Dynamic microtubules at the vegetal cortex predict the embryonic axis in zebrafish

In many animals, embryonic axis formation requires polarized transport of maternally deposited factors via the microtubule cytoskeleton. For example, parallel microtubule arrays were shown to be crucial for dorsal axis specification in frog embryos (Elinson and Ninomiya, 2003; Elinson and Rowning, 1988; Rowning et al., 1997). In Drosophila, microtubule orientation and polarity determine directional transport of bicoid and oskar transcripts in oocytes (Brendza et al., 2000; Cha et al., 2001; Glotzer et al., 1997; Pokrywka and Stephenson, 1991; Pokrywka and Stephenson, 1995; Riechmann and Ephrussi, 2001; Schnorrer et al., 2000; Weil et al., 2006; Zimyanin et al., 2008), and fushi tarazu and wingless transcripts in embryos (Bullock and Ish-Horowicz, 2001; Delanoue and Davis, 2005; Lall et al., 1999; Wilkie and Davis, 2001). Disruption of the microtubule cytoskeleton has severe consequences on oocyte polarity and embryonic patterning in these animals (Elinson and Pasceri, 1989; Lecuyer et al., 2007; Steinhauer and Kalderon, 2006).

In zebrafish embryos, the microtubule cytoskeleton has been shown to have several important functions. Epiboly movements, yolk syncytial layer formation and axonal transport are all processes that require an intact microtubule cytoskeleton (Du et al., 2012; Guzik and Goldstein, 2004; Hsu et al., 2006; Lachnit et al., 2008; Solnica-Krezel and Driever, 1994; Strahle and Jesuthasan, 1993; Takesono et al., 2012). Parallel microtubule arrays in early zebrafish embryos were first described by Jesuthasan and Strahle (Jesuthasan and Strahle, 1997), by Tubulin immunostaining in fixed embryos. However, the dynamics and behavior of these arrays were not investigated further, and the polarity of early microtubule arrays with respect to the embryonic axes was also unknown.

Here, we use transgenic reporters marking the microtubule cytoskeleton and plus-ends of microtubules to investigate microtubule dynamics and polarity in early zebrafish embryos. We show that there are two dynamic microtubule populations in the vegetal cortex that are aligned with or oriented along two of the three embryonic axes. We describe the dynamics and polarity of one of these populations and show that parallel arrays are oriented towards future embryonic dorsal. Thus, the formation and orientation of parallel arrays at 20 minutes post-fertilization (mpf) is the earliest visible marker of future dorsal. Our findings show that microtubules at the vegetal cortex are dynamically remodeled and predict the early embryonic axis in zebrafish.

All experiments were carried out according to the institutional animal care and use guidelines, using standard procedures (Westerfield, 2007). Transgenic zebrafish lines expressing the following microtubule markers were used for live imaging: (1) Tg(ef1α:dclk-GFP) transgene contains two doublecortin domains (microtubule-binding domain) of zebrafish dclk1 fused to gfp, and under the control of the Xenopus ef1α promoter; (2) Tg(zpc:Mmu.Eb1-GFP) fish was generated by injecting a Tol2 plasmid containing a part of zona pellucida protein C promoter (Onichtchouk et al., 2003), and murine EB1-GFP coding sequences (Mimori-Kiyosue et al., 2000), together with Tol2 transposase mRNA into 1-cell stage embryos (Kawakami et al., 2004); (3) Tg(Xen.ef1α:Hsa.EB3-GFP) fish were generated by injecting a Ds plasmid containing the Xenopus ef1α promoter (Johnson and Krieg, 1994) and human EB3-GFP sequences (Stepanova et al., 2003), together with Ac transposase mRNA into 1-cell stage zebrafish embryos (Emelyanov et al., 2006).

