Drosha mediates destabilization of Lin28 mRNA targets

Abstract

Lin28 plays important roles in development, stem cell maintenance, oncogenesis and metabolism. As an RNA-binding protein, it blocks the biogenesis primarily of let-7 family miRNAs and also promotes translation of a cohort of mRNAs involved in cell growth, metabolism and pluripotency, likely through recognition of distinct sequence and structural motifs within mRNAs. Here, we show that one such motif, shared by multiple Lin28-responsive elements (LREs) present in Lin28 mRNA targets also participates in a Drosha-dependent regulation and may contribute to destabilization of its cognate mRNAs. We further show that the same mutations in the LREs known to abolish Lin28 binding and stimulation of translation also abrogate Drosha-dependent mRNA destabilization, and that this effect is independent of miRNAs, uncovering a previously unsuspected coupling between Drosha-dependent destabilization and Lin28-mediated regulation. Thus, Lin28-dependent stimulation of translation of target mRNAs may, in part, serve to compensate for their intrinsic instability, thereby ensuring optimal levels of expression of genes critical for cell viability, metabolism and pluripotency.

Introduction

Lin28 is an evolutionarily conserved RNA-binding protein that plays pleiotropic roles in diverse physiological and pathological processes including development, pluripotency, metabolism and cancer (reviewed in refs. 1 and 2). Mechanistically, Lin28 inhibits the biogenesis of let-7 miRNAs by binding to the terminal loops of let-7 precursors, leading to inhibition of processing and induction of precursor degradation (reviewed in refs. 1, 3 and 4). While let-7 negatively regulates the expression of genes involved in cell growth, differentiation and metabolism, Lin28 antagonizes this activity by both directly stimulating the translation of a subset of these genes and by blocking let-7 production (reviewed in refs. 2 and 5). In human and mouse embryonic stem (ES) cells and embryonal carcinoma (EC) cells, Lin28 was shown to selectively bind to a subset of mRNAs, thereby promoting their translation.^6-10 This mode of action of Lin28 is believed to contribute critically to the mechanism by which Lin28 impacts pluripotency, proliferation and survival of ES and EC cells (reviewed in ref. 2). It has been estimated that ~5% of polyadenylated mRNAs in human ES cells are bound and regulated by Lin28, including mRNAs encoding metabolic enzymes and ribosomal proteins whose expression levels are known to be coupled to cell growth and metabolism.⁹ Importantly, this set of transcripts overlaps significantly with those enriched for glucose, insulin and diabetes-related genes identified using gene set enrichment analysis (GSEA).¹¹

The molecular basis underlying Lin28 mRNA target recognition and regulation has been investigated. All targets validated to date contain Lin28-binding sites, designated Lin28-responsive elements (LREs) in their coding regions or in their 5′- or 3′-UTRs (reviewed in ref. 2). It has been proposed that Lin28 selectively binds to target mRNAs through recognition of LREs, while recruiting co-factor RNA helicase A (RHA) to promote translation (reviewed in ref. 2). Recently, using bioinformatic approaches combined with in vitro binding and in vivo reporter gene analysis, a unique sequence and structural motif shared by multiple LREs has emerged. This motif is characterized by a small “A” bulge flanked by two G:C base-pairs embedded in a complex secondary structure (ref. 12 and see below). Remarkably, the critical roles of the “A” bulge in Lin28 binding and regulation of translation were underscored by the loss of both activities through single nucleotide substitution or deletion at this position.¹²

The RNase III enzyme Drosha is a component of the Microprocessor, which also contains an RNA-binding protein, designated DGCR8, which is essential for processing primary miRNAs (pri-miRNAs) to precursor miRNAs (pre-miRNAs)^13-16 (reviewed in ref. 17). While Drosha acts as the catalytic engine, DGCR8 assists Drosha by recognizing the RNA substrate.¹⁸ The activity of the Microprocessor is modulated by an autoregulatory feedback loop.^19-23 The 5′-UTR of the DGCR8 mRNA contains a hairpin that mimics a pri-miRNA and that is cleaved by the Microprocessor, resulting in destabilization of DGCR8 mRNA. In contrast, DGCR8 stabilizes Drosha via protein-protein interaction.¹⁹ It has been proposed that when there is excess Microprocessor activity, it leads to cleavage and destabilization of DGCR8 mRNA, thereby decreasing DGCR8 protein levels, which in turn destabilizes Drosha, thereby lowering Microprocessor activity.^19-24

