During oogenesis in Drosophila melanogaster, the mother packs her eggs with mRNA and protein products required for early events in embryogenesis. These events include, in order of occurrence, specification of germ cells and somatic cells, establishment of the primary embryonic axes and coordinates, gastrulation and the establishment of the three basic germ layers, initiation of segmentation, and the formation of proneural clusters from which the nervous system arises (Nakao and Campos-Ortega, 1996; Campos-Ortega and Hartenstein,1997; Cabrera,1990; Brennan et al.,1997; Rusconi and Corbin,1998; Morel et al.,2003; LeComte et al.,2006; http://flybase.org). The zygote might also contribute to these events, as zygotic transcription is known to initiate before 1 hr AEL (Liang et al.,2008; de Renzis et al.,2007; Lecuyer et al.,2007) although it becomes the primary source of gene expression only by about 4 hr after egg laying (h AEL) at room temperature (22°C). The last event that is primarily under maternal control, proneural cluster formation, is completed just before 4 hr AEL. Therefore, we will refer to 0–4-hr AEL embryos as early-stage embryos. Lateral inhibition, the process that specifies neuronal and epidermal precursor cells within each proneural cluster, is initiated at about 4 hr AEL. This process and all subsequent embryogenesis events are primarily under the control of zygotic products (Nakao and Campos-Ortega, 1996; Campos-Ortega, and Hartenstein,1997; Cabrera,1990; Brennan et al.,1997; Rusconi and Corbin,1998; LeComte et al.,2006; http://flybase.org). We will refer to embryos older than 4-hr AEL but younger than 16-hr AEL, in which lateral inhibition and parallel or subsequent major events are taking place, as mid-stage embryos. As some maternal products persist well beyond 4-hr AEL, it is possible they also contribute to developmental regulation in mid-stage embryos (Liang et al.2008; de Renzis et al.2007; Lecuyer et al.,2007).
Notch signaling is a basic, evolutionarily conserved developmental pathway. It regulates developmental events based on inter-cellular communication. The critical components of the Notch signaling pathway are the cell surface receptor Notch (N) and its cell surface–anchored ligand Delta (Dl). When Dl that is expressed on one cell binds N that is expressed on the neighboring cell, the N intracellular domain (Nintra) is proteolytically cleaved and translocated to the nucleus where, in association with the transcription factor Suppressor of Hairless, Nintra activates the transcription of target genes such as the Enhancer of split Complex (E(spl)C) genes. Cells with a low level of N signaling and a high level of Dl activity commit to one fate or developmental program (often the default choice) while cells with a high level of N signaling and a low level of Dl activity commit to the alternate fate or developmental program. After commitment, the cells require N signaling for maintenance of their fate or program. This process is repeated in all differentiation events at all stages of development (Shellenbarger and Mohler,1978; Cabrera,1990; Heitzler and Simpson,1991; Fortini et al.,1993; Struhl et al.,1993; Lieber et al.,1993; Rusconi and Corbin,1998; Artavanis-Tsakonas et al.,1999; Mumm and Kopan,2000; Brou et al.,2000; Lieber et al., 2002; Schweisguth,2004; Ahimou et al.,2004; LeComte et al.,2006).
Many genetic screens and studies have identified RNA-binding proteins as important regulators of N signaling (Norga et al.,2003; Kankel et al.,2007; Penn and Schedl,2007; Okabe et al.,2001; Okano et al.,2002). A number of these proteins target mRNA 3′ processing or 3′ UTRs. One example is the Drosophila poly(A) polymerase (PAP) gene hiiragi that is responsible for the addition of poly(A) tail to mRNAs (Juge et al.,2002; Murata et al.,2001). Another example is the hephaestus gene encoding Polypyrimidine Tract Binding Protein (PTBP), which regulates mRNA 3′ processing (Dansereau et al.,2002). In general, proteins and sequences involved in mRNA 3′ processing and 3′ UTR-based regulation determine if or how much mRNA is processed at the nearby poly(A) site and how much protein is produced from the mRNA. A poly(A) site is composed of the consensus hexamer (AAUAAA) or the non-consensus hexamer (such as UAUAAA, AUUAAA, or AAUAUA), the Downstream Sequence Element (DSE), Upstream Sequence Elements (USEs), and the cleavage site. When these sequences emerge behind the Pol II transcription complex, the basic components of mRNA 3′ processing (some of which are associated with the transcription complex) bind their cognate sequences and form the functional complex that cleaves the mRNA at the cleavage site and adds the poly(A) tail to the cleaved mRNA. The poly(A) tail is stabilized and the mRNA exported to the cytoplasm for translation. In general, USEs determine whether the mRNA 3′ processing complex forms at a poly(A) site and the DSE determines the efficiency of cleavage and polyadenylation at this site. USEs continue to play important roles in mRNA stability, nuclear export, cytoplasmic polyadenylation (if involved), and translation efficiency (Wells et al.,1998; Wilusz et al.,2001a,b; Li et al.,2001; Soller and White,2001; Lejeune et al.,2002; Reed and Hurt,2002; Wickens et al.,2002; Kornblihtt et al.,2004; Bentley2005; Beaudoing et al.,2007).
mRNA 3′ processing and polyadenylation is required for most if not all genes. However, genes like hiiragi that encode the poly(A) polymerase show a special relationship to N signaling so much so that mutant alleles even phenocopy classic N-signaling phenotypes (Murata et al.,2001). The roles of mRNA 3′ processing genes recovered in N signaling–based genetic screens are unknown except for mushashi, which is known to repress translation of Numb, a negative regulator of N signaling (Okabe et al.,2001; Okano et al.,2002). Others are known just as negative regulators (for example, hephaestus or sex lethal; Dansereau et al.,2002; Penn and Schedl,2007) or positive regulators (for example, hiiragi; Murata et al.,2001). It is not known whether these genes function independently or as part of a specific regulatory mechanism.
