cells. They are characterized by expression of distinctive cell surface markers permitting consistent enrichment. Chronic myeloid leukemia (CML) was the first cancer shown to be derived from a cancer stem cell.86 In the chronic phase of the disease the fusion protein BCR-ABL is present at high concentrations in hematopoietic stem cells (HSCs). After transition to blast crisis BCR-ABL was found to be amplified in granulocyte-macrophage progenitors (GMPs), which gain self renewal activity through aberrant activation of the Wnt/β-catenin pathway.87 The first tumor initiating cells (TICs) in a solid tumor were isolated from breast cancer.88 These cells efficiently formed anchorage-independent mammospheres. Mammospheres are known to be derived from mammary epithelial stem cells. It has been shown that a single mammosphere can give rise to an entire mammary ductal tree when implanted in mice.89 When isolated from breast cancer these CD24low CD44high cells have tumor initiating activity as indicated by the fact that they can differentiate and are highly tumorigenic when injected into immunodeficient mice. In addition to breast cancer, TICs have now been identified in melanoma, brain, liver, lung, head and neck, ovarian, pancreas, prostate and colon cancers.90-100
EMT-like processes occur as part of embryonic development, wound healing101 and during carcinogenesis102 when tumor cells undergo a change from a differentiated to a dedifferentiated, more invasive tumor.101,103 During the process of EMT cells lose epithelial features and acquire mesenchymal characteristics, including vimentin filaments and a flattened morphology. They may start migrating and expressing proteases, gaining invasive activity that allows them to pass through the underlying basement membrane. Adherens junctions and desmosomes become at least partially dissociated. At the same time, a massive cytoskeletal reorganization takes place and a remodeling of the actin microfilament mesh occurs. An analysis estimated that changes in expression levels of about 4,000 genes (representing about 10% of the human genome) could be associated with EMT.104 The final effectors controlling the cell phenotype, including the cytoskeleton and the cell-cell and cell-substratum adhesion systems, appear to be similar or identical in all forms of EMT-related differentiation processes. A family of E-box-binding, zinc finger-containing transcription factors have been associated with EMT. Members of this family, Snail, Slug, ZEB1 and ZEB2/ SIP-1, have been extensively studied with respect to their function both in tumor progression and in embryogenesis.105-110 EMT plays a profound role during embryonic development111,112 in the formation of the primary mesoderm from upper epiblast epithelium, in neural crest cell formation from part of the ectoderm, and in palatal formation. In adults EMT occurs during adult placenta formation, and in the formation of fibroblasts during inflammation113 and wound healing.114 The reverse of EMT, mesenchymal to epithelial transition (MET) also occurs. One instance of MET is the formation of the nephron epithelium in the developing kidney from nephric mesenchymal cells.115 Another instance of EMT/MET is carcino-genesis. Many of the EMT inducing transcription factors such as Snail1, Snail2, ZEB1, ZEB2, TWIST1, FOXC2 and Goosecoid have been associated with tumor invasion and metastasis.112 While EMT during cancer progression is well characterized, MET is more difficult to observe in vivo. In one study, GFP labeled mammary cancer cells were shown to exit the primary tumor and were demonstrated to lose their mesenchymal nature when forming micrometastases in the lungs.116 An example of re-epithelialization of cancer cells during metastasis formation was also shown for the bladder cancer cell line, TSU-PrI,117 which showed increased expression of epithelial markers during metastatic deposits in scid mice. The expression of the epithelial marker E-cadherin is often detected in distant metastases in cells that may have passed through EMT.118 A number of reports recently identified the miR-200 family of miRNAs as a fundamental marker and regulator of EMT.45,48,119 miR-200 family members were found to be downregulated during TGFβ induced EMT induction of MDCK cells, whereas their expression was highly enriched in the epithelial cell lines of the NCI60 cell lines. Most importantly, altering their expression in cancer cells forced a number of cell lines to undergo either EMT or MET.45,48
