Spiralian quartet developmental potential is regulated by specific localization elements that mediate asymmetric RNA segregation

Spiralian embryogenesis is defined by the characteristic asymmetry and chiral geometry of the early cell divisions (see Fig. 1). It is also distinguished by similarities in the fate maps of the blastulae produced by these divisions, even across long phylogenetic distances (Henry and Martindale, 1999; Lambert, 2010). Spiralian development is found in several phyla in the large clade of bilateral animals called the Lophotrochozoa. These include the annelids, molluscs, nemerteans, sipunculans and platyhelminthe flatworms. Recent phylogenetic evidence suggests that the last common ancestor of the Lophotrochozoa had spiral cleavage (Dunn et al., 2008), raising the possibility that the taxon Lophotrochozoa should revert to the earlier taxonomic term Spiralia (Schleip, 1929). These results indicate that spiralian development was present in the stem group of one of the three great clades of bilateral animals. This highlights the importance of understanding spiralian development for reconstructing the early evolution of development in protostomes and, by extension, in the bilateral metazoans.

The strong conservation of the cleavage pattern and fate map in these embryos allows homologous cells to be identified in distantly related groups. One of the hallmarks of spiralian development is the regionalization of the animal-vegetal axis by the production of three micromere quartets. The first two cleavages divide the embryo into four macromeres. The micromeres are produced when the four macromeres divide synchronously towards the animal pole in three successive divisions, producing the first, second and third quartet micromeres (Fig. 1A). Fate-mapping experiments have shown that across spiralian phyla, each quartet generates similar sets of fates (Ackermann et al., 2005; Boyer et al., 1996; Damen and Dictus, 1994; Hejnol et al., 2007; Henry and Martindale, 1998; Henry et al., 2004; Huang et al., 2002; Maslakova et al., 2004; Render, 1997; Weisblat and Shankland, 1985). Typically, the first quartet gives rise to the head ectoderm, including neural structures such as eyes, as well as to the majority of the primary ciliated band. For instance, in molluscs, the first quartet cells generate the ciliated lobes of the head (the velum), as well as anterior neural structures including the eyes (Fig. 1B). The second quartet usually gives rise to a posterior portion of the ciliated band and trunk ectoderm, whereas the third quartet normally generates ventral structures and foregut. Most of the mesoderm, including the heart and larval retractor muscle, is derived from the 4d cell, the fourth micromere formed by the D macromere.

Embryological experiments in Ilyanassa indicate that each quartet is born with a distinct developmental potential (Sweet, 1998). After transplantation, members of the first, but not the second, quartet can develop into eyes, consistent with the normal fate of the first quartet. Conversely, second, but not first, quartet cells can produce shell material. These landmark experiments showed that the spiralian micromeres acquire quartet-specific developmental potentials at, or shortly after, birth, but the molecular basis of this is not known. In the spiralian Platynereis dumerilii, asymmetries in the levels of nuclear β-catenin are required in different quartets for normal development. However, this appears to be a general property of divisions in the early embryo and does not provide quartet-specific cues (Schneider and Bowerman, 2007). In Ilyanassa, a number of quartet-specific RNAs have been discovered (Kingsley et al., 2007; Lambert and Nagy, 2002). In all cases examined thus far, these RNAs are inherited in a quartet-specific manner after first localizing to macromere centrosomes (Fig. 1A). These observations suggest that segregated RNAs might be involved in quartet-specific developmental potential, although none of these RNAs has been functionally implicated in development and the mechanisms that restrict them to particular cells are not known.

The present study focuses on the first quartet micromeres in Ilyanassa. First, we show that a particular RNA that is segregated to the first quartet, IoLR5, is required for the normal development of structures derived from this group of cells, i.e. the velum and the eyes (Clement, 1967; Render, 1991). Then, we define a narrow interval when centrosomal localization of RNA can occur, and show that this event is necessary for segregation to the first quartet. Finally, we identify a discrete secondary structure in the IoLR5 RNA that mediates localization to the centrosome and segregation to the first quartet. Our characterization of this structure and a similar one from another first quartet RNA suggests a model for how first quartet specificity is achieved.