Drosha-mediated mRNA destabilization is not limited to DGCR8. An increase in the level of ~100 mRNAs in response to Drosha or DGCR8 loss (but not Dicer or Ago2 loss) was observed in siRNA-knockdown and microarray studies using HeLa cells.¹⁹ Similarly, Karginov et al. detected Drosha-dependent and Ago2-independent cleavage events within a cohort of mRNAs in transcriptome-wide RACE (rapid amplification of cDNA ends) studies using conditional Drosha/Ago2-null mouse ES cell lines.²³ In Drosophila S2 cells, Drosha depletion by dsRNA, followed by tiling microarray analysis, identified six putative Drosha-regulated mRNAs based on their upregulation in response to Drosha (but not Dicer) depletion. All six mRNAs harbor a predicted miRNA-mimetic hairpin in their 5′-UTR, 3′-UTR or coding region.²² Except for DGCR8, none of the above studies had provided evidence for in vitro direct mRNA cleavage by a recombinant Drosha or Microprocessor. However, two recent studies showed that many mRNAs were destabilized in a Drosha-dependent manner, and that a number of these mRNAs contain stem-loop structures that could be cleaved directly by Drosha in vitro.^25,26 Collectively, these studies suggest the existence of a widespread Drosha-mediated mechanism in controlling mRNA stability in vivo. However, it has yet to be determined whether the same (i.e., the Microprocessor) or different Drosha-containing complexes are involved in destabilizing these various classes of mRNAs, given that at least two distinct Drosha-containing complexes have been identified in human cells.¹³

In this report, we provide evidence that a subset of Lin28 mRNA targets are destabilized in a Drosha-dependent manner, and that the LREs may contribute to this regulation. Given the predominant nuclear and cytoplasmic localization of Drosha and Lin28, respectively, we postulate that cytoplasmic Lin28-mediated stimulation of translation may, in part, serve to compensate for the nuclear loss of these mRNAs, thereby ensuring proper levels of expression, which is critical for the regulatory balance required for normal (or malignant) cell growth, metabolism and pluripotency.

Results

LRE-containing reporter mRNAs are unstable

Our laboratory has been testing the functionality of Lin28-responsive elements (LREs) using a well-characterized firefly luciferase (FFL) reporter system. Typically, a putative LRE is placed at the 3′-UTR of the reporter gene (Fig. 1A), and the resulting construct is transfected into HEK293 cells (which do not express endogenous Lin28) in absence or presence of increasing amounts of Lin28 expressed from a co-transfected plasmid (FL-Lin28).⁸ Luciferase activities are measured 18–24 h post-transfection, and results are presented after normalization against luciferase mRNA levels. When an LRE responds positively for translational stimulation, a Lin28 dose-dependent increase in the luciferase activity is seen (arbitrarily setting the first point, without FL-Lin28 co-transfection, hence in absence of Lin28 expression as 1).^8,9,12 A puzzling theme that repeatedly drew our attention was that whenever an LRE proved to be Lin28-responsive, the first luciferase activity point (in absence of Lin28 expression) was always lower than for those containing no LRE or a mutant inactive LRE. Such a property has since become a useful first indication of an active LRE, as illustrated in Figure 1B. Shown are luciferase activities of reporter genes containing no LRE (FFL) or LREs from five validated Lin28 targets (Oct4, Hmga1, Rps19, Rps13 and Her2).^8,9,12 The schematic structures of the wild-type (Oct4–95, Hmga1–129 and Rps19–106) and the “A” bulge point mutation counterparts (Oct4–14T, Hmga1–53T and Rps19–3xT) are shown in Figure 1C. These experiments were performed in absence of FL-Lin28 co-transfection (i.e., no Lin28 expression) and the luciferase activities are presented without normalization against luciferase mRNA levels. All wild-type (WT) LREs (Oct4–95, Hmga1–129, Rps19–106, RPs13-ORF, Her2–200) exhibit a luciferase activity that is ~20% lower than that of the parental FFL, which is arbitrarily set as 1 (Fig. 1B, top panel, compare left first bar with bars of the third, fifth, seventh, eighth and ninth). Strikingly, such a decrease in luciferase activity is not seen with all three mutant (MUT) counterparts (Oct4–14T, Hmga1–53T and Rps19–3xT) (compare left first bar with bars of the second, fourth and sixth). This pattern of luciferase activity is mirrored by that of corresponding mRNA levels (Fig. 1B, bottom panel), suggesting that presence of WT LREs may negatively affect the stability of the reporter mRNAs. To test this possibility, RNA stability analysis was performed using the three construct pairs: Oct4–14T/Oct4–95, Hmga1–53T/Hmga1–129 and Rps19–3xT/Rps19–106. Thus, HEK293 cells were transfected with the indicated individual construct, followed by measurement of luciferase mRNA levels at 0, 2 and 4 h after actinomycin D (ActD) treatment, 18 h post-transfection. All three WT LREs (Oct4–95, Hmga1–129 and Rps19–106) conferred a faster decay rate to the host luciferase mRNA compared with their mutant counterparts (Oct4–14T, Hmga1–53T and Rps19–3xT) (Fig. 1D), suggesting that presence of WT LREs destabilizes the reporter mRNA. Notably, the same point mutations that abolished Lin28 binding and regulation of translation¹² also blunted this destabilization effect, although Lin28 per se is not involved in decay of LRE-containing mRNAs (see Discussion).