Evidence is mounting that the dedifferentiation process, EMT, may result in the generation of stem cells. The reversion of mesenchymal cells such as fibroblasts to iPS colonies expressing cadherins, mimics the MET characteristic of malignant transformation suggesting that reprogramming and tumor progression may act on similar dedifferentiation programs. Recently it was reported that poorly differentiated breast, glioma and bladder cancer cells express an ES cell signature.120 In addition, targets that are transcriptionally activated by Oct4, Sox2, Nanog and c-myc are frequently upregulated in these cancers. Breast cancers expressing an ES cell signature often belong to the basal type, which is characterized by poor clinical outcome. The fact that Oct4, Sox2 and Nanog can be used to reprogram somatic cells into pluripotent stem cells makes it likely that the combined expression of these genes in cancer cells could also contribute to the dedifferentiation observed during cancer progression. In fact, ectopic expression of the master regulator, Oct4, is sufficient to induce tumor growth in mice.121 The key regulators of stem cell pluripotency have been detected in various human cancers.122-125 Recently more direct evidence of a connection between stem cells and EMT was provided by the demonstration that induction of EMT in human mammary epithelial cells resulted in upregulation of a number of stem cell markers.126,127 In addition, cells isolated according to their stem cell markers had undergone EMT. These cells were found to overexpress classical EMT markers such as FOXC2, ZEB2, TWIST and SNAIL. Functionally, it was shown that induction of EMT resulted in the generation of cancer stem cells with the formation of markedly increased numbers of mammospheres by these cells.
There are additional parallels between the induction of dedifferentiation during the process of EMT and the generation of iPS cells. In both processes the achievement of substantial dedifferentiation requires treatment for weeks rather than days or hours. We and others observed that the induction of EMT required that cells be treated for 10 to 20 days.45,48,126 The induction of EMT in HCT116 cells required that cells be repeatedly transfected with miR-200 for up to three weeks.45 During this time cells gradually acquired a mesenchymal phenotype. Although the reason for this delay is unclear, it is consistent with the observation that TGFβ induced EMT requires a similarly long time in certain cells.48,126 The delay in the acquisition by cells of miRNA-induced changes is remarkably similar to the timing observed in reprogramming somatic fibroblasts to become iPS cells. It was shown that for complete reprogramming to occur the four factors Oct4, Sox2, Klf4 and c-Myc had to be expressed in the cells for at least 12 days.128 Current thinking is that a series of stochastic events leads to the formation of iPS cells.129 The notion is that ectopic expression of the four factors may trigger a series of epigenetic events, such as chromatin modification or remodeling, that results in only about 0.5% of the infected mouse embryonic fibroblasts giving rise to iPS cells within 3−4 weeks. Similar mechanisms may be involved in EMT.
Both let-7 and miR-200 family members are differentially expressed in tumor cell lines representing the two differentiation stages we called Type I and Type II cells.42,45,130 We suggested that Type I cells have a less differentiated, mesenchymal phenotype and represent more advanced cancer, whereas Type II cells have a more differentiated, epithelial phenotype and represent less advanced cancer. Both let-7 and miR-200 are almost absent in Type I cells consistent with the assumption that these cells represent more advanced forms of cancer.45 Interestingly the regulation of expression of both miRNA families involves negative feedback loops (Fig. 1). Hammond and colleagues demonstrated that early during embryonic development at a time when mature let-7 levels are undetectable, the primary transcripts of let-7 are highly expressed. Based on this observation, it was postulated that the amount of let-7 would be regulated at a posttranscriptional level.131 Three studies shed some light on this mechanism by showing that the let-7 targets, Lin28 and Lin28B, are both inhibitors of let-7 processing and their expression is restricted to early embryonic development. Both proteins were shown to bind to the loop region of let-7 precursors, which results in blocking of processing of let-7 at either the Drosha132,133 or the Dicer level134 (Fig. 1A). Recently, it was shown that Lin28 and Lin28B act as posttranscriptional repressors through an additional mechanism by mediating terminal uridylation of let-7 precursors, which results in their degradation.135 Hence, the two proteins seem to suppress let-7 expression by acting at multiple levels.
