Smicl is required for phosphorylation of RNA polymerase II and affects 3′-end processing of RNA at the midblastula transition in Xenopus

Smicl is a protein that interacts with transcription factors and forms a complex with the Cleavage and polyadenylation specificity factor (CPSF)(Collart et al., 2005a). We have previously shown that Smicl regulates the expression of Chordinin the early Xenopus embryo, but Chordin is unlikely to be the only target of Smicl, because ectopic expression of Chordincannot rescue completely the effects of an antisense morpholino oligonucleotide (MO) directed against Smicl(Collart et al., 2005b). In this paper we use microarray analysis to search for additional targets of Smicl, both in an effort to understand its role in early development as well as to obtain more insights into its mode of action. This analysis has identified many genes whose onset of expression at the midblastula transition(Newport and Kirschner, 1982a; Newport and Kirschner, 1982b; Schier, 2007) requires Smicl activity and whose activation is correlated with the translocation of a tagged form of Smicl from cytoplasm to nucleus. Analysis of one of these targets, Xiro1, shows that Smicl influences 3′-end processing of the Xiro1 transcript.

The effects of Smicl on zygotic gene expression are first detected at the midblastula transition (MBT), which in Xenopus occurs after 12 cell divisions and involves many coordinated changes in cell behaviour. In particular, the cell cycle loses its synchrony and, with the introduction of G1 and G2 phases, becomes longer (Kimelman et al., 1987; Lemaitre et al.,1998). In addition, several proteins translocate from cytoplasm to nucleus (Dreyer, 1987),maternal mRNA is degraded (Bushati et al.,2008; Giraldez et al.,2006), the apoptosis programme is engaged(Peng et al., 2007), cells become motile (Kimelman et al.,1987) and karyomeres (individual chromosomes surrounded by a nuclear envelope) disappear (Lemaitre et al., 1998). Perhaps most significantly, the MBT marks the onset of bulk transcription in the embryo, driven by RNA polymerase II (RNA polII)(Newport and Kirschner, 1982a; Newport and Kirschner,1982b).

RNA polII consists of two large subunits and several smaller subunits. The largest subunit, Rpb1, contains an unusual C-terminal domain (CTD) that contains repeats of the heptapeptide YSPTSPS, which can be phosphorylated on serines 2 and 5 (Hirose and Ohkuma,2007). The CTD acts as a docking platform for factors involved in co-transcriptional processes, and the positions of phosphorylated serines determine which transcription factors and RNA processing factors are recruited(Egloff and Murphy, 2008). Importantly, before the MBT in Xenopus, serine 2 of the CTD heptapeptide is not phosphorylated, but it becomes hyperphosphorylated thereafter (Palancade et al.,2001). Serine 2 phosphorylation is important for overcoming an early elongation block in Xenopus(Peterlin and Price, 2006) as well as for splicing and polyadenylation(Hirose and Manley, 1998; Hirose et al., 1999).

CPSF is required for pausing and termination of transcription, and it does this by binding through its 30 kDa subunit (CPSF30) to the body of Rbp1(Nag et al., 2007). Because Smicl affects 3′-end processing and has a zinc finger domain similar to that of CPSF30, we asked whether it could form a complex with Rpb1. Our work shows that Smicl interacts with Rpb1 and is required for phosphorylation of the Rpb1 CTD, at least between MBT and mid-gastrula stages. These observations link the onset of zygotic gene expression in the Xenopus embryo with the translocation of Smicl from cytoplasm to nucleus, the phosphorylation of Rpb1 and the 3′-end processing of newly transcribed mRNAs.

The open reading frame (ORF) of Venus was cloned 5′ of Smicl-HA in pCS2 (Collart et al., 2005b)using restriction enzymes ClaI and EcoRI, to generate a full-length Smicl construct with N-terminal Venus and C-terminal HA tags. Smicl-HA in pCS2 was also digested using restriction enzymes ClaI and SacI to generate a similarly tagged Smicl construct containing just the last 338 amino acids of the ORF (which include the zinc fingers). FLAG-Rpb1 and FLAG-Rpb1ΔCTD constructs were as described(Rosonina and Blencowe,2004).

