Regulated alternative polyadenylation is an important feature of gene expression, but how gene transcription rate affects this process remains to be investigated. polo is a cell‐cycle gene that uses two poly(A) signals in the 3′ untranslated region (UTR) to produce alternative messenger RNAs that differ in their 3′UTR length. Using a mutant Drosophila strain that has a lower transcriptional elongation rate, we show that transcription kinetics can determine alternative poly(A) site selection. The physiological consequences of incorrect polo poly(A) site choice are of vital importance; transgenic flies lacking the distal poly(A) signal cannot produce the longer transcript and die at the pupa stage due to a failure in the proliferation of the precursor cells of the abdomen, the histoblasts. This is due to the low translation efficiency of the shorter transcript produced by proximal poly(A) site usage. Our results show that correct polo poly(A) site selection functions to provide the correct levels of protein expression necessary for histoblast proliferation, and that the kinetics of RNA polymerase II have an important role in the mechanism of alternative polyadenylation.
Transcription by RNA polymerase II (Pol II) is a complex and multistep process that consists of promoter binding, initiation, elongation and termination. For most eukaryotic genes, the nascent transcript is recognized by RNA processing enzymes, as it is being synthesized by Pol II. For instance, excision of non‐coding introns to generate faithfully spliced messenger RNA (mRNA) occurs long before the polymerase has completed transcription of the whole gene (Moore and Proudfoot, 2009). Similarly, the nascent transcript is released from the transcribing polymerase by the action of a large processing protein complex (Shi et al, 2009) coupled to the addition of a poly(A) (pA) tail to the terminal 3′ end of the pre‐mRNA and to transcriptional termination (for reviews, see Hirose and Manley (2000), Proudfoot (2004), Bentley (2005), Buratowski (2005) and Richard and Manley (2009)).
Although it was previously believed that the critical step for regulating transcription was at the stage of promoter binding and transcription initiation, recent advances have revealed that many genes are regulated at the phase that follows the promoter‐paused Pol II entering into elongation (Core et al, 2008; Gilmour, 2009). At this stage, the speed at which Pol II transcribes a transcriptional unit may have an impact on the regulation of gene expression and may determine pre‐mRNA processing (Caceres and Kornblihtt, 2002). In Drosophila, the C4 mutation in the RpII215 gene, which encodes the largest subunit of Pol II, changes amino acid 741 from Arg to His (Chen et al, 1993). In vivo this mutation induces Pol II α‐amanitin resistance with a 50% slower transcription elongation rate than its wild‐type counterpart (Coulter and Greenleaf, 1985; Chen et al, 1993) and reads less efficiently through DNA pause elements (Chen et al, 1996). In human cells, a mutation equivalent to C4 has been shown to affect the alternative splicing pattern of the human fibronectin (FN) exon 33 (E33) EDI. Here, the ‘slow’ Pol II acts in concert with splicing factors to induce the inclusion of the alternative E33 (de la Mata et al, 2003, 2010). Additionally, in Drosophila C4 embryos resplicing of the Hox gene Ultrabithorax is stimulated (de la Mata et al, 2003, 2010). Thus, a ‘kinetic coupling’ model in which the transcription elongation rate determines the outcome of the alternative splice site selection was put forward (Caceres and Kornblihtt, 2002).
The 3′ end processing complex is precisely regulated and specifically recognizes sequences in the nascent transcript that comprise the pA signal of the gene, which is typically AAUAAA, and a GU‐ or U‐rich downstream RNA sequence element (for reviews, see Colgan and Manley (1997), Minvielle‐Sebastia and Keller (1999), Wahle and Ruegsegger (1999), Zhao et al (1999), Edmonds (2002) and Shi et al (2009)). Additional sequences positioned either upstream (Carswell and Alwine, 1989; Gilmartin et al, 1992; Chen and Wilusz, 1998; Moreira et al, 1998; Hall‐Pogar et al, 2005; Hu et al, 2005; Venkataraman et al, 2005; Danckwardt et al, 2007) or downstream of the pA signal (Bagga et al, 1995; Arhin et al, 2002; Dalziel et al, 2007) may act as auxiliary elements by binding to components of the 3′ end processing complex or to proteins that can influence the efficiency of this complex and consequently modulate the overall efficiency of mRNA 3′ end formation.