Adult Ilyanassa obsoleta were collected from the Bass River on Cape Cod, MA, the Peconic Bay in Riverhead, NY, Ocean Isle Beach, NC, or ordered from the Marine Resources Center at the Marine Biological Labs in Woods Hole, MA, USA. Animal care and embryo collection have been described (Collier, 1981; Gharbiah et al., 2009).

The IoLR5 clone isolated from a previously described in situ screen was 2055 bp (Kingsley et al., 2007). Northern blots on RNA from many individuals identified two transcripts of ∼3.5 and ∼3.0 kb. 3′ rapid amplification of cDNA ends (RACE) on a cDNA pool primed from a ligated linker generated an additional 1395 bp; 5′ ligation-mediated RACE yielded no additional bases. The new 3′ fragment contains a large microsatellite (∼850 bp in the allele recovered in the 3′ RACE), which appears to be a continuous (TC)_n repeat; we sequenced 100 bp of repeat from either side before the signal deteriorated, and we were unable to cut this region using a mixture of four four-cutter restriction enzymes that would be likely to cut if there were any significant length of non-repeat sequence in the region. PCR was performed on genomic DNA from many individuals, using primers flanking the microsatellite, resulting in bands of ∼850 and ∼400 bp. Consequently, the final clone is ∼3450 bp (EU087578.1) and the 3.0 kb band seen on the northern blot is likely to be due to allelic variation in the microsatellite. We have identified a single intron using genomic PCR with primers along the transcript, but we have no evidence for alternatively spliced transcripts. We made a morpholino-resistant IoLR5 construct (IoLR5MOres) using fusion PCR (Szewczyk et al., 2006) to alter seven nucleotides in the morpholino binding site. IoTis11+IoLR5LE was made by inserting IoLR5LE at base-pair 47 in the original IoTis11 cDNA clone.

We raised rabbit antibodies against a synthetic peptide corresponding to part of the IoLR5 protein (QMRMRERNYRVG) (Pocono Rabbit Farm and Laboratory, Canadensis, PA, USA). On western blots of Ilyanassa total protein, the affinity-purified antibody (at 1:1000 dilution) recognized two distinct bands at ∼12 kDa and ∼14 kDa, close to the 9.6 kDa predicted size of the IoLR5 protein. When animals were injected with IoLR5MO1, both bands on the western blot were diminished in strength, as was staining in whole-mount embryos.

Embryos were fixed and processed for in situ hybridization as previously described (Kingsley et al., 2007). Antibody staining and counterstaining with DAPI and phalloidin-Alexa Fluor 488 (Molecular Probes) were carried out as described (Lambert and Nagy, 2001) using a 1:800 dilution of the affinity-purified IoLR5 antibody.

Zygotes were injected with the indicated concentration of IoLR5MO1 (5′-GCTGGTGATACAACTCTTGATGACA-3′) or the Standard Control MO (5′-TATAAATTGTAACTGAGGTAAGAGG-3′) (Gene Tools), as previously described (Gharbiah et al., 2009; Rabinowitz et al., 2008). Embryos were reared for 7-15 days in 22 μM-filtered artificial seawater (FASW; Instant Ocean), then anesthetized and fixed as described (Lambert and Nagy, 2001). Larvae were individually scored at 400× with a Zeiss Axioplan 2 microscope. Capped and tailed IoLR5MOres mRNA was generated with the mScript Kit (Epicenter) and co-injected with IoLR5MO1 at ∼450 ng/μl. For each manipulation, at least five embryos were scored from each of three or more different capsules to control for between-capsule variation.

For localization element (LE) mapping experiments, fragments were amplified by PCR using a forward primer incorporating a T3 RNA polymerase site. DNA templates were phenol/chloroform extracted and ethanol precipitated before in vitro transcription with either digoxigenin-11-UTP (Roche) or ChromaTide Alexa Fluor 488-5-UTP (Molecular Probes) for 2-4 hours. Transcribed RNAs were treated with DNase I (Roche) for 20 minutes and those longer than 500 bases were purified using the MEGAclear Kit (Ambion), whereas those shorter than 500 bases were precipitated with sodium acetate and ethanol. Bright-field images were taken on a Zeiss Axioplan 2 microscope with a Canon 300D digital SLR camera and Qioptiq coupler (Fairport). Fluorescently labeled RNAs were imaged with a Leica SP5 confocal microscope and z-stacks were projected using Leica software.