Embryos of Xenopus laevis were obtained by artificial fertilisation. They were maintained in 10% normal amphibian medium (NAM)(Slack, 1984) and staged as described (Nieuwkoop and Faber,1975). Embryos were injected at the one-cell stage with 80 ng Smicl antisense morpholino oligonucleotide (Smicl MO) or control morpholino(CoMO) (Collart et al., 2005b),C-terminally HA-tagged XtSmicl in pCS2(Collart et al., 2005b) was linearised with Asp718 and sense RNA was generated with SP6 RNA polymerase. RNA injections were performed in embryos at the one-cell stage. Use of a fluorescein-labelled Smicl MO confirmed that injected MOs distribute evenly within the embryo (data not shown).

Microarray analysis was carried out essentially as described, using a microarray that represents about one-third of the genes expressed during early development (Chalmers et al.,2005; Ramis et al.,2007). Microarrays were scanned using an Axon 4000B scanner and GenePix Pro software (Axon). The microarray results were imported into Acquity analysis software (Axon) and normalised using Lowess normalisation. Data files were then created as described (Ramis et al., 2007). The complete datasets were deposited in the NCBI Gene Expression Omnibus (GEO) data repository(http://www.ncbi.nlm.nih.gov/geo/),with Accession number GSE4952.

Differential expression was validated by real-time RT-PCR using the LightCycler 480 (Roche). Reverse transcription was carried out using Transcriptor First Strand cDNA Synthesis Kit (Roche) followed by real-time PCR using the LightCycler 480 SYBR Green I Master kit (Roche) following the manufacturer's instructions. Primers specific for Ornithine decarboxylase (ODC) and Chordin were as described(Piepenburg et al., 2004). Primers to amplify Xiro1 intron sequences are: forward,5′-CAGCTAAGTTCAGCCCAAGG-3′; reverse,5′-GCGTTTATCGGACAACGATT-3′. Primers to amplify Chordinintron sequences are: forward, 5′-CACTGTTGAAGCCAAGCAAA-3′;reverse, 5′-GAGGCTGCATTGCTCTTCTC-3′. Details of other primers are provided in Table 1. The normalised target concentrations (in arbitrary units) were calculated from the real-time PCR efficiencies and the crossing point of the target and of the reference gene ODC, as described(Pfaffl, 2001). Correlations between microarray and real-time RT-PCR results were performed using the Windows SPSS version 11.5 software (SPSS, Chicago, USA).

An aliquot of 4 μg of total RNA was ligated for 30 minutes at 37°C with 0.4 μg of a 3′ amino 5′ phosphorylated oligonucleotide P1(5′-P-GGTCACCTTGATCTGAAGC-NH₂-3′) in a volume of 10μl using T4 RNA ligase (New England Biolabs). The reaction was placed at 70°C for 15 minutes. The whole 10 μl ligation reaction was used in a 50μl reverse transcription reaction using Superscript III (Invitrogen),according to the manufacturer's directions using 0.4 μg primer P2(5′-GCTTCAGATCAAGGTGACCTTTTT-3′) as described(Graindorge et al., 2006). Of the resulting cDNA, 1 μl was used in a 50 μl PCR reaction. As a reverse primer, P2 was used in all reactions. As forward primers, we used P3, specific for our positive control Eg3:5′-AAGTGACTATGCAATTTGAGCTAGAAGTAT-3′(Graindorge et al., 2006), and designed primers specific for MCM10 (P4:5′-ACAGCATTGCAGAATCATGG-3′), FBXO43 (P5:5′-TCTTGCACCTGATGTTTGTG-3′) and Xiro1 (P6:5′-CCGTGTTCCATTTCAGACCT-3′). The PCR reaction mixture contained 1X buffer (Invitrogen), 0.2 μM of each primer (forward and reverse), 200 μM dNTPs, 1U of Platinum Taq Polymerase (Invitrogen) and 1.5 mM MgCl₂,in the presence (P4) or absence (P3, P5, P6) of 10% DMSO. The amplification programme consisted of a preincubation step for denaturation of the template cDNA (95°C for 5 minutes), followed by either 40 (P3), 45 (P4) or 50 (P5,P6) cycles consisting of a denaturation step (95°C for 30 seconds), an annealing step (56°C for 30 seconds for P3, P4, P5 or 59°C for 30 seconds for P6) and an extension step (72°C for 30 seconds), and then one final extension step (72°C for 7 minutes). The amplified products were digested with PvuII (P3), BstUI (P6) or PsiI (P4 and P5) to verify the specificity of the PCR products. PCR reactions and digests were separated on a 12% non-denaturing polyacry-lamide gel and stained with ethidium bromide.

