Disclosure of potential conflicts of interest is found at the end of this article.
Author contributions: S.P.: collection and assembly of data, data analysis and interpretation; L.-L.C.: data analysis and interpretation, financial support; X.-X.L.: collection and assembly of data; L.Y.: data analysis and interpretation; H.L.: conception and design, financial support; G.G.C.: conception and design, financial support; Y.H.: conception and design, data analysis and interpretation, financial support.
First published online in STEM CELLSEXPRESS January 7, 2011.
Lin28 inhibits the expression of let-7 microRNAs but also exhibits let-7-independent functions. Using immunoprecipitation and deep sequencing, we show here that Lin28 preferentially associates with a small subset of cellular mRNAs. Of particular interest are those for ribosomal proteins and metabolic enzymes, the expression levels of which are known to be coupled to cell growth and survival. Polysome profiling and reporter analyses suggest that Lin28 stimulates the translation of many or most of these targets. Moreover, Lin28-responsive elements were found within the coding regions of all target genes tested. Finally, a mutant Lin28 that still binds RNA but fails to interact with RNA helicase A (RHA), acts as a dominant-negative inhibitor of Lin28-dependent stimulation of translation. We suggest that Lin28, working in concert with RHA, enhances the translation of genes important for the growth and survival of human embryonic stem cells. STEM CELLS 2011;496–504
Highly expressed in human embryonic stem cells (hESCs), Lin28 facilitates the reprogramming of fibroblasts to induced pluripotent stem cells (iPSCs) by increasing the number of reprogrammed clones . This is consistent with a role for Lin28 in cell growth and survival. Also consistent is the recent report that Lin28 knockout mice were severely underdeveloped and nonviable . These effects likely result from several distinct molecular functions. Lin28 inhibits the biogenesis of a group of microRNAs, among which are the let-7 family microRNAs shown to participate in the regulation of expression of genes involved in cell growth and differentiation [3, 4]. This protein binds to the loop regions of microRNA precursors and blocks their processing into mature microRNAs [5–7]. In addition, Lin28 induces uridylation of the precursors and promotes their degradation [8–10]. On the other hand, Lin28 alters cell fates during neurogliogenesis via mechanisms distinct from those mediated by let-7 and causes changes in gene expression before any effect on let-7 could be detected . Likewise, a mutant Lin28 that permits let-7 production could still completely inhibit gliogenesis . Moreover, Zhu et al.  have recently demonstrated that transgenic mice that overexpress Lin28 exhibit overgrowth and delayed onset of puberty. However, no decrease in the level of let-7 was observed in the hypothalamic-pituitary-gonadal axis that plays a critical role in controlling development and reproduction. Therefore, mechanisms other than let-7-mediated pathways must also play important roles in Lin28-dependent gene regulation. During muscle cell differentiation, Lin28 binds to insulin-like growth factor (IGF)-2 mRNA and stimulates its translation . It also selectively binds to mRNAs of the key pluripotency factor Oct4 and a subset of cell cycle-related factors and promotes their expression at the post-transcriptional level [13–15]. Lin28-responsive elements (LREs) have been mapped to the 5′-, 3′-untranslated regions, or open reading frames (ORFs) of mRNA targets [12–15]. Recently, post-transcriptional regulation mediated by Lin28 has been shown to require a functional interaction with RNA helicase A (RHA) .