Poly(A) sites are thus integral parts of the gene, signalling where in the precursor RNA cleavage and polyadenylation should occur. However, a large percentage of genes possess multiple pA sites that can be alternatively utilized. It is estimated that over half of human genes contain more than one pA signal and that alternative polyadenylation may be a major player in gene regulation (Lutz, 2008; Neilson and Sandberg, 2010; Lutz and Moreira, 2011). In some cases the choice of alternative pA sites is coupled with alternative splicing, which results in the production of different transcripts that are translated into different protein products with diverse cellular functions, for example, the membrane or secreted forms of immunoglobulin in B cells (Peterson and Perry, 1986; Lou et al, 1996; Takagaki et al, 1996; Edwalds‐Gilbert et al, 1997; Peterson, 2007). In contrast, the presence of alternative pA sites in the 3′‐most exon (tandem pA signals) directs the synthesis of mRNAs that share the same coding sequence but have different 3′ untranslated regions (UTRs). This is usually coupled with the presence of regulatory elements in the 3′UTR that interact with RNA binding proteins or miRNAs, modulating mRNA stability, translation, mRNA transport and/or mRNA localization (Lewis et al, 1995; Huang and Carmichael, 1996; Jacobson and Peltz, 1996; Wickens et al, 1997; Preiss et al, 1998; Ghosh et al, 2008; Merritt et al, 2008; Newnham et al, 2010). Importantly, two recent genome‐wide studies, in tumour cells and in activated versus resting T cells, have found that in general shorter 3′UTRs are indicative of enhanced cell proliferation (Sandberg et al, 2008; Mayr and Bartel, 2009). The shortening of the 3′UTRs resulting from alternative polyadenylation leads to the loss of regulatory elements in this region, such as miRNA target sites. Conversely, during mouse embryonic development, alternative polyadenylation was shown to be responsible for a progressive lengthening of mRNA 3′UTR (Ji and Tian, 2009). It was also recently shown that differences in the 3′UTR due in part to alternative polyadenylation, define molecular signatures that can distinguish tumour subtypes (Singh et al, 2009). In Drosophila the enhancer of rudimentary (e(r)) (Gawande et al, 2006) and the suppressor of forked (su(f)) are two examples of genes regulated by alternative polyadenylation (Audibert and Simonelig, 1998). e(r) produces two mRNAs due to two pA signals present in the 3′‐most exon, but only the longer e(r) mRNA is specifically expressed in the female germline, which requires co‐expression of the female‐specific Sex‐lethal (Wojcik et al, 1994; Gawande et al, 2006). su(f) contains three pA signals, the first one within intron 4, leading to the production of a truncated transcript, and the other two in the 3′UTR. It was suggested that the Su(f) protein regulates its own accumulation via a negative feedback loop that stimulates 3′ end formation of the truncated su(f) RNA (Audibert and Simonelig, 1998).
Polo was the first described member of an important family of cell‐cycle kinases, the Polo‐like kinases (Plks), and is involved in many critical steps in the cell cycle, including mitotic entry, centrosome organization, spindle formation, chromosome segregation and cytokinesis (Sunkel and Glover, 1988; Llamazares et al, 1991; Moutinho‐Santos et al, 1999; Glover, 2005). polo contains a proximal and a distal pA signal in the 3′UTR (referred herein as pA1 and pA2, respectively). Despite evidence that pA site selection in the 3′ most exon of eukaryotic genes has a role in gene expression, the physiological impact of this mechanism on the development of an organism and the role of Pol II kinetics on alternative pA site choice is still poorly understood.
In this study, we show that in RpII215 mutant flies, which display reduced Pol II elongation kinetics, Pol II occupancy along polo is altered and polo pA1 is used 3.5‐fold more efficiently than in wild‐type flies. An increase in proximal pA site usage was also observed for five other alternatively polyadenylated transcripts in Drosophila. Moreover, we show that in flies carrying a deletion of the polo pA2 signal, Polo is translated at low levels and as a result precursor cells of the abdominal epidermis do not proliferate during metamorphosis. This causes a defect in abdomen development and lethality. Consequently, we demonstrate that in transgenic flies polo pA2 signal is necessary for proper protein levels production, cell proliferation and viability in the living organism. Importantly, Pol II kinetics has a critical role in the selection of alternative pA sites in Drosophila.
Pol II kinetics affects polo pA site usage
The polo gene, represented in Figure 1A, produces two mRNA species as previously described (Llamazares et al, 1991). The 3′ end of these two mRNAs was mapped by sequencing cDNAs isolated from a Poly(A)+ 0–24 h embryonic mRNA library, and they were shown to differ in their 3′UTR due to alternative usage of two pA signals localized in this region (Llamazares et al, 1991). The pA1 signal is the fly variant ATTAAA present in 10–32% of the Drosophila genes analysed, while the pA2 signal is the variant AATATA present in 5–10% of the Drosophila genes (Figure 1A; Graber et al, 1999; Retelska et al, 2006).
To investigate the role of Pol II kinetics on pA site usage of polo, we used the RpII215 Drosophila homozygous strain (referred herein as slow Pol II), which contains a C4 mutation in the largest subunit of Pol II reducing its transcription rate by 50% (Chen et al, 1996). We performed ChIP analysis on these flies, using an α‐Rpb3 Pol II antibody to investigate Pol II distribution along the polo gene. Pol II occupancy was measured at nine positions within the polo transcription unit, the intergenic region and the promoter region of the next downstream gene, snap (Figure 1A), from chromatin sheared to an average size of 200–300 bp. Each position corresponds to a 125‐ to 150‐bp amplicon separated by a minimum of 60 nucleotides (probes 7–8 and 8–9); the remaining probes are separated by >160 nucleotides. Quantification of the chromatin immunoprecipitation (ChIP) was done by real‐time PCR using SybrGreen fluorescence (see Materials and methods). In wild‐type flies, Pol II occupancy is reduced after pA1, suggesting that more pA1 than pA2 transcripts are being produced in adult flies (Figure 1A, left panel). This is in agreement with the quantification made in adult flies by real‐time PCR (Figure 1E, w1118) as well as with previously published results (Llamazares et al, 1991). The drop in Pol II occupancy that is seen after pA1 is indicative of polo termination, before it increases again over the closely spaced tandem gene snap (Figure 1A, left panel). However, in RpII215 flies this pattern is altered, with lower but more even Pol II levels across the polo‐snap locus (Figure 1A, right panel), which may be due to the mutation in the RNA polymerase II. In particular, no drop in signal was observed following pA1 (probe 7). The lower levels of Pol II over polo‐snap do not reflect a decrease in the levels of polo transcripts (Figure 1B) or Polo protein (Figure 1C) produced. We then determined whether this altered Pol II distribution affects polo pA site selection, by 3′RACE and RT–qPCR in RpII215 adult flies. 3′RACE analysis suggests that pA1 is used more efficiently than pA2 in RpII215 flies (Figure 1D). In order to quantify this effect, we measured the ratio between the total levels of polo transcripts (polo pA1+polo pA2) and polo pA2 mRNA levels by RT–qPCR. We observed that in RpII215 adult flies, this ratio is increased 3.5‐fold in comparison with the wild‐type flies (Figure 1E), indicating that pA1 is used more efficiently than pA2 in the RpII215 mutant. Our results clearly establish that in vivo the Drosophila slow Pol II affects polo pA site selection.