Fluorescent Alexa 568 UTP-labeled wnt8a RNA was synthesized from linearized pCS2+wnt8a:wnt8a UTR plasmid DNA, injected into mature eggs from Tg(ef1α:dclk-GFP) females, and fertilized with wild-type sperm (Gilligan et al., 2011; Gore et al., 2005). Fertilized embryos were dechorionated and embedded in 0.7% low melting agarose. RNA localization was examined from 15 mpf onwards, and imaged using a 40× lens with a Numerical Aperture (N.A.) of 1.3 on a custom-built spinning disk system 1 (Nikon Eclipse Ti microscope fitted to a Yokogawa CSU-X1 confocal scan unit, and a Photometrics CoolSnap HQ² camera, controlled by Metamorph 7.7, Molecular Devices).

Transgenic embryos or unfertilized, activated eggs were dechorionated in 30% Danieau’s solution, embedded in 0.6-1.1% low melting agarose on a No. 1.5 cover-slip and imaged at the vegetal pole at 28.5°C. To study the structure and dynamics of microtubules, Tg(ef1α:dclk-GFP) activated eggs and embryos were imaged using a 100× lens (N.A. 1.4), a 63× lens (N.A. 1.4) or a 40× lens (N.A. 1.3) on a custom-built spinning disk system 2 (Zeiss Axiovert 200M microscope fitted to a Yokogawa CSU21 confocal scan unit, and a Hamamatsu C4742-80-12AG camera, controlled by Metamorph 7.6, Molecular Devices). z-stacks were acquired with step sizes of 0.3-0.6 μm at 15- to 90-second intervals, starting at ∼0.2 hpf and imaged for ∼1 hour.

To examine the orientation of parallel microtubules, Tg(Xen.ef1α:Hsa.EB3-GFP) embryos were manually dechorionated, and imaged at a single z-plane at 0.8-second intervals on the spinning disk system 2 using a 100× lens (N.A. 1.4). Tg(zpc:Mmu.Eb1-GFP) embryos were dechorionated with pronase (1 mg/ml in E3 medium), and imaged at 4-second intervals on a LSM700 confocal microscope (Carl Zeiss) using either a 40× (N.A. 1.3) or a 63× lens (N.A. 1.4). Embryos were kept in agarose and imaged at gastrula stages to identify the position of the shield.

The three-dimensional (3D) structure of microtubules at the vegetal cortex was rendered by Imaris 7.3 or 7.4 (Bitplane), exported as .tif files, and processed using MBF-ImageJ (http://www.macbiophotonics.ca/imagej/) core 1.46 and the Resize plugin (Munoz et al., 2001). Images in Fig. 1A-D and Fig. 5A-H were 3D deconvoluted by AutoQuant X2.1.3 (Media Cybernetics) using default settings (adaptive PSF, ten iterations, medium noise) and visualized by Imaris. To track cortical granule movement, a two-dimensional (2D) maximum intensity z-projection was generated using Metamorph. Cortical granules (apparent as dark spherical areas lacking fluorescent Dclk2-GFP signal in videos) were tracked using MTrackJ plugin (Meijering et al., 2012) in ImageJ 1.46. We tracked 45 cortical granules in five embryos, from ∼14 mpf to 40 mpf. Measurement data were exported to a tab-delimited file and analyzed using Microsoft Excel. To study movement of EB1-GFP, 50 comets at ∼20 mpf, and 30 comets at ∼ 35 mpf, were tracked in each embryo using the Manual Tracking plugin in ImageJ 1.43u. The position of each comet was determined at three consecutive time points, and slope of the track was calculated by the LINEST function in Microsoft Excel. Deviation angles between EB1-GFP tracks and the embryonic shield were calculated, assigned into classes of 10°, ranging from –180° to +180°, and plotted as a histogram.

To examine the correlation between parallel arrays and the dorsal organizer, parallel microtubules at the vegetal cortex of Tg(ef1α:dclk-GFP) embryos mounted in 0.9-1.1% low melting agarose were imaged at ∼20-40 mpf using a 63× objective lens with N.A. 1.25 on a wide-field Zeiss Axiovert 200M microscope. Embryos were allowed to develop and were imaged during gastrulation to identify the position of the organizer using a 10× objective lens (N.A. 0.3). The deviation angle between parallel arrays and the organizer was measured by drawing a line representing the direction of parallel arrays, and determining its position on the gastrula stage image relative to the embryonic shield, by ROI Manager. The angle between the line and the shield was measured using the ‘angle’ tool of ImageJ 1.46. Measurements were exported to a tab-delimited file and analyzed using Microsoft Excel. Deviation angles were calculated and assigned into classes of 10°, ranging from –90° to +90° and plotted as a histogram.