Figure 1. Presence of LREs destabilizes reporter mRNA. (A) Schematic diagram of firefly luciferase reporter construct, with blue box representing coding region and thin blue line representing 3′-UTR where a LRE was inserted. (B) Results of reporter gene assays. Parental reporter plasmid (FFL) or plasmids containing individual wild-type (WT) or mutant (MUT) LREs were each transfected into HEK293 cells, followed by measurement of luciferase activities (top panel) and mRNA levels (bottom panel) 18 h post-transfection. Top, relative luciferase activities plotted with FFL luciferase activity arbitrarily set as 1. Bottom, relative luciferase mRNA levels with FFL mRNA level arbitrarily set as 1. Numbers are mean ± SD (n = 3). ** p < 0.01 (compared with FFL levels). (C) Schematics of secondary structures of three LREs predicted by mFOLD.¹² The characteristic “A” bulges are depicted in red with corresponding point mutations (“A” replaced by “U”) labeled in blue. The point mutations were predicted not to affect the secondary structures of the LREs.¹² The sizes of the LREs Oct4–95, Hmga1–129 and Rps19–106 in nucleotides are 95, 129 and 106, respectively. (D) Results of RNA stability analysis. HEK293 cells were transfected with each indicated reporter construct. Actinomycin D was added to the cells to a final concentration of 5 ug/ml to stop transcription 18 h post-transfection. Total RNAs were isolated at 0, 2 and 4 h time points after ActD treatment and luciferase mRNA levels determined by RT-qPCR analysis. Results are presented after normalization against internal control β-actin mRNA levels with 0 time point luciferase mRNA levels arbitrarily set as 1. Numbers are mean ± SD (n = 3). * p < 0.05; ** p < 0.01 (compared with corresponding time points in the WT group). T1/2, half-life of indicated mRNA.

Drosha contributes to the destabilization of reporter mRNAs

As Drosha has been implicated in mRNA destabilization,^19-23,25 we asked whether Drosha might contribute directly to destabilizing LRE-containing mRNAs. To address this question, we downregulated Drosha expression in HEK293 cells by transfecting a siRNA specifically targeted to human Drosha (siDrosha). Approximately 80% of Drosha knockdown at the RNA level was achieved 72 h after transfection into HEK293 cells with siDrosha compared with control siRNA (siCon) (Fig. 2A, left column, compare yellow bar with blue bar). The level of DGCR8 mRNA was predictably increased in response to Drosha downregulation (middle column), while the level of non-targeted β-tubulin mRNA was not altered (right column). Consistent with this level of RNA change, the DGCR8 protein level also went up (Fig. 2B, middle blot, compare lane two to lane one) when the Drosha protein level went down (top blot, compare lane two to lane one). Recapitulating our earlier observations (Fig. 1B, bottom panel), in siCon-transfected HEK293 cells, the level of luciferase mRNA containing Oct4–95 was ~20% lower than that containing the mutant counterpart Oct4–14T (Fig. 2C, left panel, first column from left, compare pink bar with red bar). Similar observations were made with Hmga1 (middle panel, first column from left) and Rps19 (right panel, first column from left). Importantly, when Drosha was downregulated (Fig. 2A and B), the difference in the level of various luciferase mRNAs containing wild-type vs. mutant Oct4 LRE was abolished (Fig. 2C, left panel, middle column, compare pink bar with red bar). This response is exemplified by both LREs from Hmga1 (middle panel, middle column) and Rps19 (right panel, middle column).