To extend this observation to other alternatively polyadenylated genes in Drosophila, we tested seven other previously described (GenBank, NCBI) or in silico predicted (Joel Graber, personal communication) genes, with a similar genomic structure to polo, that is containing two pA signals in the 3′UTR: abd‐B (Abdominal‐B, a member of the bithorax complex in Drosophila; Sanchez‐Herrero et al, 1985), lace (a homologue of the LCB2 subunit of serine palmitoyltransferase; Adachi‐Yamada et al, 1999), stlk (Ste20‐like kinase), CG6024 (Artero et al, 2003), cycD (Cyclin D; Finley et al, 1996), drep‐3 (DNA fragmentation factor‐related protein 3; Inohara and Nunez, 1999) and stg (String, a Cdc5‐like phosphatase; Edgar and O'Farrell, 1989). Using RT–qPCR with specific probes for these genes, we measured the ratio between total RNA levels and longer transcript levels (produced by distal pA signal selection) in RpII215 and w1118 flies. Five out of the seven genes analysed showed an increase in total/distal RNA levels in RpII215 flies in comparison with the wild type similar to that found for polo (Figure 1F). CG6024 shows a three‐fold increase and Abd‐b, stlk and lace show a 1.5‐ to 2‐fold increase. Drep‐3 and stg show a slight increase in this ratio. These results clearly show that Pol II elongation kinetics affects alternative pA site selection in other genes and suggests that it may be a general effect in Drosophila.
polo pA2 is required for Drosophila viability and abdominal development
Having established that slow Pol II kinetics correlate with increased pA1 site usage, we then addressed the question of the function of the polo distal pA2 site in vivo. We therefore determined the consequence of mutation/deletion of each of the two tandem pA sites located in the polo 3′UTR (Figure 2C). Flies that express polo tagged with gfp at the N‐terminus containing either one of the two pA signals were generated, using a previously described transgene gfp‐polo (Moutinho‐Santos et al, 1999). Point mutations were introduced in the pA1 signal (ATTAAA to GTTAAC), thus generating the transgene ΔpA1 that allows the expression of the longer polo transcript (Figure 2A). A second transgene, ΔpA2, was generated by deletion of pA2 signal and the entire downstream region (Figure 2A). The point mutations in pA1 and the deletion of pA2, inactivating pA1 and pA2, respectively, were therefore sufficient to allow the selective expression of only the longer or shorter polo mRNA in each transgenic line as shown by northern blot analysis in Figure 2B (open and black arrows, respectively). In this northern blot, polo pA1 and polo pA2 mRNAs are detected at similar levels in brains of third instar larvae in a wild‐type background (Figure 2B) in contrast to what was observed in adult flies. This difference may reflect the nature of the developmental stage used and is in agreement with previous results (Llamazares et al, 1991). All transgenic lines express GFP–Polo protein throughout fly tissues (Supplementary Figure S1).
To study the role of polo pA signals in vivo, the viability of homozygous lines for the transgenes was analysed in a polo9 mutant background, which has been described as the strongest hypomorphic allele identified for polo (Donaldson et al, 2001). This mutant was generated by insertional mutagenesis of a P‐element into the 5′UTR of polo, 176 nucleotides upstream of the ATG codon, where it interferes with polo transcription (Donaldson et al, 2001). polo9 homozygous third instar larval brains have barely detectable levels of Polo protein that are not sufficient for the larvae to develop beyond this developmental stage, and consequently die (Donaldson et al, 2001). We first determined if the polo mRNA and protein levels contributed by the transgenes were sufficient to rescue the third instar larval lethal phenotype displayed by polo9 individuals in the transgenic flies gfp‐polo;polo9, ΔpA1;polo9 and ΔpA2;polo9. As can be seen, the gfp‐polo and ΔpA1 transgenes rescue polo9 phenotype and the flies reach adulthood (Table I). In contrast, nearly all ΔpA2;polo9 individuals died during pupariation (Table I). These results clearly show that the longer polo mRNA produced using pA2 is required for normal Drosophila development beyond the late third instar larval stage, more precisely, during metamorphosis. Strikingly, analysis of the few ΔpA2;polo9 adult escapers indicated that these individuals have strong defects in abdominal morphology, with the complete absence or incorrect formation of tergites (Figure 2D, panel 7) while all other adult structures appear normal. Panels 2–4 show abdomens prepared from transgenic flies, in a polo9/TM6B background, containing the transgenes plus one copy of polo (TM6B). Panels 5–7 show transgenic flies in a polo9 homozygous background. Transgenic gfp‐polo;polo9 adult flies have normal tergites as expected, while ΔpA1;polo9 present a mild abdominal phenotype (Figure 2D, panels 5 and 6). The possibility that the phenotype may arise as a result of the insertion of transgenes into the same genomic region in both transgenic lines was excluded (Supplementary Figure S2). In spite of the almost undetectable endogenous Polo in polo9 (Donaldson et al, 2001), we quantified the remaining polo mRNA contribution by this allele, as well as the levels produced by each transgene, in all the transgenic polo9 homozygous lines, by RT–qPCR (gfp‐polo;polo9, ΔpA1;polo9 and ΔpA2;polo9; Figure 2E). Using specific primers for the gfp transgene (light grey bars) or primers that amplify both endogenous polo and gfp‐polo mRNAs (dark grey bars), it is clear that the transgenes are principally responsible for polo mRNA levels (Figure 2E). The small difference between the bars corresponds to the endogenous polo contribution. Moreover, ΔpA2;polo9 flies show higher mRNA levels than ΔpA1;polo9 flies. This clearly indicates that the phenotype presented by ΔpA2;polo9 individuals (Figure 2D, panel 7) is due to the lack of polo pA2 transcript and that selection of the pA2 signal is essential for normal development of the abdomen and viability of the fly.
