Nuclear pre-mRNA 3′-end processing regulates synapse and axon development in C. elegans

Nuclear processing of the 3′ end of pre-mRNA is crucial for the generation of the 3′ untranslated region (UTR), mRNA maturation and subsequent export into the cytoplasm (Mandel et al., 2008). This pre-mRNA processing event entails multi-step reactions defined as sequence recognition, cleavage, and the addition of a polyadenine tail at the 3′ end, collectively referred to as nuclear polyadenylation (NpolyA). NpolyA activities require a megadalton protein complex that acts on pre-mRNA cis-regulatory elements. In mammalian cells, catalysis of 3′ cleavage requires cleavage and polyadenylation specificity factor 73 (CPSF-73; CPSF3), and polyadenylation is catalyzed by polyA polymerase (PAP). These core proteins are components of a much larger polyadenylation machinery that includes four major subcomplexes: Cleavage factor I (CFI), Cleavage factor II (CFII), Cleavage and polyadenylation specificity factor (CPSF) and Cleavage stimulation factor (CstF) (Mandel et al., 2008; Shi et al., 2009). These subcomplexes act in concert to control the recognition, specificity and activity of the NpolyA process, which also coordinates with transcription and splicing (Moore and Proudfoot, 2009). Constitutive NpolyA is essential for cell viability and growth (Bloch et al., 1978; Maftahi et al., 1998; Ohnacker et al., 2000; Wang et al., 2005; Xu et al., 2006; Zhao et al., 1999). Both loss- and gain-of-function alterations in cis and trans factors in NpolyA lead to a variety of diseases, including thrombophilia, thalassemias, cancer and muscular dystrophy (Danckwardt et al., 2008; Mayr and Bartel, 2009).

Recently, genome-wide RNA analyses have revealed that extensive regulation must occur at the 3′ ends of pre-mRNA. In human tissues, it is estimated over 50% of transcripts have alternative 3′ terminal exons, 3′ UTRs and polyadenylation sites (Wang et al., 2008). Regulation via alternative NpolyA can be particularly prevalent in the nervous system. For example, isoforms of neuronal proteins that differ as the result of alternative polyadenylation, such as Homer1, can be rapidly produced in an activity-dependent manner (Flavell et al., 2008; Niibori et al., 2007). The neuronal RNA-binding protein Nova shows preferential binding at the 3′ ends and regulates alternative polyadenylation of many transcripts (Licatalosi et al., 2008). However, the trans-acting regulators of pre-mRNA 3′-end processing and their roles in specific cellular and developmental processes remain poorly understood.

Here, using genetic modifier screens in C. elegans, we have uncovered a role of the novel protein SYDN-1 and the NpolyA processing factor PFS-2 in neuronal development. The PFS2 (Polyadenylation factor subunit 2) protein family is characterized by the presence of seven highly conserved WD repeats (Smith et al., 1999), and was originally identified from yeast and mammalian cells based on the polyadenylation activity (Preker et al., 1997; Shi et al., 2009). Yeast Pfs2p directly interacts with proteins in subcomplexes of NpolyA and possibly acts as a scaffold for the 3′-end processing machinery (Ohnacker et al., 2000). Yeast Pfs2p is essential for viability. By contrast, an Arabidopsis PFS2 protein, FY, is dispensable for cell viability and controls floral timing (Simpson et al., 2003; Xu et al., 2006). These observations suggest that PFS2 proteins are involved not only in constitutive NpolyA, but also in specific regulatory processes. We find that loss of function of sydn-1 disrupts axon and synapse morphology in C.elegans, and that reduction of pfs-2 and NpolyA activities restores these defects to normal. Our studies show that SYDN-1 is a negative regulator of NpolyA and support a conclusion that the activity levels of NpolyA are crucial for neuronal development in a context-dependent manner.

Strains were maintained as described (Brenner, 1974). For transgene, strain and allele information, see Table 1. All double mutants were constructed following standard procedures, and the genotypes were confirmed by allele-associated single-nucleotide polymorphisms (SNPs) using the primers listed in Table S1 in the supplementary material.

Enhancers were obtained by mutagenizing CZ455 rpm-1(ju44);juIs1[P_unc-25SNB-1::GFP] with 50 mM EMS. F2 progeny were first screened for severe uncoordination, followed by examination of the SNB-1::GFP-encoding transgene. Candidate mutants were backcrossed to N2 to confirm enhancement. We obtained 14 enhancer mutations from 20,000 haploid genomes, including seven syd-2 alleles, four sad-1 alleles and two syd-1 alleles. sydn-1(ju541) defined a new locus.