In contrast to let-7, miR-200 family members are regulated at the transcriptional level. miR-200 negatively regulates the expression of the E-box binding transcription factors ZEB1 and ZEB2,42,45 (Fig. 1B). Interestingly however, the promoters of both miR-200 clusters, one on chr 1 and the other on chr 12, contain E-boxes and are negatively regulated by ZEB1 or ZEB2.53,136 Both the let-7 and the miR-200 miRNA families, are comprised of multiple members and each member is predicted to regulate a largely overlapping set of targets. In order to alter the expression of these targets without having to regulate expression of each miRNA individually, regulatory mechanisms are in place such that reducing the amount of only one abundant member of each family is thought to result in the reduction of the total amount of miRNA below a threshold at which suppression of expression of the inhibitors is abrogated. This results in the expression of inhibitors and the consequent shutting down of the expression of all miRNA family members. These kinds of negative feedback regulatory mechanisms ensure that the differentiation processes regulated by entire families of miRNAs behave in a switch-like fashion. This phenomenon is especially obvious for the epithelial marker E-cadherin and the mesenchymal marker vimentin. Cells either express large amounts of either of these markers or express none. Very little intermediate expression is observed.
In a recent study, breast TICs were isolated from patients prior to and after receiving chemotherapy.137 Cells after chemotherapy were found not only to be more resistant to drugs, in vitro, but were also to express surface markers characteristic of TICs. These cells were devoid of let-7 expression but, when plated under conditions favoring differentiation, strongly upregulated let-7 over a course of 5 days. Interestingly, all miR-200 family members were also upregulated, albeit slightly delayed compared to let-7, indicating that stem-like cancer cells lack both let-7 and miR-200 expression. This finding is not restricted to cancer cells. A recent study using a deep sequencing approach identified expression of 14 miRNAs (9 known and 5 unknown) in undifferentiated normal human embryonic stem cells (HEC).81 Chromatin IP data demonstrated binding of Nanog, Sox2 and Oct4 to the promoters of 9 of the expressed miRNAs. It was also shown that both let-7 and miR-200 were upregulated when HECs were induced to differentiate. The miRNA showing the highest upregulation was let-7e (927 fold). In addition, in an analysis to identify ES-associated transcriptional regulators, not only were Oct4, Sox2, Nanog and Lin28 identified, but also identified was the miR-200 target, ZEB2.120 These data suggest that both let-7 and miR-200 are involved in regulating normal and cancer stem cells.
If one views cancer formation as a general dedifferentiation phenomenon, one would not be surprised if cancer cells displayed global downregulation of miRNAs. Consistently a large number of miRNAs that are downregulated during various forms of cancer have emerged.138 However, some miRNAs are upregulated during tumor progression. Among them are many that exhibit tissue specific expression. Prominent examples are miR-10b and miR-335 which promote breast cancer metastases.139,140 However, expression of certain miRNAs is altered during tumor progression in a more universal fashion. miR-21 has been shown to be upregulated during EMT141 and in many human cancers when compared to matching normal tissue.142 In addition, two large miRNA families have emerged as universal markers of advanced cancer. The first of these is let-7, the loss of which occurs early during neoplastic transformation, setting in motion a process that is similar to reversed embryogenesis.32 The second, miR-200, the fundamental regulator of EMT is often deregulated as tumors acquire invasive behavior. In studies of various human cancers including lung, colon, ovarian, gastric cancer, leiomyoma and melanoma, let-7 family members have been described as being downregulated during cancer progression.13-16,42,143-148 The situation with respect to miR-200 is more complex. The early stage of the process of metastasis is similar to EMT, but at later stages following extravasation, metastasizing cells start to settle down in their target tissue and undergo a differentiation process similar to MET.111 This, in theory, should again result in upregulation of miR-200. It may therefore