MATERIALS AND METHODS
Antibodies, siRNAs, and Plasmids
The antibodies specific for high mobility group AT-hook1 (HMGA1) (Santa Cruz, sc-8982; Santa Cruz, CA, http://www.scbt.com/index.html), CD63 (Santa Cruz, sc-15363), ribosomal protein S13 (RPS13) (Protein Tech Group Inc., 16680-1-AP; Chicago, IL, http://www.ptglab.com/), eukaryotic translation elongation factor 1 gamma (EEF1G; Abcam, ab72368; Cambridge, MA, http://www.abcam.com/), Lin28 (Abcam, ab46020), Oct4 (Santa Cruz, sc-5279), RHA (Abcam, ab54593), β-tubulin (Abcam, ab6046), β-actin (Abcam, ab8226), Flag (Santa Cruz, sc-807; Stratagene, 200472; Santa Clara, CA, www.stratagene.com), and rabbit pre-immune serum (SouthernBiotech, 0040-01; Birmingham, AL, www.southernbiotech.com) were purchased. The siLin28 (Dharmacon, ON-TARGETplus SMARTpool, L-018411-01; Lafayette, Colorado, www.dharmacon.com), siLin28-2 (an equal molar mixture of two siRNAs, J-018411-09 and J-018411-11), siCon (Dharmacon, D-001810-10-05), and the plasmid-expressing Flag-Lin28 were previously described . Flag-Lin28ΔC was created by cloning a PCR fragment containing part of the human Lin28 coding region (aa 1–176, relative to the translational start site) into pFLAG-CMV-2 (Sigma, E7398) at the NotI and BamHI sites. The luciferase reporter constructs Oct4-R2, Oct4-R4 , and H2a  were previously documented. Constructs Oct4-95 (Gene ID: NM_002701), HMGA1-ORF (Gene ID: NM_002131), RPS13-ORF (Gene ID: NM_001017), EEF1G-R3 (Gene ID: NM_001404), and Oct4-70 (Gene ID: NM_002701) were made by inserting PCR fragments containing nucleotides 516–610, 323–613, 33–488, 811–1,140, and 541–610 (relative to the transcriptional start sites) of the corresponding genes, respectively, at the NotI and XhoI sites of the firefly reporter vector .
Cell Culture and Transfection
The culture and transfection of the hESC line H1 (WA01, WiCell), embryonal carcinoma (EC) line PA-1, and HEK293 cells were carried out as preciously described .
Ribonucleoprotein Particle Immunoprecipitation and RT-qPCR
Ribonucleoprotein particle (RNP) immunoprecipitation (IP) experiments were carried out essentially as described . To prepare samples for deep sequencing, IP was scaled up 10-fold. The real-time PCR primers are listed below. β-actin forward: 5′-ATCAAGATCATTGCTCCTCCTGAG; β-actin reverse: 5′-CTG CTTGCTGATCCACATCTG; β-tubulin forward: 5′-CGTGTTCG GCCAGAGTGGTGC; β-tubulin reverse: 5′-GGGTGAGGGCAT GACGCTGAA; Lin28 forward: 5′-CGGGCATCTGTAAGTGGT TC; Lin28 reverse: 5′-CAGACCCTTGGCTGACTTCT; Oct4 forward: 5′-GTGGAGGAAGCTGACAACAA; Oct4 reverse: 5′- GC CGGTTACAGAACCACACT; firefly luciferase forward: 5′-GCT GGGCGTTAATCAGAGAG; firefly luciferase reverse: 5′-GTG TTCGTCTTCGTCCCAGT; Renilla forward: 5′-GCAAATCAGG CAAATCTGGT; Renilla reverse: 5′-GGCCGACAAAAATGATC TTC; HMGA1 forward: 5′-CAGCGAAGTGCCAACACCTAAG; HMGA1 reverse: 5′-CCTTGGTTTCCTTCCTGGAGTT; RPS13 forward: 5′-CTCTCCTTTCGTTGCCTGAT; RPS13 reverse: 5′CCCTTCTTGGCCAGTTTGTA; eukaryotic translation initiation factor 4A (EIF4A) forward: 5′-TGCTTAACCGGAGATAC CTGTC; EIF4A reverse: 5′-GTCCCTCATGAACTTCTTGGTC; CD63 forward: 5′-CCCGAAAAACAACCACACTGC; CD63 reverse: 5′-GATGAGGAGGCTGAGGAGACC; EEF1G forward: 5′- AGCGGAAGGAGGAGAAAAAG; EEF1G reverse: 5′-GACC AGCCGTCCTTATCAAA.
Deep Sequencing Analysis
Lin28 and preimmune IP RNA samples from H1 cells were used for deep sequencing analysis, and the sequencing libraries were prepared according to the manufacturer's instructions (Illumina, P/N 1004814; San Diego, CA, http://www.illumina.com/applications.ilmn). Briefly, RNAs extracted from IP samples were treated by two successive rounds of oligo-dT selection. The poly(A)+ RNAs were fragmented using divalent cations under elevated temperature, followed by first and second strand cDNA synthesis with random hexamer priming. The cDNA fragments were cleaned up, end-repaired, and phosphorylated at their 5′ ends. After a nontemplated 3′ end addition of A residues, Illumina adapters were ligated to both ends, and ∼300-bp fragments were isolated and amplified by PCR using Illumina adapters. The libraries derived from Lin28 IP and preimmune IP samples were individually used for sequencing on an Illumina GAII platform using a single-read protocol. Approximately 10 million reads were obtained from each IP sample, and these sequences were aligned to the human genome using Bowtie . For both preimmune and Lin28 IP libraries, ∼4.7 to ∼6.0 million reads were uniquely aligned. The sequencing reads were uniquely aligned to the human hg18 genome and splice junction index using Bowtie  that allows up to two mismatches. Wiggle track files were generated from Bowtie output files by a custom bowtie2wiggle script and loaded onto the UCSC genome browser (2006; http://www.genome.ucsc.edu) for visualization. Gene expression levels were determined by calculating quantitative RPKM scores (Reads Per Kilobase of gene model per Million mapped reads) as described . The raw data can be accessed at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=nrktvaqeewuiyrs&acc=GSE23109. mRNAs that were enriched by at least 2.5-fold in the Lin28 IP compared with preimmune were selected as significant Lin28 targets.