We used the standard ratio of labeled to unlabeled nucleotide (1:3) for the digoxigenin-labeled RNAs, but we did not observe localization of Alexa Fluor 488-labeled RNAs unless we used a ratio of 1:12.5. This might be due to the fluorescent label interfering with the proper secondary structure of the LEs or with protein interactions. We could not detect the localization of fluorescent transcripts that were shorter than 500 bases, probably because of the low incorporation of label. At the 4-, 8- and 24-cell stages, the fluorescently labeled RNAs we tested had the same localization pattern as digoxigenin-labeled transcripts. At the 2-cell stage, only fluorescently labeled transcripts localized to centrosomes.

Capsules were divided so that half of the zygotes were injected with 1.0 mM IoLR5StemBlockMO (5′-CACTGTCCCCTCTAGGCAACATAGG-3′) and 1% Oregon Green 488 70,000 MW lysine-fixable dextran (Invitrogen) and the other half were injected with 1.0 mM IoLR5MO1 and 100 μM Sulforhodamine 101 (Invitrogen). Embryos were then pooled for fixation and in situ hybridization, following which the two sets could be distinguished by the presence of red versus green fluorescence from the co-injection solutions. Co-processing allowed for qualitative comparison between the two pools. Other experiments showed that Oregon Green dextran does not affect centrosomal localization of the endogenous RNA.

A cDNA clone named IoLR5 (Ilyanassa obsoleta localized RNA 5) was recovered during a screen for localized RNAs during early cleavage stages (Kingsley et al., 2007). Northern blot hybridization identified transcripts of ∼3.5 and ∼3.0 kb (Fig. 1C). Using RACE PCR we extended the original 2055 bp IoLR5 cDNA to 3.5 kb (Fig. 1D). The two bands on the northern blot appear to be due to allelic size variation in a microsatellite (see Materials and methods).

IoLR5 RNA is present in the unfertilized egg, is localized to the centrosomes at the 2-cell and 4-cell stages, and then segregates into the first quartet cells (Kingsley et al., 2007) (as summarized in Fig. 1). At 16.5 hours after egg laying (AEL) (∼70 cells), the transcript was abundant in 1a¹²¹, 1b¹²¹ and 1c¹²¹ (Fig. 1E and see Fig. S1 in the supplementary material). The RNA was found transiently in 1d¹²¹ progeny and appeared to become more abundant in the four trochoblast cells (1q², the main cells of the ciliated band); these observations suggest that there is zygotic IoLR5 transcription, at least in these lineages (Fig. 2A). At ∼2 days AEL, IoLR5 mRNA was detected in two B quadrant cells at the ventral end of the apical plate and in groups of four cells on the left and right sides of the head that were derived from 1a¹²¹ and 1c¹²¹, respectively (Fig. 2C). A time course from 24 to 48 hours AEL showed that the IoLR5-expressing cells (Fig. 2C) were the descendants of 1a¹²¹, 1b¹²¹ and 1c¹²¹ (not shown). At 3 days AEL, these two groups of cells continued to divide and RNA was also observed in one cell on each side that was located dorsal to these clusters. The mRNA signal became increasingly weak and variable and by 4-5 days AEL the IoLR5 mRNA was no longer detectable by in situ hybridization.

Although the IoLR5 transcript is 3450 bases long, there is only one open reading frame (ORF) longer than 50 codons, which encodes a protein of 87 amino acids and has a strong translational start sequence (Fig. 1D) (Kozak, 1987). This protein has no significant homology to known proteins as assessed by BLAST searches of database sequences (Altschul et al., 1990). We raised and validated an anti-IoLR5 antibody (see Materials and methods) and examined expression by western blots. The protein was first detected at 24 hours AEL, peaked at 2 days AEL and began to decline by 4 days AEL, until it was no longer detectable on the western blot after 6 days AEL.

