Notch1 is asymmetrically distributed from the beginning of embryogenesis and controls the ventral center

During the past two decades, many studies, mostly performed in amphibian and fish embryos, have demonstrated that an antagonistic interplay between two signaling centers, a ventral center (VC) and a dorsal center (the gastrula Spemann–Mangold's organizer, SMO), pattern the dorsoventral (DV) and anteroposterior body axes in vertebrates. The VC, which has been described at the gastrula stage, secretes BMP4 and Wnt8a (De Robertis, 2009; Thisse and Thisse, 2015). These morphogens induce ventral-posterior fates in the embryo, whereas the dorsal center secretes antagonists and expresses transcriptional repressors of the ventral morphogens, thus protecting the dorsal region from their ventralizing and posteriorizing activities (De Robertis, 2009; Thisse and Thisse, 2015). The SMO derives from the ‘blastula chordin- and noggin-expressing center’ (BCNE), which appears when zygotic transcription begins. The BCNE secretes the BMP antagonists that confer its name, and also contains the precursors of the brain. Its formation is triggered by a Wnt/β-catenin cascade activated as a consequence of the relocation of dorsal determinants after fertilization (Kuroda et al., 2004). β-Catenin accumulates in dorsal nuclei and activates the transcription of Siamois-related homeobox (Sia) genes. These homeodomain proteins are the first transcription factors expressed in the BCNE and, in turn, both repress (probably indirectly) bmp4 transcription and transiently activate the expression of the BMP antagonists in the prospective anterior neuroectoderm. This double inhibition of BMP signaling, which is initiated at blastula stages, is crucial for brain formation (Wessely et al., 2001; Kuroda et al., 2004; De Robertis and Kuroda, 2004; Ishibashi et al., 2008). Importantly, whether an event of asymmetric distribution of ventral determinants analogous to that triggered by Wnt dorsal signaling occurs in the earliest stages of embryogenesis has not yet been determined, as molecules involved in ventral development have not shown evidence of an early uneven allocation in the initial DV axis (see below).

Knockdown experiments have demonstrated that the most important BMP signals for ventral development are BMP4 and BMP7 (Reversade et al., 2005). In Xenopus, bmp7.2 transcripts (but not bmp4 transcripts, which begin to accumulate during gastrulation) are stored to a significant level in eggs (Nishimatsu et al., 1992; Session et al., 2016). During early cleavage stages, bmp7.2 mRNA is ubiquitously distributed in the animal hemisphere, and zygotic bmp4 and bmp7.2 mRNAs show a uniform expression in ectoderm and mesoderm in late blastulae. During gastrulation, bmp4 and bmp7.1 transcripts accumulate at high levels in the VC, and bmp7.2 mRNA retains a uniform distribution in the entire mesoderm (Hemmati-Brivanlou and Thomsen, 1995; Hawley et al., 1995; Kirmizitas et al., 2017). BMPs act by activating the expression of the Ventx family of ventral transcriptional repressors (Gawantka et al., 1995; Onichtchouk et al., 1996; Shapira et al., 1999; Rastegar et al., 1999; Henningfeld et al., 2000). The BMP signaling cascade directly activates vent-2B transcription at the promoter level (Henningfeld et al., 2000).

Wnt8a transcripts are not maternally stored in Xenopus. They begin to accumulate in the ventral region at late blastula (Session et al., 2016; Smith and Harland, 1991), being detectable by in situ hybridization (ISH) earlier than bmp4. Overexpression of wnt8a after the mid-blastula stage, by injection of plasmid DNA, gives rise to ventralized phenotypes (Christian and Moon, 1993), whereas its knockdown results in dorsalization (Li et al., 2006). It has been suggested that wnt8a transcription probably depends on maternal BMP signaling, as a dominant-negative BMP receptor suppressed wnt8a expression (Hoppler and Moon, 1998). Both Wnt8a and BMP signaling are cooperatively required for ventrolateral mesoderm development, by promoting the expression of Ventx genes (Hoppler and Moon, 1998). Also, in zebrafish, both Wnt8a and zygotic BMP are necessary for the maintenance of Ventx transcript levels during gastrulation; however, Wnt8a is required earlier for this task, at the blastula/gastrula transition, whereas zygotic BMP is the main positive regulator of Ventx genes at mid- to late gastrula (Ramel and Lekven, 2004). Wnt8a expressed after the mid-blastula transition operates through canonical β-catenin/Tcf signaling by stabilizing cytosolic, free β-catenin, which activates gene expression in a Tcf-dependent fashion, as shown in Xenopus (Darken and Wilson, 2001). Moreover, Wnt8a induces transcription of Ventx genes by the canonical β-catenin/Tcf pathway, as shown in zebrafish (Ramel and Lekven, 2004).