We report here evidence for an mRNA 3′ processing-based mechanism that regulates N signaling at the earliest stages of development. We show that N and Dl mRNAs in early-stage embryos and in mid-stage embryos are processed at alternative poly(A) sites. The resultant shorter or longer mRNAs have different protein-producing potentials, which might explain the pattern of N and Dl protein expression in these embryos. Our data raise the possibility that the manner in which N signaling is activated and regulated changes during embryogenesis and this change is likely effected through 3′ UTR-binding and -processing factors that target alternative poly(A) sites in the N and Dl mRNAs.
Differences in N and Dl Protein Levels Are Correlated With Differences in mRNA sizes
N and Dl mRNAs expressed in 0–4-hr AEL embryos are shorter than their counterparts expressed in 4–12-hr AEL embryos (Fig. 1). The longer N and Dl mRNAs are the predominant mRNAs at all stages from 4- to 16-hr AEL by which time the major embryogenesis events are completed (data not shown). Others have also observed the shorter Dl mRNA in early-stage embryos, as early as ∼1-hr AEL (Vassin et al., 1987; Kopczynski et al., 1988; Alton et al., 1989; Haenlin et al., 1990). The shorter N mRNA is also expressed at very early stages (see Supp. Fig. S1, which is available online; please refer to the Experimental Procedures section for distribution of embryonic stages within each of our samples). We will refer to the longer N mRNA by its FlyBase designation RA, the shorter N mRNA as RB, the longer Dl mRNA as RA1 (which is equivalent to RA in FlyBase), and the shorter Dl mRNA as RA2 (RA2 is not the same as Dl RB in FlyBase, which is much shorter). N RB and Dl RA2 are not annotated in Flybase but evidence for them can be found in transfrags (tiled transcription of Drosophila genomic fragments at different stages of embryogenesis) and the EST data sets (accessible through the FlyBase).
To determine if the difference in sizes of N and Dl mRNAs between 0–4- and 4–12-hr AEL embryos has the potential to affect protein level, we examined levels of N and Dl proteins in these embryos. We found complementary patterns of protein expression. N protein level was lower in 0–4-hr AEL embryos compared to the level in 4–12-hr AEL embryos. On the other hand, Dl protein level was higher in 0–4-hr AEL embryos compared to the level in 4–12-hr AEL embryos. Figure 2A shows N or Dl protein levels in representative embryos of 0–4- and 4–12-hr AEL stages revealed by in situ immuno-labeling. A more elaborate series is shown in Supp. Figure S2. At these stages of embryogenesis, N and Dl proteins are expressed more or less uniformly in all cells and, therefore, signals are also more or less uniformly distributed. The antibodies used are very specific to N or Dl proteins and control embryos that were not treated with primary antibodies were devoid of signals (data not shown). Thus, signals observed in our immuno-labeling experiments were derived from N or Dl protein. To verify our immuno-labeling data, we performed Western blotting analysis. Figure 2B shows N and Dl protein levels in 0–4- and 4–12-hr AEL embryos revealed by Western blotting. They confirmed the complementary protein expression pattern: whereas N expression in 0–4-hr embryos is lower than the level in 4–12-hr AEL embryos, Dl expression in 0–4-hr embryos is higher than the level in 4–12-hr AEL embryos.
The above data raised the possibility that the difference in N and Dl protein levels between early-stage 0–4-hr and mid-stage 4–12-hr AEL embryos is linked to N and Dl mRNAs of different sizes. Dl RA1 and Dl RA2 are thought to differ in the length of their 3′ UTR based on Northern blotting analyses (Vassin et al., 1987; Kopczynski et al., 1988; Alton et al., 1989; Haenlin et al., 1990). The basis for the size difference between N RA and RB was unknown, which we proceeded to determine.
N mRNAs RA and RB Differ in 3′ UTR Length Just as Dl mRNAs RA1 and RA2
We initially determined the basis for the size difference between N RA and RB by probing Northern blots for ∼ 500-bp-long to 2-kilo-base (kb)-long segments all along the length of N RA (the longer mRNA). We used 0–3- and 3–15-hr AEL embryos so that the shorter N RB would be present at a high level in the early stage sample. We also loaded more of 0–3-hr AEL embryonic RNA to help us unambiguously determine a negative result: a probe that does not detect N mRNA RB. Figure 3 shows a portion of our analyses indicating that N RA and RB differ in the length of the 3′ UTR: N RB was detected by probes to the 5′ UTR and the coding region but not by probes to the terminal (term) portion of the 3′ UTR. These data suggested that N RA and RB differed in the length of 3′ UTR, with N RB lacking more than 500 bp of the terminal portion of the 3′ UTR.