Figure 2. Drosha is involved in destabilizing LRE-containing reporter mRNA. HEK293 cells were transfected with siCon, siDrosha or siDicer. Forty-two hours post-siRNA transfection, a second transfection with reporter constructs (see below) was performed. RNA and protein were then extracted 18 h after the second transfection and levels determined by RT-qPCR (A, C and D) and western blotting (B and E), respectively. (A and D) Relative levels of indicated endogenous mRNAs are presented with levels from siCon-transfected cells arbitrarily set as 1. (B and E) Western blot results using specific antibodies indicated on the right. Beta-tubulin was used as loading control. (C) Effects of Drosha or Dicer knockdown on luciferase reporter mRNA levels. HEK293 cells were first transfected with siCon, siDrosha or siDicer, and second with each indicated report construct as described above. RNAs were extracted 18 h after the second transfection and analyzed by RT-qPCR. Relative luciferase mRNA levels are presented with mutant LRE-containing luciferase mRNA levels arbitrarily set as 1. Numbers are mean ± SD (n = 3). * p < 0.05; ** p < 0.01.

The above observation is likely a direct effect of Drosha and not an indirect effect of miRNA depletion due to Drosha knockdown, because luciferase mRNA level restoration was not observed when Dicer, another essential factor for miRNA biogenesis, was specifically downregulated (Fig. 2C, first column from right in all three panels). The efficacy and specificity of Dicer knockdown were confirmed at both the RNA and protein levels. The reduction of Dicer at the RNA level was typically ~80% in siDicer- vs. siCon-transfected HEK293 cells (Fig. 2D, left column, compare pink bar with red bar), while no appreciable change occurred in the β-tubulin mRNA control (right column). Under conditions where the Dicer protein level was decreased by ~80% (Fig. 2E, top blot, compare lane two to lane one), there was an increase in the protein level of c-Myc (middle blot), a known target of let-7 inhibition,⁵ further confirming effective downregulation of Dicer expression. Together, these results strongly support the hypothesis that a Drosha-containing complex may be in place to cleave and destabilize mRNAs in an LRE-specific manner, and that the same point mutations that abrogate Lin28 binding and translational regulation also abrogate this Drosha-dependent activity.

Drosha contributes to the destabilization of endogenous Lin28 mRNA targets

The above studies suggest the intriguing possibility that endogenous Lin28 mRNA targets may be substrates of Drosha-dependent destabilization. To test this hypothesis, we specifically reduced Drosha expression in PA-1 cells, a well-characterized human EC cell line that endogenously expresses high levels of both Lin28 and Oct4,^8,9,12 and analyzed effects on expression of endogenous Lin28 mRNA targets. When Drosha RNA was reduced by ~90% (Fig. 3A, first column from left, compare purple bar with blue bar) and Drosha protein by ~85% (Fig. 3B, top blot, compare lane two to lane one), increases in the RNA levels of five LRE-containing Lin28 targets (Oct4, Hmga1, Her2, Rps19 and Rps13) were also observed, albeit to varying extents (Fig. 3A, second through sixth columns from left). Similar results were obtained when another Drosha siRNA (siDrosha-2, previously called siRNA D1²⁷) targeted to a different region of Drosha was used (Fig. S1). This siRNA was previously shown to both efficiently and specifically knockdown Drosha expression in human cells.²⁷ Consistent with RNA level changes, the protein levels of Oct4 (Fig. 3B, second blot from top), Hmga1 (third blot from top) and Her2 (fourth blot from top) also increased. Increases in protein levels of Rps19 (Fig. 3B, third blot from bottom) and Rps13 (data not shown) were not detected, perhaps due to the high intrinsic stability of these ribosomal proteins. As expected, under the same conditions, DGCR8 RNA (a known mRNA target of Drosha, Fig. 3A, fourth column from right) and pri-miR21 RNA (a known pri-miRNA target of Drosha, Fig. 3A, first column from right) increased accordingly, while the levels of nucleolin and β-tubulin mRNAs did not change (Fig. 3A, second and third columns from right). The protein level of DGCR8 also increased (Fig. 3B, second blot from bottom, compare lane two to lane one).