Gene Ontology Analysis
Gene ontology (GO) terms of Lin28 IP mRNA targets were identified using the Funcassociate 2.0 software , where these mRNAs were used as the “query” set and all human genes as the “gene space” set.
Sucrose Gradient Polysome Fractionation
These were carried out essentially as described previously . Briefly, PA-1 cells (3 × 107) were harvested, washed with PBS, and resuspended in 0.5 ml of freshly prepared extraction buffer (100 mM KCl, 0.1% TritonX-100, 50 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES) , pH 7.4, 2 mM MgCl2, 10% glycerol, 1 mM dithiothreitol (DTT), 20 U/ml Protector RNase [Roche], 1× complete mini EDTA-free protease inhibitor cocktail [Roche]). After incubation on ice for 10 minutes, the lysate was centrifuged at 1,300g at 4°C for 10 minutes to remove insoluble materials. The supernatant was applied onto the top of a 15%–55% (wt./wt.) linear sucrose gradient and centrifuged at 150,000 g for 3 hours in a Beckman ultracentrifuge. Fractions (0.2 ml each) were collected and used for RNA extraction or protein analysis. In the case of polysome IP, pooled polysome fractions in a total of ∼4 ml were divided into two tubes and incubated with protein A sepharose beads prebound with either anti-Lin28 antibody or preimmune IgG at 4°C overnight. Bound RNAs were extracted and used in reverse transcription and quantitative polymerase chain reaction (RT-qPCR) analysis.
These were carried out basically as previously described . Briefly, the indicated firefly luciferase reporter plasmids were each transfected into HEK293 cells, with or without cotransfection of Flag-Lin28 or Flag-Lin28ΔC. The Renilla reporter was included in all transfections for normalization purposes. Transfection was carried out in a 48-well plate scale. The amount of total plasmid DNA per well was 400 ng that included 100 ng of firefly luciferase reporter, 2 ng of Renilla, and the indicated amounts of Flag-Lin28 or Flag-Lin28ΔC.
To examine the interaction between Flag-Lin28 (or Flag-Lin28ΔC) with RHA, 8 × 106 HEK293 cells were transfected with 6 μg of Flag-Lin28, Flag-Lin28ΔC, or empty vector in a 6-cm plate scale. Cells were collected 48 hours later by manual scraping using a rubber policeman and pelleted by centrifugation. Cell pellet was washed once with PBS and resuspended in 400 μl of gentle lysis buffer (10 mM Tris-HCl at pH 7.5, 10 mM NaCl, 10 mM EDTA, 0.5% TritonX-100, 1 mM phenylmethylsulfonyl fluoride (PMSF) , 1× protease inhibitor cocktail [Calbiochem], 1 mM DTT, and 10 μg/ml of RNase A [Roche]) and incubated on ice for 15 minutes. Insoluble materials were removed by centrifugation at 13,400 g in a microcentrifuge at 4°C for 15 minutes. NaCl was added to the cleared lysate to a final concentration of 250 mM, and 350 μl of the lysate incubated with 10 μl of protein A sepharose beads prebound with 10 μg of monoclonal anti-Flag M2 antibody at 4°C overnight. The next day, beads were washed and bound fractions eluted with 3× SDS-sample buffer by heating at 95°C for 5 minutes. Proteins were resolved by SDS-polyacrylamide gel electrophoresis (PAGE), followed by Western blot analysis.
