miRNA Processing and Human Cancer: DICER1 Cuts the Mustard

See allHide authors and affiliations

Science Translational Medicine  30 Nov 2011:
Vol. 3, Issue 111, pp. 111ps46
DOI: 10.1126/scitranslmed.3002493


Numerous studies have implicated microRNAs (miRNAs) in cancer initiation and progression. In contrast, only recently has attention been focused on the pathway that generates these regulatory molecules. The identification of neoplasia-associated germline mutations in DICER1 has focused translational research on components of the miRNA processing pathway. Deciphering of the many links between miRNA processing perturbations and cancer will likely provide insights into mechanisms of cancer control.

Like microRNAs (miRNAs), the miRNA processing pathway (Fig. 1) has been implicated in diseases that affect a variety of tissue types from nearly every human organ studied. Considering that a single mutation in any component of the miRNA processing machinery has the potential to modify the entire miRNA profile of a cell or tissue, it is not surprising that a broad range of cancers has been associated with deregulation of this crucial regulatory pathway. In this Perspective, we focus on DICER1, a miRNA processing protein that plays a central role in epigenetic modulation of gene expression.

Fig. 1. Processing primer.

The canonical pathway for miRNA production in humans. Step 1: An RNA polymerase II enzyme is responsible for reverse transcription of the gene, and the resulting molecule is a single-stranded RNA with a stem-loop secondary structure called the primary miRNA (pri-miRNA). Step 2: The pri-miRNA is processed within the nucleus by RNASEN and DGCR8, which combine to form the microprocessor complex. Step 3: The resulting RNA molecule is termed preliminary miRNA (pre-miRNA). Step 4: Exportin-5 binds pre-miRNA and shuttles the molecules through nuclear pores to the cytoplasm, a process that requires Ras-related nuclear protein (RAN), a guanosine triphosphatase. Step 5: In cooperation with its cofactor TRBP, DICER1 functions in a manner analogous to the nuclear microprocessing complex by measuring the necessary length of the miRNA and cutting the pre-miRNA to mature 20- to 22- nt strands. Step 6: The resulting mature miRNA is then passed on to a member of the Argonaute family and other participating proteins, including Gemin3, Gemin4, and GW-182. Step 7: Argonaute binds the miRNA and forms RISC, a group of proteins that is directed to the mRNA transcript to be silenced and either cuts the transcript to prevent translation or recruits other proteins to block translation.



miRNA biogenesis is initiated in the nucleus and terminated in the cytoplasm (Fig. 1). To process a miRNA molecule, a precursor must first be reverse-transcribed from miRNA-encoding genes (Fig. 1, step 1) by miRNA polymerase II. The resulting molecule is a single-stranded RNA with stem-loop secondary structure. In step 2, this primary RNA (pri-miRNA) is processed within the nucleus by ribonuclease III (RNASEN, also known as DROSHA) and the product of the DGCR8 gene (DiGeorge critical syndrome region gene 8, also known as PASHA), which together form the microprocessor complex. RNASEN cuts the pri-miRNA stem, leaving a shorter stem-loop structure called the preliminary miRNA (pre-miRNA) (Fig. 1, step 3) (1). The production of pre-miRNA marks the end of the nuclear component of miRNA biogenesis, necessitating pre-miRNA transport to the cytoplasm. In step 4, the exportin-5 karyopherin protein (XPO-5) binds pre-miRNA and shuttles the molecules through nuclear pores into the cytoplasm (1). The pre-miRNA is then cleaved (Fig. 1, step 5) into a 20- to 22-nucleotide (nt) strand of mature miRNA by the endonuclease DICER1. In cooperation with its cofactor, transactivation response (TAR) RNA–binding protein (TRBP) (1), DICER1 functions in a manner parallel to the microprocessor complex by measuring the necessary length of the miRNA and cutting the pre-miRNA to meet the specification (1). The resulting mature miRNA is then passed on to a member of the Argonaute family of proteins (Fig. 1, step 6) and, together with cofactors that include GEMIN3 (2), GEMIN4 (2), and the glycine-tryptophan protein of 182 kD (GW-182) (3), Argonaute binds the miRNA (Fig. 1, step 7) and forms the RNA-induced silencing complex (RISC). This complex represents a group of proteins that is directed by the miRNA to the mRNA transcript to be silenced and either cuts the transcript to prevent translation or recruits other proteins to block translation (1).