Embryos or unfertilized, activated eggs were fixed and immunostained for α-Tubulin as described (Topczewski and Solnica-Krezel, 1999). Briefly, embryos were dechorionated and fixed in microtubule assembly buffer [80 mM KPIPES (pH 6.5), 5 mM EGTA, 1 mM MgCl₂, 3.7% formaldehyde, 0.25% glutaraldehyde, 0.5 μM taxol, and 0.2% Triton X-100] for 6 hours at room temperature (RT). Fixed embryos were dehydrated and kept in methanol at –20°C overnight, rehydrated in phosphate buffered saline (PBS) containing 0.1% NP40, incubated in 100 mM NaBH₄ in PBS for 6-16 hours at RT, and washed extensively in tris-buffered saline (TBS). After blocking in TBS containing 2% BSA and 5% normal goat serum for 30 minutes at RT, embryos were incubated with mouse monoclonal antibody against human α-tubulin (Sigma-Aldrich, clone B-5-1-2, 1:400 dilution) at 4°C overnight, washed with five changes of TBS at RT, and incubated with Alexa488-conjugated goat anti-mouse secondary antibody (Invitrogen, 1:4000) for 2-3 hours at RT. Embryos were washed in TBS at 4°C overnight, cleared, and mounted in 100% glycerol for imaging on a Leica SP5 confocal microscope using 100× (N.A. 1.35), 40× (N.A. 1.25) or 20× lenses (N.A. 0.7). To detect nuclear β-Catenin, 256-cell and 1000-cell stage embryos were fixed in 4% paraformaldehyde in PBS and processed for fluorescent immunohistochemistry using a rabbit polyclonal anti-β-Catenin antibody (C2206, Sigma) and Alexa488-conjugated goat anti-rabbit secondary antibodies (Molecular Probes). Images of β-Catenin- and DAPI-stained embryos were acquired using a Zeiss LSM 5 Exciter upright confocal microscope. Approximately 15-25 optical sections at 1.76 μm intervals were examined per embryo. We detected β-Catenin-positive nuclei in many sections. However, owing to intense membrane and cytoplasmic β-Catenin staining that obscured nuclear staining upon z-projection of all sections obtained, three serial confocal sections for each embryo were selected, z-projected using LSM Image-browser software, and cropped using Adobe Photoshop.

Unfertilized eggs were activated upon placing in egg water. Activated wild-type eggs and fertilized embryos were incubated in 1 μM thapsigargin (Fermentek) from 5 minutes post-activation (mpa), dechorionated, and fixed at 25 mpa or 25 mpf, respectively. Fixed eggs and embryos were processed for immunostaining with α-Tubulin antibodies to detect parallel arrays. For nocodazole-bead treatments, Affi-Gel blue beads (Bio-Rad) were soaked in 25 μg/ml nocodazole (Sigma-Aldrich) for at least 30 minutes at RT and embedded in agarose, abutting Tg(ef1α:dclk-GFP) embryos. Microtubule images were acquired with a 40× lens (N.A. 1.3) on the custom-built spinning disk system 1.

We obtained embryos from Tg(ef1α:dclk-GFP) fish and observed microtubule dynamics at the vegetal cortex soon after fertilization. Several bundles of microtubules are observed ∼15-30 μm from the cortex (Fig. 1A-D; supplementary material Movie 1). These bundles are perpendicular to the vegetal cortex, and move towards the animal pole. Perpendicular bundles are transient, detected only in the first 10-30 mpf and are not observed subsequently. As perpendicular bundles are oriented along the animal-vegetal axis, they are likely to transport maternal factors, such as sqt (ndr1 – Zebrafish Information Network) RNA (Gore et al., 2005; Gore and Sampath, 2002), through the yolk to the blastoderm. Because these arrays project deep into the yolk, visualizing this population in live embryos is difficult beyond 75-100 μm from the vegetal cortex with current live imaging methodologies. It is not known whether similar perpendicular bundles are present in other chordate embryos.