Ventx genes start their expression at mid-blastula transition (Session et al., 2016). There are different subclasses of these genes in Xenopus (Scerbo et al., 2012). Their transcripts show similar expression patterns, being uniformly expressed in the animal hemisphere of blastulae and later restricted to the VC during gastrulation (Gawantka et al., 1995; Onichtchouk et al., 1996; Rastegar et al., 1999; Shapira et al., 1999; Nordin and LaBonne, 2014). Ventx proteins promote ventralization, and exert their action by directly repressing organizer-specific genes (Gawantka et al., 1995; Onichtchouk et al., 1996, 1998; Shapira et al., 1999, 2000; Sander et al., 2007). It was proposed that the early widespread expression of Ventx repressors establishes an anti-organizer activity throughout the embryo, which is later overcome on the dorsal side because of local activation of Wnt/β-catenin signaling triggered by cortical rotation after fertilization (Shapira et al., 2000). It was also suggested that Ventx proteins protect the prospective ventral region from premature commitment towards dorsal fates rather than directly promoting ventral differentiation (Scerbo et al., 2012).

The transmembrane receptor Notch is involved in multiple developmental programs in metazoans. When the pathway is activated, Notch can stimulate or inhibit cell proliferation, cell death, specific cell fates or differentiation in a context-dependent fashion. In the canonical pathway, the main transcriptional mediator of Notch signaling is known as CSL (from the names in vertebrates, fly and worm: CBF1, Su(H) and Lag1, respectively). The Notch receptor is activated by binding to a DSL ligand (characterized by a Delta/Serrate/LAG-2 motif) present on the surface of a neighboring cell. In the absence of such signaling, a repressor complex containing CSL inhibits the transcription of Notch-target genes. Binding of the DSL ligand to the mature Notch receptor triggers a sequence of proteolytic cleavages (involving the enzyme complex γ-secretase), leading to the release of the Notch intracellular domain (NICD). NICD translocates to the nucleus and binds CSL. The co-repressors bound to CSL are displaced, and co-activators are recruited in a new complex, which activates the transcription of Notch-target genes (Kopan and Ilagan, 2009).

We have previously shown that overexpression of the Notch1 intracellular domain (NICD1) resulted in ventralized phenotypes with a loss of cephalic structures in Xenopus laevis (Xla) embryos (Acosta et al., 2011). This was accompanied by downregulation of the genes encoding the BMP antagonists Chordin and Xnr3 in the BCNE. Conversely, attenuation of notch1 with a translational-blocking morpholino (Notch1 Mo) enhanced dorsoanterior development. Strikingly, when we injected su(H)1^DBM mRNA to impair the canonical, CSL-dependent Notch pathway, we did not observe an expansion of the chordin or Xnr3 domains on the animal hemisphere at blastula s9. We found that Notch1 restricts the expression of these BMP antagonists to the BCNE center by destabilizing maternal β-catenin rather than by a mechanism that relies on CSL-mediated transcription. Gain of function of Notch1 abolished the dorsalizing activity of β-catenin by reducing its steady state levels before mid-blastula transition, even if the β-catenin protein lacked the glycogen synthase kinase-3β (GSK-3β) phosphorylation sites. Notch1 Mo increased the levels of endogenous β-catenin protein, and those of the β-catenin form resistant to GSK-3β-mediated degradation. Thus, we proposed that Notch1 has ventralizing properties, and that early ventral Notch1 activity contributes to the restriction of the BCNE center to the dorsal side of the blastula. This occurs through a non-canonical mechanism via the destabilization of β-catenin in a process that does not require β-catenin phosphorylation by GSK-3β. Through restricting the extent of the BCNE within the animal hemisphere, Notch1 controls the size and the anteroposterior pattern of the brain (Acosta et al., 2011). This negative regulation of β-catenin by Notch1 strongly resembles that described in the wing imaginal disc in Drosophila (Hayward et al., 2005, 2008; Sanders et al., 2009; Muñoz-Descalzo et al., 2010) or in mouse embryonic day 14 embryonic stem cells (ESCs) (Kwon et al., 2011), in which membrane-bound Notch negatively titrates transcriptionally active hypophosphorylated β-catenin through a physical post-translational interaction that leads to its lysosomal degradation without the intervention of the GSK-3β-dependent activity of the β-catenin destruction complex. Notably, Notch1 is the predominant Notch receptor involved in regulating the levels of active β-catenin in ESCs (Kwon et al., 2011).