Although the canonical pathway is described in Fig. 1, processing steps once thought to be essential to miRNA biogenesis can be bypassed in special cases: miRNA production can proceed from pre-miRNA to miRNA by using Argonaute2 (Ago2) slicer activity instead of DICER1 (4, 5). This deviation from the canonical flow was demonstrated only for a small number of miRNAs, but the detection of this shortcut highlights the relatively limited understanding of the absolute requirement for each component of miRNA biogenesis and of the dynamic interactions among them.


Numerous studies have attempted to associate the concentrations or functions of miRNA pathway components with cancer prognosis (Fig. 2). For example, in lung cancer, low amounts of DICER1 mRNA, as measured with quantitative polymerase chain reaction (PCR) assays, are associated with a poorer prognosis when compared with cancers in patients with elevated DICER1 mRNA concentrations (6, 7). Low DICER1 expression was also noted in advanced breast tumors, in which quantitative PCR demonstrated that decreased expression is associated with aggressive, metastasis-prone tumors and that higher expression correlates with longer survival time (79). A pattern of down-regulation of both RNASEN and DICER1 is also associated with poor-prognosis skin, lung, breast, and ovarian cancers, among others (Fig. 2) (611).

Fig. 2. Predicting prognosis.

Shown are mRNA expression trends in poor-prognosis, advanced-stage cancers compared with tumors with better prognoses. Blue indicates higher gene expression in tumors compared with normal tissue samples; purple indicates lower expression expression in tumors compared with normal tissue samples; gray indicates no significant difference in gene expression in tumors versus normal tissue samples. BCC, basal cell carcinoma; AML, acute myeloid leukemia.


Analysis of prostate cancers shows overexpression of XPO-5, DICER1, and AGO2 in metastatic lesions (12), whereas RNASEN is up-regulated in advanced-stage cervical cancers (13) and metastasis-prone, poor-survival esophageal cancers (14) (Fig. 2). DICER1 is overexpressed in precursor lesions of lung cancers (adenomatous hyperplasias) as compared with advanced tumors (15), in some oral cancers (16), and in a cohort of acute myeloid leukemias (17), but no link has been found between expression and outcome in the latter two cancers. Curiously, mucoepidermoid cancers arising in the throat or upper esophagus exhibit, through immunohistochemistry, both over- and under-expression of DICER1 when compared with normal tissues from the throat and esophagus (18). Bladder cancers also exhibit a dual pattern of expression, in which DGCR8 was found to be down-regulated in premalignant tissue and up-regulated in malignant tissue, whereas the opposite is true for RNASEN, DICER1, and XPO-5 (19).

These divergent results across different cancers suggest that there are tissue-specific effects associated with aberrant expression patterns of miRNA pathway genes, but the wide variation in the findings may also be explained in part by technical differences among the studies. Notably, the results of immunohistochemical studies should be interpreted with caution given that a more recent study was unable to show a relationship between DICER1 mRNA levels and antibody-staining intensity in formalin-fixed paraffin-embedded tissues for many of the available antibodies to DICER1; at the mRNA level, no correlation was found between DICER1 or RNASEN expression and outcomes in breast cancers uniformly treated with anthracycline-based chemotherapy, but these authors noted a down-regulation in certain tumor subgroups, including triple-negative breast cancers (20). Also, melanomas, breast and ovarian cancers, and some cancer cell lines display copy number alterations (both gains and losses) in miRNA-encoding genes and in RNASEN, DGCR8, XPO-5, DICER1, and AGO (21).


DICER1 currently stands as the sole member of the miRNA pathway in which germline mutations have been found to predispose the carrier to human disease. Mutations in DICER1 were first identified in 11 of 11 children with the rare pediatric tumor pleuropulmonary blastoma (PPB), a neoplasm of the mesenchyme tissue of the lung or pleural cavity; all of these patients had a family history of PPB or the related familial tumor–dysplasia syndrome, which together are referred to as PPB-FTDS (22) [Online Mendelian Inheritance in Man (OMIM) no. 601200]. PPB-FTDS includes a variety of diverse tumor types, principally PPB; cystic nephroma (CN, a rare, benign tumor in the kidney); embryonal rhabdomyosarcoma (ERMS, a neoplasm of the connective tissue); ovarian sex cord-stromal tumors, especially Sertoli-Leydig cell tumors (SLCT), which originate from gonadal stromal tissue; and multinodular goiter (MNG, benign nodules of thyroid tissue) (23).