Concurrently with the perpendicular bundles, we observed a second population of microtubules at the vegetal pole, located more cortically. These are parallel microtubule arrays, which are also transient and dynamic. Parallel arrays grow rapidly from ∼15 mpf (Fig. 1E-I), elongate, bundle, and begin to dissociate from ∼25 mpf (Fig. 1J,K). Parallel arrays cannot be detected by ∼30 mpf (Fig. 1L), and are replaced by a non-directional meshwork of microtubules (supplementary material Movie 2). Thus, both perpendicular and parallel arrays of microtubules are transiently detected at the vegetal cortex and are observed only within the first 30 mpf.

Immunostaining using antibodies to detect α-Tubulin shows that parallel microtubule arrays become asymmetric (Fig. 2A,B) with respect to the vegetal pole of the embryo. Parallel arrays span a region encompassing a ∼45° sector of the embryo and extend nearly a third of the distance from the vegetal pole towards the animal pole of the embryo. Thus, parallel microtubule arrays cover an extensive region of the early embryo, and we found their spread to be similar to the expression domain of asymmetrically localized maternal wnt8a transcripts (Fig. 2C; supplementary material Fig. S1) in early embryos (Lu et al., 2011). Therefore, we investigated whether wnt8a RNA colocalizes with the parallel arrays. Alexa568 UTP-labeled fluorescent wnt8a RNA was injected into eggs from Tg(ef1α:dclk-GFP) females, which were then fertilized and observed at the vegetal pole. Punctate aggregates of fluorescent wnt8a RNA can be detected localizing with the Dclk2-GFP-marked parallel arrays by ∼20 mpf (Fig. 2D; n=6 embryos), and by 35 mpf, similar to endogenous wnt8a RNA, injected fluorescent RNA is detected asymmetrically on the vegetal lateral cortex and in the blastoderm (Fig. 2I). Furthermore, fluorescent wnt8a punctae are not observed outside of the region covered by the parallel arrays (see bottom left area of Fig. 2D). Thus, wnt8a RNA colocalizes with parallel microtubule arrays.

To determine whether the asymmetric localization of the parallel arrays corresponds to the embryonic axes, we observed the arrays as they formed in real time in agarose-embedded Tg(ef1α:dclk-GFP) embryos, and marked the orientation along which parallel arrays were aligned (Fig. 2F,H, blue line). We later marked where the dorsal organizer, the shield (Fig. 2G,H, red arrow), formed in each embryo during gastrulation and determined the difference in angle (δ) between orientation of the parallel arrays in early embryos and the position of the shield (Fig. 2H). In most embryos (>90%, n=23), the shield forms within 30° of the orientation of parallel arrays (Fig. 2K). Thus, orientation of parallel arrays corresponds to future embryonic dorsal.

To determine the directionality of parallel arrays, we used two transgenic lines labeling the growing (‘plus’) ends of microtubules, Tg(zpc:Mmu.Eb1-GFP) and Tg(Xen.ef1α:Hsa.EB3-GFP), and observed the growing tips of microtubules by live imaging (supplementary material Movies 3-6). Comets of EB1-GFP plus-ends were tracked over time and the direction of the plus-ends versus the position of the dorsal shield at gastrula stages was determined (Fig. 2E,J). Within the first 30 mpf, most plus-ends are oriented directionally, towards future embryonic dorsal (Fig. 2L). The angle of deviation (δ), between the direction of plus-ends and the shield, is <20° for >80% of the tracks (Fig. 2L; supplementary material Movie 3). However, after 30 mpf, when parallel arrays dissociate, EB1-GFP comets move in all directions and the deviation angle shows a range (Fig. 2M; supplementary material Movie 4). Similar findings were obtained using embryos expressing a different plus-end binding protein, EB3-GFP (supplementary material Movies 5, 6). We tracked the speed of the comets and find that the average EB3-GFP particle speed before 30 mpf (when parallel arrays are present and comets move directionally), is 0.38 μm/second (n=40 comets), and this increases to 0.62 μm/second (n=30 comets) when the EB3-GFP comets move in all directions. Therefore, the plus-ends of parallel arrays at the vegetal cortex of zebrafish embryos are oriented towards future embryonic dorsal.

