All's Well that Ends Well: Alternative Polyadenylation and its Implications for Stem Cell Biology

Introduction

The remarkable ability of stem cells to facilitate tissue formation during embryonic development and regeneration in adulthood is contingent upon the regulation and interplay of a plethora of molecular pathways. The advent of microarray and high-throughput sequencing technologies have permitted a global evaluation of these pathways in different cell populations [1–3]. Not only have these platforms created an opportunity to study changes in transcript expression, but they have also revealed a tremendous diversity in transcriptional landscape [4–6]. Of particular interest has been polyadenylation site (PAS) choice because of its potential, by virtue of 3’ untranslated region (3’ UTR) modifications, both to influence the regulation of the transcript and to result in the production of truncated protein isoforms. In this review, we highlight recent findings that point to APA as a mechanism that influences stem cell behavior.

APA as a mechanism of transcript regulation

Polyadenylation occurs post-transcriptionally with cleavage at the 3’ end of the transcript and the addition of a series of adenosine molecules. In APA, the transcript contains more than one PAS which allows for cleavage at multiple locations and, hence, the production of distinct transcript variants [5]. Numerous methods have been used to detect APA globally, and recent estimates suggest that over 30% and 50% of genes are alternatively polyadenylated in mice and humans, respectively [7,8]. Table 1 provides a comprehensive summary of relevant methods and studies [2,7–36].

Depending upon the position of the alternative PASs, the quantity or the structure of the protein produced can be altered [4,37,38]. If all alternative PASs are located in the 3’ UTR (referred to as “UTR-APA” [5] or the “development of tandem UTRs” [28]), the resulting transcripts share an identical protein coding sequence. However, the transcript variants that utilize the more proximal PASs harbor shorter 3’ UTRs (Figure 1A). By contrast, if an alternative PAS is located in an internal intron or exon of the coding region (referred to as “coding region APA” [5] or “3’ exon switching” [28]), the alternative transcript codes for a truncated protein and alternate 3’ UTR [37] (Figure 1B).

Because the coding region is altered in the latter case, those protein isoforms may exhibit functional differences. Interestingly, a recent investigation of tyrosine kinases demonstrated that coding region PAS choice may result in a negative feedback system to fine-tune signaling mediated by the encoded proteins [39]. In particular, APA in the coding region of these receptors may produce soluble isoforms that contain the ligand binding domain but not the transmembrane region or kinase domain. These isoforms thus act as decoy receptors that compete for ligand binding but do not activate signaling [39]. Likewise, a study of a tRNA synthetase showed that APA can create protein isoforms that interfere with the activity of the full-length counterparts [40]. Conversely, CR-APA enhances Cyclin D1 function, as the truncated form is constitutively active [41,42].

Regarding UTR-APA, a global study suggested that transcripts with shorter 3’ UTRs tend to be more highly expressed than their longer counterparts [22]. Several mechanisms can explain why changes in 3’ UTR length may affect protein abundance. One of the best-characterized processes is that of microRNA (miR)-mediated degradation. In studies of myogenic [43,44], hematopoietic [28], and cancer [45] cells, transcripts bearing shorter 3’ UTRs contained fewer miRNA-binding sites, thus allowing these transcripts to evade miRNA-mediated degradation. Transcripts are also subject to length-dependent degradation by the nonsense-mediated decay (NMD) pathway [46,47]. In NMD, Upf1 binds to the 3’ UTR in a length-dependent manner, thus eliciting degradation of longer transcripts more rapidly [48].

The 3’ UTR contains elements that affect not only transcript degradation but also stability. In a genome-wide computational analysis of sequence and stability data, a number of motifs regulating mRNA stability in the 3’ UTR were reported [49]. The truncation of the 3’ UTR by APA, and hence the removal of these motifs, could thus affect the steady-state level of the expressed transcripts. Elements in the 3’ UTR have also been shown to affect transcript localization [50] and translational efficiency [51], allowing for additional levels of regulation of protein abundance.

APA not only may have an effect on the polyadenylated transcript itself, but also can influence transcription of neighboring genes. In one study of embryogenesis, production of a longer transcript of one gene, Mest, inhibited transcription of an overlapping antisense gene, Copg2 [52]. Intriguingly, because Mest is imprinted, transcription of Copg2 was inhibited only on the allele from which Mest was expressed. Thus, Mest expression allowed for mono-allelic expression of Copg2, making it appear as though Copg2 was also imprinted [52].

APA in Stem Cell Biology

A growing body of evidence demonstrates that APA is active in stem cell populations and affects genes that are critical for stem cell function. We will discuss recent findings that relate to adult stem cell populations, as well as embryonic progenitors and induced pluripotent stem (iPS) cells.