pA2 is required for abdominal histoblast proliferation at metamorphosis onset and for correct polo protein expression
The high levels of lethality for ΔpA2;polo9 observed during the pupa stage suggested that polo pA2 was essential at this stage of development. Additionally, the phenotype presented by ΔpA2;polo9 is in agreement with that found in mutants for arrowhead (awh) and escargot (esg) genes, two other genes previously reported as necessary for the formation of the adult abdomen (Hayashi et al, 1993; Curtiss and Heilig, 1995). In esg mutants, the histoblasts—small diploid cells that arise during the last stages of embryogenesis (Roseland and Schneiderman, 1979; Madhavan and Madhavan, 1980)—undergo DNA endoreplication resulting in polyploid cells which fail to proliferate, leading to adult flies with an abdominal phenotype very similar to ΔpA2;polo9 flies (Hayashi et al, 1993). Upon entering metamorphosis, the histoblasts undergo very rapid proliferation and from 15 h after pupa formation they expand, replacing all larval polytene epidermis cells (Ninov et al, 2007), starting to form a complete adult epidermis layer. Therefore, we analysed the histoblasts in pupae collected 26–27 h after pupa formation (Figure 3A). At this stage, histoblasts are already proliferating, replacing larval epidermis cells and forming over 50% of the adult abdomen epidermis (Roseland and Schneiderman, 1979). As expected, in gfp‐polo;polo9 and ΔpA1;polo9 pupae normal histoblast proliferation can be clearly seen: histoblasts are the small cells (delimited by the dotted line in Figure 3A). Larval epidermis cells (delimited by solid line in Figure 3A) are larger and easily distinguished from histoblasts. However, in ΔpA2;polo9 pupa epidermis only the large polytene larval cells are observed and histoblasts could not be detected in pupae at this stage (Figure 3A). Moreover, this phenotype is only observed in a polo9 homozygous background (Supplementary Figure S3). In conclusion, our results show that in the absence of the polo pA2 transcripts, histoblasts are not detected during metamorphosis.
The absence or incorrect formation of the tergites in ΔpA2;polo9 flies could result either from a failure in the specification/formation of abdominal histoblasts before the entrance to the pupal stage, or from the inability of these cells to proliferate during this developmental stage. To distinguish between these two possibilities, ΔpA2;polo9 third instar larvae were dissected and immunostained with anti‐Headcase (Hdc) antibody, used as a marker for abdominal histoblasts, showing that histoblasts are clearly present in gfp‐polo;polo9, ΔpA1;polo9 and ΔpA2;polo9 larvae (Supplementary Figure S4A). Furthermore, all histoblast cells are strongly stained for Cyclin B, as expected from G2‐arrested cells (Supplementary Figure S4A). Moreover, quantification of the number of histoblasts in each nest shows that these larvae have similar number of cells in each histoblast nest for the different abdominal segments (Supplementary Figure S4B). Taken together, these results indicate that ΔpA2 abdominal histoblasts are correctly formed during embryogenesis and are able to maintain the G2 arrest until the larvae stage. Consequently, the longer polo transcript derived from pA2 usage is essential for the proliferation of these cells at the onset of metamorphosis.
Having determined that deletion of polo pA2 was detrimental to the fly, we went on to investigate the expression of GFP–Polo obtained from the different transgenes in late pupal development, using pupae in a heterozygous background (polo9/TM6B), to allow proliferation of the histoblasts. We observed a significant decrease of the GFP–Polo signal in histoblasts of ΔpA2 pupae (Figure 3B; histoblast cells are delimited by a dotted yellow line). Quantification of the fluorescent levels throughout the whole pupa epidermis, using the same individuals, showed that while gfp‐polo and the ΔpA1 transgenes express similar levels of GFP–Polo a 30% decrease in fluorescence is observed for the ΔpA2 transgene (Figure 3C). To elucidate if this is due to a decrease in the mRNA produced, we quantified the mRNA by RT–qPCR. As can be seen in Figure 3D, there is no significant difference between the mRNA levels produced by each transgene; thus, the decrease in protein levels observed in ΔpA2 flies is unlikely to be due to a simple reduction in the mRNA levels. These results thus suggest that the appropriate levels of Polo protein in abdominal histoblasts may only be achieved when the distal pA2 site is used.