Lin28 Knockdown Affects hESC Growth
Both human and mouse ESCs proliferate rapidly and have a unique cell cycle thought to be biologically coupled to pluripotency [19, 20]. This, coupled with the additional evidence discussed above suggesting that Lin28 is involved in stem cell proliferation, led us to ask whether Lin28 might play a direct role in the growth and survival of hESCs. Therefore, we inhibited Lin28 expression using a Lin28-specific siRNA (siLin28) [15, 21]. siLin28 reduced Lin28 expression to 8% and 12% of the control at the RNA (Fig. 1A) and protein (Fig. 1B) level, respectively. Importantly, we observed a concomitant decrease in the number of viable cells (Fig. 1C, left panel) and an increase in apoptosis, which was indicated by an elevated level of caspase 3/7 activity (Fig. 1C, right panel). To rule out possible nonspecific (i.e., off-target) effects of siLin28, we also used another siRNA (siLin28-2; ) targeted to a different region of Lin28 mRNA and obtained similar results (Supporting Information Fig. S1). Taken together with our previous findings in mouse ESCs that reducing Lin28 expression slows cell growth and overexpressing Lin28 accelarates cell growth , our results support the conclusion that Lin28 is important for the growth and survival of hESCs. We cannot conclude, however, that Lin28 is absolutely essential for hESC viability in vivo. Under our cell culture conditions, we see significant cell death; however, Lin28 knockout mice, though nonviable and weighing less than 20% of wild-type mice at birth , suggest that Lin28 deficiency severely compromises cell growth but is not obligatory under all conditions.
Genome-Wide Identification of Lin28 mRNA Targets
How might Lin28 exert its biological effects? Most likely, both let-7-dependent and let-7-independent pathways are involved. To investigate the contribution of mRNA targets that might be regulated by Lin28, we developed a genome-wide approach. Thus, we isolated Lin28-containing RNPs from hESCs by IP, followed by identification of associated mRNAs using cDNA synthesis and high throughput deep sequencing with the Illumina platform. The detailed procedures are outlined in “Materials and Methods” and the full list of mRNAs enriched by Lin28 IP is presented in Supporting Information Table S1. Strikingly, we found Lin28 to be highly selective in recognition of mRNAs. Only a small subset (1,259 genes/4.8% of cellular mRNAs) were enriched more than 2.5-fold in the Lin28 IP as compared with the control preimmune IP. We selected the top (lowest p values) 268 genes with at least 2.5-fold enrichment in Lin28 IP versus preimmune IP (Supporting Information Table S2) and carried out GO analysis to classify those mRNA targets into different groups. As shown in Table 1 and Supporting Information Table S3, the top mRNA cluster selected by Lin28 represents genes encoding RNP proteins (including several essential splicing factors), followed by genes participating in translation (including ribosomal proteins and key translation initiation and elongation factors) and genes involved in cellular metabolism. In contrast, many genes are strikingly under-represented in the Lin28 IP samples. Genes relating to membrane receptor activity, DNA-binding and transcription (Oct4 is an exception) are rarely associated preferentially with Lin28. For example, none of the total of 1,271 genes in the human genome that encode G-protein-coupled receptors is among the top 268 selected genes, whereas many ribosomal protein mRNAs are apparent Lin28 targets. We note that most of the genes enriched in the Lin28 IP are consistent with a role of Lin28 in regulating cellular growth and metabolism. Supporting Information Figures S2–S10 show examples of the data obtained.
Table 1. GO analysis
Abbreviations: GO, gene ontology; N, number of genes in the query with this attribute; P, single hypothesis one-sided p value of the association between attribute and query adjusted by fraction of 1,000 null-hypothesis simulations having attributes with this single hypothesis p value or smaller; RNP, ribonucleoprotein particle; X, number of genes overall with this attribute.
Our IP and sequencing studies did not involve protein-RNA cross-linking prior to isolation and deep sequencing. We omitted this step for technical reasons as it severely reduced RNA yields, making deep sequencing unreliable. However, several lines of evidence suggest that our target list largely reflects bona fide Lin28 targets. First, in every case, so far, where we have chosen a target from the list for further validation, this has been successful (see below and data not shown). Second, many of the most enriched targets fall into only a few functional categories, as evidenced by our GO analysis (Table 1). Third, Lin28 is clearly not associating preferentially only with the most abundant mRNAs in the extracts. Many highly abundant mRNAs are not enriched at all, whereas a number of moderately abundant and relatively low abundance mRNAs are highly selected.