not surprise that there is no general consensus as to whether or not miR-200 is upregulated or downregulated in advanced or metastasizing cancers. MiR-200 has been reported to be upregulated,146,147 downregulated48,149-151 or to remain unchanged152,153 in human cancers. In tumors from primary ovarian cancer (OvCa) we found a positive correlation between the expression of E-cadherin and miR-200. Goodall and colleagues reported an intriguing loss of expression of miR-200 in advanced breast cancer.48 In clear cell carcinoma both miR-141 and miR-200c were the most downregulated miRNAs.149 In OvCa the situation seems to be more complex. Three recent studies reported that miR-200 is a marker for more advanced OvCa. When OvCa tumors were compared to normal ovaries or ovarian surface epithelial cells miR-200 was found to be upregulated.146-148 In some of these reports the miR-200 family was found to be upregulated among all four histologic types of OvCa as compared to normal tissue. However, another recent study reported downregulation of miR-200 in OvCa as compared to normal tissue.154 Finally, to further confound the issue, in another recent analysis of the expressed miRNAs in OvCa none of the miR-200 miRNAs were found to be changed.152
An apparent upregulation of miR-200 during OvCa could have two reasons: (1) MiR-200 becomes upregulated during advanced OvCa. It was shown previously that early during OvCa progression cancer cells upregulate E-cadherin.155 (2) However, in some of the studies whole ovaries were taken rather than the single cell layer of epithelium on the surface of the ovaries from which the cancer is derived. The majority of cells inside of ovaries are of stromal nature and we demonstrated that stromal cells are completely devoid of miR-200 expression.45 It is therefore reasonable to assume that this mesenchymal cell mass contains less miR-200 than the epithelial cells or the cancer cells which were derived from the epithelial layer. Alternatively, ovarian surface epithelial cells were isolated and cultured before the analysis146 and the culturing or immortalization of these cells might have changed their miRNA composition. While such technical issues may have contributed to the contradicting results, it could be in the nature of miR-200 to be both up or downregulated during cancer depending on the stage of progression. Early when cancer cells acquire invasive behavior miR-200 may be downregulated, but it may also be upregulated again during the re-epithelialization of distal metastases when cells undergo MET. Such stage specific regulation of miR-200 has been demonstrated for liver cancer.156 miR-200 was found to be down-regulated in comparison to normal liver tissue with benign lesions, but was upregulated in advanced carcinomas. To clarify the role of miR-200 in cancer progression it will be necessary to improve the methods of isolating both cancer cells and normal cells and/or to develop better methods to quantify miRNA expression levels in single cells.
In humans let-7 has been shown to act as a tumor suppressor, most likely through targeting a number of genes with oncogenic activity such as RAS13-16 or high mobility group (HMG) A2.38-40,42 Based on the evolutionary conserved role let-7 plays during development, its main function may lie in suppressing early embryonic genes some of which may have activities in stem cells. Using a genome-wide bioinformatics approach we recently identified 12 let-7 regulated oncofetal genes (LOGs) that are suppressed by let-7 in most adult tissues, but which are re-expressed in various forms of cancer when let-7 expression is lost. In the next section I discuss the function of the top three LOGs in cancer and in stem cells.
Stem cells gradually lose their proliferative activity and capacity for self renewal with age.157 Recently, in a screen to identify genes that are selectively expressed in young versus old HSCs, Morrison and colleagues identified 26 genes that were differentially expressed between young and old stem cells.158 More recently using higher stringency they focused on HMGA2 as the only factor that is highly expressed in HSC and fetal neuronal stem cells but not expressed in any adult CD45+ cell.159 HMGA2 expression declined in aging stem cells and was undetectable in stem cells from 2-year-old mice. It was found that HMGA2 does not generally promote the growth or proliferation of cells but rather selectively promotes self renewal of stem cells.
