**Mechanisms of Lin28-Mediated miRNA and mRNA Regulation—A Structural and Functional Perspective**

**Florian Mayr 1,2 and Udo Heinemann 1,2,\***


*Received: 31 May 2013; in revised form: 22 July 2013 / Accepted: 25 July 2013 / Published: 9 August 2013*

**Abstract:** Lin28 is an essential RNA-binding protein that is ubiquitously expressed in embryonic stem cells. Its physiological function has been linked to the regulation of differentiation, development, and oncogenesis as well as glucose metabolism. Lin28 mediates these pleiotropic functions by inhibiting *let-7* miRNA biogenesis and by modulating the translation of target mRNAs. Both activities strongly depend on Lin28's RNA-binding domains (RBDs), an N-terminal cold-shock domain (CSD) and a C-terminal Zn-knuckle domain (ZKD). Recent biochemical and structural studies revealed the mechanisms of how Lin28 controls *let-7* biogenesis. Lin28 binds to the terminal loop of pri- and pre-*let-7* miRNA and represses their processing by Drosha and Dicer. Several biochemical and structural studies showed that the specificity of this interaction is mainly mediated by the ZKD with a conserved GGAGA or GGAGA-like motif. Further RNA crosslinking and immunoprecipitation coupled to high-throughput sequencing (CLIP-seq) studies confirmed this binding motif and uncovered a large number of new mRNA binding sites. Here we review exciting recent progress in our understanding of how Lin28 binds structurally diverse RNAs and fulfills its pleiotropic functions.

**Keywords:** Lin28; *let-7* miRNA; miRNA processing; RNA-binding protein; cold-shock domain; zinc-knuckle domain; TUTase; oncogene; stem cell

#### **1. Introduction**

Lin28 (cell lineage abnormal 28) is a conserved RNA-binding protein in higher eukaryotes that regulates several important cellular functions associated with development, glucose metabolism, differentiation and pluripotency. It was first described as a heterochronic gene in *Caenorhabditis elegans (C. elegans)*, since mutations within *lin-28* disturbed the developmental timing of the worm and accelerated differentiation of hypodermal seam cells and vulva stem cells [1,2]. Subsequent experiments revealed that Lin28 is expressed early in nematode embryonic and larval development, but its expression is down-regulated by *lin-4* and *let-7* miRNA as differentiation proceeds [2,3].

A similar expression pattern and physiological function was also shown for *Drosophila*, *Xenopus* and mammalian Lin28 [4]. The human paralogs Lin28a (routinely termed simply Lin28) and Lin28b encode for basic 23- or 28-kDa proteins that are highly expressed in embryonic stem cells (ESC) but are down-regulated upon differentiation of ESCs into embryoid bodies [5]. Reciprocally, Yu and colleagues used Lin28a, in conjunction with Oct4, Sox2 and Nanog, to reprogram adult human fibroblasts to induced pluripotent stem cells (iPSCs) [6]. A knockdown of Lin28a expression in mouse ESCs led to loss of Oct4 and Nanog expression, indicating an impaired self-renewal potential [7]. Increased Lin28a/Lin28b expression was reported in various hepatocellular and other carcinomas and was associated with poor patient prognosis [8–13]. Recently, Lin28a was linked to the regulation of developmental and metabolic processes. After ectopic overexpression of Lin28a mice developed a bigger size and delayed sexual maturation, whereas Lin28 knockout mice were smaller and died shortly after birth [14]. In addition, Lin28a overexpression was associated with increased insulin sensitivity and glucose metabolism, while a depletion of Lin28a resulted in insulin resistance and glucose intolerance [15].