We observed several dense cytoplasmic granules (Fig. 3A, dotted circles) at the vegetal cortex, which appear to move within the first 30 mpf when parallel arrays are present (supplementary material Movie 7). By tracking individual granules in a time course (Fig. 3A-C, numbered granules), we find that they move rapidly and directionally initially (0.17 μm/second, n=29 granules at ∼20 mpf; Fig. 3D, Phase 1), and after 30 mpf, the granules slow down (0.02 μm/second, n=38 granules at 35 mpf; Fig. 3D, Phase 2), and their movement, if any, is random (supplementary material Fig. S2). Phase 1 correlates with the presence of parallel arrays, and in Phase 2, parallel arrays disappear, giving way to a dense non-directional microtubule meshwork. We find that individual granules move >70 μm in Phase I, and some granules move up to 110 μm, whereas in Phase 2, the granules typically move <8 μm (Fig. 3E) and their movement is non-directional. The distance of 70-100 μm traveled by granules along parallel arrays at the vegetal cortex (in embryos with a diameter of ∼500 μm) corresponds to a ∼20° displacement of the granules along the cortex. These results suggest that vegetal cytoplasmic granules in zebrafish embryos undergo a cortical rotation-like shift, similar to that observed in early Xenopus embryos.

To determine whether granule movement is dependent on parallel microtubule arrays, we transiently disrupted microtubules in a small region of the cortex by local application of a bead soaked in nocodazole (Fig. 4A-D, schematic). The extent of disruption of microtubules in bead-treated embryos was determined by examining the Dclk2-GFP parallel arrays (Fig. 4E-H). We tracked the movement of individual granules in bead-implanted embryos (Fig. 4E-H, yellow asterisks). Granules in embryos treated with control DMSO beads move rapidly and directionally, and their net displacement is at least 82 μm (mean 99 μm; n=8 granules from two embryos; Fig. 4A,E,I,M,S; supplementary material Movie 8). Granules in close proximity to nocodazole beads (Fig. 4B,D) do not move directionally (Fig. 4J,L,N,P; supplementary material Movies 9, 10), exhibit sudden changes in direction (see granule trajectories in Fig. 4N), and the net distance traveled is also reduced to ∼25 μm (n=24 granules from two embryos; Fig. 4S). Granules located further from nocodazole beads (Fig. 4C) move directionally (Fig. 4K,O; supplementary material Movie 10) together with parallel arrays that are visible (Fig. 4G). Thus, directional movement of vegetal cortical granules is dependent on parallel microtubule arrays.

To determine how the dorsoventral axis is affected by transient nocodazole-bead treatment and parallel array disruption, we examined the nuclear accumulation of the early dorsal marker β-Catenin. In DMSO-bead treated embryos, in which normal parallel arrays formed (supplementary material Fig. S3A), we detected β-Catenin in many dorsal nuclei (Fig. 4Q; supplementary material Fig. S3B-D) whereas in nocodazole-bead treated embryos with disrupted parallel arrays (supplementary material Fig. S3E), we did not detect any nuclear β-Catenin accumulation (Fig. 4R; supplementary material Fig. S3F-H). Therefore, local disruption of parallel microtubule arrays affects the directional transport of vegetal cytoplasmic granules and disrupts specification of the dorsal axis.