We selected six genes (representing several different functional categories from the GO analysis) for IP/RT-qPCR validation: CD63 (a transmembrane protein), EEF1G, EIF4A, HMGA1 (a chromosome binding protein), Oct4 (known Lin28 target as a positive control), and RPS13 (a ribosomal protein). β-actin mRNA was a negative control for Lin28 binding. Each of the selected mRNAs was enriched by more than 2.5-fold in the Lin28 IP versus preimmune IP samples from hESCs (Fig. 2A). Similar results were obtained using RNPs isolated from human EC PA-1 cells (Fig. 2B). The ESCs and EC cells share many properties including Lin28 expression, cell surface antigen expression, proliferation characteristics, the ability to self-renew and differentiate, and the expression of core transcription factors that control their undifferentiated state . Taken together with the GO analysis, we suggest that many or most of the candidate genes selected by our analyses are likely to be in vivo targets of Lin28.
Downregulation of Lin28 Leads to Decreased Levels of Proteins Expressed from Target Genes
If binding to target mRNAs reflects a mode of gene regulation by Lin28 then we would hypothesize that lower Lin28 expression would alter the expression of proteins encoded by these targets. To determine whether Lin28 influences the expression of its target genes, we performed siRNA knockdown experiments. When the level of Lin28 protein in siLin28-transfected cells was reduced to 15% of that seen in siCon transfected cells (Fig. 3A, top panel, compare lane 2 with lane 1), a concomitant decrease in the Oct4 protein level (52% of siCon-transfected cells, second from the top blot, compare lane 2 with lane 1) was also observed. Although this likely results directly from Lin28 interaction with the Oct4 mRNA, it remained possible that the observed effect was indirect, due to changes in hESC growth or pluripotency, or let-7 expression levels resulting from Lin28 knockdown. However, downregulation of Lin28 also led to decreased protein levels of the other selected target genes, whereas the level of β-actin protein was not affected (Fig. 3A, compare lanes 2 with lanes 1 of the indicated genes). Similar results were obtained when PA-1 cells were used (Fig. 3B). Given that the protein level changes were larger between siLin28 and siCon-transfected cells compared with their respective mRNA level changes, we conclude that the differences between the mRNA and protein level changes observed most likely result from impaired translation due to reduced Lin28 levels.
Lin28 Inhibition Induces Shifts of Target mRNAs from Polysomal to Nonpolysomal Fractions
If Lin28 stimulates the translation of target mRNAs, we would expect an enrichment of these mRNAs in polysomes that contain Lin28, whereas nontarget mRNAs such as β-actin would not be enriched. Oct4 mRNA was enriched in Lin28-containing polysomes by greater than threefold, whereas β-actin mRNA was not enriched at all (Supporting Information Fig. S11A and ). Similarly, EEF1G, HMGA1, and RPS13 mRNAs were also enriched at least twofold in Lin28-containing polysomes (Supporting Information Fig. S11A). Importantly, the fold enrichments observed did not reflect the steady-state levels of the respective mRNAs in the polysomes (Supporting Information Fig. S11B), indicating that association of these mRNAs with Lin28 in polysomes is specific and that Lin28 likely plays a role in modulating the translation of these mRNAs.
As in most cases an increased polysome association of an mRNA indicates an increase in translation efficiency, we next asked whether downregulation of Lin28 would shift target mRNAs from polysomes to nonpolysome fractions. Thus, PA-1 cells were transfected with siLin28 or siCon, followed by sucrose gradient fractionation of cytoplasmic extracts collected 48 hours after transfection. Total RNAs were isolated from polysome or nonpolysome fractions (which included RNP, 40S, 60S, and 80S fractions), and polysome distributions of indicated mRNAs analyzed. We observed significant decreases in polysome association of the putative target mRNAs in siLin28-transfected cells versus siCon-transfected cells. The decreases were 65%, 40%, 85%, and 56% with EEF1G, HMGA1, Oct4, and RPS13 mRNAs, respectively, whereas polysome association of β-actin mRNA decreased by only 13% (Fig. 3C, top panel). Given that the steady-state mRNA levels were essentially unchanged (Fig. 3C, bottom panel), we conclude that the decreased polysome association of the target mRNAs was most likely due to reduced translation.