Transiently transfected HEK293T cells were solubilised in lysis buffer containing 1% NP40, 150 mM NaCl, 20 mM Tris pH 7.5, 2 mM EDTA, 50 mM NaF, 1 mM sodium pyrophosphate, supplemented with protease inhibitors (Roche Molecular Biochemicals). Cell lysates were cleared by centrifugation and precipitations were performed by overnight incubation at 4°C with anti FLAG M2 agarose affinity gel (Sigma). Unbound proteins were removed by washing four times with cold lysis buffer. Bound proteins were harvested by boiling in sample buffer,and they were resolved by SDS-polyacrylamide gel electrophoresis. Flag-tagged and HA-tagged proteins were visualised after western blotting using rat monoclonal anti-HA-peroxidase-coupled high-affinity antibody (3F10) (Roche)and goat polyclonal anti-FLAG-peroxidase coupled antibody (Abcam) and the SuperSignal West Dura Extended Duration Substrate kit from Thermo Scientific(Pierce).

Embryos were fixed overnight in MEMFA (3.7% formaldehyde, 100 mM MOPS, 2 mM EGTA, 1 mM MgSO₄, pH 7.4) at 4°C, and the vitelline membrane was removed. After gradually dehydrating and rehydrating the embryos in methanol, they were washed in phosphate buffered saline (PBS) and bleached in 2 ×SSC with 2% formamide and 1.5% H₂O₂. Embryos were washed 2 × 10 minutes in PBS and 2 × 30 minutes in PBS with 0.1% BSA and 0.2% Triton X-100 (PBSbt) at room temperature. They were blocked for 1 hour at room temperature in PBSbt with 10% serum (PBSbts) and incubated overnight at 4°C in a 1/250 dilution of the anti-HA-peroxidase-coupled rat monoclonal antibody (Roche) in PBSbts. Embryos were washed for 1 hour in PBSbts, 4 × 1 hour in PBSbt and 10 minutes in PBS and incubated in Tris-buffered saline (TBS) with 0.066% 3,3′-Diaminobenzidinetetrahydrochloride (DAB) (SIGMA) for 10 minutes. H₂O₂ was added to a final concentration of 0.024% and staining was observed after 1 minute.

To detect Rpb1 and phospho S2 Rpb1, we used mouse monoclonal antibody 8WG16(ab817) and a rabbit polyclonal antibody (ab5095) from Abcam.

Smicl maintains expression of Chordin in the early gastrula of Xenopus laevis (Collart et al.,2005b), but it is unlikely that this gene is the only embryonic target of Smicl, because injection of Chordin mRNA cannot completely rescue the phenotype of embryos injected with a Smicl antisense MO(Collart et al., 2005b). Further insight into the role and mode of action of Smicl requires the identification of additional targets, and to this end we performed microarray analysis on RNA derived from control embryos at the early gastrula stage and from embryos injected with Smicl antisense MOs. The microarray slides used in our experiments were designed from transcript sequences derived from a large-scale Xenopus tropicalis EST project(Gilchrist et al., 2004). These X. tropicalis microarrays also work in X. laevis(Chalmers et al., 2005; Ramis et al., 2007).

Xenopus laevis embryos from three different spawnings were injected at the one-cell stage with either Smicl MO, which inhibits Smicl function, or the control CoMO (Collart et al., 2005b) and were cultured to the early gastrula stage before RNA isolation. Some embryos were cultured to later stages to confirm that their development was impaired as described previously(Collart et al., 2005b). RNA from each spawning was hybridised with dye-swapped technical replicates,making six microarray slides in total. Oligonucleotides were considered to be differentially expressed when: (1) they showed at least a twofold difference(sample versus control) in expression levels in four out of the six microarrays; and (2) they were significantly different (q=0). In embryos in which Smicl was downregulated, 95 oligonucleotides fulfilled these criteria(Table 2): 33 genes were upregulated and 62 were downregulated.