The vegetal cortex of zebrafish embryos has dense cytoplasmic granules that appear to undergo cortical rotation-like movement similar to that observed in Xenopus embryos. In Xenopus, it is thought that sperm entry and fertilization trigger movement of the cortex (Gerhart et al., 1989; Gerhart et al., 1981; Vincent et al., 1986), which then shifts by 30° relative to the point of sperm entry (Vincent et al., 1986). However, in zebrafish, the sperm entry point is fixed, and is a defined channel in the chorionic membrane, the micropyle, located at the animal pole of the egg (Hart and Donovan, 1983). So, we tested whether fertilization is even necessary for formation of parallel arrays. We artificially activated squeezed mature eggs and examined microtubule dynamics at the vegetal cortex. Remarkably, parallel arrays form in unfertilized, activated zebrafish eggs at ∼20 mpf (Fig. 5A-H; supplementary material Movie 11). Parallel microtubule arrays in unfertilized, activated eggs become elongated and dense, and, similar to arrays in fertilized embryos, they dissociate by 34 mpf. Thus, parallel arrays form independently of sperm entry and fertilization.

So what might trigger parallel array formation? Waves of Ca²⁺ transients have been observed in early zebrafish eggs (Lee et al., 1999; Leung et al., 1998), and disruption of Ca²⁺ signaling is thought to affect embryonic dorsoventral patterning (Westfall et al., 2003). Therefore, we tested whether Ca²⁺ signaling is required for parallel array formation. Activated unfertilized eggs were treated with the drug thapsigargin, which inhibits intracellular Ca²⁺ pumps, and formation of arrays was observed in treated eggs. Unactivated eggs do not show any parallel arrays (Fig. 5I). In contrast to control DMSO-treated eggs (n=27; Fig. 5J,O), thapsigargin-treated eggs either show reduced/abnormal bundles (16%, n=44; Fig. 5K,O) or no parallel arrays at all (84%, n=44; Fig. 5L,O). Therefore, formation of parallel arrays depends on Ca²⁺ signaling.

Ca²⁺ signaling has been shown to be a trigger for cortical granule exocytosis in zebrafish. The cortical microtubule network in maternal brom bones/hnRNP1 mutant embryos, which are defective in dorsoventral patterning, is disorganized, presumably owing to failure of granule exocytosis, and can be rescued by providing Ca²⁺ (Mei et al., 2009; Wagner et al., 2004). Thus, in our drug-treated embryos, and in the Westfall et al. experiments (Westfall et al., 2003), the reduced/abnormal parallel arrays caused by Ca²⁺ inhibition probably lead to dorsoventral defects. To test this, we examined nuclear accumulation of β-Catenin in dorsal cells of thapsigargin-treated embryos. We find that in comparison to control DMSO-treated embryos, which show numerous β-Catenin-positive nuclei (Fig. 5M; supplementary material Fig. S4A-C), there were no detectable β-Catenin-positive nuclei in thapsigargin-treated embryos (Fig. 5N; supplementary material Fig. S4D-F). Thus, disruption of parallel arrays by interfering with Ca²⁺ signaling leads to disruption of dorsal axis formation.

Our results show that parallel arrays form autonomously even in activated eggs. This had been suggested from observations of sub-cortical arrays in activated frog eggs (Elinson and Rowning, 1988) and in vegetal fragments of amphibian embryos (Elinson and Palecek, 1993), and it has been suggested that the sperm aster only biases the direction of parallel arrays rather than triggers formation of parallel arrays (Gerhart, 2004). However, as the explants/fragments involved surgical manipulations that have the potential to change intracellular Ca²⁺ fluxes and microtubule polymerization, and thereby indirectly affect microtubule dynamics, it was not entirely clear whether sperm entry and fertilization were necessary for parallel array formation in this organism (Houliston and Elinson, 1991). Thus, our in vivo finding that formation of parallel arrays in zebrafish is independent of sperm entry and fertilization resolves this long-standing question. Our findings are also consistent with the observation that haploid zebrafish embryos still have a normal dorsal axis (Streisinger et al., 1981).