To identify the molecular mechanism responsible for the decrease in Polo protein when pA2 is deleted, we used a Drosophila cell line to assay for the stability of the endogenous polo mRNAs. Upon treatment with actinomycin D and measurement of RNA decay by RT–qPCR, we observe that both polo RNAs and polo pA2 RNA levels are maintained for at least 6 h (Figure 4A), suggesting that both transcripts are equally stable. As a control for the actinomycin D treatment, we used the Vha68‐1 (vacuolar H+ATPase subunit 68‐1) RNA which displays a half‐life of ∼30 min, as previously shown (Behm‐Ansmant et al, 2006). We then tested if the different 3′UTRs present in each transcript were responsible for the different levels of protein produced. The 3′UTR ending at pA1 was fused downstream of a Luciferase reporter gene to produce the pA1 plasmid (Figure 4B). In a construct containing the whole 3′UTR fused to luciferase, the pA1 signal was mutated (ATTAAA–GTTAAC) to produce plasmid pA2 (Figure 4B); the resultant plasmids, pA1 and pA2, were transfected into Kc cells. The mRNA 3′ends formed from each construct were confirmed by 3′RACE and to distinguish translational output from mRNA turnover, luciferase translational activity was normalized to luciferase‐reporter mRNA levels to obtain the translation efficiency (defined in Supplementary Figure S5). Significantly, we found that pA1 causes a three‐fold decrease in translation as compared with pA2 (Figure 4C). These results clearly indicate that the longer polo mRNA is more efficiently translated than the shorter transcript. This suggests that the correct levels of Polo are defined by pA2 signal selection, in agreement with the in vivo results obtained with the transgenic fly lines.
Finally, we asked if the levels of utilization of pA1 and pA2 could be regulated by in vivo overexpression of Polo using the previously established GAL4/UAS system (Brand and Perrimon, 1993). Using this system, it was previously shown that Polo is overexpressed by 2.5‐fold which does not affect its physiological function in mitosis (Martins et al, 2009). Thus, using a neuroblast driver (MZ1061‐Gal4) we tested the correlation between the expression of each transcript and protein levels in a tissue‐specific manner. For this, MZ1061‐Gal4; gfp‐polo:polo9/TSTL females were crossed with MZ1061‐Gal4;UAS‐polo:polo9/TSTL males and the polo9 homozygous brain progeny was analysed; in these flies, polo cDNA with the full 3′UTR is transcribed from an upstream inducible element (Mirouse et al, 2006). As shown in Figure 5A, Polo is overexpressed 2.3‐fold relative to the wild‐type control, levels that are in agreement with previous studies (Martins et al, 2009). Using RT–qPCR, we then measured the ratio of pA1 to pA2 mRNAs in the polo9 homozygous F1 generation taking strain MZ1061;gfp‐polo:polo9/TSTL as a control. Significantly, overexpression of Polo results in a 2.5‐fold increase in the total/pA2 mRNA ratio (Figure 5B). We further determined the amount of GFP–Polo protein levels by western blot analysis and observed a four‐fold decrease in GFP–Polo in the flies where polo was overexpressed (Figure 5C). Taken together, these results show that overexpression of polo induces a switch from pA2 to pA1 mRNA and a consequent decrease in the protein levels. These results suggest the existence of a potential auto‐regulatory feedback loop mechanism.
Recent studies have shown that in cancer cells alternative polyadenylation in general selects shorter 3′UTRs and consequently there is an increase of protein production (Mayr and Bartel, 2009). This was shown to be partly due to the presence of miRNA target sites in the extended 3′UTRs. We have searched in the available databases and prediction programs for miRNAs that could potentially target polo transcripts and found dme‐mir‐8 and dme‐mir‐1016 (Stark et al, 2007). The predicted miRNAs were quantified and tested in Kc cells. However, we could rule out an effect of dme‐mir‐8 and dme‐mir‐1016 overexpression on polo 3′UTR (Supplementary Figure S6).
Expression levels of the essential cell‐cycle regulatory gene polo are critical to Drosophila development. We show in these studies that Polo protein levels are regulated by the relative usage of two alternative pA sites (pA1 and pA2) present in the polo gene terminal exon. Unexpectedly, we show that pA2 is essential for fly viability whereas the pA1 site is non‐essential. The molecular explanation for this specificity is that polo mRNA utilizing pA2 is translated more efficiently than polo pA1 mRNA thereby generating sufficient protein for cell proliferation. Thus, we show in these studies that deletion of pA2 displays a lethal phenotype (Table I; Figure 2D). Our studies further suggest that levels of Polo protein are auto‐regulated by the selective usage of these two pA sites. Thus, polo overexpression results in a switch in pA site selection from pA2 to pA1 with consequent downregulation of Polo expression (Figure 5).