Whole-mount staining showed that IoLR5 protein is in the same cells as the RNA at 1-2 days AEL and persists after the RNA is no longer detectable, allowing us to follow the fates of IoLR5-expressing cells into organogenesis (Fig. 2B,D). The bilaterally paired dorsal IoLR5 cells do not divide in the interval that we examined (to 7 days AEL). These cells are located directly above the eyes (once these structures appear), in regions that are likely to develop into the tentacles. The paired ventral IoLR5-expressing cells also do not divide in this interval, and they become the ventral-most cells of the apical plate, the region of large cells in the middle of the developing head. The ventral and dorsal pairs of cells both maintained IoLR5 protein expression through 5.5 days AEL. Unlike the dorsal and ventral pairs of cells, each lateral cluster continues to divide and by 3 days AEL, nine IoLR5-expressing cells were apparent on each side (Fig. 2E,E′). By 4.5 days AEL, each cluster of nine cells had split such that the ventral portion (4-5 cells) was a few cell diameters from the dorsal group on each side (Fig. 2F,F′). Around 5-6 days AEL, the ventral group of cells was completely internal and was part of a densely packed mass of cells that form the cerebral ganglion. Only a subset of the dorsal group of cells was internalized. These cells also appeared to join the growing cerebral ganglion, while the remaining cells stayed on the velar surface ventral to the developing eye, until at least 7 days AEL (Fig. 2H-K). IoLR5 staining was also observed in two long processes that extend from the cerebral ganglia into the distal region of the foot (Fig. 2G,G′). These processes included at least two cell bodies, and appeared to be axons. Indeed, they are in the location reported for the pedal connectives (Dickinson and Croll, 2003). These results show that IoLR5-expressing cells generate several regions of head ectoderm and provide a major contribution to the cerebral ganglion, suggesting that IoLR5 is involved in the development of the nervous system in the head.

To determine the functional role of IoLR5 we injected a translation-blocking antisense morpholino oligonucleotide (IoLR5MO1) into zygotes and reared embryos to the veliger stage. Injection of IoLR5MO1 resulted in reproducible, dose-dependent phenotypes in the larvae. At 0.25 mM, there was no visible effect on development and at 1.5 mM or higher all animals arrested during, or immediately after, gastrulation (not shown). At 0.5 to 1 mM, animals suffered increasingly frequent defects in the velar lobes and eyes (head structures; Fig. 3). Less frequent defects in such tissues as the larval retractor muscle, intestine and statocysts were seen at similar levels across the range of concentrations (Fig. 3G, Table 1). Animals injected with a 1 mM Standard Control MO always exhibited wild-type development.

The eye phenotypes consisted of no eyes, one eye, or synopthalmia (in which the eyes are abnormally close; Fig. 3A-E′). At 1 mM IoLR5MO1, 86% (n=22) lacked one or both eyes and 95% had some kind of eye defect when synopthalmia was included. Surprisingly, injected animals that had only one eye were about three times more likely to be missing the right eye than the left (Table 1). These defects are very similar to those observed after ablation of first quartet cells. For instance, ablation of 1b causes synopthalmia, and ablation of 1a or 1c causes a smaller velar lobe or the lack of an eye on the left or right side, respectively (Clement, 1967).

Two lines of evidence indicate that the IoLR5MO1 phenotype is the result of a specific knockdown of IoLR5 translation. First, on western blots, the IoLR5 protein in IoLR5MO1-injected animals was either completely absent or strongly diminished compared with control animals (Fig. 3F). Second, the eye phenotypes were largely rescued when IoLR5MO1 was injected together with an IoLR5 mRNA that included seven mismatches in the morpholino target site (IoLR5MOres). In this experiment, 6% of IoLR5MO1-injected animals had two wild-type eyes (n=80), but when IoLR5MOres was co-injected with IoLR5MO1, 62% of the animals developed two wild-type eyes (n=157) (Fig. 3H). Injection of IoLR5MOres by itself had no effect on development and none of the rescue animals developed more than two eyes (not shown), indicating that the transcript is not sufficient for ectopic eye development.