Posttranscriptional regulation by APA during development has been reported across multiple species. A recent study of zebrafish development demonstrated that embryos express an abundance of transcripts in which multiple PASs are used [15]. Over a thousand transcripts displayed significant changes in PAS choice during development [15]. Moreover, there was a global trend towards increased usage of distal sites and, hence, an increase in average 3’ UTR length [15]. A separate study of zebrafish confirmed this trend, noting that while 3’ UTR length decreases in early development during the transition from zygote to blastula, it increases sharply in gastrulation [19]. Similar changes have been observed in mouse [30] and Drosophila [53] in which the average 3’ UTR length increases in somatic tissues during embryonical and postnatal development [30,53]. Interestingly, analysis of microarray data for a number of cells before, during, and after reprogramming revealed that iPS cells display shortened 3’ UTR lengths compared with the somatic cells from which they were derived [29].

APA is important for progenitor function not only during development but also in the adult. In a study of muscle stem cells, or satellite cells, our laboratory found that APA controls expression of a critical myogenic regulator, Pax3, whose transcript contains miR-206 binding sites in the 3’ UTR [44]. Curiously, while Pax3 transcript is highly expressed in satellite cells of both limb muscles and the diaphragm, the protein is expressed only in diaphragmatic satellite cells. Because miR-206 is expressed equally in both satellite cell populations, the differential expression of Pax3 protein cannot be explained by differential miR-206 expression. Further investigation revealed that satellite cells in limb muscles use, primarily, the distal Pax3 PASs, thus generating transcripts that contain miR-206 binding sites and allowing for miR-206-mediated suppression of Pax3 protein expression. Satellite cells in the diaphragm, however, primarily use the proximal PAS and thus express Pax3 transcripts with shorter UTRs that are devoid of miR-206 binding sites, allowing these cells to evade miR-206 regulation of Pax3 [44].

A study of muscle differentiation revealed that a critical differentiation regulator, brain-derived neurotrophic factor (BDNF), is regulated by APA [54]. The muscle cell line, C2C12, produces two distinct BDNF transcript variants that differ in 3’UTR length. Production of the shorter form allows BDNF to escape miR-206 mediated repression during differentiation [55]. A recent report of BDNF in neurons also illustrated a role of the 3’ UTR in regulating transcript localization and translation. In resting neurons, the long variant is sequestered in ribosome-free ribonucleoprotein particles and hence translationally repressed. During neuronal activation, the long variant is released and robustly translated [56]. Conversely, the short form is consistently translated, allowing for a basal production of the protein [56].

APA is also active in neural stem cells where the RNA-binding protein HuR, which primarily facilitates the initial proliferation stage, is surprisingly also expressed later during neuronal differentiation along with the pro-differentiation proteins HuB, HuC, and HuD. Two recent studies found that the binding of these proteins and HuR itself to the proximal PAS of nascent HuR mRNA induced polyadenylation at a distal site. The longer variant was less stable, resulting in decreased protein production compared with the shorter form. Thus, the cell could control output of HuR by regulating transcript polyadenylation [57,58]. Although HuR 3’ UTR length increased, sequencing analysis of neural stem/progenitor cells and differentiated neurons revealed that global 3’ UTR length decreases in differentiation. The mechanisms for this trend remain unknown [17].

Spermatogenesis is also characterized by the production of a multitude of polyadenylation variants of germ cell-specific transcripts [33]. One of these variants is the shortest known variant of the testis-brain RNA-binding protein (TB-RBP), a protein that facilitates intra- and intercellular mRNA transport [59,60]. In the testis, the cleavage stimulation factor-64 (Cstf-64) promotes production of the shortest TB-RBP isoform, which is the most efficiently translated variant [59]. As with neural differentiation, spermatogenesis is also characterized by a global decrease in average 3’ UTR length [33].

Within stem cell populations, patterns of expression of factors known to regulate APA further support the idea that changes in 3’ UTR length are important in stem cell function. Cleavage and polyadenylation specificity factor-160 (Cpsf-160) as well as Cstf-64 are proteins that bind to PAS sites directly to mediate cleavage and polyadenylation [38,61], and as just noted, Cstf-64 influences APA specificity in the testis. Intriguingly, these factors have been shown to be associated with shorter UTRs. In the mouse, both factors are highly expressed in the male germ cells [62], which display 3’ UTR length decreases postnatally [15,30,33,35]. Moreover, they are also upregulated with a corresponding trend towards shorter 3’ UTR variants during iPS cell generation [29] and in cancer cells [45]. In contrast, these factors are downregulated during C2C12 differentiation when 3’ UTR length increases [30]. Another factor, polypyrimidine tract binding protein (PTB), which has been implicated in alternative splicing and PAS selection [63,64], is also downregulated during myogenic differentiation [65].

Perspectives

A number of questions remain unanswered in the burgeoning field of APA and stem cell regulation. Certainly, further studies of stem cell populations are needed to understand the extent to which APA occurs in different stem cell types and to uncover changes in polyadenylation patterns during quiescence, activation, and differentiation. General patterns that have emerged in the study of pluripotent stem cells include a lengthening of the 3’ UTR in embryonic progenitors during development and a decrease in 3’ UTR length of somatic cells during reprogramming (Figure 2A). Although there are few large-scale analyses on the 3’ UTR length of adult stem cells, the data on proliferating and differentiated somatic cells suggest that adult stem cells will globally express transcripts with shorter 3’ UTRs upon exit from quiescence and entry into the cell cycle and will then express longer 3’ UTRs upon differentiation (Figure 2A, B).