Since many genes in flies and other eukaryotes possess alternative pA sites in terminal exons, we investigated the possibility that general transcription kinetics might also have a role in alternative polyadenylation. To do this, we employed the well‐characterized C4 fly mutant, which has a 50% decrease in Pol II elongation rate (Greenleaf et al, 1979). Interestingly, we show that this results in increased utilization of the polo pA1 site (Figure 1D and E). In particular, in C4 flies, polo pA1 is used 3.5‐fold more efficiently than in the wild type. We further demonstrate that the effect of Pol II kinetics in alternative pA site selection is likely to be general as analysis of five additional Drosophila genes possessing alternative terminal exon pA sites showed a similar switch to upstream pA site usage in the C4 mutant. This clearly indicates that Pol II kinetics has an important role in alternative polyadenylation site selection. Significantly, human Pol II carrying the equivalent mutation to the Drosophila C4, when transfected into human cells, affects alternative splicing of a FN minigene. Inclusion of the alternative EDI exon in this system increases ∼4‐fold (de la Mata et al, 2003), a similar level to that obtained for alternative polo pA signal selection (Figure 1E). C4 flies also show a difference in Pol II occupancy across the polo gene (Figure 1A). Thus, in wild‐type flies Pol II levels are reduced downstream of pA1, while in C4 flies this pattern is disrupted. Presumably, the slow Pol II enables the pA1 signal on the nascent transcript to be exposed to the polyadenylation machinery for a longer time before Pol II transcribes pA2. Therefore, pA1 will be processed before pA2 is transcribed pointing to a mechanism that relies on the rule of ‘first come, first served’ (Figure 6A). When the Pol II elongation rate is higher, as in wild‐type flies, it transcribes through pA1 and pA2 more efficiently, so that the polyadenylation machinery processes both pA signals on the nascent transcript (Figure 6A). This resembles the mechanism described above for EDI alternative splicing where slow Pol II preferentially includes the alternative EDI exon which is normally excluded, because it allows the machinery time to assemble on the spliceosome (de la Mata et al, 2003, 2010). Our results now indicate that both alternative polyadenylation and alternative splicing depend on Pol II kinetics. In view of recent findings that highlight the importance of alternative pA signal selection (Sandberg et al, 2008; Ji and Tian, 2009; Mayr and Bartel, 2009; Singh et al, 2009), our results now suggest a general molecular mechanism for this process.
The physiological consequences of correct polo pA signal choice in vivo as shown in this study are profound: pA2 is essential for abdominal histoblast proliferation, development of the adult epidermis and viability of the transgenic flies. The lethality and strong abdominal phenotype observed in gfp‐poloΔpA2;polo9 flies are due to the fact that these flies lack polo pA2 mRNA and consequently cannot express sufficient levels of Polo protein (from polo pA1 mRNA) for flies to survive the pupa stage. This phenotype is in agreement with earlier studies showing that polo1/polo2 individuals express low levels of Polo protein (Herrmann et al, 1998) and present an abnormal development of the abdomen (Sunkel and Glover, 1988). We anticipated that the C4 mutant would display a similar phenotype to gfp‐polo ΔpA2 flies since it downregulates pA2 usage. However, we observe a developmental defect similar to the so‐called ‘Ubx effect’ (Greenleaf et al, 1980; Mortin et al, 1988). We note that slow Pol II elongation will impact on many genes so that any phenotype observed is likely to derive from complex genetic effects as previously reported (Greenleaf et al, 1980). We also note that polo pA2 mRNA is still produced at significant levels in the C4 mutant (Figure 1D and E), which may produce sufficient protein for development of the abdomen, as opposed to the ΔpA2 transgenic flies that lack polo pA2 mRNA (Figure 2B) and show a decrease in Polo protein production in the histoblasts (Figure 3B and C).
Two genes in humans (hap and Bzw1) where alternative distal pA site usage results in increased levels of protein production have been previously described (Qu et al, 2002; Yu et al, 2006). In T cells and cancer cell lines, proximal pA site selection in the 3′UTR results in a relief from microRNA repression with the same final effect of an increase in protein production (Sandberg et al, 2008). However, a considerable proportion of genes do not follow this pattern, pointing to the existence of other regulatory elements in the different 3′UTRs (Mayr and Bartel, 2009). Presumably, polo is one such gene as we show that its expression is not regulated by dme‐mir‐8 and dme‐mir‐1016 overexpression (Supplementary Figure S6). Instead regulated polo expression relies on the fact that the choice of pA1 by the transcriptional/processing machinery leads to a decrease in the translation of Polo (Figure 4). This indicates that Polo protein expression is modulated by pA signal selection and by translational control through the 3′UTR, suggesting the presence of regulatory elements in the different 3′UTRs (P Pinto and A Moreira, unpublished results). As Polo is a master regulator of the cell cycle, we predict that the consequences of this type of control will be critical to the cell.
Histoblasts have a very high proliferation rate and go through very rapid cell cycles at the onset of metamorphosis (Madhavan and Schneiderman, 1977; Roseland and Schneiderman, 1979; Madhavan and Madhavan, 1980; Ninov et al, 2007, 2009), therefore would be predicted to be especially sensitive to Polo levels. The lack of polo pA2 mRNA in ΔpA2 flies leads to a reduction in Polo protein levels with the consequent block of histoblast proliferation. This will in turn result in a subsequent failure to correctly develop the adult epidermis. Consistent with our studies, loss of Polo was previously shown to cause G2 arrest (Rehwinkel et al, 2005).