Figure 3. Drosha negatively affects the steady-state level of endogenous Lin28 mRNA targets. PA-1 cells were transfected with siCon, siDrosha or siDicer. RNA and protein were extracted 72 h post-transfection and levels determined by RT-qPCR (A and C) and western blot analysis (B and D), respectively. In (A and C), RNA levels from siCon-transfected cells were arbitrarily set as 1. In (B and D), β-tubulin was used as loading control. Results are representative of three independent transfection experiments.

The elevated mRNA levels in the five Lin28 targets as well as DGCR8 were likely not a result of an indirect effect of miRNA depletion, since Dicer knockdown did not lead to such effects (Fig. 3C). The effectiveness of Dicer knockdown was confirmed by immunoblot analysis (Fig. 3D, top blot, compare lane two to lane one), which showed an increase in the c-Myc protein level in the Dicer-knockdown group when compared with the control group (middle blot, compare lane two to lane one).

Next, ActD time course experiments were performed using PA-1 cells to determine whether the observed Lin28 mRNA target level change was a result of increased mRNA stability. As shown in Figure 4A, when Drosha was downregulated (Fig. 3A and B), the decay rates of Oct4, Hmga1, Her2, Rps19 and DGCR8 mRNAs were all reduced, while those of β-tubulin and nucleolin (data not shown) mRNAs were not, suggesting that the increased levels of Lin28 mRNA targets in response to Drosha knockdown (Fig. 3A) were likely a result of increased mRNA stability.

Figure 4. Drosha negatively affects the stability of endogenous Lin28 mRNA targets. (A and D) PA-1 cells were transfected with siCon, siDrosha, empty vector or FL-DGCR8. Act D was added to a final concentration of 5 ug/ml 48 h post-transfection. RNAs were harvested at 0, 1, 2 and 4 h time points and levels determined by RT-qPCR analysis. Results are presented after normalization against internal control GAPDH mRNA levels with 0 time point mRNA levels arbitrarily set as 1. Numbers are mean ± SD (n = 3). (B and C) PA-1 cells were transfected with FL-DGCR8 or empty vector. Protein and RNA were isolated 48 h post-transfection and analyzed by western blot (B) and RT-qPCR (C), respectively. Results are representatives of two independent transfection experiments.

To provide further evidence that Drosha might be involved in destabilization of endogenous Lin28 mRNA targets, we performed reciprocal experiments by increasing the level of Drosha expression in PA-1 cells. Although we were unable to increase Drosha protein level by transfecting Drosha expressing vectors (data not shown), we were able to do so by transfecting a tagged DGCR8 expression vector (FL-DGCR8) (Fig. 4B, top blot, compare lane two to lane one), which resulted in a ~2-fold increase in the protein level of endogenous Drosha (middle blot, compare lane two to lane one), consistent with the notion that Drosha is stabilized by DGCR8 through protein-protein interaction.¹⁹ Thus, when the Drosha protein level was increased, the levels of Oct4, Hmga1 and Her2 mRNAs were reduced (Fig. 4C, first through third columns from left, compare red bars with blue bars), suggesting decreased mRNA stability of these gene products. This was further supported by the increased decay rate of Hmga1 mRNA in the FL-DGCR8-transfected cells (Fig. 4D). The levels of Rps19 and Rps13 mRNAs did not significantly change, likely due to higher mRNA stability compared with those of Oct4, Hmga1 and Her2 (Fig. 4A, compare bottom left panel with top three panels). Based on these results, here we propose that a Drosha-associated complex is likely involved in the destabilization of Lin28 mRNA targets that contain LREs.

Discussion

A serendipitous observation that reporter genes bearing LREs display lower steady-state mRNA levels than those with mutant LRE counterparts (Fig. 1B) prompted us to investigate the underlying molecular basis for this phenomenon. By RNA stability analysis we demonstrate that the lower level of mRNAs with wild-type LREs is caused by the faster decay rates of those mRNAs in comparison to their mutant mRNA counterparts (Fig. 1C and D). Remarkably, the same LRE mutations previously shown to abolish Lin28 binding and translational regulation are found in this study to also abrogate destabilization (Fig. 1C and D). Using an RNAi approach, we demonstrate that this destabilization is Drosha-dependent and Dicer-independent (Fig. 2), consistent with Drosha being actively involved in destabilizing LRE-containing mRNAs through a direct rather than an indirect (i.e., via miRNA depletion) mechanism. Furthermore, we provide evidence that a group of endogenous Lin28 mRNA targets, including Oct4 mRNA, is destabilized in a Drosha-dependent manner (Figs. 3 and 4). Based on these observations, we propose that a subset of Lin28 mRNA targets is likely also substrates of Drosha-targeted degradation.