We confirmed Northern blot results with RT-PCR. The primers used are shown in Figure 4A. cDNA templates were prepared using embryonic RNA from 0–1-hr AEL embryos (in which N RA was expected to be either absent or present at a low level and N RB was expected to be present at a high level) or 3–6-hr AEL embryos (in which N RB was expected to be either absent or present at a low level and N RA was expected to be present at a high level). First, we examined the level of N RA in the two samples using 3′UTR 5+ and 3′UTR 2− primers that are specific to this mRNA. The expected ∼350-bp fragment amplified to detectable levels at a much earlier cycle with RNA from 3–6-hr AEL embryos compared with RNA from 0–1-hr AEL embryos (Fig. 4B). This result indicates that N RA is present at a high level in later-stage embryos and present at a very low level in early embryos, which is consistent with our Northern blotting analysis. Next, we determined the termini of N RA and RB by performing 3′ RACE using the same RNA samples and a series of primer sets all along the length of the 3′ UTR of N RA. We located the terminus of N RB within the ∼200-bp RT-PCR fragment amplified from the 0–1-hr AEL RNA sample using 3′UTR 4+ and 3′ RACE primers (Fig. 4C, lanes 1–2). The same primers amplified an ∼800-bp fragment from the 3–6-hr AEL RNA sample, which was expected due to the high expression of N RA in this sample (Fig. 4C, lanes 3–4). We sequenced the ∼200- and ∼800-bp fragments and found that N RB ends after a non-consensus poly(A) site hexamer UAUAAA, about 570 bases upstream of the end of N RA near the consensus poly(A) site hexamer AAUAAA. Figure 4A shows the location of the two poly(A) sites and sequence conservation across the 12 sequenced Drosophila genomes (Karolchik et al.,2003; http://genome.cse.ucsc.edu). The actual sequence that is conserved across the two N poly(A) sites is shown in Supp. Figures S3 and S4. We find that the poly(A) sites and a large number of 3′ UTR sequence elements are highly evolutionarily conserved suggesting their functional significance.
We took an identical RT-PCR approach to determine the structures of Dl RA1 and RA2 using the same embryonic RNA samples used in RT-PCR analysis of N mRNAs (shown in Fig. 4B,C). We found that Dl R2 ends after the non-consensus poly(A) site hexamer AAUAUA, about 700 bp upstream of the end of Dl R1 near another non-consensus poly(A) site AUUAAA (Fig. 5A, top portion). To confirm whether we had indeed identified the difference between Dl RA1 and RA2, we performed Northern blotting analysis using probes for the Dl-coding region and the putative Dl mRNA region specific to Dl RA1. Here we used RNA samples that we knew would contain detectable amounts of both Dl RA1 and RA2 in the early-stage embryos, as it would help in clearly verifying the mRNA structures. Results confirmed our analysis (and those of others) that Dl RA1 and RA2 differ in the length of 3′ UTR, with the latter missing about a third in the 3′ most terminal portion (Fig. 5B). Just like N 3′ UTR, Dl 3′ UTR includes highly evolutionarily conserved sequence elements spanning the two poly(A) sites as well as many regions upstream and downstream of these poly(A) sites (Fig. 5A, bottom portion). The actual sequence that is conserved across the two Dl poly(A) sites is shown in Supplementary Figures 5 and 6.
N mRNA RB and Dl mRNA RA1 Have Lower Protein-Producing Potential
Our Northern blotting and RT-PCR data indicated that N mRNA in 0–4-hr AEL embryos is processed at a poly(A) site with a non-consensus hexamer to yield N RB and N mRNA in 4–15-hr AEL embryos is processed at a poly(A) site with the consensus hexamer to yield N RA. Since the consensus hexamer-containing poly(A) site has high affinity for the mRNA 3′-processing complex, N RA was expected to produce more protein than N RB. If that were the case, it could explain the higher level of Notch protein in 4–15-hr AEL embryos compared to the level in 0–4-hr AEL embryos. Since both Dl RA1 and RA2 are processed at poly(A) sites with non-consensus hexamers, there was no a priori expectation that one would produce more protein than the other. However, the higher level of Delta protein in 0–4-hr AEL embryos compared to the level in 4–15-hr AEL embryos suggested that Dl RA2 would produce more protein than Dl RA1.
We directly tested protein-producing potentials of different N and Dl mRNAs in cultured Drosophila S2 cells. These cells are ideal for the purpose because almost all cells in flies express N and Dl and there is strong developmental feedback regulation or cross-regulation by N signaling or other pathways. These regulations would confound interpretation of experimental results. Drosophila-cultured S2 cells do not express endogenous N or Dl but can be made to express these proteins. Such S2-N and S2-Dl cells are excellent model systems for studying the molecular and biochemical aspects of N and Dl function as they reproduce all N signaling aspects observed in flies. S2 cells show minimal, if any, activities of other pathways interacting with N signaling and minimal feedback regulation (Fehon et al.,1990; Klueg et al.,1999; Wesley and Saez,2000; Mishra-Gorur et al.,2002; Wesley and Mok,2003; Ahimou et al.,2004; Bardot et al.,2005, Mok et al.,2005 are but a few examples). The goal of our experiments was to determine the basic activity of N RA relative to N RB and of Dl RA1 relative to Dl RA2. We encountered dominance of the distal N mRNA RA poly(A) site, which was expected due to the presence of the consensus hexamer. To properly assess the protein-producing potential of the proximal N RB poly(A) site (with the non-consensus hexamer), we simply deleted the N RA poly(A) site and USEs specific to it. With Dl, we encountered dominance of the proximal RA2 poly(A) site forcing us to resort to mutagenesis of this site's hexamer for assessment of the distal RA1 poly(A) site.