A number of stem cell populations display molecular and functional heterogeneity, though the full characterization of this phenomenon and its consequences for tissue regeneration and aging have yet to be understood [66]. We postulate that APA-directed 3’ UTR length changes may mediate heterogeneity within a population. The global analyses of APA discussed previously have shown that multiple transcript variants can be expressed simultaneously in a cell population. However, this population-based heterogeneity could arise as a result of different patterns at the single cell level with individual cells expressing different isoforms from their neighbors (Figure 3A).

Likewise, changes in 3’ UTR length that are observed in populations during state transitions can occur as a result of different patterns at the single cell level. We propose two (non-mutually-exclusive) models by which a stem cell population could exhibit global length changes during a transition from one state to another (Figure 3B). In this example, we illustrate a hypothetical gene with only two transcript variants, a long form and a short form. In one case (the “conversion” model), a global decrease in UTR length could occur if all cells are capable of expressing both transcripts and the result of the transition is that each cell expresses a higher proportion of the short transcript. In the other case (the “selection” model), in which each cell is capable of expressing only one form, global decrease in UTR length in the population would occur when there is a selection or selective expansion of those cells expressing the short form (Figure 3B). The development of technologies such as single-cell PCR [67–69] and single-cell RNA sequencing [70,71] will surely herald advances in our understanding of APA and stem cell heterogeneity.

Although many studies have addressed the functional consequences of 3’ UTR length changes for individual genes, the effects of global changes in 3’ UTR length remain elusive. One potential explanation stems from pivotal work in the area of miRNA sponges where it was shown that one could use decoys containing target sites to sequester miRNA molecules and inhibit their activity [72,73]. Interestingly, a number of studies have suggested that the abundance of mRNAs can also dilute the activity of the targeting miRNAs and siRNAs [72,74,75]. Consequently, if multiple transcripts are targeted by the same miRNA, changes in the abundance of any one of those transcripts, which have been termed competing endogenous RNAs (ceRNAs), could influence the abundance of any of the other transcripts [75–80]. We hypothesize that global changes to 3’ UTR length could have the same effect. In this model, mRNAs with longer 3’ UTRs would act as ceRNAs. If a cell state change were to occur such that 3’ UTR lengths were globally decreased, then miRNA availability would be enhanced overall.

Another area for further research is in the regulation of APA. Global studies have suggested that PAS usage varies widely across different tissues with some polyadenylation variants expressed in a tissue-specific manner [24,35]. It is thought that a number of factors involved in polyadenylation and splicing may play a role in specifying the unique polyadenylation signatures of different cells [5,38]. Various techniques have been utilized recently to identify factors that influence polyadenylation on a global level. In one study, high-throughput sequencing of RNA isolated by crosslinking immunoprecipitation (HITS-CLIP) was used to identify the RNA binding sites of a neuron-specific splicing factor, Nova. Because this protein was also bound to regions in 3’ UTRs, a novel role for this protein in APA was revealed [81]. In another study, Systematic Evolution of Ligands by Exponential Enrichment (SELEX) technology was used with RNA sequencing to pinpoint binding motifs for two other regulators of APA, epithelial splicing regulatory protein (Esrp) 1 and Esrp2 [82]. Conversely, in a third study, a reporter for APA and an RNAi-based screen were used to identify a number of factors, including Poly(A)-Binding Protein Nuclear 1 (PABPN1), that regulate APA [83]. Furthermore, recent studies demonstrating that widespread RNA methylation occurs in the 3’ UTR have raised the intriguing possibility that RNA methylation can play a role in PAS selection [84,85]. The application of these techniques to stem cell populations will allow for an understanding of how these factors interact to produce the diversity in PAS usage amongst different populations and stem cell states.

Some of these tools have also revealed how changes to 3’ UTR processing could induce tissue dysfunction and disease. For example, mutations in PABPN1 cause oculopharyngeal muscular dystrophy [86], and such mutations correlate with global changes in 3’ UTR lengths and pervasive gene dysregulation [83]. Moreover, recent investigations have shown that tumor cells exhibit decreased 3’ UTR length [12,27,45]. The shorter 3’ UTR length of the proto-oncogene IGF2BP1/IMP-1 and consequent enhancement in protein expression suggested a role for APA in cancer progression [45]. These studies may have relevance to the stem cell field to the extent that those global changes are characteristic of cancer stem cells that may exist in many tumors [87]. A thorough understanding of the extent and patterns of APA in stem cell populations will likely lead to new insights into the regulation of stem cell behavior and heterogeneity as well as tissue maintenance and disease.