Interestingly, we also show that, in flies where Polo is overexpressed, the shorter mRNA produced by pA1 usage is more abundant than the longer mRNA produced by pA2 selection (Figure 5A and B). This suggests that Polo controls its own expression levels by an auto‐regulatory loop, where higher levels of Polo lead to preferential recognition of pA1. As this mRNA is not efficiently translated (Figure 4C), this will lead to a decrease in the protein levels produced (Figure 6B). It was previously shown that upregulation of su(f) leads to an auto‐regulatory feedback loop through the recognition of a weak intronic pA site (Audibert and Simonelig, 1998). This results in the production of a truncated polypeptide leading to a shut down in gene expression. In the case of polo, a more subtle process is evident, as pA1 selection generates a functional transcript, even though it is not efficiently translated into protein.
Taken together, our results suggest that Pol II kinetics have an important role in pA site selection. Moreover, our results reinforce the view that tight regulation at the level of pA signal selection is necessary for the cell. The importance of precisely choosing the correct pA signal to control cell viability during development is underlined by our studies.
Materials and methods
ΔpA1 transgene: pXB7ΔpA1 was constructed by site directed mutagenesis (adapted from Makarova et al (2000)) using pXB7 (Moutinho‐Santos et al, 1999) as a template. pXB7ΔpA1 was digested with BstAPI and EcoRI and the 1576‐bp fragment was subcloned into pCR‐gfp‐polo to generate the pCR‐gfp‐poloΔpA1 plasmid. pCR‐gfp‐poloΔpA1 was digested with CciNI and XhoI and the 7717‐bp DNA fragment was subcloned into pW8 to generate pW8‐gfp‐poloΔpA1 (named ΔpA1). ΔpA2 transgene: constructed by PCR amplification with the pshA1 genomic and not pA1 oligonucleotides. The PCR product was subcloned into pW8‐gfp‐polo. For luciferase assays, polo 3′UTR (ΔpA1 and ΔpA2) was PCR amplified from the above constructs, subcloned into pCR®II‐TOPO (Invitrogen) and subsequently subcloned into the EcoRI restriction site immediately downstream of the firefly coding region in the pAc5.1C‐F‐Luc plasmid. miRNA coding regions were PCR amplified from genomic DNA (∼400 bp surrounding the mature miRNA) and subcloned into pCR®II‐TOPO (Invitrogen). miRNA reporters were subcloned into KpnI and XhoI restriction sites of pAc5.1A vector. All constructs used in this study were fully sequenced.
Total RNA isolation and northern blot analysis
Total RNA was extracted with Trizol (Invitrogen) according to the manufacturer's protocol. Northern blot was performed as described (Sambrook and Russell, 2001). Quantifications were made by densitometry using a GS‐800 densitometer (Bio‐Rad).
w1118 and RpII215, carrying the RNA Pol II 215 C4 allele, were obtained from the Bloomington Stock Centre. UAS‐polo and polo9/TM6C flies were kindly given by Jean‐Louis Couderc (INSERM, France) and David Glover (University of Cambridge, UK), respectively. For the overexpression studies, MZ1061;gfp‐polo:polo9/TSTL females were crossed with males MZ1061;UAS‐polo:polo9/TSTL and the polo9 homozygous progeny was analysed (Mirouse et al, 2006). The strain MZ1061;gfp‐polo:polo9/TSTL was used as a control. All stocks were grown at 25°C using standard culture conditions and media.
For the gfp‐polo; polo9/TM6B, ΔpA1; polo9/TM6B and ΔpA2; polo9/TM6B crosses were performed to rescue the polo9/polo9 allelic combination. From these crosses, flies homozygous for polo9 were quantified over the total number of flies.
Adult abdomen preparation
Flies were cut between the thorax and the abdomen and all appendages were removed. Abdomens were incubated in a lactic acid:ddH2O (3:1) solution overnight at 60°C, mounted in a fresh solution and incubated overnight at 60°C. The images were acquired with a Leica DM LB microscope (Leica Microsystems) with a 3 CCD color camera vision module (DONPISHA) and treated with the Adobe Photoshop 7.0 Software.
Pupa epidermis preparation
The dissection of 26–27 h pupae was performed according to Gompel and Carroll (2003). The epidermis was fixed in 1.85% (v/v) formaldehyde in fixation buffer (0.1 M PIPES pH 6.9, 1 mM EGTA pH 6.9, 1% (v/v) Triton X‐100 and 1 mM MgCl2) for 10 min at RT. After fixation, the epidermis was washed twice for 5 min at RT in PBT (1 × PBS, 0.03% (v/v) Triton X‐100) and mounted in 6 μl of VectaShield with DAPI (Vector Laboratories). The images were acquired with the AxioImager Z1 microscope (Carl Zeiss), with a b/w CCD camera Axiocam MR ver. 3.0 (Carl Zeiss) and treated using the Adobe Photoshop 7.0 Software. The number of histoblasts present in each nest represents the average number of cells in two pupae of each genotype.
Immunostaining of third instar larvae abdominal histoblasts
The protocol used was adapted from Weaver and White (1995) and Sullivan et al (2000). Third instar larvae were incubated in 1 × PBS at 55°C for 10 s, cut along the dorsal–ventral axis and all internal tissues were carefully removed. The cuticles were then fixed in 100 μl of 3.7% (v/v) formaldehyde in PBT (1 × PBS, 0.1% (v/v) Triton X‐100) for 20 min at RT and washed three times in 100 μl of PBT. Blocking was performed for 1 h, at RT, in 200 μl of PBT with 10% (v/v) FBS. The cuticle was incubated ON at 4°C with 200 μl of α‐Hdc antibody (1:5), kindly provided by Robert White (Weaver and White, 1995), and α‐CycB antibody (Jacobs et al, 1998) (1:3000) in PBT with 10% (v/v) FBS. The following day, the cuticles were washed three times for 10 min in 200 μl of PBT and mounted in 20 μl of VectaShield with DAPI (Vector Laboratories). The images were acquired with an inverted Laser Scanning Confocal Microscope Leica SP2 AOBS SE (Leica Microsystems) and deconvolved and projected onto a single plan using the Huygens Deconvolution Software (SVI). The images were treated using the Adobe Photoshop 7.0 Software.