On the molecular level, Lin28a and Lin28b act as both negative regulator of *let-7* miRNA biogenesis and post-transcriptional regulator of mRNA translation. Both activities strongly depend on Lin28's two RNA-binding domains (RBDs): an N-terminal cold-shock domain (CSD) and a C-terminal Zn-knuckle domain (ZKD) composed of two tandemly arranged retroviral-type CCHC Zn knuckles. The individual domain combination of both RBDs is unique in animals with the RBDs being highly conserved. The human Lin28 paralogs share an overall sequence identity of 65% (FFAS, [16]) and contain low-complexity regions at the N-terminus, a putative bipartite nucleolar localization sequence (NoLS) as well as a C-terminal nuclear localization signal (NLS) in the case of Lin28b [17] (Figure 1). Lin28a and Lin28b can localize to both cytosol and nucleus [2,4,17–21] and interact with primary (pri-) or precursor (pre-) *let-7* miRNAs thereby preventing their maturation [20,22,23]. In addition, binding of Lin28a to messenger ribonucleoprotein complexes containing translation initiation (eIF3B, eI4E) and elongation factors (EF1α, EF1α2), poly(A) binding proteins, Igf2bps and RNA helicase A was reported in various studies [18,20,24,25]. Under stress conditions, Lin28a was shown to localize to cytoplasmic stress granules and P-bodies where mRNA translation is temporally stalled [18]. Since a mutation in Lin28a's ZKD caused Lin28a to accumulate in the nucleus, it was suggested that Lin28a exits the nucleus in a complex with bound RNA and thus regulates the post-transcriptional processing of its target RNAs [18].

**Figure 1.** Domain organization of Lin28. (**A**) *Caenorhabditis elegans* (*Cel*) and human (h) Lin28a/Lin28b contain two RNA-binding domains (RBDs): an N-terminal cold-shock domain (CSD) and a C-terminal Zn-knuckle domain (ZKD) comprised of two retroviral type CCHC Zn knuckles (ZnK). Additionally, Lin28 harbors low-complexity sequences, Lys/Arg (K/R)-rich stretches, bipartite nuclear localization signals (NLS) or putative nucleolar localization sequences (NoLS); (**B**) Sequence alignment of hLin28a and hLin28B. Amino acids belonging to CSD or ZKD are shaded in blue or green, respectively.

#### **2. Lin28 Blocks** *let-7* **Processing**

The opposing expression pattern of Lin28 and *let-7* miRNA became initially apparent when studying *C. elegans* larval development [2–4,26,27]. At an early stage in larval development both Lin28 and pri-*let-7* are present, however, no levels of either pre-*let-7* or mature *let-7* can be detected, indicating a regulation at a post-transcriptional level [28]. As larval development proceeds, a heterogenic cascade involving *lin-4* miRNA and the *let-7* sisters *mir-48/84/241* lead to a relief of pri-*let-7* processing inhibition and to a subsequent down-regulation of Lin28 expression (reviewed in [29]). This inverse relationship between Lin28 and *let-7* miRNA is also present in mammalian cells, where Lin28a/b are mainly expressed in undifferentiated cells, and mature *let-7* is only detectable upon differentiation or tissue development [5,20,23]. Furthermore, levels of pri-*let-7* remain constant throughout the entire differentiation or development process suggesting a negative regulation of *let-7* biogenesis by Lin28a/b in stem or progenitor cells [30–32]. Purification of pre-*let-7* bound complexes and subsequent analysis via mass spectrometry revealed that both human Lin28 paralogs specifically associate with pri- or pre-*let-7 in vivo* [20,22,23]. Moreover, *in vitro* purified Lin28a could inhibit pri- and pre-*let-7* processing by Drosha and Dicer by binding to the double-stranded stem close to the Dicer cleavage site and the pre-element (preE, terminal loop or hairpin) [33]. Mutations within Lin28's CSD and ZKD impaired pre-*let-7* binding and inhibition of Dicer processing, suggesting a competitive relationship between Lin28 and Dicer [33–35]. Moreover, recent studies provided evidence that Lin28a/b can induce a structural change within pre-*let-7*'s preE, thereby leading to an opening of the double-stranded stem including the Dicer cleavage site [34–36].