The X. laevis homologues of the X. tropicalis cDNAs recognised by the oligonucleotides were identified by BLAST searches(Table 2). Genes were manually classified according to the annotation of their human homologues (NCBI databases, http://www.ncbi.nih.gov/),because gene ontology annotation for Xenopus species is not yet available (Fig. 1A). Apart from the unknowns, the largest group is involved in regulation of transcription(20%).

Microarray results were validated by real-time RT-PCR using the same samples that were used for microarray experiments. Primers were designed for the X. laevis homologues of all differentially expressed transcripts,with the exception of nine cDNAs for which X. laevis homologues could not be identified (Table 2). Of the genes tested, 76% (66) were confirmed as being differentially expressed(Table 2). Bilateral correlation analysis of the results obtained by microarray hybridisation and those obtained by real-time RT-PCR showed a Pearson correlation of 0.909(P<0.0005) (Fig. 1B).

Smicl is expressed both maternally and zygotically during Xenopus development (Collart et al., 2005b). Direct targets of Smicl are likely to be among the first to be up- or downregulated by loss of Smicl function. To find such targets, we studied the temporal expression patterns of 78 of the genes identified in our microarray analysis in control embryos and in embryos injected with a Smicl MO. We found that the genes could be divided into four categories, an example of each of which is shown in Fig. 2; for the complete results, see Fig. S1 in the supplementary material. In category 1 (41 genes),normal embryos exhibited low maternal mRNA levels that increased at the onset of zygotic transcription (between stages 7 and 8.5). Loss of Smicl function caused a reduction in zygotic transcripts(Fig. 2A,E). In category 2 (19 genes), maternal transcript levels normally decreased after the MBT, but the degradation of these gene products was delayed in Smicl-depleted embryos(Fig. 2B). Category 3 (four genes) comprise genes with expression levels in Smicl-depleted embryos that were first upregulated and later downregulated compared with control embryos(Fig. 2C). And category 4 (14 genes) consists of the genes with RNA expression levels in this experiment that proved to be only slightly affected by depletion of Smicl(Fig. 2D). We do not yet understand the transient upregulation observed in the four category 3 genes,which include Chordin. One possibility is that loss of Smicl causes the downregulation of a repressor of Chordin expression.

The expression profiles of the genes in our categories 1 and 2 indicate that absence of Smicl affects the time of onset of zygotic gene expression and of the degradation of maternal mRNAs. These events usually occur at the MBT,and, interestingly, we observed that an HA-tagged form of Smicl translocated from the cytoplasm to the nucleus at precisely this time(Fig. 3A-F). Similar behaviour was observed in dissociated animal pole cells using a Venus-tagged form of Smicl and confocal microscopy: at the MBT the distribution of Smicl changed from diffuse and cortical to well defined and nuclear(Fig. 3G,H). A similar transition was observed in a construct consisting of just the Smicl zinc finger domain (Fig. 3I).

To ask whether the nuclear translocation of Smicl at the MBT requires new transcription, we injected embryos with α-amanitin as well as with RNA encoding HA-tagged Smicl. Such embryos divided normally to stage 9, but the transcription of GS17 and Siamois was prevented, confirming that transcription was blocked (Fig. 3J). This inhibition of transcription had no effect on the nuclear translocation of Smicl after the MBT (Fig. 3K,L). This is discussed below.