Suppressors of sydn-1 were isolated by mutagenizing CZ4767 sydn-1(ju541);syd-2(ju37);juIs76[P_unc-25GFP] with 50 mM EMS. Mutants that showed suppression of the Unc behavior, but retention of Syd-2 behavior, were further examined for suppression of sydn-1 using P_unc-25 GFP. We obtained ten suppressors from 10,000 haploid genomes.

The ju541 and ju608 mutations were mapped to chromosomes I and II, respectively, using SNPs between Hawaiian (CB4856) and N2 strains (Wicks et al., 2001) (Fig. 5A, Fig. 6A). We constructed the dapk-1(ju4) sydn-1(ju541) dpy-5(e61);juIs1 strain to further map ju541 between snp_Y39G10AR[4] and snp_Y39G10AR[6], placing ju541 in a 4 kb interval of the 5′ region of Y39G10AR.17. ju608 was first mapped between pas11767 and pas43942 and to the left of unc-52.

DNA manipulations followed standard procedures. Plasmids were generated using Gateway technology (Invitrogen, Carlsbad, CA, USA): pCZGY165(P_dpy-7 SYDN-1(cDNA)), pCZGY147(P_rgef-1SYDN-1(cDNA)), pCZGY180(P_rgef-1FLAG::SYDN-1(cDNA)), pCZGY748(P_unc-25SYDN-1(cDNA)), pCZGY650(P_pfs-2PFS-2(genomic)::GFP), pCZGY754(P_pfs-2 PFS-2(genomic)::mCHERRY). Germline transformation was performed following standard procedures (Mello et al., 1991), using 50-75 ng/μl P_ttx-3GFP or P_ttx-3RFP as a co-injection marker.

Transgenic GFP reporters in live animals were observed using a 63× Plan-Apochromat objective and Chroma filter sets on a Zeiss Axioplan2 microscope. Ectopic axon branches were quantitated as the number of neurons displaying branches divided by the number of neurons scored per animal. The complexity of ectopic branches always correlated with an increased total number of branches in a given animal. Confocal images of juIs1[P_unc-25SNB-1::GFP] were obtained from the central region of the dorsal nerve cords of 1-day-old adults using a 63× Plan-Apochromat objective. SNB-1::GFP z-stacks were constructed from four to eight 0.4 μm merged planes. z-stacks of SNB-1::GFP soma expression were constructed from nine to sixteen 0.4 μm merged planes of motoneuron cell bodies. Fluorescence intensity in the soma was determined using ImageJ (rsb.info.nih.gov/ij). A circle was drawn covering the entire soma and the average soma intensity minus background was calculated.

Whole-mount staining was conducted as described previously (Finney and Ruvkun, 1990). Samples were processed in parallel. Primary antibodies were: rabbit anti-FLAG (1:1000) and mouse anti-GFP (1:500) from Invitrogen; rabbit anti-ELKS-1 (1:500) and rabbit anti-SNT-1 (1:2000), gifts from Michael Nonet (Deken et al., 2005; Nonet et al., 1993); and rabbit anti-CPF-2 (1:500), a gift of Tom Blumenthal (Evans, 2001). DAPI stain and goat anti-rabbit Alexa Fluor 594 and goat anti-mouse Alexa Fluor 488 secondary antibodies were from Invitrogen.

Confocal images of immunostained PFS-2::GFP, FLAG::SYDN-1, ELKS-1 and SNT-1 were collected using a Plan-Apochromat 100× objective on a Zeiss LSM 500 or 700 microscope with 488 nm and 594 nm lasers. Laser output was set to 40% and transmission was optimized for minimum bleed-through and optimal detection. Detector gain was tuned to minimize pixel saturation and maximize detection range. Single 0.4 μm confocal planes were captured for each channel, merged using LSM software and exported as a TIFF file. The ELKS-1 immunostaining area, distance and intensity were calculated above threshold using the Region Measurement function of MetaMorph. Colocalization images of PFS-2::GFP and FLAG::SYDN-1 with DAPI were obtained using a 63× Plan-Apochromat objective and Chroma filter set on an Axioplan2 microscope and merged using Adobe Photoshop CS. For quantification of PFS-2::GFP and FLAG::SYDN-1, confocal z-stacks were constructed from six to fifteen 0.4 μm merged planes and analyzed using ImageJ.

Total RNA was extracted from mixed-stage animals using Trizol and reverse transcribed using an oligo(dT) primer and SuperScript III First-Strand Synthesis System for RT-PCR according to the manufacturer's protocol (Invitrogen). We performed 5′ and 3′ RACE reactions using the 5′/3′ RACE Kit (Roche Applied Science, Indianapolis, IN, USA); for primers, see Table S1 in the supplementary material. No SL1 or SL2 splice leaders were detected in the 5′ RACE products. The full-length sydn-1 cDNA was 2057 bp, starting from ATG through 320 bp of the 3′ UTR, and was confirmed by DNA sequence analysis.