We generated N constructs that contained the 5′ UTR and the coding sequence with either (1) the complete 3′UTR corresponding to N RA plus ∼100 bp of adjacent 3′ sequence that included the DSE or (2) the 3′ UTR corresponding to N RB plus ∼100 bp of the adjacent downstream sequence. These constructs were expressed using the heat-shock promoter in order to assess expression at different time points after a single induction, which would indicate relative stabilities of mRNA. The constructs were transfected into S2 cells. We extracted DNA from these cells, used them to transfect bacteria, and assessed the number of bacterial colony-forming units. We found the transfection efficiency to be the same with the different constructs (data not shown). Experiments with these transfected S2 cells showed that N RA produces a much higher amount of the full-length N protein than N RB (Fig. 6A, lanes 1–4). This difference was not due to any significant difference in the stability of the two N mRNAs as their levels were more or less comparable (Fig. 6A, lanes 5–8, top). Total protein and RNA used in lanes 1–4 and 5–8, respectively, were extracted from the same population of cells and, therefore, can be directly compared. A probe directed against the terminal 3′ UTR detected only N RA and not N RB, which lacks the sequence, verifying the expression of predicted mRNAs in each lane (Fig. 6A, lanes 5–8, middle). The low amount of N protein in N RB lanes was not due to ligand-independent N signaling–related turnover as the expression of the N-signaling target gene in S2 cells, E(spl)C m3 (Mishra-Gorur et al.,2002), was lower in samples with N RB compared to samples with N RA (Fig. 6A, lanes 5–8, bottom). E(spl)C m3 expression is generally above the background level in N-expressing S2 cells because of the low level of Nintra production even in the absence of ligands; this level increases upon treatment with S2-Dl cells (Mishra-Gorur et al.,2002; Wesley and Mok,2003; Mok et al.,2005). These S2 cell data indicate that N RA has a higher protein-producing potential than N RB, which is consistent with embryo data showing that the level of N protein is high when N RA is predominant and low when N RB is predominant.
For analysis of the relative activities of the Dl mRNAs, we mutated the proximal RA2 poly(A) site AAUAUA to AAGAGA, which was expected to suppress mRNA 3′ processing at this site. We created constructs with the Dl 5′ UTR, the coding region, and the full-length 3′ UTR plus ∼300-bp downstream region containing the DSE that was either (1) wild type or (2) with the mutated proximal RA2 poly(A) site. We cloned these constructs downstream of the heat-shock promoter and performed experiments just as described for N. Northern blot analysis showed that the wild type proximal Dl RA2 poly(A) site is dominant as we barely detected Dl transcripts processed at the distal Dl RA1 poly(A) site (Fig. 6B lane 1). The construct with the mutated proximal RA2 poly(A) site suppressed mRNA 3′processing at this site and promoted processing at the distal RA1 poly(A) site (Fig. 6B, lane 2). The signal from Dl RA2 in lane 1 was comparable to the combined signals from Dl RA1 and RA2 in lane 2. Western blotting analysis showed that the samples in which Dl mRNA was predominantly processed at the proximal RA2 poly(A) site contained a higher level of Dl protein compared with the samples in which the Dl mRNA was processed predominantly at the distal RA1 poly(A) site (Fig. 6B, lanes 3,4). Thus, our S2 cell experiments indicate that the early-stage N RB has a lower protein-producing potential than the mid-stage N RA and that early-stage Dl RA2 has a higher protein-producing potential than the mid-stage Dl RA1.
The Two N and Dl poly(A) Sites Are Bona Fide Alternative Polyadenylation Sites
With N, the weaker poly(A) site is located upstream of the stronger poly(A) site. As a consequence, the weaker poly(A) site would emerge first behind the transcription complex and would present the opportunity for processing before the emergence of the stronger poly(A) site. If the two poly(A) sites are indeed alternate polyadenylation sites, we predicted two transcripts from constructs containing the full-length N 3′ UTR and some downstream sequence (to include the DSE). Furthermore, if the two poly(A) sites are bona fide alternate polyadenylation sites, we predicted that the full-length N 3′ UTR and the adjacent downstream region would be sufficient for directing mRNA processing at the upstream poly(A) site to yield RB-like transcript and the downstream poly(A) site to yield the RA-like transcript.
To test our predictions, we placed the full-length N 3′ UTR and about 100 bases downstream sequence next to a GFP reporter–coding sequence. This GFP-N3′UTR fragment was cloned into a pBluescript vector downstream of the act5C promoter to obtain the actGFP-N3′UTR construct. We sequenced the GFP-N3′UTR part and determined there were no sequence errors. The actGFP-N3′UTR construct was transfected into S2 cells and Northern analysis performed. The same RNA sample was probed for the GFP-coding sequence, the full-length N 3′ UTR, and the N 3′ UTR terminus specific to N RA. Results are shown in Figure 7. Aliquots of total RNA from the same untransfected S2 cells sample were loaded in lanes 1, 4, and 7. Aliquots of total RNA from the same actGFP-N3′UTR sample were loaded in lanes 2, 5, and 8. Aliquots containing a lower amount of total RNA from another actGFP-N3′UTR sample were loaded in lanes 3, 6, and 9. Different amounts of two actGFP-N3′UTR samples are loaded to show not only the robustness of transcript production but also properly resolved bands. Blots from the top of the gel to the bottom are shown to indicate the strength of mRNA 3′ processing. As expected, there is no signal from any of the probes in lanes with untransfected S2 cells RNA (Fig. 7, lanes 1, 4, 7). The GFP probe showed a strong band of the size expected from the transcript ending near the end of Notch RA, ∼2,150 bp, and a weaker band of the size expected from the transcript ending near the end of Notch RB, ∼1,550 bp (Fig. 7, lanes 1–3). As expected, both of these bands were detected by the full-length N 3′UTR probe but only the larger one was detected by the N 3′UTR probe specific to N RA (Fig. 7, lanes 4–9). Additional smeary bands observed with the GFP probe (marked with double asterisk) are products of incomplete extension (terminating before the 3′UTR) commonly observed with the over-expression of any gene in S2 cells or in flies. The faint ∼4-kb signal (marked with an asterisk) derives from the low level of plasmid DNA that is normally recovered from total RNA extractions of transfected S2 cells (we avoided poly(A) selection as it might bias recovery if RA and RB transcripts carry poly(A) tails of different lengths). As is apparent, 3′ processing is robust at poly(A) sites of N RA and RB, as there is hardly any signal below the bands. GFP RB RNA appears diffuse as it migrates close to the ribosomal RNAs. Similar results were obtained in five repetitions of the experiment using three independently transfected constructs. These results indicate that the N 3′ UTR promotes 3′ processing at two positions, near the distal poly(A) site where N RA ends and near the proximal poly(A) site where N RB ends. The higher level of GFP RA might be due to the RA poly(A) site with the consensus hexamer being stronger than the RB poly(A) site with the non-consensus hexamer. There was no indication that GFP RB was less stable than GFP RA.