EB1-GFP and EB3-GFP comet movement shows that parallel microtubule arrays are oriented with their plus-ends towards dorsal. Thus, parallel microtubules can directionally transport dorsal determinants from the vegetal pole. These include maternal wnt8a RNA, as we have shown, and it is likely that the maternal linker/adapter protein Syntabulin is also transported along parallel microtubule arrays (Lu et al., 2011; Nojima et al., 2010). It is also possible that Wnt8a protein is transported. In Xenopus, vesicle-like organelles are enriched in future dorsal cells after cortical rotation, and a Dsh-GFP fusion protein associates with the vesicle-like structures (Miller et al., 1999). GBP-GFP and KLC-GFP particles also exhibit directed translocation during cortical rotation (Weaver et al., 2003). In zebrafish, we observed directional movement of cortical granules along parallel microtubule arrays. It is possible that the transported cortical granules contain components of the Wnt pathway.

This also raises the question of how general cortical rotation-like processes are. There is evidence for a rotation-like activity in actinopterygii and acipenserids, including sturgeons (Abraham et al., 1995; Dettlaff, 1962). Hence, this process has been previously observed in animals other than Xenopus and Rana (Elinson and Rowning, 1988; Gerhart et al., 1989; Houliston and Elinson, 1991). A rotation-like process has been suggested in lampreys as well (Clavert, 1962). Although it has been reported that eggs of the teleost Medaka do not manifest cortical rotation (Abraham et al., 1995), our results suggest that an apparently similar movement takes place in zebrafish. The expression patterns of maternal wnt8a RNA and Syntabulin (also known as Tokkaebi) protein in early zebrafish embryos (Lu et al., 2011; Nojima et al., 2010) also support such a movement. Although we did not find any direct evidence for movement of the cortex per se, asymmetry of the parallel arrays and the directional movement of the vegetal yolk cytoplasm that accompanies it might underlie the transport and displacement of various maternal factors towards future embryonic dorsal in zebrafish.

We find that parallel microtubule arrays are spread over an extensive area spanning ∼250 μm in the yolk cell vegetal region. Maternal wnt8a transcripts, which we find colocalizing with the parallel arrays, are also distributed over a fairly broad domain of ∼250 μm (Fig. 2C). Therefore, the maternal Wnt8 signal from the yolk, which is likely to be diffusible, can potentially access at least one-fifth to one-sixth of the blastoderm margin (Fig. 6). Additionally, there is also wnt8a expression in the blastoderm (Lu et al., 2011). However, β-Catenin accumulates in the nuclei of a very small number of dorsal cells ((Dougan et al., 2003; Lim et al., 2012). Based on antisense knockdown experiments, Sfrp1a and Frzb were suggested as the factors that negatively regulate Wnt8 activity in the blastoderm to restrict β-Catenin accumulation (Lu et al., 2011). The expression pattern of maternal sfrp1a and frzb transcripts is not spatially restricted (B. Thisse and C. Thisse, personal communication), and zygotic sfrp1 and frzb RNA are enriched later in dorsal cells (Peng and Westerfield, 2006; Tendeng and Houart, 2006). These transcripts encode secreted Wnt antagonists, which should be freely diffusible as well (Lee et al., 2006; Lin et al., 1997; Mii and Taira, 2009; Pera and De Robertis, 2000; Ploper et al., 2011; Wang et al., 1997). Therefore, it is not clear whether the uniform, maternal Sfrp1a and Frzb activity is sufficient to restrict the response of the broad, maternally expressed dorsal Wnt signal to a few marginal cells. It is possible that additional factors and mechanisms function to limit the range of maternal Wnt activity and to spatially restrict nuclear β-Catenin accumulation to dorsal cells. Identification of these factors will help clarify how maternal wnt8a, transported by the vegetal microtubule parallel arrays, is regulated to form future dorsal structures.

We thank members of the Hibi laboratory, Sampath laboratory, the Singapore fish community, Mohan Balasubramanian, Claire Canning, Meredith Calvert, Mithilesh Mishra, and Shijie Tao for discussions and suggestions; Hideaki Nojima for generating the Eb1-gfp transgenic line; the TLL sequencing, imaging and fish facilities and the NUS CBIS facility.