The finding of germline mutations in children with PPB prompted recent studies that identified germline mutations in individuals with tumors that fall within the PPB-FTDS spectrum, even when PPB itself was absent from family members (2426). Notably, DICER1 is situated at the genetic locus associated with familial multinodular goiter (MNG1), which was previously localized to chromosome 14q (27). In the largest family analyzed (20 affected individuals and no other reported illnesses), goiter segregates completely with a single–amino acid substitution in DICER1, which is predicted to be deleterious (26). Furthermore, DICER1 is the genetic explanation for the previously identified link between SLCT and MNG (28), with deleterious germline DICER1 mutations found in all three families that exhibit both conditions (26).

Another family contained a female who developed both Wilms tumor (WT) at 8 years of age and bilateral SLCT at 12 years of age (25). Overall, however, only 4 of 297 children with WT have been found to carry germline DICER1 mutations, suggesting a very modest contribution to the incidence of WT (24, 25, 29), especially when family members do not display any of the other diseases more typically associated with DICER1 mutations.

There have also been four cases of cervical embryonal rhabdomyosarcoma (cERMS), another rare embryonal tumor, occurring in families with DICER1 mutations (29). In addition to these tumors, there have been single cases of other DICER1-associated tumor types, such as intraocular medulloepthelioma (IM), an exceptionally rare embryonal tumor of the eye, which occurred in a child with PPB and a DICER1 mutation (25); together with an earlier report from the PPB registry that documented four instances of IM among PPB families, this finding suggests a strong etiological link between DICER1 and IM (30). No loss of heterozygosity of the wild-type allele has been detected in several types of tumors studied thus far. The importance of this observation is explained below.

The tissue-specific phenotypes of deleterious germline DICER1 mutations remain unexplained. At least one family exclusively exhibits the benign phenotype of MNG, whereas others exhibit only PPB, CN, or SLCT. Also, the penetrance of DICER1 mutations is extremely variable; some mutations are highly penetrant, as in an MNG family in which the disease phenotype was present in all 20 individuals with DICER1 mutations (26), when compared with another study in which only 8 of 25 individuals who harbor a DICER1 mutation exhibit a disease phenotype (MNG, SLCT, or CN) (25). In this study, the genomic DNA of 823 individuals with PPB spectrum tumors and 781 cancer cell lines derived from 50 different cancer types were sequenced, and the data revealed DICER1 mutations in 44 individuals from 19 families; of these, 26 presented with diseases that are likely or very likely to be associated with the DICER1 mutation present, demonstrating moderate penetrance for these mutations.

The contribution of DICER1 mutations to the different tumor types is variable, with a high proportion of individuals with PPB (11 of 14) and SLCT (4 of 7) harboring DICER1 mutations as compared with a low proportion of individuals with WT displaying DICER1 mutations (1 of 243) (25). Furthermore, there is no obvious relationship between the mutation type or location along the DICER1 gene and the type of disease observed. The question of whom to screen for DICER1 mutations remains unanswered, but certainly one should be strongly suspicious if any of the conditions described above (except perhaps WT) are present, even in the absence of a family history of such disorders.


Mouse models suggest that Dicer1 may function as a haplo-insufficient tumor suppressor—that is, a gene product whose normal function is to inhibit or control cell division in which one active gene copy alone does not produce enough product to prevent tumor formation and proliferation. In parallel studies, two research groups demonstrated that loss of a single allele of Dicer1 in mice fosters enhanced tumor proliferation (and even cancer initiation) in models of retinoblastoma (a tumor caused by mutations in the tumor suppressor–encoding Rb-1 gene) and lung cancer caused by a mutation in the Kras protooncogene; however, both studies also showed that loss of the second allele did not result in increased tumor proliferation, but rather, stunted tumor growth almost entirely (31, 32). This is an especially interesting result considering a recent study that showed that retinal cell degeneration in age-related macular degeneration results from a Dicer1 deficiency because Dicer1 protects retinal cells from the degenerative effects of toxic Alu RNA accumulation (Alu sequences are repetitive DNA elements, originally defined by the Alu restriction endonuclease) (33). It is possible that the retinal degeneration observed by Lambertz et al. (31) might be caused, not by an miRNA dysregulation effect, but instead by the loss of Dicer1 function as an Alu RNA–depletion agent. Results such as these weaken the notion that DICER1 dysfunction in cancer is mediated solely by dysregulation of miRNA processing.