Target Genes Contain LREs in Their Coding Regions
We have previously mapped a 369-nt long LRE within the coding region of Oct4 mRNA (called Oct4-R2) that allows for Lin28-dependent stimulation of translation in a reporter system . To determine whether EEF1G, HMGA1, and RPS13 mRNAs also contain LREs, we initiated mapping using a luciferase reporter . As an additional positive control for luciferase stimulation, we included a 393-nt-long fragment derived from the ORF of the mouse histone H2a gene shown to stimulate the translation in a Lin28-dependent fashion . As a negative control, we used Oct4-R4, a fragment derived from the Oct4 3′UTR . We identified LREs in all three genes, all of which mapped to the coding regions (Fig. 4A). We next assessed the activity of shorter derivatives of the Oct4-R2 element. We obtained a 95-nt-long sequence that retains the full activity of the 369-nt-long R2 fragment (Fig. 4A, 4B). However, further deletion of either end of the fragment completely abolished activity (Fig. 4B). In similar experiments using elements derived from HMGA1, RPS13, and EEF1G, we have likewise been unable to identify any LRE shorter than 95-nt, consistent with an idea that Lin28 recognition may involve RNA structural features rather than simple sequence motifs.
RHA Participates in Lin28-Dependent Stimulation of Translation
Lin28 interacts specifically with RHA and downregulation of RHA expression impedes Lin28-dependent stimulation of translation in a reporter system . The Lin28 protein contains two types of RNA-binding motifs: a cold shock domain (CSD) and a pair of retroviral-type cys-cys-his-cys (CCHC) zinc fingers (Fig. 5D) [24, 25]. Inactivation by point mutations of either CSD or CCHC domain led to the loss of the ability of Lin28 to associate with mRNA . These same mutations do not affect its interaction with RHA (data not shown). However, a 35-aa deletion at the carboxyl terminus of Lin28 dramatically diminishes its ability to interact with RHA (Fig. 5A, 5D) but not RNA (Supporting Information Fig. S12C). As shown in Figure 5A, while approximately 4% of endogenous RHA was coimmunoprecipitated with the wild-type Flag-Lin28, only ∼0.3% of RHA was precipitated with the mutant Lin28 (top panel, compare lane 2 with lane 1). Importantly, the mutant Lin28 not only reduced its ability to stimulate translation (Supporting Information Fig. S12A, S12B), but also exhibits an inhibitory effect in the presence of wild-type Lin28 (Fig. 5B, top panel), whereas the expression levels of RHA and Flag-Lin28 were not altered as a result of the mutant expression (Fig. 5B, bottom panel). When the mutant Lin28 was expressed in PA-1 cells, we expectedly observed an inhibition of translation of the endogenous target mRNAs as judged by the polysome shift analysis. In the presence of Flag-Lin28ΔC expression, the association of EEF1G, HMGA1, Oct4, and RPS13 mRNAs with polysomes decreased by 81%, 42%, 40%, and 43%, respectively, compared with those in empty vector transfected cells (Fig. 5C and Supporting Information Fig. S13A). Importantly, Flag-Lin28ΔC expression also leads to decreased cell viability (Supporting Information Fig. S13B), consistent with decreased translation of mRNAs important for cell growth and survival. Taken together, these results thus suggest that Lin28-mediated translational stimulation occurs through the concerted interaction and activities of two RNA-binding proteins.
A major biological function of Lin28 is to support the rapid growth of ESCs. This is supported by the following evidences: (a) during reprogramming, Lin28 increases the number of human iPSC colonies, consistent with its ability to promote the proliferation and survival of reprogrammed cells ; (b) in mouse ESCs inhibition of Lin28 slows cell proliferation, whereas overexpression of Lin28 accelerates cell proliferation ; (c) in transgenic mice, overexpression of Lin28 leads to increased cell proliferation ; and (d) downregulation of Lin28 in hESCs or expression of a mutant Lin28 that is incapable of RHA interaction in PA-1 cells results in decreased numbers of viable cells, likely a combined result of decreased cell proliferation and survival (this report).