In the same manner as above, we created an actGFP-DlRA3′UTR construct by placing the Dl 3′ UTR and ∼100-bp downstream sequence next to actin promoter GFP-coding sequence. We performed experiments in the same manner described for actGFP-N3′UTR construct. A high level of GFP RA2 transcript was produced relative to the level of GFP RA2 transcript (Fig. 7B). The high level of GFP RA2 transcript is likely due to the proximal RA2 poly(A) site being dominant. There was no indication that GFP RA1 transcript was less stable than GFP RA2 transcript. These experiments with actGFP-DlRA3′UTR indicate that the information necessary for producing Dl RA1 and RA2 RNAs is encoded within the Dl 3′ UTR and downstream sequences. Thus, the two poly(A) sites in N and Dl genes are bona fide alternate poly(A) sites and the information required for their use is encoded in the 3′ UTR and the downstream sequence. As the proximal poly(A) site is used more efficiently in one instance (Dl) and not in the other (N), apparently it is the poly(A) site itself, and not the order of poly(A) sites within the 3′ UTR, that determines usage efficiency. The consensus or non-consensus hexamers of the two poly(A) sites in N and Dl genes are highly conserved. Of the 48 hexamer sequences (12 species X 2 poly(A) sites X 2 genes), only five have changed. However, in four of these cases there is a gain of non-consensus poly(A) site in the immediate vicinity. The sequences in the vicinity of the hexamers are also highly conserved. This level of evolutionary sequence conservation strongly suggests functional significance for the use of alternate N and Dl poly(A) sites during Drosophila embryogenesis.
High Expression of N mRNA RA Severely Disrupts Early-Stage Embryogenesis
Mutations within the endogenous N or Dl gene that block processing at one of the poly(A) sites or in its regulators would be ideal for determining the functional significance of mRNA processing at alternative poly(A) sites. Unfortunately, these are unknown at this time for both N and Dl genes. With the N gene, there is an indication that obtaining a mutation that blocks processing at the RA poly(A) site might be difficult. We have recently shown that the DSE of the N RA poly(A) site acts as a negative regulator of mRNA 3′ processing and polyadenylation. This site is mutated in the temperature-sensitive Nnd1-dse allele that is wild type at temperatures less than 25°C and mutant at higher temperatures (Shellenbarger and Mohler,1975; Shepherd et al.,2009). Nnd1-dse embryos reared at 18°C for 4-hr AEL and then transferred to 29°C produce a high level of polyadenylated N RA, a high level of Nintra, and manifest severe developmental defects due to excessive N signaling at the lateral inhibition and subsequent stages. Since a mutation in the DSE is unlikely to directly affect N protein cleavage, which produces Nintra, and a high level of polyadenylated mRNA is generally associated with increased translation, we infer that the high level of polyadenylated N mRNA leads to a high level of Notch protein that is rapidly converted to Nintra (Shepherd et al.,2009). Data in this report suggest that the N RA poly(A) site might be under even stronger negative regulation during oogenesis and/or in early-stage embryos as the N RA level is lower in these stages compared to the level in mid-stage embryos. In addition to the difficulties presented by the negative regulation of the N RA poly(A) site, there are difficulties associated with the requirement of Dl for producing eggs and the inability of RNAi to knock out the shorter mRNA without also knocking out the larger mRNA. All these difficulties present significant hurdles to directly address the functional significance of the switch in N or Dl poly(A) site usage between early-stage and mid-stage embryos. However, the Nnd1-dse allele enabled us to ask a more limited question, at least for the N gene: is there a consequence to a high level of N RA expression in early-stage embryos that normally express very low levels of this mRNA. If the answer is no, we could conclude that probably there is not much functional significance to alternative N poly(A) site switching during early-stage and mid-stage embryogenesis.
To address the question, we transferred adult Nnd1-dse female and male flies to 29°C, collected embryos at 29°C, and continued to rear them at 29°C. These embryos produce a higher level of N RA in early-stage embryos compared to the level in wild-type embryos collected and reared in an identical manner (Fig. 8A, lanes 1,2). N RA continues to be expressed at a high level even at later stages (Fig. 8A, lanes 3,4). A probe specific to N RA was used for N blots in lanes 1–4 (Fig. 8). N-coding region-specific probe showed that the over-expressed N mRNA in Nnd1-dse embryos is primarily N RA (Fig. 8A, lanes 5,6). Western blotting analysis using antibodies specific to the ankyrin repeats (Anks Ab; Lieber et al.,1993) and the carboxyl terminus (Cterm Ab; LeComte et al.,2006) of the N-intracellular domain showed that Nintra accumulates to a high level in 0–4-hr AEL Nnd1-dse embryos (Fig. 8B). The Cterm Ab has a higher affinity for Nintra and it clearly reveals the relative levels of Nintra in wild-type and Nnd1-dse embryos. We do not observe a significant accumulation of the full-length N protein, possibly due its rapid conversion to Nintra (Shepherd et al.,2009). Nintra continues to be expressed at high levels even in mid-stage Nnd1-dse embryos (Shepherd et al.,2009).