Quantitative (q) RT-PCR was conducted using SYBR Green on a StepOne Real-Time PCR System, and data were analyzed using StepOne v2.1 (Applied Biosystems, Foster City, CA, USA). The normalized transcript levels were determined by first setting the threshold of product production below saturation, and then determining the cycle at which the threshold was met for each sample. The value for lin-15A transcripts was divided by that of ama-1. The ratio of lin-15A to ama-1 was then normalized to that of the N2 strain. At least two different RNA samples were used per genotype.

A pfs-2 RNAi plasmid (pCZGY649) was constructed by cloning 500 bp of pfs-2 genomic DNA, amplified from cosmid R06A4, into pCZGY115 (equivalent to the L4440 vector). All other RNAi clones were obtained from the Ahringer Library (Geneservice, Cambridge, UK) (Fraser et al., 2000). Feeding RNAi was conducted as described (Kamath et al., 2001). For titrated pfs-2 RNAi, overnight HT115 cultures expressing pfs-2 double-stranded RNA were diluted 1:10 to 1:200 with HT115 culture. RNAi treatment was performed in triplicate and conducted in two independent experiments at 15°C.

A mixture of L4 to young adults from four strains – N2, CZ333 juIs1, CZ4574 sydn-1(ju541);juIs1 and CZ4741 sydn-1(ju541) – were fixed following a high-pressure freeze procedure as described (Weimer et al., 2006). Sections (50 nm) were collected and imaged at 10,000× magnification on a JEOL-1200 EX transmission electron microscope fitted with a Gatan Orius SC600 digital camera. Axon measurements were collected from digital micrographs of ventral and dorsal nerve cords using ImageJ. Axon diameter was calculated from the longest cross-sectional axis.

Strains containing lin-15(n765ts) were maintained at 15°C. L1 larvae were shifted to 22.5°C in parallel. Three days later, adults were scored as Muv if they displayed one or more vulval protrusions. sydn-1;lin-15(n765ts) animals were 100% Muv, and the number of vulval protrusions was scored.

Complete loss of function in several synaptogenesis genes, such as rpm-1, syd-2 and syd-1, leads to distinct defects in synapse development yet comparatively mild effects on the behavior of the animal (Zhen et al., 2000; Zhen and Jin, 1999; Hallam et al., 2002). Synergistic interactions have been observed among these synapse development mutations. For example, syd-2;rpm-1 double mutants are severely uncoordinated, with few synapses (Liao et al., 2004; Nakata et al., 2005). Such synergistic interactions indicate that multiple, partly redundant signaling pathways regulate synapse development.

To identify additional pathways in synaptogenesis we mutagenized rpm-1 animals and searched for double mutants displaying enhanced locomotion defects, which are associated with severe disruptions of synapses, and isolated sydn-1(ju541) (see Materials and methods). sydn-1 single mutants were mildly uncoordinated (Fig. 1A) and exhibited abnormal synapse and axon morphology. We quantified the synaptic defects in sydn-1 using an SNB-1::GFP marker expressed in GABAergic motoneurons. In wild-type (WT) animals, SNB-1::GFP showed evenly spaced, round, fluorescent puncta in the dorsal and ventral cords. In sydn-1 mutants, SNB-1::GFP formed fewer puncta and these were abnormally shaped (Fig. 1B). The GFP intensity in the soma was not altered compared with WT, suggesting that sydn-1 is unlikely to have affected the overall expression levels of the transgene (Fig. 1C). The cholinergic motoneurons revealed similar effects, although these were less severe than for GABAergic neurons (Fig. 1D). Other synapse markers, for the active zone (SYD-2::GFP, SYD-1::GFP) and endosomes (YFP::RAB-5, CFP::RAB-7), were also abnormal (see Fig. S1A in the supplementary material; data not shown). We then verified that these defects were not simply a transgene effect by immunohistochemical staining of two presynaptic proteins. The fluorescence intensities for the synaptic vesicle protein SNT-1 and the active-zone protein ELKS-1 were noticeably reduced in the nerve ring and nerve cords in sydn-1 mutants, as compared with WT (Fig. 2A,B). The number of ELKS-1-stained puncta in the dorsal nerve cords was significantly reduced in sydn-1 mutants (Fig. 2C). These analyses reveal that multiple aspects of sydn-1 mutant synapses are abnormal.