In the present work, we addressed whether the Notch ventral activity that antagonizes hypophosphorylated β-catenin (Acosta et al., 2011) could be due to an asymmetric distribution of Notch1 protein and/or mRNA, and whether Notch1 is able to control the development of the VC. Through immunodetection studies, we found that Notch1 protein is consistently enriched in the ventral side from the beginning of embryogenesis in Xla embryos. ISH and RT-qPCR analysis also showed a ventral enrichment of notch1 mRNA, suggesting that the asymmetry of the protein might be due to a mechanism that controls mRNA localization. We also observed an asymmetric distribution of Notch1 protein in zebrafish from the beginning of embryogenesis. By the mid-blastula stage, an opposite distribution of Notch1 and nuclear β-catenin in both species is established. To our knowledge, the accumulation of Notch1 mRNA and protein shown here is the earliest localized sign of ventral development reported in vertebrates. By performing gain- and loss-of-function experiments in Xenopus, we show that Notch1 is required for the normal expression of genes essential for a wholly functional VC.

In this work, to study the distribution of Notch1 protein during early embryogenesis, we first performed a whole-mount immunofluorescence (IF) analysis in left and right halves of Xenopus embryos collected from Nieuwkoop and Faber stage (s) 1 to s8 and then corroborated the results by IF on cryostat sections. Left and right halves were cut from batches of regularly cleaving embryos in which the paler pigmentation zone was bisected by the first cleavage furrow, as the paler region predicts the prospective dorsal side with 70% accuracy in Xla (Klein, 1987). Notably, we found a clear and very consistent asymmetric distribution of Notch1 protein, with the highest levels ventrally and the lower levels dorsally. This was seen in all stages analyzed, in both left and right halves, in cryostat sections (81%, n=28) (Fig. 1A-D′,I-M′,P; Table S1) and in whole-mount IF (89%, n=136, in a total of six independent batches; Figs S1-S4). Control experiments showed that the primary antibody specifically recognizes Notch1 protein in Xenopus embryos (Fig. S5A-F) and that the secondary antibody did not show non-specific fluorescence (Fig. S5G-H′). The ventral enrichment of Notch1 protein was confirmed by western blot (WB) of protein extracts of ventral and dorsal halves obtained at s3 and s7 (Fig. 2A).

To compare the distribution of endogenous Notch1 and β-catenin proteins, we performed whole-mount double IF in left and right halves. Analysis by whole-mount IF in three of the previous batches suggested that, at the earliest stages, the highest levels of total β-catenin protein in some embryos were relatively displaced in relation to the region containing the highest levels of Notch1 (for example, compare white and light blue asterisks in Fig. S1B,C,F,G and in Fig. S2B-D; white asterisks and light blue arrows in Fig. S2F-H,N-P; Table S2). In thin cryostat sections, β-catenin IF was low and noisy at the earliest stages analyzed. We were unable to detect a consistent biased distribution for total β-catenin in the DV axis at early cleavage stages (Fig. 1E-L, Table S1). Cryostat sections of s7 embryos confirmed the ventral enrichment of Notch1 protein (Fig. 1D,D′,P; Fig. S3, Table S1). As we used an antibody that recognizes total β-catenin, a complex pattern was observed at s7, with membrane-associated and cytoplasmic pools, and some immunopositive nuclei that begin to be distinguished (Fig. S3). We obtained mixed results for total β-catenin IF quantification at this stage (Fig. 1, Fig. S3, Table S1). However, in some s7 embryos, it appeared that the enrichment of Notch1 IF in the apical region of the ventral-most animal cells was complemented with lower levels of β-catenin IF, compared with the basal region of these cells, or with dorsal cells exhibiting a more uniform distribution of β-catenin (see cryosections in Fig. 1D,H,L and Fig. S3E-N). In other embryos, we noticed the weakest β-catenin IF in the nuclear region of the ventral-most cells with the highest levels of Notch1 IF (whole-mount hemisection, Fig. S3A-D″). In addition, it is known that dorsal accumulation of transcriptionally active nuclear β-catenin is best detected at s8 (Schohl and Fagotto, 2002). The higher cell number and smaller cell size helped to obtain more DAPI⁺ nuclei in the same plane at this stage. Analysis of images with a high enough resolution to quantify IF intensity at the cellular level (see cryosection in Fig. 1M-O,Q-U) and analysis of the IF patterns in whole-mount hemisections (Fig. S4) showed a consistent opposite distribution of both proteins at s8, with dorsal enrichment of total and nuclear β-catenin and ventral enrichment of total and nuclear Notch1 (Tables S1 and S2).