The precise nature of the putative Drosha-associated complex is unclear. We were unable to detect cleavage products after exposing in vitro LRE-containing transcripts to immunoaffinity purified Drosha and DGCR8, despite the observation that pri-miR16 hairpin transcripts were readily cleaved by Drosha under the same conditions (data not shown). A parsimonious explanation for these observations is that an, as-yet-unknown, cofactor(s) necessary for Drosha to process LRE-containing transcripts is missing in the in vitro processing analysis. This contention is supported by several observations. First, a large number of mRNAs have been identified as putative Drosha targets, yet only a few have been validated to contain pri-miRNA-like hairpins that can be cut by Drosha in vitro.^19,22,23,25 The majority of the putative targets, however, do not contain canonical pri-miRNA hairpins, but can be folded into extensive secondary structures, suggesting the existence of alternative cofactors recognizing these structural motifs that might confer upon Drosha a specificity for these particular targets.²³ Second, at least two distinct Drosha-associated complexes have been isolated from human cells.¹³ With Drosha being the catalytic engine for both the small and large complexes, the small complex is the Microprocessor that is essential for pri-miRNA processing both in vivo and in vitro. The large complex, on the other hand, contains multiple classes of RNA-binding proteins, including RNA helicases, double-stranded RNA-binding proteins, hnRNPs and the Ewing′s sarcoma family of proteins, and exhibits a negligible activity in processing pri-miRNAs based on in vitro analysis.¹³ It has been proposed that the large complex likely functions in processing RNAs other than pri-miRNAs.¹³ In line with this concept, here, we postulate that a multi-protein complex contains Drosha as its catalytic engine, but that this complex is distinct from the Microprocessor and is likely responsible for destabilizing LRE-containing mRNAs in vivo.

The observation that the same mutations in the LREs that negatively affect Lin28 binding and stimulation of translation also abrogate Drosha-dependent mRNA destabilization is especially intriguing and suggests that the two processes may be mechanistically linked. While a putative Drosha-associated complex may actively target LRE-containing mRNAs for degradation within the nucleus (Drosha is known to localize in the nucleus), Lin28 may act in the cytoplasm to stimulate translation of the mRNAs that have escaped Drosha attack, thereby ensuring their optimal expression levels. It is noteworthy that Lin28 does not seem to directly affect the stability of LRE-containing mRNAs by competing with nuclear Drosha for binding to the LREs, as Lin28-knockdown in human ES and EC cells did not lead to destabilization of LRE-containing mRNAs,^8,9 nor did ectopic Lin28 expression in HEK293 cells increase their steady-state levels.⁸ However, it is not unprecedented that the same RNA sequence/structural motif is targeted by different regulatory factors in different subcellular compartments. For example, the terminal loop (TL) of pri/pre-let-7 miRNA plays a central role in regulating let-7 biogenesis. The RNA-binding protein KSRP binds to a conserved sequence/structural motif in the TL and promotes pri-let-7 processing by Drosha in the nucleus.²⁸ hnRNP A1 antagonizes this effect by competing for binding to the same motif in the TL of pri-let-7.²⁹ Lin28 is yet another important regulatory factor in this process: in the cytoplasm, Lin28 interacts with the same motif in the TL of pre-let-7, leading to inhibition of Dicer processing and induction of uridylation and degradation of pre-let-7 (reviewed in ref. 2). During pri-miRNA processing, DGCR8 acts as an essential cofactor by specifically recognizing the RNA substrate as well as by directing Drosha to cut the RNA at precise positions.¹⁸ It is tempting to postulate that an as-yet-unknown cofactor(s) may play an analogous role by specifically recognizing LREs and facilitating Drosha cleavage of the respective mRNAs in the nucleus. Identification and characterization of this putative cofactor(s), together with associated complex, will prove important in better understanding not only the regulation of LRE-containing mRNAs by Drosha, but also of mechanisms of Drosha-mediated mRNA decay. Future experiments will also include capture of in vivo Drosha decay intermediates following nuclear exonuclease knockdown so that a direct role of Drosha-mediated cleavage of LRE-containing mRNAs can be formally established.