We also examined axon morphology in the sydn-1 mutants. Although most axons were grossly normal in their trajectory or axon path, close observation revealed small protrusions (∼1-2 μm) from the neuron processes (Fig. 1E,F). For example, in sydn-1 mutants, 6% of GABAergic motoneurons (n=132 animals) displayed ectopic protrusions compared with none in control animals (n=51) (Fig. 4C). These protrusions extended from variable positions of the dorsal or ventral processes, or commissures. Additionally, in sydn-1 mutants the GFP-labeled axons often displayed noticeably reduced fluorescence intensity, which was most obvious in the commissures and dorsal processes of the ventral cord neurons, as well as in longitudinal axons of mechanosensory, HSN and PVQ neurons (Fig. 1E; data not shown). The neuronal morphology phenotypes and locomotion deficit of sydn-1 mutants were visible in young larvae and did not progressively worsen in aged adults, indicating an early defect in development (see Fig. S1B in the supplementary material).

To further evaluate the architecture of the nervous system, we analyzed the ultrastructure of the ventral cord of animals fixed using a high-pressure freeze procedure. The axon membrane bilayer appeared grossly normal in sydn-1 mutants and the overall left-right asymmetry of the nerve bundle was maintained. Although the left and right bundles still had the same gross number of main axons (Fig. 3A), there were more small axons in sydn-1 (Fig. 3). Since we did not observe an increase in neuronal number using various GFP reporters (Fig. 1; data not shown), we suspect that the increased number of axon profiles in the sydn-1 ventral cord is likely to be due to the small axonal protrusions that extend off from the normal axon processes (Fig. 1E).

To determine whether the behavioral synergy between sydn-1 and rpm-1 reflected synergistic effects at the synapse, we examined synaptic morphology in sydn-1;rpm-1 double mutants. In these animals, the SNB-1::GFP puncta in the GABAergic motoneurons were almost completely absent (Fig. 4A). Immunostaining of ELKS-1 and SNT-1 showed severe reduction in all major areas of the nervous system (Fig. 2C,D; see Fig. S2 in the supplementary material). Such enhanced synaptic defects are consistent with sydn-1 and rpm-1 acting in parallel in synapse development.

The altered synapse morphology in sydn-1 mutants partially resembled the phenotype of the synapse development mutants syd-1, syd-2 and sad-1 (Crump et al., 2001; Hallam et al., 2002; Zhen and Jin, 1999). We tested genetic interactions of sydn-1 with these genes. All the double mutants between these three genes and sydn-1 exhibited enhanced locomotion defects that were independent of transgenes. The synapse morphology showed more severe abnormalities than each single mutant by SNB-1::GFP fluorescence or by immunostaining of SNT-1 and ELKS-1 (Fig. 2, Fig. 4A). More strikingly, we observed an elaboration of axon protrusions in the double mutants (Fig. 4B,C). The aberrant axon morphologies were seen in different types of neurons to varying degrees. For example, both GABAergic and cholinergic motoneurons in sydn-1;syd-2 animals showed a similar degree of excess axon branching and protrusions, whereas the mechanosensory neurons showed only small axon protrusions (data not shown). These effects on axon morphology were not seen in rpm-1;syd-2 or syd-1;syd-2 double mutants (Fig. 4C; data not shown). The observed synergistic interactions indicate that sydn-1 acts in a novel pathway parallel to those regulated by rpm-1, syd-2, syd-1 or sad-1.

To further explore the effects of sydn-1 on synapses, we examined interactions between sydn-1 and genes that function in synaptic transmission. unc-18 (mammalian Munc18; syntaxin-binding protein) enables synaptic vesicle docking and promotes exocytosis (Weimer et al., 2003). The GABA receptor unc-49 is required for a postsynaptic GABA response and locomotion (Bamber et al., 1999), but not for morphological development of the synapse (Gally and Bessereau, 2003). sydn-1;unc-18 and sydn-1;unc-49 double mutants showed additive deficits in locomotion and did not display strong synergy in axon or synapse morphology phenotypes (Fig. 4C; data not shown). The CaMKII homolog UNC-43 has many roles in the development and function of the synapse (Reiner et al., 1999) and has also been reported to interact with the netrin axon-guidance pathway (Wadsworth, 2002; Wang and Wadsworth, 2002). We analyzed genetic interactions of sydn-1 with both unc-43(n498sd) gain-of-function and unc-43(n498n1196) loss-of-function mutants and did not observe strong enhancement of the neuronal or behavioral phenotypes of sydn-1 mutants (Fig. 4C).