When we examined developmental events taking place in 0–4-hr AEL Nnd1-dse embryos, we found many of them to be severely disrupted. For example, immuno-staining with the neuronal cell marker Hunchback showed that in Nnd1-dse embryos, neurogenesis is precociously initiated, excessive, and aberrantly patterned (Fig. 8C). This is the stage at which the proneural clusters are just beginning to form in wild-type embryos (Cabrera,1990; Rusconi and Corbin,1998). Soon after, we find developmental mosaics with some regions manifesting excess neuronal cells and other regions manifesting loss (Fig. 8D). Other early-stage processes such as germ-cell specification and gastrulation were also disrupted (data not shown). Over all, early-stage events (taking place in 0–4-hr AEL embryos) were severely disrupted in 57.02% of Nnd1-dse embryos compared with 9.02% of wild-type embryos at 29°C (n = 1,000). In a sizeable fraction of these embryos, it was impossible to even characterize the defects. A sample of such Nnd1-dse embryos and comparable-stage wild-type embryos is shown in Figure 8E.
All early-stage (0–4-hr AEL) mutant phenotypes described above are inferred to be due to the Nnd1-dse allele based on the following observations. One, Nnd1-dse allele was recently re-isolated and the procedure employed would have randomized ∼80% of the genomic (autosomal) background (Shepherd et al.,2009). Two, Nnd1-dse is a temperature-sensitive allele that is wild type at the permissive temperature (<25°C) and mutant at the restrictive temperatures (>27°C). Accordingly, we did not observe early-stage mutant phenotypes when Nnd1-dse parental flies and embryos were reared at the permissive temperature. Three, we did not observe early-stage mutant phenotypes when Nnd1-dse parental flies and embryos under 3-hr AEL were reared at the permissive temperature. When the embryos were shifted to the restrictive temperature after 3-hr AEL, mutant phenotypes related to lateral inhibition and subsequent processes were observed in Nnd1-dse embryos (Shepherd et al.,2009). Four, when Nnd1-dse embryos were reared continuously at the restrictive temperature, the few adult Nnd1-dse flies that emerged displayed all the classic N phenotypes, such as wing notching and thick veins, and none recognizable as belonging to other pathways. While the classic N phenotypes might appear to be due to the loss of N function, they might be due to the opposite N function at an earlier stage, comparable to the mosaic phenotypes observed in embryos (shown in Fig. 8C,D). Mosaic phenotypes appearing to be consistent with both gain and loss of N signaling have also been reported during wing development at larval and pupal stages of Nnd1-dse flies (Royet et al.,1998). Five, the basis for the mutant phenotypes linked to the Nnd1-dse allele has been genetically mapped to the N locus and localized to the 3′ UTR and downstream sequence of the N gene (please see extended discussion of this matter in Shepherd et al.,2009). These five observations strongly support our inference that the phenotypes we observe in early-stage (0–4-hr AEL) embryos are due to the Nnd1-dse allele.
Our data showing that a high level of N RA expression in early-stage embryos is associated with severe disruption of embryogenesis suggests that there might be functional significance to why the N RA poly(A) site is strongly suppressed in early-stage embryos. We emphasize that this is merely a suggestion, as there are caveats regarding the Nnd1-dse allele. As mentioned above, we have inferred that the increased level of N RA in Nnd1-dse embryos leads to the increased level of Nintra based on current knowledge of mRNA 3′ processing and translation. We cannot rule out the effect of Nintra, which is independent of N RA or vice versa. In addition, Nnd1-dse/wt heterozygote embryos are morphologically wild type (data not shown), indicating that Nnd1-dse behaves as a recessive allele when paired with the wild-type allele. We do not know why but it is possible that the wild type allele in Nnd1-dse/+ embryos facilitates compensatory negative regulation as it contains the wild-type DSE required for negative regulation of N RA expression and activity (Shepherd et al.,2009). Although the high level of Nintra in Nnd1-dse embryos (at the restrictive temperature) argues in favor of this allele being a gain-of-function allele when on its own, the early-stage embryogenesis phenotypes are not easily classifiable as gain or loss of function. Thus, the genetic behavior of Nnd1-dse is complex, which makes interpretations difficult. Complex behavior is not unusual for an N allele but Nnd1-dse is different than other well-known complex alleles such as some Abruptex alleles and the N60g11 allele. The latter alleles behave as dominant gain-of-function alleles in the presence of the wild-type N allele but as loss-of-function alleles when homozygous (Xu et al.,1990; Brennan et al.,1997).
Our data show that during Drosophila embryogenesis N and Dl mRNAs are processed at evolutionarily conserved alternative poly(A) sites that contain either the consensus hexamer (N RA) or one of the non-consensus hexamers (N RB, Dl RA1, or Dl RA2). Early-stage embryos (0–4-hr AEL) and mid-stage embryos (>4-hr AEL) prefer one site or the other. N and Dl mRNA processed at the two poly(A) sites have different protein-producing potentials. The early-stage N RB and the mid-stage Dl RA1 produce less protein than the mid-stage N RA and the early-stage Dl RA2, respectively. The expression pattern of these mRNAs and their protein-producing potentials could explain the difference in N and Dl protein levels between early-stage (0–4-hr AEL) and mid-stage (>4-hr AEL) embryos.