We have used IP and deep sequencing to identify Lin28 target mRNAs and show that Lin28 selectively enriches mRNAs from several distinct classes. The high selectivity of Lin28 for ribosomal protein mRNAs (Table 1 and Supporting Information Table S2) is striking and fits well with a role for this protein in growth stimulation, as rapid ESC growth requires coordinated production of ribosomes. Ribosome biogenesis is connected to cell cycle control , and its perturbation can lead to cell cycle arrest and/or apoptosis [28, 29]. Many but not all ribosomal protein mRNAs are selected by Lin28 (Table 1 and Supporting Information Tables S1–S3). This may reflect the fact that ribosomal mRNA expression is regulated by both common and distinct mechanisms, depending on the cell type and environmental conditions . Lin28 also selects mRNAs that encode proteins involved in metabolism (Table 1 and Supporting Information Table S2). Not coincidentally, Zhu et al.  observed that transgenic mice overexpressing Lin28 not only grew bigger but also manifested increased glucose metabolism. A direct stimulation of translation of the related mRNA targets by Lin28 could well be the basis for many of the effects observed, especially in tissues and cells where let-7 expression was not affected . It is important to note that a fraction of the mRNA targets we have identified may also be regulated by let-7. Experiments using mutant forms of Lin28 and/or let-7 microRNAs that allow the uncoupling of the two activities (i.e., translational stimulation vs. inhibition of let-7 processing) are underway to address this issue.
Lin28 most likely exerts its biological effects by binding to its RNA targets. However, how this protein recognizes RNA is still unclear. Although Lin28 has been reported to affect let-7 processing at a variety of steps, the mechanism(s) by which Lin28 does this are still somewhat controversial and key RNA-protein interactions likely involve not only specific sequence motifs but also structural elements [3, 5, 6]. It is also clear that the reported Lin28 binding elements show only relatively low or modest affinities for the protein in vitro [3, 5]. We show here that most Lin28 mRNA targets so far examined (including Oct4, RPS13, HMGA1, EEF1G, and histone H2a) contain LREs in their coding regions (Fig. 4; ). Coding regions are the sequences expected to be most highly conserved between human and mouse. In addition, our sequence mapping studies suggest that Lin28 is not recognizing short sequence motifs (Fig. 4). Thus, although we initially hoped to find that Lin28 binds to short sequence motifs like many other RNA binding proteins (such as SR proteins  and Nova ), results have revealed that this is not likely to be the case. Our current hypothesis is that Lin28 recognizes its targets either (a) along with one or more additional cellular factors and/or (b) via binding to larger structured regions of RNA. This second model is consistent with the mechanism by which Lin28 affects let-7 processing [3, 5–7, 33]. Also, there are a number of RNA-binding proteins that recognize only longer RNA sequences/structures. These include the human immunodeficiency virus (HIV-1) Rev protein , the essential mRNA export factor NXF1 , and RHA [36, 37].
Finally, we note from our GO analysis that there are several additional and possibly quite important classes of Lin28 targets. A number of chromosomal proteins (such as HMGA1 and chromatin modifying protein 2A (CHMP2A); see Supporting Information Table S2) may be regulated by Lin28 and these may participate in the establishment or maintenance of the unique chromatin structure found in pluripotent cells. Importantly, knockdown of Lin28 leads to lower levels of HMGA1 protein in both hESCs and in PA-1 cells (Fig. 3). Also, another enriched class of Lin28 targets include mRNAs encoding RNA-binding proteins, such as a number of proteins involved in pre-mRNA splicing or RNA metabolism (such as poly(U)-binding-splicing factor 60 (PUF60), mago nashi homolog (MAGOH), ras-related nuclear protein (RAN) and the arginine methyltransferases protein arginine methyltransferase 1 (PRMT1) and PRMT2; see Supporting Information Table S2). It will be interesting to investigate alterations in alternative splicing patterns, mRNA trafficking and turnover in response to either upregulation or downregulation of Lin28 expression.
Our genome-wide studies show that Lin28 associates preferentially with a relatively small subset of cellular messages. These targets primarily fall into several functional classes, most of which represent genes that are important for cell growth and survival. We also find that Lin28 recognizes specific sequence elements within its target mRNAs and, in concert with RNA helicase A, enhances the translation of these mRNAs into proteins. These findings open new avenues of future work on the role of Lin28 in pluripotency and stem cell function.
We thank Caihong Qiu and Yinghong Ma for culturing human embryonic stem cells. This work was supported by the Connecticut Stem Cell Research grants 09SCAYALE14 to Y.H., 06SCB08 to G.G.C., 09SCAUCHC16 to L.-L.C., and 06SCD01 to H.L.
DISCLOSURE OF POTENTIAL CONFLICTS OF INTEREST
The authors indicate no potential conflicts of interest. The contents of this material are solely the responsibility of the authors and do not necessarily represent the official views of the State of Connecticut, the Department of Public Health of the State of Connecticut, or Connecticut Innovations, Incorporated.