Aberrant axon branching is often associated with cytoskeletal defects. We tested interactions between sydn-1 and genes that function in axon outgrowth and branching or in cytoskeleton regulation. The Enabled homolog UNC-34 is a regulator of the actin cytoskeleton and is required for axon outgrowth and guidance (Gitai et al., 2003; Withee et al., 2004; Yu et al., 2002). Abelson-interacting protein 1 (ABI-1) interacts with multiple cytoskeleton regulators, including UNC-53, to regulate dendrite outgrowth and morphogenesis (Schmidt et al., 2009). The conserved protein UNC-119 (mammalian Unc119) inhibits axon branching (Knobel et al., 2001). We found that double mutants of sydn-1(ju541) with unc-34, abi-1 or unc-119 displayed overall additive behavioral phenotypes and that the axonal phenotypes were no more severe than for the strongest single mutant (Fig. 4C). In summary, although our genetic double-mutant analysis cannot be exhaustive, these results reveal that sydn-1 function becomes crucial when synapses are improperly formed.

We positionally cloned sydn-1 and determined that sydn-1(ju541) contains a 1 kb deletion in the gene Y39G10AR.17 (Fig. 5B). Using RACE and RT-PCR we verified the predicted gene structure and obtained full-length sydn-1 cDNA. SYDN-1 comprises 578 amino acids and has a highly proline-rich C-terminus that contains multiple regions of four to 11 consecutive prolines. Apparent orthologs of SYDN-1 are present in other nematode species. For example, the C. briggsae protein (BP:CBP31994) exhibits an overall 58% identity and 71% similarity with SYDN-1.

The 1 kb deletion in sydn-1(ju541) removes 193 bp of the promoter and most of the first exon (Fig. 5B). By RT-PCR analysis, we found that the remaining exons 2-5 were transcribed in sydn-1(ju541) animals, possibly via a cryptic promoter in the first intron. Transgenic expression of WT sydn-1 genomic DNA fully rescued the behavior, axon and synapse defects in sydn-1(ju541) (Fig. 5C,D). Overexpression of the mutant genomic region from sydn-1(ju541) did not rescue the mutants, nor did it cause any discernable abnormalities in WT. These results show that ju541 is unlikely to express any active SYDN-1 protein and that the mutant phenotypes are caused by sydn-1 loss of function.

We determined the sydn-1 cell-type requirement by tissue-specific expression. Transgenic expression of sydn-1 from a pan-neuronal promoter (P_rgef-1), but not from an epidermal promoter (P_dpy-7), fully rescued the locomotion and neuronal phenotypes of sydn-1 mutants. Expression of the sydn-1 cDNA under the control of a GABAergic motoneuron promoter (P_unc-25) rescued the GABAergic neuron synaptic and axonal morphology phenotypes, but not the locomotion behavior (Fig. 5C,D). These data show that SYDN-1 can act cell-autonomously to promote synaptogenesis and that the movement deficits are likely to result from defects in multiple neuron types.

To determine the subcellular localization of SYDN-1 in neurons, we expressed functional FLAG-tagged SYDN-1 under the control of the rgef-1 pan-neuronal promoter in sydn-1 animals (Fig. 5C-F). Immunostaining with anti-FLAG showed that FLAG::SYDN-1 was localized to the nucleus and was frequently enriched where DAPI staining was weak (Fig. 5E, arrows). Within the nucleus, FLAG::SYDN-1 localization appeared either as a speckled pattern or a small subnuclear ring (Fig. 5F) (speckled pattern, 47±5%; ring pattern, 53±5%; average ± s.e.m.; n=1108 nuclei from 39 animals). DAPI staining was frequently observed outside of the subnuclear ring of SYDN-1, suggesting that SYDN-1 is not localized close to the nuclear envelope. Thus, SYDN-1 defines a nuclear component that regulates axon and synapse development.

To better understand the molecular function of SYDN-1 we conducted a genetic suppressor screen (see Materials and methods). Among the suppressor mutations, we focused on a strong recessive suppressor: allele ju608. The abnormal synaptic puncta of sydn-1(ju541) animals were completely suppressed by ju608 (Fig. 6C). Similarly, the increased axon protrusions in sydn-1;syd-2 animals were completely eliminated by ju608, as were the enhanced locomotion defects (Fig. 6D,F). ju608 single mutants did not display synaptic defects and did not suppress the phenotypes of syd-2 (Fig. 6C; data not shown). Animals were grossly WT in movement, size and behavior, but showed a significant reduction in brood size (see Fig. S3A in the supplementary material).