Having shown that IoLR5 RNA is required for the development of first quartet derivatives, we turned to the question of how it is specifically segregated to these lineages. We first addressed whether centrosomal localization per se is required for asymmetric segregation by examining the fate of RNAs when centrosomal localization was bypassed.

We found that localization did not occur if RNA was injected after the interval when localization normally occurs. Preliminary experiments showed that an in vitro synthesized RNA corresponding to bases 1-2055 of IoLR5 and labeled with Alexa Fluor 488 could recapitulate localization to 4-cell stage centrosomes and segregation to the first quartet micromeres when it was injected into the zygote. To examine whether later-injected RNA would also be localized, the RNA was injected at various time points from late 2-cell stage until 15 minutes into the 4-cell stage. RNA localization was examined every 10 minutes until the mid 8-cell stage, to ensure that only embryos with normal division were included in the analysis. Embryos with representative staining patterns were fixed, stained and imaged by confocal microscopy. When RNA was injected during the 2-cell stage, it always localized to the centrosome (n=16; Fig. 4A,B). Of the embryos injected after the end of cytokinesis, around half (14/30) had a strikingly different pattern of localization, in which RNAs localized to a shallow zone on the surface of the centrosome (Fig. 4F). The remainder of the injections in this class showed no localization (16/30; Fig. 4J). To determine whether these two patterns were related to the timing of injection, we performed additional injections that were staged based on the distinct morphology of the embryo in the first 5 minutes of the 4-cell stage (Fig. 4E). We found that the majority (21/24) of injections in this interval showed localization to the surface of the centrosome (as in Fig. 4F), with the remainder showing no localization (as in Fig. 4J), and that injections 10 minutes later showed no localization (n=33). The shallow localization suggests that the RNA-containing pericentriolar matrix accumulates progressively on the outside surface of the centrosome, with minimal mixing. The presence of both shallow localization and no localization in the embryos injected during the first 5 minutes of the 4-cell stage indicates that transport to the centrosome ends during this interval.

To evaluate the importance of centrosomal localization for subsequent RNA segregation, we followed the behavior of injected RNAs later in the cell cycle. When RNA did localize to the centrosome it moved to the cortex during prophase and was predominantly asymmetrically segregated into the first quartet micromeres, as observed for IoLR5 RNA injected into zygotes (Fig. 4C-D′). When shallow localization was observed (as in 4-cell + 0-5 min injections), the labeled RNA moved to the cortex as above but as individual speckles of signal, and only a portion of the RNA was segregated (Fig. 4G-H′). When RNA did not localize during interphase (as in 4-cell + 10-15 min), it did not move to the cortex at prophase and was not segregated asymmetrically (Fig. 4K-L′); instead, the RNA remained cytoplasmic until the next interval, when it localized to the micromere and macromere centrosomes at 8-cell stage interphase. This demonstrates that the RNAs remain competent to localize even if they miss the window for transport at the third cleavage (see Fig. 7 and Fig. S3 in the supplementary material).

It might be that segregation requires an event that is concurrent with centrosomal localization but not dependent on it, but this seems less probable than a model in which localization to the centrosome per se is required for RNA segregation. The centrosome could be the site of an essential processing or packaging event, or it might be that the transport only occurs when RNAs are anchored to the pericentriolar matrix material.

The large number of RNA localization and segregation patterns in the Ilyanassa embryo suggests that this system is under very intricate control. One way that this could be generated is by specific RNA localization elements (LEs) (see Martin and Ephrussi, 2009; St Johnston, 2005). To test whether RNA segregation in Ilyanassa relies on an LE, we adapted the fluorescence assay described above to map regions involved in centrosomal localization. Digoxigenin-labeled RNA corresponding to the original clone (bp 1-2055) could recapitulate centrosomal localization after injection into zygotes (compare Fig. 5B with 5A), whereas an antisense transcript of IoLR5 bases 1-2055 failed to localize (not shown). Labeled IoAlpha-Tubulin mRNA did not localize, consistent with its endogenous transcript (Fig. 5C,D). Together, these results show that this approach is effective and specific.