Heo and colleagues revealed an additional inhibition mode of *let-7* miRNA processing, which irreversibly targets pre-*let-7* to a decay pathway [7,37]. They demonstrated that Lin28a/b induce oligo-uridylation of pre-*let-7*'s 3' overhang. Oligo-uridylated pre-*let-7* is resistant to Dicer cleavage given that Dicer normally recognizes a 2-nt 3' overhang in miRNAs via its PAZ domain. Thus, Dicer is unable to recognize the elongated 3' overhang and to process pre-*let-7*. Furthermore, it was reported that oligo-uridylated RNAs recruit 3'–5' exonucleases and are targeted for decay [38,39]. Indeed, oligo-uridylated pre-*let-7* was more rapidly degraded than unmodified pre-*let-7* [37]. Recently, Chang and colleagues identified the 3'–5' exonuclease Dis3l2 that catalyzes the decay of oligo-uridylated pre-*let-7* in mouse ESCs [40]. Consistent with this, a knockdown of Dis3l2 in mouse ESCs caused an accumulation of uridylated pre-*let-7*. Oligo-uridylation of pre-*let-7* is catalyzed by the non-canonical poly(A) polymerase TUT4 (terminal uridyl transferase 4/Zcchc11) and to a minor extent by TUT7 (Zcchc6) in a Lin28-dependent manner [7,41,42]. Interestingly, these enzymes catalyze mono-uridylation of pre-miRNAs with a 1-nt 3' overhang (like most pre-*let-7* family members) in the absence of Lin28, thereby enhancing Dicer-mediated processing [43]. However, in the presence of Lin28, pre-*let-7* and other miRNAs containing a GGAG motif within their preE were subjected to oligo-uridylation. Upon mutation of this motif, both Lin28 binding and oligo-uridylation were impaired, indicating that the GGAG motif is essential for these processes [7].

In *C. elegans* a similar mechanism for inhibiting pre-*let-7* processing has been reported [44]. The poly(U) polymerase PUP-2 was shown to oligo-uridylate pre-*let-7* in a Lin28-dependent fashion, thereby suppressing premature expression of mature *let-7* during larval development. In addition, subsequent RNA and chromatin immunoprecipitation assays revealed a specific interaction between Lin28 and pri-*let-7* that co-transcriptionally inhibits pri-*let-7* processing by Drosha [28]. An interaction between Lin28 and endogenous pri-*let-7* was also described for human ESCs and neuronal stem/progenitor cells [28]. Here, a highly expressed RNA-binding protein called Musashi1 (Msi1) selectively recruits Lin28a to the nucleus and synergistically blocks the cropping step of pri-*let-7* [45]. Moreover, it was suggested that Lin28b predominantly localizes to the nucleolus where it sequesters pri-*let-7*, thereby preventing Drosha processing in the nucleus [17]. Thus, Lin28a/b seem to obviate precocious expression of mature *let-7* during early development and differentiation by interfering with both the Drosha and Dicer complexes and by targeting pre-*let-7* towards degradation. Conversely,

upon differentiation of stem or progenitor cells, *let-7* ensures constant down-regulation of Lin28 by binding to the 3' UTR of Lin28 and its promoting transcription factor c-Myc [20] (Figure 2).

**Figure 2.** Lin28/*let-7* regulatory axis. In undifferentiated cells, Lin28 is highly expressed and blocks the biogenesis of *let-7* miRNA. By binding to the pre-element of pri- or pre-*let-7*, neither Drosha nor Dicer can process the corresponding *let-7* precursor. In addition, Lin28 recruits TUT4/TUT7 to pre-*let-7* and promotes its 3'-end oligo-uridylation. Oligo-uridylated pre-*let-7* cannot be cleaved by Dicer and thus serves as a signal for the cellular 3'–5' exoribonuclease Dis3l2. Upon differentiation, Lin28 expression is reduced, which leads to increased levels of mature *let-7*. The latter silences gene expression of proto-oncogenes (Ras, c-Myc, Hmga2), cell cycle progression factors (Cyclin D1 and D3, Cdk4), components of the insulin-PI3K-mTOR pathway and Lin28 itself, thereby establishing a positive feedback loop. Besides its role in differentiation, a Lin28/*let-7* regulatory network is apparently involved in several cellular processes such as proliferation, oncogenesis, development and physiology, as well as metabolism (recently reviewed in [42]).