Quantification of GFP levels in 26–27 h after pupa formation pupae epidermis
The images were acquired using a Laser Scanning Confocal Microscope Leica SP2 AOBS SE (Leica Microsystems) and GFP fluorescence measured using the Leica Confocal Software, version 2.61 Build 1538 LCS Lite (Leica Microsystems). For each genotype, two pupae were analysed. For each sample, one plane was chosen and several GFP measurements were made throughout the epidermis. The GFP levels were quantified using the following formula: mean × area/number of cells, where the mean is the average level of GFP fluorescence measured for each sample.
cDNA synthesis and RT–qPCR
cDNA synthesis was performed using either oligodT or random hexamers with Superscript III following the manufacturer's instructions (Invitrogen). cDNA was then quantified in an iQ5 Bio‐Rad real time PCR machine with iQ™ Sybr® Green Supermix. rp49 was used for normalization in all assays.
Western blotting was performed as previously described (Moutinho‐Santos et al, 1999). For Polo, MA294 antibody was diluted 1:80 and α‐tubulin was detected using mAB DM1A (Sigma) diluted 1:8000. Quantifications were made by densitometry using a GS‐800 densitometer (Bio‐Rad).
mRNA stability assays
Kc cells were treated with actinomycin D (5 μg/ml), total RNA was isolated at designated time points and cDNA synthesis was performed using random hexamer primers with SuperScript III (Invitrogen) following the manufacturer's guidelines, followed by qPCR. mRNA half‐life was calculated from the decay curves.
3′UTR and miRNA functional assays
Kc cells were transfected in 6‐well plates using Effectene transfection reagent (Qiagen) with 100 ng of the Firefly luciferase reporter plasmid (F‐Luc) containing the SV40 late polyadenylation signal (control) or the whole polo 3′UTR fused downstream to the Firefly luciferase reporter, 400 ng of the Renilla luciferase transfection control plasmid (R‐Luc) and 500 ng of the plasmid (pAc5.1/V5‐HisA, Invitrogen) expressing the miRNA primary transcripts. Three days after transfection, Firefly and Renilla activities were measured using a Dual‐Luciferase reporter assay system (Promega). Total RNA was also isolated and northern blot (miRNA assay) or RT–qPCR (3′UTR assay) on oligodT synthesized cDNA was performed with Quantace Sensimix NoRef in a Rotor‐Gene 6000 (Corbett LifeScience) using Firefly‐ and Renilla‐specific primers as described above.
ChIP was performed as previously described (Schwartz et al, 2003). Each value for ChIP experiments was derived from IPs of at least three independent adult fly samples. Extracts were sonicated at 4°C on the medium‐sonication setting of Bioruptor® (Diagenode) with 20 s bursts/30 s cool off, to achieve average fragment sizes of 200–300 bp. For the immunoprecipitations, 4 μl of rabbit anti‐Rpb3 was used (Adelman et al, 2006). Primer sequences are available upon request.
Supplementary data are available at The EMBO Journal Online (http://www.embojournal.org).
Conflict of Interest
The authors declare that they have no conflict of interest.
Supplementary Figure 1
Supplementary Figure 2
Supplementary Figure 3
Supplementary Figure 4
Supplementary Figure 5
Supplementary Figure 6
We are grateful to John Lis (Cornell University, USA) for the α‐Rpb3 Pol II antibody, David Glover (University of Cambridge, UK) for the polo9 allele, Jean‐Louis Couderc for the UAS‐polo flies (INSERM, Clermont‐Ferrand, France), Robert White (University of Cambridge, UK) for the Headcase antibody and Elisa Izaurralde (Tübingen, Germany) for the Luc plasmid. We are grateful to Joel Graber for sharing unpublished data, Fatima Gebauer, Martine Simonelig, Simon Lee and James Castelli‐Gair for critical reading of the manuscript and to Fernando Casares, Enrique Martin‐Blanco and Paulo Pereira for useful discussions. We thank J. Saraiva for technical support and all the members of the laboratories for their valuable suggestions on the work. The Wellcome Trust UK funds the NJP laboratory. AM and CES laboratories are funded by European Regional Development Fund (FEDER), Programa Operacional Ciência, Tecnologia e Inovação 2010 (POCI and POCTI 2010) and the Fundação para a Ciência e Tecnologia (FCT). PABP, TM, MOF and TH are recipients of studentships from FCT‐POCTI. PAC is supported by a postdoctoral grant from FCT.
Author contributions: The overall study was conceived, designed and coordinated by AM, with important contributions from NJP, CES, PABP, TH, PAC and AMC. PABP, TH, MOF, PSW, TM and RD performed the experiments. AM, CES and NJP analysed the data. AM and PABP wrote the paper with substantial contribution from NJP.
- Copyright © 2011 European Molecular Biology Organization