We mapped the localization activity of the original clone by testing various fragments, and found that localization required a region around base 1260 (Fig. 5E). In silico folding of the entire IoLR5 RNA with RNAstructure 4.6 (Mathews, 2006) predicted a large, 118-base stem-loop at bases 1194-1312 (Fig. 5P). This structure was assigned the highest probability of any in the sequence, and was the only structure found in each of the top 20 secondary structure predictions, suggesting that the 118-base stem-loop is likely to be present in vivo. We tested this small fragment and discovered that it localizes to the centrosome (Fig. 5F). We then deleted this 118-base region from IoLR5(1091-1802) and this caused the RNA to remain distributed throughout the cytoplasm (Fig. 5I). These results demonstrate that this region is necessary for the localization of IoLR5 to 4-cell centrosomes and is sufficient to localize itself. To test whether the predicted secondary structure is sufficient to localize a normally unlocalized RNA, we inserted the stem-loop structure into the IoTis11 RNA, which does not normally localize to 4-cell stage centrosomes (Kingsley et al., 2007) (Fig. 5L). The resulting RNA localized to 4-cell stage centrosomes, demonstrating that the stem-loop is sufficient to confer localization (Fig. 5L,M). We refer to this 118-base region as the IoLR5LE.

To examine whether the proper secondary structure of the LE is important for its role as a centrosomal localization signal, we took advantage of the fact that morpholinos can be used to disrupt RNA secondary structure [e.g. in splicing or miRNA biogenesis (Draper et al., 2001; Kloosterman et al., 2007)]. We designed a morpholino targeted to a region near the center of the IoLR5LE stem-loop (IoLR5StemBlockMO; Fig. 5P). Injection of this molecule into zygotes caused a significant impairment of localization of the endogenous IoLR5 transcript (Fig. 5O), whereas injection of the translation-blocking morpholino did not affect localization (Fig. 5N). This novel approach suggests that the secondary structure of the IoLR5LE is important for centrosomal localization. Despite the partial inhibition of centrosomal localization at the 4-cell stage, these animals still developed into apparently wild-type veligers.

We followed the behaviors of the LE and deletion constructs after centrosomal localization. The IoLR5LE localized to the cortex during prophase and was asymmetrically segregated into the 8-cell stage micromeres (Fig. 5F-H). By contrast, the LE deletion transcript remained cytoplasmic throughout the 4-cell stage and was never asymmetrically segregated (Fig. 5I-K). These data demonstrate that the IoLR5LE is necessary for all phases of the segregation process. Although the IoLR5LE was capable of recapitulating the movements of the endogenous RNA, it did so with decreased specificity. When the endogenous RNA segregates during 4-cell stage cytokinesis, all of the transcripts are delivered to the first quartet and none is detectable in the macromeres. For the injected RNAs, centrosome localization during interphase was very specific; however, not all of the injected transcripts became asymmetrically segregated as some were still visible in the macromeres at the 8-cell stage (see Fig. 5H, Fig. 7 and see Fig. S3 in the supplementary material).

During the third cleavage, the IoLR1 RNA is temporally and spatially coincident with IoLR5 (Kingsley et al., 2007). The original cDNA clone was 3.6 kb and the complete transcript is ∼9.6 kb, based on northern blotting (not shown). We tested several fragments of the original 3.6 kb clone for localization to 4-cell stage centrosomes (Fig. 6A-C,E) and found a 1 kb region that could localize. We folded this fragment using RNAstructure 4.6 and three large stem-loops were found to be highly probable. Each of these was individually tested for 4-cell localization, but only one localized (Fig. 6C,E,F). When this region was deleted from the 1 kb fragment, centrosomal localization was abolished (Fig. 6D), showing that this 88-base sequence is necessary for localization of IoLR1. We refer to this region as the IoLR1LE.

We attempted to test whether the two LEs use the same machinery through competition experiments. However, we were unable to competitively inhibit centrosomal localization of either endogenous or exogenous IoLR5 RNA, even at a 1000-fold molar excess of injected IoLR5LE RNA (not shown). We were also unable to competitively inhibit localization of other first quartet RNAs with IoLR5LE. These results suggest that the centrosomal transport machinery is not limiting during the localization process.