We mapped ju608 to the right arm of chromosome II, and determined that ju608 contains a single G-to-A nucleotide transition in the third exon of pfs-2 (Fig. 6A,B). The suppression activities of ju608 were fully rescued by transgenic expression of WT pfs-2 (Fig. 6D-F). PFS-2 is a member of a conserved family of proteins that are characterized by the presence of seven WD repeats at the N-terminus (Fig. 6G), which have been shown to be crucial for yeast Pfs2p function in NpolyA (Ohnacker et al., 2000). The pfs-2(ju608) mutation causes an Arg147Gln substitution at a completely conserved residue in the first WD repeat (Fig. 6G). To determine the nature of the ju608 mutation, we conducted pfs-2 RNAi using the enhanced RNAi strain sydn-1;syd-2;eri-1;juIs76[P_unc-25GFP]. pfs-2 RNAi caused complete sterility (see Fig. S3B in the supplementary material). In order to test the effects of pfs-2 knockdown in neurons, we treated the animals using a titrated pfs-2 RNAi approach (see Materials and methods), and observed a complete suppression of the neuronal and behavioral defects in the viable animals (Fig. 6H). Furthermore, overexpression of the full-length pfs-2(ju608) mutant genomic DNA in sydn-1(ju541) animals did not cause any suppression of the mutant phenotypes (Fig. 6C, bar chart). These data indicate that pfs-2 is essential for embryo viability and adult fertility and that ju608 is a recessive partial loss-of-function (plf) mutation.

All Pfs2p homologs in higher eukaryotes have an extended C-terminus that shares little sequence homology among them (Fig. 6G). We tested the role of the C-terminus of PFS-2 by transgenic expression of a truncated genomic DNA containing only the WD-repeat region of PFS-2 and observed that this expression dramatically reduced the rescuing activities in sydn-1;pfs-2 animals (Fig. 6F). This suggests that the C-terminal extension is required for the function of PFS-2.

The genetic suppression of sydn-1(lf) by pfs-2(plf) suggests that PFS-2 might be negatively regulated by SYDN-1. To address the basis of this genetic interaction, we first performed qRT-PCR, and found that SYDN-1 and PFS-2 do not regulate one another's transcript levels (see Fig. S3C in the supplementary material). We then examined the subcellular localization of both proteins. Expression of a GFP-tagged transgenic PFS-2 (P_pfs-2PFS-2::GFP) fully rescued the suppression activity of pfs-2(ju608) (Fig. 6E,F), and PFS-2::GFP was localized to nuclei in neuronal and non-neuronal tissues (Fig. 7A). Colocalization studies of neuronally expressed FLAG::SYDN-1 and PFS-2::GFP revealed similar subnuclear patterns, such that within a single nucleus both proteins appeared either as a speckled or a ring pattern (Fig. 7A,B) (n>100 nuclei), with varying degrees of overlap in subnuclear speckles.

We next asked whether the localization of the two proteins depend on each other. FLAG::SYDN-1 nuclear localization was not detectably altered in pfs-2(ju608) mutants (Fig. 7C). By contrast, the intensity of PFS-2::GFP was noticeably increased in sydn-1 mutants compared with WT (Fig. 7C). The frequency of the ring versus speckled pattern of PFS-2::GFP in sydn-1 was about the same as in WT (for WT, 79±1% speckled and 21±1% ring, n=274 nuclei; for sydn-1, 73±1% speckled and 27±1% ring, n=499 nuclei; average ± s.e.m.; P>0.9, Student's t-test).

To test the specificity of the observed effects of sydn-1(lf) on PFS-2, we examined the localization of the C. elegans ortholog of CstF-64, CPF-2 (Evans et al., 2001). Immunostaining of CPF-2 showed nuclear ring and speckled patterns, and the overall nuclear intensity was comparable in WT and sydn-1 mutants (see Fig. S4 in the supplementary material). These analyses suggest that SYDN-1 is likely to negatively regulate the activity of PFS-2 by influencing its abundance.

The above observations raise the question of whether suppression of SYDN-1 is mediated by changes in NpolyA activities. We performed RNAi of other NpolyA factors in sydn-1 mutants. Knockdown of these genes caused lethality and adult sterility in the sydn-1;syd-2;eri-1;juIs76 strain (data not shown), confirming multiple previously reported RNAi observations (e.g. Kamath et al., 2003; Simmer et al., 2003). We observed partial suppression of the sydn-1;syd-2;eri-1;juIs76 behavior and axon protrusion phenotypes among the escaper adult progeny from animals treated with RNAi targeting the 3′-end processing factors cpf-1, cpf-2, cpsf-2, cids-1 and cids-2, and a putative PAP, t15h9.6 (Fig. 8A). By contrast, RNAi knockdown of three other PAPs (F43G6.5, pap-1, gld-2) did not show significant effects on Sydn-1 defects, consistent with studies implicating these genes in other types of RNA regulation (Cui et al., 2008; de la Mata and Kornblihtt, 2006; Maciolek and McNally, 2007; Meinhart and Cramer, 2004; Wang et al., 2002). These observations show that aberrant synaptic and axonal development in sydn-1 mutants can be ameliorated by specific reduction of nuclear pre-mRNA 3′-end processing activities.