The DICER1 haplo-insufficiency model represents a previously unknown and perplexing mechanism, considering the standard behavior for tumor suppressors. Classical tumor suppressor genes function by producing a protein that inhibits tumor growth, and one functioning copy of these genes is sufficient for tumor suppression. However, when the second copy is lost, the suppressing function is lost, and tumor growth is enhanced. If the proposed hapli-insuffiency mouse model is correct and also applies to humans, promotion of tumor growth would be conditional on the presence of one wild-type DICER1 allele, representing a unique situation among tumor suppressors. This mechanism is supported by the absence of loss of heterozygosity of the wild-type allele—that is, loss of the normal function (in this case, loss of tumor suppression) of the second allele of a gene in the context of one inactive allele—in human tumors that arise in DICER1-mutation carriers (22, 24, 26). In DICER1-related tumors, there currently are no data to indicate that the second wild-type allele has been inactivated.

The haplo-insufficiency hypothesis is challenged by the observation that the amounts of DICER1 mRNA and the corresponding protein in lymphocytes do not correlate with allelic status (26); under a model in which half-dosage is permissive for tumorigenesis, one would expect that individuals with only one functioning DICER1 allele (as a result of a deleterious germline DICER1 mutation) would, on average, have lower levels of mRNA and protein in their lymphocytes than those of individuals with no mutations in DICER1. These findings suggest that operational feedback loops between DICER1 and miRNAs that can themselves feed back and inhibit DICER1 messenger transcript translation, such as let-7 and miR-103/107 (34, 35), may render a strict haplo-insufficiency model overly simplistic.

Perhaps a germline mutation of a DICER1 allele serves as an initial contributing hit in a model in which a single DICER1 mutation is coupled with tissue-specific effects, such as gain- or loss-of-function of other cancer genes, to foster proliferation (as was seen in the Rb-1 and Kras- driven mouse retinoblastoma and lung cancer models, respectively). Moreover, second hits in DICER1 itself have not been ruled out. In this way, DICER1 mutations would create a sensitized background that lowers the threshold needed for tumors to develop. However, the universal importance of a gene such as DICER1 coupled with the extreme rarity of most of the diseases to which mutations in the gene have been linked is paradoxical, because a sensitized background should, in theory, increase the likelihood of tumor development in general and thus be associated with common cancers. Perhaps such an effect is developmental stage–dependent; many of the DICER1-associated tumors are embryonal or blastoma-type pediatric tumors. Thus, alterations in the DICER1 gene dose may be most influential when cells are in an undecided fate state.


Although DICER1 is the only part of the miRNA processing pathway for which germline mutations have been identified, single-nucleotide polymorphisms (SNPs) and somatic mutations in other pathway components have been associated with disease or disease progression. An allele of the RNASEN gene that contains six specific SNPs was associated with poor prognosis after lung cancer resection This result was reinforced when an association was found between the same RNASEN haplotype and reduced survival in a several cohorts of patients with early-stage resectable lung cancers (36).

Furthermore, eight distinct RNASEN SNPs (different from the six SNPs in the previously mentioned study) were associated with primary tumor recurrence in head and neck cancers, with the likelihood of recurrence increasing with the number of risk-associated SNPs present (37). In this study, high-risk individuals (more than nine risk-associated SNPs, including the eight within RNASEN) showed an average time of ~28 months to recurrence, compared with more than 93 months among low-risk groups (zero to four risk-associated SNPs). However, this study used a smaller sample set (150 patients), with no replication in an independent patient cohort. SNPs in XPO-5 were also found to be associated with modified risk of esophageal cancer (38). Although these studies demonstrate potentially interesting links between intragenic SNPs and disease state or outcome, none of the studies investigated functional differences associated with the presence of the SNPs at the molecular level.

Germline mutations in genes that regulate DNA mismatch repair cause chromosomal instability and can result in Lynch syndrome, which is characterized by high lifetime risks of colorectal, endometrial, and other cancers. Failure of DNA mismatch repair processes can cause tumor-promoting insertions and deletions (indels) in microsatellite DNA—short (one– to six–base pair), simple DNA sequence repeats in the human genome—which is a manifestation of the underlying genetic instability. Somatic frameshift mutations in XPO-5 in microsatellite-unstable endometrial cancer lead to an inability to transport pre-miRNA from the nucleus into the cytoplasm, resulting in a near-complete loss of miRNA (39). Upon the screening of 337 tumors, somatic mutations in XPO-5 were found in 26% of colorectal cancers that arose as a part of Lynch syndrome, 22% of sporadic colon tumors, 28% of sporadic gastric tumors, and 13% of sporadic endometrial tumors. A study of the colorectal cancer cell line HCT-15, which harbors a somatic mutation in XPO-5, demonstrated a complete lack of miRNA in the cytoplasm, and these cells rapidly formed tumors when injected into mice, a trait that was greatly diminished with the introduction of a vector that expressed wild-type XPO-5 (39).