We have focused on the control of localization at the 4-cell stage, but examination of the behavior of LEs at other stages is informative regarding how the specificity and complexity of localization are achieved. The endogenous IoLR5 and IoLR1 RNAs are both localized to centrosomes at the 4-cell stage, but at the 2-cell stage only IoLR5 is localized to the centrosomes (Kingsley et al., 2007). We mapped the sequences required for centrosomal localization of IoLR5 at the 2-cell stage, and found that this required the same region as for 4-cell localization, i.e. IoLR5LE (Fig. 7A,B). We next tested 2-cell stage localization for IoLR1(1059-2045), which contains the LE for 4-cell stage localization. Surprisingly, this injected RNA was occasionally capable of very weak localization at the 2-cell stage (Fig. 7C). Since the IoLR1 fragment only had weak and variable localization, we did not attempt fine-scale mapping for this region.

The endogenous IoLR5 and IoLR1 transcripts are highly specific for the first quartet and are not found in the second or third quartet micromeres. Since our injected RNAs were not completely segregated, we were able to follow injected RNAs later in development and test for localization in cells in which these messages are not normally found. We discovered that the injected RNAs can localize to macromere centrosomes at the 8-cell stage (Fig. 7D,E). Furthermore, at the 24-cell stage, when the first, second and third quartets are all present, the injected RNAs were centrosomally localized in all cells of the embryo (Fig. 7F,G), even though the endogenous RNAs are never observed in many of these cells. Similarly, the IoLR5LE RNA was centrosomally localized in the 1q¹ cells at the 24-cell stage, even though the endogenous message is always cytoplasmic at this stage. The injected RNAs were not appreciably degraded, which might be important for establishing the normal pattern of endogenous transcripts. These results indicate that transcript abundance is tightly regulated to ensure the proper distribution of segregated RNAs and they hint that other mechanisms are in place that can prevent transcript localization, even when an LE is present.

One of the hallmarks of spiralian development is the patterning of the animal-vegetal axis by the production of tiers of cells, called quartets, with different developmental potentials, but the mechanisms that control this are unknown. Our results show that a segregated RNA is functionally important for quartet-specific developmental potential. The IoLR5 mRNA is specifically segregated into the first quartet of micromeres, which generate the larval head. Inhibiting IoLR5 translation impairs the development of larval head structures, particularly the eyes and velum, demonstrating the functional role of the protein in development. There is no IoLR5 expression in the cells that become the eyes, indicating that the effect is non-cell-autonomous. The eyes develop in the optic ganglia, which arise in contact with the cerebral ganglia where IoLR5 protein is found (Gibson, 1984; Lin and Leise, 1996). Thus, our results suggest that the proper development of the eyes or optic ganglia might require previously unappreciated signaling cues from proneural or neural tissues.

Our experiments suggest a model of how first quartet-specific localization arises. We characterized an RNA element that mediates localization to the centrosome at the 4-cell stage and segregation to the first quartet. We found a similar element in another RNA that is segregated to the first quartet, and we propose that these two elements define a class of RNA sequences that mediate localization to the first quartet and thus contribute to first quartet identity. These elements are also capable of mediating segregation of injected RNAs to other quartets, even though the endogenous messages are only segregated into the first quartet. Thus, our data indicate that specific first quartet localization is achieved by ensuring that no significant population of a first quartet RNA molecule remains in the macromeres to be segregated in the next mitosis. This seems to be achieved by highly efficient transport mechanisms and, in some cases, by RNA decay. The high capacity and efficiency of this transport are supported by several lines of evidence. First, for most first quartet RNAs, such as IoLR5 and IoLR1, we cannot detect any RNA in the macromeres immediately following division. Second, we have found that centrosomal localization can occur as late as 5 minutes after the beginning of the 4-cell stage, but that endogenous IoLR5 and IoLR1 RNAs are localized well before this, during cytokinesis (see Kingsley et al., 2007). Finally, we attempted to saturate the transport machinery by injecting high molar excesses of IoLR5LE RNA, but were unable to competitively inhibit endogenous IoLR5 or IoLR1 RNA, or injected IoLR1 RNA, from localizing to the centrosome. RNA decay also appears to play a role in maintaining quartet specificity. In the case of IoEve mRNA, which is segregated to the first quartet, a fraction of RNAs remain in the macromeres, and first quartet specificity is achieved by decay of the message in the macromeres (Lambert and Nagy, 2002). Thus, first quartet-specific segregation appears to be controlled by a combination of specific RNA secondary structures and regulation of transcript abundance.