Our previous work in tissue culture identified Smicl as a molecule that may be involved in 3′-end processing(Collart et al., 2005a). To address this point we studied three Smicl targets identified in our microarray experiments. We chose Xiro1, a category 1 gene that is among the earliest zygotically transcribed genes and is therefore likely to be a direct target (Fig. 2A,E), and two category 2 genes, FBXO43 and MCM10, with regulation that is more likely to be indirect. To investigate lengths of poly (A) tails we used an RNA ligation assay (Graindorge et al.,2006) with minor modifications(Fig. 4A). The Eg3gene, with RNA that undergoes deadenylation between fertilisation and the 64-cell stage (Graindorge et al.,2006), was used as a positive control. Fig. 4B confirms that the assay could detect the deadenylation of Eg3 mRNA during early cleavage stages, and Fig. 4C shows that overexpression of Smicl lengthened the Xiro1 mRNA poly(A) tail whereas depletion of Smicl, by injection of an antisense MO, shortened it. To facilitate comparison of the PCR products produced in the polyadenylation assay, the assay was performed at the very onset of transcription of Xiro1, when transcript levels in the three different samples are still similar. This activation of Xiro1 occurred slightly earlier than that of GS17, the onset of expression of which in our experiments defined the MBT. Smicl had no effect on the lengths of the poly(A)tails of FBXO43 or MCM10. These experiments show that Smicl regulates 3′-end processing of Xiro1 mRNA.

To ask whether the early reduction in transcript levels caused by loss of Smicl is associated with this decrease in polyadenylation, perhaps by affecting the stability of the mature gene product, we studied transcript levels of Xiro1 using primers specific for the mature mRNA and for the unprocessed mRNA. We have previously shown that levels of Chordinare reduced in Smicl morpholino-injected embryos and we therefore used Chordin as a positive control.

Our results show that transcript levels of Xiro1 as well as Chordin were downregulated in embryos injected with Smicl MO and,significantly, that this is true for unprocessed as well as processed transcripts at the early gastrula stage(Fig. 5A). These observations suggest that downregulation also occurs at the level of transcription, and that the lower levels of Xiro1 mRNA observed are not solely a consequence of RNA instability resulting from a shortening of its poly(A)tail.

The same assay was used to analyse levels of Xiro1 and Chordin after overexpression of Smicl. Although Smicl causes upregulation of Chordin (Collart et al., 2005b), it had no effect on levels of Xiro1 at the early gastrula stage (Fig. 5B). Xiro1 proves to not be unusual in this respect: of all the Smicl target genes, shown in Fig. S1 in the supplementary material,only Chordin and FoxI1 mRNA levels were upregulated in response to Smicl (data not shown).

We have identified 41 category 1 targets, all of which require Smicl for the onset of zygotic gene expression. To ask whether all 41 are regulated by a common mechanism, we first asked whether Smicl is brought to their promoters by a common transcription factor. To this end we analysed sequences 500 base pairs 5′ of the transcription start sites of the genes and searched for motifs from the JASPAR CORE database(http://jaspar.genereg.net/). We found the best match for each motif in each sequence using the nmscan tool from the NestedMICA suite (Down and Hubbard, 2005) and then compared the distribution of scores for each motif to that from 1000 randomly selected Xenopus genes. In this way we found that the NF-Y motif (core sequence CCAAT) was significantly over-represented (P=0.007: an empirical figure based on re-sampling of scores from the random genes). This sequence is the binding site for the transcription factors NF-Y and CBTF in early frog embryos. NF-Y consists of three subunits and is the predominant Y-box binding protein in Xenopus oocyte nuclei (Li et al.,1998). CBTF is the main CCAAT binding factor at midblastula stages(Ovsenek et al., 1991). It is expressed maternally, and its activity is regulated by its p122 subunit(CBTF^p122). This protein is perinuclear during early embryogenesis,but moves from cytoplasm to nucleus at stage 9, before the detection of CBTF activity in the nucleus (Orford et al.,1998). These observations suggest that Smicl might regulate gene expression at the MBT by interacting with NF-Y or CBTF^p122, but our co-immunoprecipitation experiments have revealed no such interaction under our experimental conditions (data not shown).

In a further attempt to identify a general mechanism by which Smicl might affect gene expression at the MBT, we noted that CPSF was required for pausing and termination of transcription, and that it did this by binding, through its 30 kDa subunit, to the body of RNA polymerase II(Nag et al., 2007). Because Smicl affects transcription as well as 3′-end processing, and because it is similar to the 30K subunit of CPSF, we asked whether it could form a complex with Rpb1, the largest subunit of RNA polymerase II. Co-immunoprecipitation experiments in HEK293T cells indicate that HA-tagged Smicl interacted with full-length Flag-tagged Rpb1(Fig. 6A). In contrast to the 30 kDa subunit of CPSF (Nag et al.,2007), Smicl did not interact with a tagged Rpb1 construct that lacked the C-terminal domain (Rpb1ΔCTD).