In a similar study by the same group, somatic mutations were detected in TARBP2 (the gene that encodes TRBP, a DICER1 cofactor) in colorectal and endometrial tumors and in gastric cancer cell lines that exhibit microsatellite instability (40). In 282 primary malignancies, 26% were found to harbor TARBP2 mutations: 43% (13/30) of colorectal tumors arose as part of Lynch syndrome (and hence microsatellite-unstable), 25% (53 of 209) of sporadic colorectal tumors, and 14% (6 of 43) of sporadic gastric tumors.

Cells with a TARBP2 mutation have a reduced capacity to generate miRNA as compared with their wild-type counterparts. Injection of human colon carcinoma TRBP-deficient cells into athymic, nude mice results in tumor growth, which is slowed when wild-type cells are introduced, implying that in this model a functional version of TRBP is required for colon tumor suppression. Because TRBP is a cofactor for and stabilizer of DICER1 (41), and lowered DICER1 expression has been linked to the development of advanced endometrial cancers (42), mutations in TARBP2 in endometrial cancers could potentially contribute to pathogenesis by reducing the capacity of DICER1 to create mature miRNAs, which in turn causes aberrations in the global miRNA profile.

Further along in the pathway (Fig. 1), somatic mutations have been reported in AGO2 and TNRC6A (which encodes the Argonaute cofactor GW-182) in microsatellite-unstable colorectal and gastric cancers (43). Here, somatic AGO2 mutations were reported in 4 of 27 gastric cancers and 6 of 41 colorectal cancers, and somatic TNRC6A mutations were found in 2 of 27 gastric cancers and 5 of 41 colorectal cancers. Although the tumors with AGO2 mutations showed no substantial difference in immunostaining with an antibody to AGO2 when compared with similar tumors with wild-type AGO2, tumors that carry a TNRC6A mutation displayed significantly lower amounts of GW-182 protein when compared with similar tumors with wild-type TNRC6A. However, no associations were detected between TNRC6A expression and clinico-pathological parameters such as tumor stage or grade or patient survival (43); hence, the clinical and biological importance of these observations remains to be fully elucidated.


Clearly, the effects of aberrations in the miRNA-processing pathway on cancer initiation and progression are complex. No consistent pattern emerges, with conflicting data arising from different studies, particularly in those that assess concentrations of pathway-component mRNAs and proteins. Nevertheless, it is clear that the miRNA processing pathway plays a crucial role in the initiation and progression of many different cancer types. The ability to therapeutically target individual pathway components involved in tumorigenesis would be an attractive proposition if it could be done with sufficient precision.

At the DNA level, the disease-association picture is clearer; studies of mutations in genes that encode components of the miRNA processing pathway have revealed unexpected relationships between perturbations in this pathway and human cancers. However, the rarity of several of the diseases associated with germline mutations in this universally important regulatory pathway, in combination with an against-the-grain model of haplo-insufficiency for genes such as DICER1, leaves many unanswered questions. Coupled with studies in patients that characterize the tissue-specific expression of miRNA processing pathway genes, development of mouse models with relevant mutations should help to clarify the connection between certain cancers and aberrant miRNA maturation, selected gene silencing, and tissue-specific effects.

The predominant obstacle to successful translational application of this new knowledge is the broad range of processes that are regulated by even a single miRNA. Nevertheless, it may be that specific tumors are exquisitely sensitive to the effect of specific miRNAs, and in those circumstances, one can envision an miRNA-based therapeutic (44). We speculate that pediatric blastoma-like tumors are the most suitable candidates for such a treatment because these tumors are most consistently associated with germline DICER1 mutations, and therefore, they are likely to be susceptible to pertubations in the miRNA pathway. Delivery of miRNAs to selected tumor sites may require novel delivery systems such as the polyethylenimine-based nanoparticles that were recently reported to be effective carriers in a mouse model of colon cancer (45). The next few years should see translational developments at the cutting edge of miRNA processing and cancer research (46).

References and Notes

  1. Acknowledgments: : We thank J. S. Reis-Filho for comments on an early draft of this manuscript. Funding: Supported by the Mendon F. Schutt Foundation, Susan G. Komen for the Cure, and the Marsha Rivkin Center for Ovarian Cancer Research. Competing interests: The authors declare no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article