A number of aspects of RNA localization in the early Ilyanassa embryo remain unclear. We do not understand why our injected RNAs can perfectly recapitulate the specificity of centrosomal localization, but are less specific later in the cell cycle and during segregation. We also do not understand how the transition to cytoplasmic distribution is controlled: injected IoLR5 is centrosomal in the 1q¹ cells at the 24-cell stage, but endogenous IoLR5 is cytoplasmic in these cells. Finally, despite the fact that we have identified sequence elements that will segregate into the second and third quartets, we do not know how localization to other quartets occurs. We have injected putatively full-length RNAs for several transcripts that are segregated to the second and third quartets and never observed localization (our unpublished results), suggesting that the mechanism used is different from that employed during first quartet localization.

For the localized RNAs that have been characterized in other organisms, such as bicoid, oskar and Vg1, translation is typically required soon after localization or fertilization of the oocyte (Kugler and Lasko, 2009; Martin and Ephrussi, 2009; St Johnston, 2005). The IoLR5 mRNA is asymmetrically segregated very early during embryogenesis, during the third cleavage, and it is reiteratively segregated through several more rounds of divisions to the 1q¹²¹ cells, before the protein is detectable. The protein is maintained in these cell lineages for several days and is eventually expressed in differentiated neurons; it is also required for development of the eyes, which form relatively late in organogenesis. The long interval of progressive segregation is striking. It suggests that for at least some factors, very fine-scale patterning is achieved by sequential segregation, rather than by a hierarchical network of patterning factors that progressively subdivide the embryo, as observed frequently in somatic patterning of other systems. Although it is uncommon for somatic RNAs to be regulated in this manner during embryogenesis, with the exception of several RNAs studied in ascidian embryos (Nishida and Sawada, 2001), this mode of patterning is very common for germ plasm. In many organisms, components of the germ plasm are segregated through several rounds of cell division to the lineage that will form the germ line (Lehmann and Ephrussi, 1994; Seydoux and Schedl, 2001). This relationship to germ plasm is intriguing because the centrosome-mediated RNA localization seen in Ilyanassa might be homologous to a process observed in oogenesis across the animal kingdom – localization to the Balbiani body (Lambert, 2009). In X. laevis, D. melanogaster and D. rerio, RNAs and other cargoes localize to this structure during oogenesis before transport to the cortex, similar to the centrosomal RNAs in Ilyanassa (Cox and Spradling, 2003; Heinrich and Deshler, 2009; Kloc et al., 2004b; Marlow and Mullins, 2008). At least in frogs, in which the ultrastructure of the developing Balbiani body has been followed, it arises from the oocyte centrosome (Kloc et al., 2004a). We propose that Ilyanassa has extended this conserved patterning mechanism from the oocyte to early cleavage divisions.

This work raises the possibility that all spiralian embryos rely on centrosome-mediated RNA segregation to pattern the micromere quartets. Consistent with this, mRNAs of several putative patterning genes have been shown to be localized to centrosomes and then asymmetrically segregated in the snail Crepidula fornicata (Henry et al., 2010). Similarly, multiple RNAs are localized to the centrosomes in the surf clam Spisula, although this work did not follow the RNAs during cell divisions to see whether they are asymmetrically segregated (Alliegro et al., 2006). Given the strong conservation of the cell fates produced by spiralian micromere quartets, it seems likely that regulatory factors that generate these fates are also conserved across this group. As such factors are identified, it will be interesting to see how their quartet specificity is established in diverse spiralian embryos.

We thank Yingli Duan, Samantha Falk and Michael Cypress for technical assistance and Xin Yi Chan, Pamela Agbu and Hyla Sweet for comments on the manuscript. This work was supported by N.S.F. grants IOB-0544220 and IOB-0844734 to J.D.L.