The core components of N signaling such as N and Dl are both tightly and finely regulated. Many mechanisms have been previously described that function at the level of the protein or transcription (Heitzler and Simpson,1991; Kooh et al.,1993; Irvine and Weischaus,1994; Doherty et al.,1996; Heitzler et al.,1996; de Celis et al.,1997; Huppert et al.,1997; Qi et al.,1999; Rothwell et al.,1999; zur Lage and Jarman,1999; Bruckner et al.,2000; Moloney et al.,2000; Parks et al.,2000; Culi et al.,2001; Lai et al.,2001; Pavlopoulos et al.2001; Mishra-Gorur et al.,2002; Panin et al.,2002; Tsuda et al.,2002; Bland et al.,2003; Ikeuchi and Sisodia,2003; La Voie and Selkoe,2003; Morel et al.,2003; Six et al.,2003; Wesley and Mok,2003; Sakata et al.,2004; Wang and Struhl,2004,2005; Wilkin et al.,2004; zur Lage et al.,2004; Emery et al.,2005; Le Borgne et al.,2005; Sapir et al.,2005; LeComte et al.,2006). Our data indicate that N signaling might also be regulated at a more primary level, the level of mRNA 3′ processing (at alternative poly(A) sites). This regulation could be important for genes such as N without a strong promoter (the N promoter remains undefined despite efforts by many labs). Regulation at the level of mRNA 3′ processing might be the reason why many RNA-binding proteins are recovered in genetic screens based on N signaling (Norga et al.,2003; Kankel et al.,2007). Among these proteins, only Sex Lethal (Sxl) is known to affect N mRNA translation, in a sex-specific manner (Penn and Schedl,2007). However, Sxl is unlikely to be involved in the alternative polyadenylation phenomenon described in this report as most (if not all) embryos produce N RB in the early-stage embryos and N RA in the mid-stage-embryos. If alternative polyadenylation were sex specific, we would have observed a 50:50 ratio between N RA and N RB or between Dl RA1 and Dl RA2. Our data provide a useful framework for examining the functions of the other RNA-binding proteins recovered in genetic screens based on N signaling.
It has been known for a long time that in early-stage and mid-stage embryos, the mRNA products of a number of genes differ in the length of their 3′ UTR, Dl being one of them. Over the years, for one reason or the other, we have examined more than 100 genes related to our studies by Northern blotting. Almost a third of them show transcripts in the early-stage and mid-stage embryos that differ in the length of their 3′ UTR. It is possible that many of these genes also manifest the N/Dl phenomenon we have described in this report. N and Dl are both required for N signaling, which is known to regulate embryogenesis events in both early-stage and mid-stage embryos. Our data suggest that the pattern of protein-producing capacities of N mRNAs would be complementary to those of Dl mRNAs, which raises some interesting possible scenarios for the regulation of embryogenesis. As the Dl mRNA produces more protein in early-stage embryos (0–4-hr AEL), it is possible that in these embryos the Dl protein is the primary target of fine-scale protein level regulation. On the other hand, the N protein might be the primary target in mid-stage embryos (after 4-hr AEL) where the N mRNA produces more protein. Our Nnd1-dse data suggest that N might not be subjected to fine-scale regulation when its level is high. Therefore, the opposite scenario is also possible wherein the fine-scale regulation requires a low level of the protein. Another possible scenario is that processing at the alternative poly(A) sites of a gene might function to adjust protein levels based on the amounts of mRNA contributed by the mother and the zygote. Recently, it has been shown that for a large fraction of genes expressed in early-stage embryos both the mother and the zygote contribute mRNA (e.g., de Renzis et al.,2007). All these possible scenarios might depend on the presence or absence of specific mRNA 3′-processing regulatory factors in early-stage and mid-stage embryos.
Sequence elements in the immediate neighborhood of a poly(A) site are necessary and sufficient to direct mRNA 3′-processing and polyadenylation. However, these processes can be regulated by sequences that are far away or events that took place early on. For example, some of the proteins that are part of the mRNA 3′-processing complex are thought to associate with the Pol II transcription complex soon after transcription is initiated (Wells et al.,1998). Thus, in vivo analysis of the regulation N or Dl expression through mRNA 3′ processing could be quite challenging because this regulation might not be effective with exogenous promoters. In vivo regulation might also depend on the interaction between mRNA 3′-processing regulators present in the early-stage and mid-stage embryos about which we know very little. Our data would provide an excellent framework for future studies aimed at systematically analyzing all the conserved N and Dl 3′ sequences, identifying RNA-binding factors involved in the regulation of mRNA 3′ processing at alternative poly(A) sites in the N and Dl genes, and determining the basis for the different translation efficiencies of the early-stage and zygotic stage N and Dl mRNAs. N RA, N RB, Dl RA1, and Dl RA2 mRNAs expressed predominantly in either the early-stage or the mid-stage embryos would serve as excellent readouts.
Fly stocks were obtained from the Drosophila Stock Center, Bloomington, Indiana. Nnd1-dse stock is described in Shepherd et al. (2009). Procedures described in Ashburner et al. (2004), Campos-Ortega and Hartenstein. (1997), and Weischaus and Nusslein-Volhard (1998) were followed for rearing flies, collecting timed embryos, and processing embryos for molecular analyses. All fly stocks were entrained to the circadian cycle so that they lay most of their eggs within a 4-hr period around dusk when reared at 22°C (Ashburner et al.,2004). This procedure minimizes age variation within embryo samples. Experimental samples were collected at 22°C (all experiments except the Nnd1-dse experiments) or at 29°C (the Nnd1-dse experiments). As an environmental chamber was used for rearing experimental stocks, temperature could be adjusted independent of the circadian cycle (the circadian clock is temperature compensated and, therefore, not affected by changes in temperature). At 22°C, the mother primarily control embryogenesis in 0–4-hr AEL and the zygote primarily controls embryogenesis at the later stages (after 4-hr AEL). Nnd1-dse stock was maintained at 18°C. Nnd1-dse and the control (yw) cages were maintained at 29°C for embryo collections. In these experiments, the times were adjusted to reflect the faster developmental rate at 29°C (Ashburner et al.,2004). However, for presentation of results the times are converted back to reflect developmental rate at 22°C for easy comparison with the other data.