#### **3. Lin28 Influences mRNA Translation**

Besides regulation of *let-7* biogenesis, Lin28a/b can interact with various mRNAs and modulate their translation. Polesskaya and colleagues revealed that Lin28a can associate with polysomes and enhance translation of a number of mRNAs in differentiating myoblasts [25]. Among the first identified mRNA target was Igf2 (insulin-like growth factor 2), a major growth and differentiation factor in muscle tissue. Further evidence was provided that Lin28 recruits Igf2 mRNA to polysomes and enhances its translation via interactions with components of the translation initiation machinery. Subsequent studies revealed a number of additional mRNA targets of Lin28a in mouse ESCs such as H2a (histone 2a), Hmga1, Cyclin A, Cyclin B, Cdk4 and Oct4 [11,46–50]. An association of Lin28a with most of these mRNAs correlated with enhanced translation, suggesting that Lin28a maintains pluripotency by stimulating the translation of corresponding cell-cycle effectors. Further genome-wide studies revealed that Lin28a facilitates translation of genes important for growth and survival in human ESCs by recruiting RNA helicase A (RHA) to polysomes [24,51,52]. Additional mutagenesis studies revealed that the C-terminal part of Lin28a is required for RHA interactions, while mutations in the ZKD only impaired the stimulatory impact on translation, but not protein-protein interactions [46].

Very recently, a number of genome-wide Lin28 RNA crosslinking and immunoprecipitation coupled to high-throughput sequencing (HITS-Clip and PAR-CLIP) studies were conducted in human and mouse ESCs as well as somatic cells [19,49,53,54]. All of these studies have in common that only a small fraction of the identified RNA targets could be traced back to miRNAs, while the majority was mapped to thousands of mRNAs and ribosomal RNAs. For example, in mouse ESCs Lin28a was predominantly bound to mRNA transcripts (42%), mainly within the CDS and 3' UTR. Furthermore, a gene ontology analysis of target RNAs, revealed a preferential interaction of Lin28a with mRNAs that are destined for the endoplasmic reticulum. Binding of Lin28a to these mRNAs was associated with a translational repression by reducing ribosome occupancy without affecting mRNA abundance [49].

On contrary, in human HEK293 cells, binding of Lin28a and Lin28b to its mRNA targets was linked to a globally enhanced protein synthesis [19,53]. As before in mESCs, both human Lin28 paralogs predominantly bound within exonic regions of mRNAs, thereby mirroring the predominant cytosolic localization of Lin28a/b in HEK293 cells. Among the top RNA targets were mRNAs encoding for splicing factors and RNA-binding proteins, cell-cycle regulators as well as Lin28 itself. Binding of Lin28b to its own mRNA, indeed, correlated with increased levels of Lin28b protein, thereby suggesting a *let-7* independent feed-forward mechanism to maintain high levels of Lin28b in proliferative cell types [19,53,54]. Apart from their own expression Lin28a/b also seem to drive expression of important cell-cycle regulators of the ERK signaling cascade, such as Cdk1, N-Ras, Ran and ERK. This would explain the strong proliferative defects observed upon Lin28b knockdown [53]. Wilbert and colleagues further detected widespread changes in protein levels of splicing factors upon down-regulation of Lin28a and Lin28b in human ESCs. Whereas Lin28a binding to hnRNP F mRNA repressed translation, binding to TDP-43 and FUS/TLS mRNA was associated with an enhanced protein synthesis of the corresponding transcript. Consistent with Lin28's impact on alternative splicing factors, up-regulation of Lin28a in somatic HEK293 cells caused dramatic changes in alternative splicing patterns [54] (Figure 3).

**Figure 3.** Lin28 binds various mRNAs and modulates their translation. Both Lin28 paralogs were shown to influence mRNA processing on several levels. In the nucleus, Lin28 could regulate splicing of bound pre-mRNAs in concert with heterogeneous nuclear ribonucleoproteins (hnRNPs). In the cytosol, Lin28 was shown to interact with an RNA helicase A (RHA) thereby modulating the translation of target mRNAs via interactions with eukaryotic translation initiation factors (eIFS), elongation factors (eEFS) and poly(A)-binding proteins (PABP). Furthermore, Lin28 was found to shuttle mRNAs to poly-ribosomes and, under stress condition, to P-bodies and stress granules, thereby providing a direct link to the miRNA decay machinery. Lin28 binding to mRNAs was typically associated with a globally enhanced protein synthesis. However, in hESCs Lin28 binding repressed translation of bound mRNAs that were destined for the ER.