The CTD of Rpb1, which is required for interaction with Smicl, is a docking platform for factors involved in various co-transcriptional events(Egloff and Murphy, 2008). Different proteins are recruited to the Rpb1 CTD at different stages of the transcription cycle in response to changes in the phosphorylation state of this domain (Egloff and Murphy,2008). Before the MBT in Xenopus, two forms of Rpb1 are present: one is hypophosphorylated in the CTD and the other is phosphorylated,but not at serine 2 of the repeated heptapeptide YSPTSPS(Palancade et al., 2001). Phosphorylation of serine 2 usually occurs at the MBT, and this hyperphosphorylated form of Rpb1 can be detected on a western blot as a band that migrates slightly more slowly than the embryonic phosphorylated form, as well as by a specific antibody (Palancade et al., 2001).

Phosphorylation of serine 2 at the CTD is required for elongation of transcription and for recruitment of factors involved in cleavage and polyadenylation (Egloff and Murphy,2008). Smicl is also involved in cleavage and polyadenylation, and it interacts with Rpb1 in a manner that requires the CTD. We therefore asked whether Smicl is required for the phosphorylation of the CTD of Rpb1. Injection of our control antisense MO proved to have no effect on the phosphorylation of the Rpb1 CTD, but injection of our Smicl MO inhibited serine 2 phosphorylation at the MBT (Fig. 6B,C). This and our other results are discussed below.

Smicl (Smad-interacting CPSF30-like) is an unusual protein that not only interacts with transcription factors(Collart et al., 2005a; Collart et al., 2005b) but also with the CPSF complex. In addition, it has a zinc finger domain that is similar to that of CPSF30 (Collart et al.,2005a). These observations suggest that Smicl may control gene expression by modulating 3′-end processing. In an attempt to test this idea we used microarray analysis to identify novel Smicl targets and have shown that at least one of these, the homeobox-containing gene Xiro1,is indeed regulated by 3′-end processing. In the course of searching for a general mechanism by which Smicl might function, we discovered that it can form a complex with Rpb1, the largest subunit of RNA polymerase II, and that it is required for serine 2 phosphorylation of the Rpb1 CTD. This modification is required for the elongation of transcription, for splicing and for polyadenylation (Egloff and Murphy,2008); in the Xenopus embryo it normally occurs at the MBT, and it coincides with the translocation of Smicl from cytoplasm to nucleus. We discuss these observations below, and speculate on the mechanism by which Smicl affects the Rpb1 CTD.

Our previous work identified Chordin as a target of Smicl, but it seemed unlikely that this gene was the only Smicl target because overexpression of Chordin does not completely rescue the loss of Smicl function (Collart et al.,2005b). We therefore used microarray analysis to compare gene expression in wild type and in Smicl-depleted embryos in an effort to identify additional Smicl targets (Fig. 1). Quantitative RT-PCR analysis at different stages of development revealed that these putative Smicl targets fell into four categories (Fig. 2A-D). Most fell within category 1, in which Smicl prevented or reduced the onset of gene expression at the MBT, or within category 2, in which loss of Smicl caused a delay in the degradation of maternal transcripts at the MBT. Interestingly,the four criteria we define can also be used to classify zebrafish genes that respond to depletion of the TATA-binding protein (TBP)(Ferg et al., 2007), and indeed 27% of our Smicl targets are present in the list of 1927 genes that are regulated by TBP. This observation is consistent with the fact that transcription and 3′-end processing are coordinated processes.

We suspect that our list of Smicl targets is not complete, and indeed Xlim5, classified as a target in the course of additional experiments(not shown), was not identified in our microarray experiment. This failure to identify all Smicl targets might have occurred because our microarray does not represent all genes expressed during early development, because our selection criteria were too stringent, or because our Smicl MO does not completely inhibit phosphorylation of Rpb1, and different genes may show different sensitivities to levels of phosphorylation.