Then, what might be the biological significance of this apparently coupled processes: Drosha-mediated destabilization and Lin28-mediated stimulation of translation of LRE-containing mRNAs? An outstanding characteristic feature shared by both ES and tumor cells is the need for rapid induction of cell growth and proliferation. The unique cell cycle of ES cells also is intrinsically tied to self-renewal and pluripotency.^30-32 The examples presented below highlight how such a coordinately regulated mechanism of control of gene expression may be useful in these cell types.

For instance, Lin28 is known to critically regulate ES and EC cell growth and survival, partly through stimulation of translation of a cohort of genes that directly control cell growth and survival (reviewed in ref. 2). These regulated genes include cell cycle regulators,⁶ the replication-dependent histone H2a,⁷ metabolic enzymes and ribosomal proteins.⁹ Moreover, it is specifically the expression level of these gene products that controls cell growth and survival; suboptimal levels of expression are detrimental to the growth and survival of ES and tumor cells. The findings reported here suggest that a subset of Lin28 mRNA targets is targeted for Drosha-mediated degradation, and that Lin28 acts to compensate for this loss, thus safeguarding proper expression levels.

Such a coupled processes may also be important for both ES self-renewal and differentiation. In ES cells, the balance between self-renewal and differentiation is highly sensitive to the dosage of core transcriptional regulators such as Oct4 (reviewed in refs. 33 and 34). A less than 2-fold increase in Oct4 expression induces mouse ES cell differentiation into primitive endoderm and mesoderm, whereas reducing Oct4 expression by half triggers differentiation into trophoblasts.³⁵ Similarly, during post-implantation development, a continuing decline in Oct4 concentration ensures the end of pluripotency, but a modest increase in Oct4 concentration above a certain threshold is sufficient to resuscitate pluripotency.³⁶ Thus, maintaining the proper level of Oct4 expression plays a critical role in determining the ES cell′s fate. Perhaps, as a result, Oct4 expression is tightly controlled at multiple levels, including transcription, translation and via posttranslational modifications that affect the activity and stability of the Oct4 protein. At each level, both positive and negative regulatory factors are involved (reviewed in ref. 33). Here we propose that Oct4 mRNA destabilization by a Drosha-associated complex may represent a novel and additional layer of regulation of Oct4 expression under both undifferentiated and differentiating conditions. Specifically, in undifferentiated ES cells, an inhibition of Oct4 expression by both Drosha at the mRNA stability level and by miR-145³⁷ at the translational level may antagonize the positive regulatory effect of Lin28, preventing Oct4 overexpression. Upon initiation of ES differentiation, Drosha-targeted Oct4 mRNA destabilization, in addition to a switch from high Lin28/low miR-145 to low Lin28/high miR-145, may contribute critically to the irreversible silencing of self-renewal and pluripotency programs. Finally, we propose that similar regulatory mechanisms may operate in many tumor cells, where a “high Lin28 state” may contribute to maintenance of the undifferentiated and highly proliferative states, characteristic of many cancer cells. As Lin28 has been implicated in directly regulating expression of glucose, insulin and diabetes-related genes,¹¹ such a coupled mechanism may also play an important role in modulating mammalian glucose metabolism.

In summary, we have found that a subset of genes positively regulated by Lin28 is also a likely target of Drosha-mediated degradation. This novel coordinately regulated mechanism of controlling gene expression may contribute to the phenotype of both ES and tumor cells as well as regulation of glucose metabolism in mammalian cells.

Materials and Methods

Antibodies, siRNAs and plasmids

Antibodies for Drosha (Cell Signaling, 3364S), DGCR8 (Novus Biologicals, NBP1–04676), Rps19 (Abcam, ab57643), c-Myc (Santa Cruz, sc-764), Dicer1 (Millipore, 04–721), Oct4 (Santa Cruz, sc-5279), Hmga1 (Santa Cruz, sc-8982), Her2 (Dako, A0485) and β-tubulin (Abcam, ab6046) were purchased. siDrosha (Dharmacon, L-016996–01), siDicer (Dharmacon, L-003483–00–0005), siCon (Dharmacon, D-001810–10–05) and siDrosha-2 (siRNA D1) (Applied Biosystems, Silencer siRNA duplexes 5′ phosphate-AUCUCCUCCUCAGGCACCAGGdTdT-3′ and 5′-CCUGGUGCCUGAGGAGGAGAUdTdT-3′) were purchased. The luciferase reporter plasmids FFL, Oct4–95, Oct4–14T, RPS19–106, Rps19–3xT, Hmga1–129, Hmga1–53T, Rps13-ORF and Her2–200 were previously described.^9,12 The FL-DGCR8 expression vector (pFLAG/HA-DGCR8, Addgene plasmid 10921) was previously described.¹⁴