Interestingly, the phenotype of the HMGA2 knock out mice is somewhat similar to that of mice deficient for the somatic stem cell factor, Bmi1.160 It was shown that the deficiency of Bmi1 could be compensated by crossing the mice with mice lacking expression of p16Ink4a and p19Arf,161 suggesting that Bmi1 suppresses expression of these cell cycle regulators. Similarly, deletion of either Ink4a or Arf partially rescued the defects of the HMGA2 deficient stem cells.159 The data establish HMGA2 as a stem cell factor that acts in young stem cells by repressing p16Ink4a and p19Arf, whereas Bmi1 acts by repressing expression of these cell cycle regulators postnatally in stem cells. The behavior of HMGA2 is similar to that of Oct4, which has also been shown to be important in embryonic stem cells, but which is entirely dispensable for the self-renewal and maintenance of adult stem cells.162 Recently it was demonstrated that in cancer stem cells let-7 regulates HMGA2,137 and it was suggested that in stem cells that lack let-7 expression, HMGA2 was involved in the maintenance of the undifferentiated state.163 Consistently, in HMGA2 containing mammospheres containing large numbers of TICs, Oct4 was found to be highly expressed.137
Evidence supporting the conclusion that HMGA2 is a stem cell factor controlled by let-7 also came from three large studies demonstrating that HMGA2 was associated with height,164-167 and that a particular single nucleotide polymorphism (SNP) associated with height mapped close to a predicted let-7 seed match in the 3'UTR of HMGA2.164 Assuming that HMGA2 in stem cells is under control of let-7, these data suggested that during aging let-7 expression in stem cells would increase. Consistent with this prediction, Nishino et al. reported that in aging stem cells the decline in HMGA2 levels was partly caused by let-7b upregulation, and confirmed that HMGA2 expression in aged stem cells was under control of let-7.159 Interestingly, it had previously been shown that HMGA2 together with INK4a contributed to induction of cellular senescence168 by suppressing expression of genes that drive proliferation. These data suggest that the low level of proliferation seen in stem cells could in part be driven by HMGA2/INK4a and be regulated by expression of let-7.
IMP-1/CRD-BP is a member of a family of RNA binding proteins including IMP-2 (=LOG6) and IMP-3, that bind to the 5'UTR of the Insulin-like Growth Factor II leader 3 mRNA species.169 IMP-1 recognizes c-myc, IGF-II, tau, FMR1, semaphorin and βTrCP1 mRNAs, as well as H19 RNA. It acts to stabilize target RNAs by shielding them from degradation.169 IMP-1 has a classic oncofetal expression pattern, with expression only early during fetal life170,171 and re-expression in many human cancers.172 Some of the mRNAs that are stabilized by IMP-1 such as IGF-II and c-myc can guide stem cell function.173,174 In addition IMP-1 has been shown to be selectively expressed in cord blood derived CD34+ stem cells but not in adult CD34+ cells.175 Consistent with a proposed function of IMP-1 in stem cells is the fact that the phenotype of mice lacking IMP-1 expression is very similar to that of mice deficient for HMGA2,171,176 which are also small and have a defect in gut development. Finally, it was previously shown that IMP-1 is one of 26 genes that were found to be selectively expressed in young, but not old, HSC.158
Interestingly Lin28B is a structural and functional homolog of Lin28.177 Recently it was found by two groups that Lin28/Lin28B act as selective inhibitors of let-7 processing in early embryonic cells132,133,178 demonstrating their function as both regulators of let-7 expression as well as being let-7 targets. Interestingly a combination of Lin28, SOX2, NANOG and OCT4 was demonstrated to reprogram somatic cells to become pluripotent stem cells.68 Lin28B as a stem cell factor responsible for pluripotency is, therefore, a good candidate gene to become upregulated in cancer cells when let-7 expression is lost. In this regard it is not surprising that Lin28B upregulation has been reported for human hepatocarcinoma cells.177 The two founding members of the miRNA family identified in C. elegans, lin-4 and let-7, have recently been linked to stem cells. It was shown that in neuronal stem cells both let-7 and miR-125 (which is the mouse ortholog of lin-4), negatively regulate expression of Lin28.134
In summary, the three let-7 targets, HMGA2, IMP-1 and Lin28 can be connected to stem cell function, thus making let-7 and its oncofetal targets fundamental regulators of stem cells, and therefore, highly relevant to cancer development and therapy.