In the following description, all times correspond to development at 22°C. Samples were collected at 1-, 2-, 3-, or 4-hr intervals (depending on the experiment) and the majority of embryos in these samples represented the middle 45, 90, 150, or 180 min, respectively. Following collection, embryos were aged to required stages. For obtaining the 4–12-hr AEL sample, we combined RNA from a 0–4-hr collection that was aged for 4 hr (i.e., the 4–8-hr sample) and RNA from a 0–4-hr sample that was aged for 8 hr (i.e., the 8–12-hr sample). For obtaining the 3–15-hr AEL sample, we combined RNAs from 0–3-hr collections that were aged for 3 hr (i.e., the 3–6-hr sample), 6 hr (i.e., the 6–9-hr sample), 9 hr (i.e., the 9–12-hr sample), or 12 hr (i.e., the 12–15-hr sample) were combined. For obtaining the 3–6-hr sample, 0–3-hr collection was aged for 3 hr. For obtaining the 2–4- and the 4–6-hr sample, 0–2-hr collections were aged for 2 or 4 hr, respectively.
The procedures followed for immunostaining, Northern blotting, Western blotting, collection of different embryonic developmental stages, and cell culture experiments are described in Lieber et al. (1992,1993), Wesley and Mok (2003), Bardot et al. (2005), Mok et al. (2005), and LeComte et al. (2006). N mRNAs are around 10 kb long, Dl mRNAs are about 5.5 kb long, and rp49 mRNAs are about 600 bp long. Furthermore, the two ∼10-kb-long N mRNAs migrate close together under a wide range of gel electrophoresis conditions. Therefore, N or Dl Northern blots are based on a long-run electrophoresis gel (21 hr at 35 V) that resolves the larger N mRNAs fairly well and rp49 Northern blots are based on a short-run electrophoresis gel (14 hr at 35 V). α466 N antibody (LeComte et al.,2006), C594.9B Dl antibody (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, Iowa), anti-Hunchback antibody (a generous gift from Paul Macdonald), and actin antibody (Abcam, Cambridge, MA) were used in immunostaining and signals developed with a Horseradish peroxidase (HRP) or Alkaline phosphatase (AP) conjugated secondary antibody. αNI N antibody (Anks Ab; Lieber et al.,1993), α466 N antibody (Cterm Ab), C594.9B Dl antibody, and anti-Hsp 70 antibody (Sigma, St. Louis, MO) were used in Western blotting.
First Choice RLM-RACE Kit (Ambion, Austin, TX) was used for RT-PCR. Standard procedures were used for other PCRs and cloning. The primers used were the 3′ RACE primer in the kit, 3′UTR5+ (5′caccaatggaaacgtataagtc3′), 3′UTR2− (5′agtttcgttttgctgtctggc3′), 3′UTR4+ (5′atcgattaaacgtttgtgggac3′). For making the actGFP-N3′UTR constructs, the KpnI-NotI GFP coding sequence fragment from pEGFP (Clontech, Palo Alto, CA) was inserted after the actin 5C promoter in the pBluescript (pBS) plasmid (a pBS plasmid with the ∼2.7-kb actin promoter EcoRI fragment was a generous gift from Simon Kidd). The Notch 3′UTR sequence, amplified with primers containing NotI sites, was cloned into the NotI site at the end of the GFP-coding sequence in pEGFP. Plasmids with the correct orientation were checked by sequencing and used to transfect S2 cells. For making the hs-N3′UTR RA and RB plasmids, an XbaI fragment containing the full-length Notch cDNA was cloned into the pBS ks− plasmid. A larger Not1-StuI fragment containing the Notch XbaI fragment plus an additional Spe I site downstream of the 3′ XbaI site was excised and cloned into NotI-StuI sites of a pCaSpeR-hs vector without the hsp70 3′UTR-containing-Bam HI-fragment (ΔBam) to obtain the Notch RNA RA-producing construct (hsN3′UTR RA). The SacII–SpeI fragment containing the Notch 3′UTR sequence from the natural Sac II site immediately after the Notch stop codon to the terminus of Notch RNA RB was PCR amplified, checked by sequencing, and used to replace the full-length Notch 3′ UTR region in hsN3′UTR RA to obtain the construct producing the Notch RNARB (hsN3′UTR ′RB′).
For mutating the upstream poly(A) hexamer of Dl, we used the QuickChange Site-Directed Mutagenesis Kit (Stratagene, La Jolla, CA) with a high-fidelity pfu turbo polymerase enzyme. A Dl cDNA clone in pBluescript was used as the template. Multiple clones were isolated, mutagenesis checked by sequencing, and cloned into the heat-shock pCaSpeR 3 vector. For generating for the actGFP-Dl3′UTR construct, we PCR amplified Dl 3′ UTR and 150-bp downstream sequence using primers with Not 1 sequence. This Not1 fragment was used to replace the N 3′ UTR in actGFP-N3′UTR construct. Correct orientation was ascertained using Kp1 and Nhe1 enzymes. Dl 3′ UTR sequence was checked for errors.
Images and figures were processed using the Photoshop (Adobe, San Jose, CA) and Canvas (Deneba) programs. Any adjustment was applied to whole images.
We thank Jingping Li, Lee-Peng Mok, and Boris Bardot for help in the initial stages of the study; the Drosophila Stock Center and Dr. Spyros Artavanis-Tsakonas (Harvard University) for fly stocks; Vermont Cancer Center for sequencing; Matt Rand and Gregory Gilmartin for comments on the initial versions of the manuscript; NIH (NINDS) grant NS43122, Department of Microbiology and Molecular Genetics at the University of Vermont (thanks to Susan Wallace), Neuroscience COBRE (P20 RR016435) at the University of Vermont (thanks to Rodney Parsons and Cindy Forehand), and the University of Vermont College of Medicine Bridge grant for funds (thanks to Wolfgang Dostmann and Markus Thali). We also thank anonymous reviewers for their helpful comments that have improved the quality of our manuscript.