Consistent with the observation that Smicl regulates gene expression at the MBT, we note that Smicl becomes concentrated in the nuclei of blastomeres at this stage (Fig. 3A-I),although we cannot exclude the possibility that there is some protein present in the nucleus before the MBT. We do not know how this nuclear accumulation occurs, but we have demonstrated that it does not require new transcription,and can therefore be uncoupled from this manifestation of the MBT(Fig. 3J-L).

To investigate the ability of Smicl to regulate 3′-end processing, we turned to Xiro1, one of the earliest transcribed category 1 Smicl targets and therefore one more likely to be a direct target of Smicl. In our first experiments, use of a modified polyadenylation assay(Fig. 4A,B) indeed showed that depletion of Smicl shortens, and overexpression of Smicl lengthens, the poly(A) tail of zygotically expressed Xiro1(Fig. 4C). These results suggest that Smicl might stabilise Xiro1 transcripts around the MBT.

We also investigated two category 2 genes, FBXO43 and MCM10, in which loss of Smicl activity causes a delay in the degradation of maternal transcripts at the MBT. For the mRNA products of these genes we saw no effect of gain or loss of Smicl on the lengths of their poly(A) tails. The delay in degradation of their maternal transcripts may therefore be indirect, resulting from a failure to activate one or more category 1 genes: zygotic transcription is required for degradation of maternal mRNA (Bushati et al.,2008; Giraldez et al.,2006).

Consistent with the idea that Smicl regulates Xiro1, we note that the expression patterns of the two genes overlap, both being activated most strongly in dorsal tissues and later in the neural plate(Collart et al., 2005b; Gâomez-Skarmeta et al.,1998). As discussed above, this regulation is likely to occur through 3′-end processing of Xiro1 mRNA(Fig. 4), although our data also show that our Smicl MO causes the loss of unprocessed Xiro1transcripts as well as mature Xiro1 mRNA(Fig. 5A). This, together with data showing that Smicl affects serine-2 phosphorylation of the CTD of Rpb1(Fig. 6), which is required for overcoming an early block in transcriptional elongation(Peterlin and Price, 2006),suggests that Smicl may indirectly affect transcription of Xiro1. However, it is possible that Smicl is required to prevent degradation of the unprocessed Xiro1 transcript, and we also note that overexpression of Smicl does not elevate expression of Xiro1(Fig. 5B). Indeed, of all the genes with expression controlled by Smicl, only Chordin and FoxI1 are upregulated in response to its overexpression (data not shown). We continue to investigate the role of Smicl in transcription.

In searching for a general mechanism by which Smicl regulates all its category 1 and 3 target genes, we first reasoned that Smicl might be brought to the promoter of its targets by a specific transcription factor, influencing 3′-end processing by substituting for the 30 kDa subunit of CPSF. It proved that the binding sites of the transcription factors CBTF and NF-Y are over-represented in the promoters of category 1 Smicl targets, but Smicl did not co-immunoprecipitate with either protein (data not shown). However, we then asked whether Smicl interacts with RNA polymerase II itself and were able to show that it forms a complex with Rpb1, the largest subunit of RNA polII(Fig. 6A). In contrast to CPSF30, which interacts with the body of Rpb1(Nag et al., 2007), it is the tail of Rbp1 that is required for complex formation with Smicl(Fig. 6A).

Smicl is also required for phosphorylation of the Rpb1 CTD at serine 2 of the repeated heptapeptide YSPTSPS, a modification that normally occurs at the midblastula transition in Xenopus(Fig. 6B,C). By regulating phosphorylation of the Rpb1 CTD, Smicl changes the docking platform for proteins involved in the elongation of transcription and RNA processing, and it therefore influences, albeit indirectly, 3′-end formation at the MBT. We do not yet know how Smicl causes the phosphorylation state of Rpb1 to change, although we note that the activity of RNA polymerase II can be regulated by small non-coding RNAs(Barrandon et al., 2008). Smicl binds and degrades small RNAs with a stem loop secondary structure(Collart et al., 2005a), and it is possible this activity is related to phosphorylation of the CTD and the modulation of RNA polymerase activity.