RNA extraction and real-time RT-qPCR

These were performed essentially as previously described.^9,12 The PCR primers for human genes are listed below. Oct4: 5′-GTGGAGGAAGCTGACAACAA (forward) and 5′- GCCGGTTACAGAACCACACT (reverse); Hmga1: 5′-CAGCGAAGTGCCAACACCTAAG (forward) and 5′-CCTTGGTTTCCTTCCTGGAGTT (reverse); Her2: 5′-AGCACTGGGGAGTCTTTGTG (forward) and 5′-CTGAATGGGTCGCTTTTGTT (reverse); Rps19 5′-AGACGTGAACCAGCAGGAGT (forward) and 5′-AGCTCGCGTGTAGAACCAGT (reverse); Drosha: 5′-CATGTCACAGAATGTCGTTCCA (forward) and 5′-GGGTGAAGCAGCCTCAGATTT (reverse); Dicer: 5′-GTACGACTACCACAAGTACTTC (forward) and 5′-ATAGTACACCTGCCAGACTGT (reverse); DGCR8: 5′-CAAGCAGGAGACATCGGACAAG (forward) and 5′-CACAATGGACATCTTGGGCTTC (reverse); Pri-miR21: 5′-GTTCGATCTTAACAGGCCAGAAATGCCTGG (forward) and 5′-ACCAGACAGAAGGACCAGAGTTTCTGATTA (reverse); β-actin: 5′-ATCAAGATCATTGCTCCTCCTGAG (forward) and 5′-CTGCTTGCTGATCCACATCTG (reverse); β-tubulin: 5′-CGTGTTCGGCCAGAGTGGTGC (forward) and 5′-GGGTGAGGGCATGACGCTGAA (reverse); Nucleolin: 5′-CAGAACCGACTACGGCTTTC (forward) and 5′-ACGCTTTCTCCAGGTCTTCA (reverse); firefly luciferase: 5′-GCT GGGCGTTAATCAGAGAG (forward) and 5′-GTGTTCGTCTTCGTCCCAGT (reverse); Renilla: 5′-GCAAATCAGG CAAATCTGGT (forward) and 5′-GGCCGACAAAAATGATC TTC (reverse).

Cell culture and transfection

Human HEK293 and PA-1 cells were cultured using standard protocols provided by the ATCC. Cell transfections with DNA and siRNA were performed as previously described.^9,12

Protein extraction and western blot analysis

These were performed as described previously.^9,12

Luciferase assays

These were done essentially as previously described.^9,12 Briefly, the indicated firefly luciferase reporter plasmids were individually transfected into HEK293 cells in a 48-well plate scale, together with Renilla reporter plasmid for control purposes. The amount of total DNA transfected per well was 150 ng that included 100 ng of reporter plasmid, 5 ng of Renilla DNA and an appropriate amount of empty vector pFLAG-CMV-2 (Sigma, E7398). Luciferase activities and mRNA levels were determined 18 h post-transfection. Relative luciferase activities and luciferase mRNA levels were presented with parental FFL luciferase activity and mRNA level arbitrarily set as 1.

Actinomycin D time-course analysis

For report RNA stability analysis (Fig. 1D), HEK293 cells were transfected with the individual reporter construct. Eighteen hours later, ActD was added to the culture medium to a final concentration of 5 ug/ml. Total RNAs were harvested at 0, 2 and 4 h time points following ActD treatment, and luciferase mRNA levels were determined by RT-qPCR. Results are presented after normalization against internal control β-actin mRNA levels with 0 time point luciferase mRNA levels arbitrarily set as 1. For endogenous mRNA stability analysis (Fig. 4A), PA-1 cells were transfected with siCon or siDrosha (or siDrosha-2), followed by addition of ActD to a final concentration of 5 ug/ml 48 h post-transfection. Total RNAs were isolated at 0, 1, 2 and 4 h time points and levels determined using RT-qPCR. Results are plotted after normalization against GAPDH mRNA levels with 0 time point mRNA levels arbitrarily set as 1.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interests were disclosed.