To further address whether pfs-2 and sydn-1 themselves directly function in NpolyA, we tested their genetic interactions with lin-15(n765ts), the mutant phenotype of which can be modulated by NpolyA activities. lin-15 encodes an operon of two transcripts, lin-15A and lin-15B, and controls vulva formation (Clark et al., 1994; Ferguson et al., 1987; Huang et al., 1994). The lin-15(n765ts) allele exhibits a temperature-sensitive multivulval phenotype (Muv) owing to the insertion of an NpolyA signal from the gene H18N23.2 into lin-15B (Fig. 8B) (Cui et al., 2008). The insertion disrupts the lin-15B transcript, but also causes a reduction of the downstream lin-15A transcript owing to abnormal NpolyA and transcriptional termination. Inhibiting NpolyA activities can suppress the Muv defect of lin-15(n765ts) (Cui et al., 2008). We found that pfs-2(ju608) almost completely suppressed the Muv phenotype of lin-15(n765ts) (Fig. 8C,E). Conversely, sydn-1(ju541) enhanced the number of multivulval protrusions (Fig. 8D,E).

To address whether these genetic interactions reflected molecular changes in NpolyA activity, we performed qRT-PCR on lin-15A transcripts. The levels of lin-15A transcript were decreased in lin-15(n765ts) compared with WT (Fig. 8F), as reported previously (Cui et al., 2008). The levels of lin-15A transcripts were increased in pfs-2(ju608);lin-15(n765ts) animals, and were decreased in sydn-1;lin-15(n765ts) animals, compared with lin-15(n765ts) (Fig. 8F). Although a role of lin-15 in neuronal tissues has been inferred (Kennedy et al., 2004), our findings of sydn-1 and pfs-2 effects on lin-15 pre-mRNA processing are likely to include changes in non-neuronal tissues. Nonetheless, these observations lend support to the proposal that pfs-2 and sydn-1 function in NpolyA and imply that elevation or misregulation of NpolyA activities underlies the neuronal phenotypes in sydn-1 mutants.

Using modifier genetics, we have uncovered two proteins that together regulate NpolyA activities in neuronal development. Our data show that NpolyA becomes crucial when synapses are improperly formed and that the novel protein SYDN-1 negatively controls NpolyA activities, in part through influencing the nuclear abundance of PFS-2.

sydn-1(lf) mutants display synaptic abnormalities and uncoordinated locomotion. The uncoordinated phenotype is due to neuronal defects because Sydn-1 phenotypes are fully rescued by cell-specific SYDN-1 expression in the nervous system. The movement defect is likely to reflect a developmental role of sydn-1 in neurons. First, synapses in sydn-1 mutants are fewer in number and those that remain often have abnormal morphology. Second, the axons of sydn-1 mutants frequently exhibit ectopic protrusions and appear to be thin in diameter. Third, the phenotypes of sydn-1 single mutants are early onset and do not progressively worsen with age. Last, synthetic defects are seen when sydn-1 is placed in combination with other synapse development mutations, but not with mutations affecting synapse transmission or general axon growth. The combination of phenotypes observed in sydn-1 mutants has not been seen in the numerous reported C. elegans synapse mutants. Our genetic double-mutant analysis, although not exhaustive, indicates that sydn-1 is a member of a novel pathway acting in parallel to known synapse development pathways and that the activity of sydn-1 is particularly sensitized to disruptions in synapse architecture. The enhanced defects in axon morphology of double mutants between sydn-1 and synaptic architecture genes further suggest a coordination in cellular differentiation between synapse assembly and axonal morphogenesis.

Via a second forward genetic screen, we found that the mutant phenotypes of sydn-1 are strongly suppressed by reduction of function in PFS-2, a member of a conserved protein family implicated in the stabilization of the nuclear 3′-end processing complex (Ohnacker et al., 2000; Shi et al., 2009; Simpson et al., 2003; Wang et al., 2005). Knockdown of several conserved NpolyA components suppresses the sydn-1 mutant phenotypes to varying degrees, resembling the effects of pfs-2(plf). The Muv suppression of lin-15(n765ts) and the increase in lin-15A transcripts by pfs-2(plf) are consistent with a role of PFS-2 in promoting NpolyA (Cui et al., 2008) (this study). Conversely, the Muv enhancement and the decrease in lin-15A transcripts by sydn-1(lf) are consistent with SYDN-1 inhibition of NpolyA. We surveyed a set of neuronal genes, including syd-2, snt-1, rpm-1 and unc-43, and did not detect any significant changes in their overall transcript levels in sydn-1 and pfs-2 mutants (data not shown). As these genes do not undergo alternative 3′-end processing, these preliminary observations are consistent with a hypothesis that the interaction between sydn-1 and pfs-2 is unlikely to result in a general effect on the stability of all neuronal transcripts, but rather plays specific roles in the control of alternative 3′-end processing, leading to transcripts that play more regulatory roles in response to synapse development signals. The genetic synergistic interactions between sydn-1 and the other synaptogenesis mutants suggest that this regulation of pre-mRNA processing is particularly crucial when synapses are improperly formed.