While loss of let-7 may precede loss of miR-200 during carcinogenesis, I propose that the two processes can be overlapping in many cases. Consistent with this hypothesis, a number of examples of crosstalk between these two pathways regulated by these miRNAs have emerged (Fig. 2). Although both HMGA2 and IMP-1 are regulated by let-7 directly, the expression of IMP-2 was also shown to be dependent on HMGA2 expression.179,180 Mutant mice lacking HMGA2 expression show a marked reduction in IMP-2 expression at day E12.5.179 HMGA2 is required for TGFβ induced EMT of mouse mammary epithelial NMuMG cells. HMGA2 was shown to regulate transcription of Snail, Slug, Twist and Id2.181 In addition RAS, which is another target of let-7 and a driving force for EMT, has been reported to induce expression of HMGA2.182,183 More recently it was demonstrated that ZEB1 and ZEB2 are regulated by HMGA2, and that the regulation of the Snail promoter by HMGA2 involves binding of HMGA2 to Smad3 and Smad4. HMGA2 was found to enhance the binding of Smad3 to the Snail promoter.184 Interestingly ZEB2 was identified through its interaction with Smad1, 2 and 5.185 These pathways could, therefore, interact and the dedifferentiation processes regulated by these pathways could lead certain cells to acquire stem cell-like properties. This hypothesis is consistent with a recent report demonstrating that signaling through the TGFβ/Smad pathway can cause the stem cell factor Nanog to be upregulated.186
Mounting evidence links two miRNA families, let-7 and miR-200, which are significantly correlated with dedifferentiation of cancer cells, to stem cells (Fig. 3). In addition to experimental evidence linking let-7 to the stem cell factors Lin28 and Lin28B, Sox2 and Klf4 are predicted to be targets of miR-200b/c/429 (TargetScan. org). Expression of both let-7 and miR-200 was correlated with more differentiated cancer using genome wide miRNA arrays or real time PCR.45,48,187 Interestingly in all of these studies additional miRNAs that were (less significantly) associated with the dedifferentiated phenotype were identified. These “runners up” can also be connected to stem cell regulation. In the study by Goodall and colleagues, in addition to miR-200, miR-205 was identified as a marker for mesenchymal cells, and miR-205 was shown to target the EMT regulators ZEB1 and ZEB2.48 In a recent genome wide analysis of the connections between the core transcriptional machinery and miRNAs, both Sox2 and Nanog were found to be bound to the promoter of miR-205, suggesting a direct link between the EMT regulator, miR-205 and ES cell regulators.63 In our own study in which we identified miR-200 as a regulator of EMT, the miRNA that followed the miR-200 family member in significance was miR-203.45 miR-203 was shown to negatively regulate stemness during skin development by suppressing expression of p63.188,189 Finally, in the screen to identify miRNAs that are preferentially expressed in Type II cells we identified miR-128a as more highly expressed in Type II cells.42 Interestingly, it was shown that the stem cell factors (and let-7 targets), Lin28 and Lin28B, bind not only to the loop region of let-7, but also to the loop region of miR-128,132 suggesting that processing of let-7 and miR-128 is coregulated. Of note, both let-7 and miR-128 are predicted to target Lin28, and miR-128 is also predicted to target Nanog (TargetScan.org). In summary, although other miRNAs have similar functions that need to be experimentally explored, I propose that let-7 and miR-200 are major guardians against inappropriate inclinations toward stemness. One of their normal roles may be to keep differentiated cells in their differentiated state.
This work was supported by NIH grant R01 GM61712. I would like to apologize to all whose work could not been cited due to space limitations.