Both SYDN-1 and PFS-2 display ring-shaped and speckled subnuclear localization patterns. In the absence of SYDN-1, PFS-2 appears more abundant, particularly in the ring-shaped conformation. We do not know the nature or significance of these distinct subnuclear conformations, nor whether the nuclei dynamically convert from one conformation to the other, or, simply, that the ring-shaped pattern is more visible and hence changes are more apparent.

An increasing number of studies have shown that dynamic and distinct nuclear compartments are likely to be associated with various pre-mRNA processing activities in many cell types (Boronenkov et al., 1998; Cardinale et al., 2007; Huang and Spector, 1991; Huang and Spector, 1992; Lamond and Spector, 2003; Mellman et al., 2008; Mintz and Spector, 2000; Politz et al., 2006). SYDN-1 could negatively regulate NpolyA activity on selected transcripts via direct interaction with PFS-2, or indirectly via additional proteins. PFS2 proteins are defined by highly conserved WD repeats, with a divergent C-terminus that is present only in higher eukaryotes. In our RNAi studies of other NpolyA proteins, none was as strong as pfs-2 in suppressing sydn-1. Conceivably, the WD-repeat domain is required for constitutive NpolyA activities and the C-terminal extension of PFS-2 possibly allows binding of additional regulatory factors in a tissue- or context-specific manner. The observation that transgenic expression of PFS-2 lacking the C-terminal extension does not rescue the suppression activities of PFS-2 is consistent with this idea.

Recent studies in C. elegans have identified multiple nuclear RNA-binding proteins that localize to subnuclear speckles in neurons, including the CELF/BrunoL ortholog UNC-75, the ELAV-like protein EXC-7, and the putative RNA-binding protein SYD-9 (Loria et al., 2003; Wang et al., 2006). Interestingly, these RNA-binding proteins show specific synergistic genetic interactions with mutations that affect synaptic transmission but not synaptic development. Such distinct genetic synergy implies that these proteins, which include SYDN-1 and PFS-2, are likely to regulate unique pools of RNA targets involved in neuronal development or function, and suggests that complex parallel signaling pathways operate at the level of pre-mRNAs.

A recent series of reports suggest that alternative pre-mRNA 3′-end processing is an important regulatory process (Lutz, 2008). Specific tissues, including those within the immune system and brain, appear to have a high tendency for biased use of polyadenylation signals (Wang et al., 2008; Zhang et al., 2005). This might be due to the tissue-specific expression of genes that encode pre-mRNA 3′-end processing machinery or differences in the sensitivity of NpolyA cis-regulatory elements (Hunt, 2008). Levels of 3′-end processing factors, such as CstF-64, can influence the production of isoform-specific mRNA for immunoglobulin genes (Takagaki and Manley, 1998). Competition between the RNA-binding protein Sex lethal and CstF-64 also results in the production of different sex-specific isoforms of the same gene through alternative polyadenylation in the Drosophila germline (Gawande et al., 2006). In C. elegans the CstF-64 homolog CPF-2 has been shown to be preferentially associated with SL2-containing small nuclear ribonucleoproteins (SnRNPs), possibly exerting regulation on transcripts within operons (Evans et al., 2001). We propose that the regulation of PFS-2 and NpolyA by SYDN-1 is likely to confer regulated NpolyA activity on selected neuronal transcripts, either via use of alternative pre-mRNA 3′-end processing sites or tissue-specific NpolyA complexes. A cohort of coordinately regulated transcripts could contribute to the temporal and spatial regulation of synapse and axon development.

We thank Min Han and Mingxue Cui for the lin-15(n765ts) strain; Mike Nonet and Tom Blumenthal for antibodies; Peg Scott and Tiep Ly for strain construction and maintenance; Benjamin Rollins, Julia Cochran and P. Clint Spiegel for technical assistance; Andrew Chisholm and Mark Palfreyman for comments on the manuscript; and members of our laboratory for discussion. Some strains were obtained from the Caenorhabditis Genetics Center, which is supported by grants from the National Institute of Health. This work was supported by a UCOP President's Postdoctoral Fellowship to H.V.E. and an NIH grant (R01 35546) to Y.J. Y.D. and A.G. are Research Associates of, and Y.J. is an investigator of, the Howard Hughes Medical Institute. Deposited in PMC for release after 6 months.