To verify the distribution of Notch1 protein in the DV axis, we performed a combined assay of Notch1 IF with ISH of known dorsal markers, such as wnt11 for early cleavage (Tao et al., 2005) and chordin for s8 (Kuroda et al., 2004). We found that, in 96% of the cases (n=23), they were clearly expressed at the opposite side in relation to Notch1 (Fig. 2B-G″, Table S3A,B). This higher frequency of ventral Notch1 location, in comparison with that found in the experiments in which we employed the pigmentation criterion described in the Materials and Methods section (Klein, 1987), is because the orientation of the embryos was more accurately assigned by the dorsal marker than by the pigmentation pattern. These results confirm that Notch1 protein is enriched in the ventral side from s1 to mid-blastula.

We conclude that Notch1 protein is consistently enriched in the ventral region from the beginning of embryogenesis in Xenopus. Opposite patterns of Notch1 and β-catenin in the DV axis are consistently evident around the mid-blastula transition, when the strongest nuclear β-catenin accumulation is found on the dorsal side of the embryo, as previously reported (Schohl and Fagotto, 2002).

In addition, we performed double IF with the same antibodies in the more transparent zebrafish embryos, which initially develop as a disc over a large yolk cell. We found an asymmetric distribution of Notch1 protein from the one-cell stage until the blastula (which was the last stage analyzed) in 88% of the embryos (Fig. 3, Table S4), which showed an opposite distribution to β-catenin IF. Although the polarity of the initial DV axis cannot be predicted from the cleavage pattern in zebrafish, accumulation of nuclear β-catenin marks the dorsal side in blastulae (Schneider et al., 1996). At this stage, total β-catenin and nuclear β-catenin were enriched in the region opposite to that in which the strongest Notch1 IF was observed (Fig. 3L-O). This indicates that the Notch1 ventral enrichment is conserved during early embryogenesis in both Xenopus and zebrafish embryos.

To begin to understand why Notch1 protein is asymmetrically distributed from the beginning of embryogenesis, we first analyzed the distribution of notch1 mRNA in pigmented left halves of Xenopus embryos. Notably, we found that notch1 mRNA was enriched in the ventral region in relation to the dorsal side from s1 to the last stage analyzed (s8) in 87% of embryos from three independent batches (Fig. 2H-M; Table S5). We also performed RT-qPCR analysis of notch1 and wnt11 mRNA on samples extracted from dorsal and ventral halves from another two independent batches of embryos (Fig. 2N). In both, wnt11 was significantly enriched in the dorsal half at s5, as has previously been described (Fig. 2N, right) (Tao et al., 2005). In contrast, notch1 mRNA levels were significantly higher in the ventral half samples than in the dorsal ones (Fig. 2N, left), confirming that it is enriched in the ventral region during early cleavage.

The previous results show that both Notch1 mRNA and protein are enriched ventrally. In addition, we were able to detect Notch1⁺ nuclei in the ventral side at cleavage/mid-blastula stages, suggesting that Notch1 protein might be ready to activate transcription. To see whether there is endogenous asymmetric Notch/CSL-dependent activity across the DV axis, we performed a CSL-luciferase reporter assay at the onset of gastrulation. Constitutively active nicd1 mRNA increased the reporter activity in three independent experiments, with significant differences between nicd1 and controls in two of the three experiments (Fig. 2O, left pair of bars), indicating that the assay is sensitive to measuring Notch/CSL-dependent activity in Xenopus embryos. The assay also detected higher endogenous CSL-reporter activity in the ventral halves compared with the dorsal halves in the three independent experiments, with significant differences in two of them (Fig. 2O, right pair of bars). These results indicate that CSL-dependent transcription is higher in the ventral side at the beginning of gastrulation.

Next, we investigated whether notch1 could control the expression of VC genes. We performed gain- and loss-of-function experiments and analyzed the expression of Xenopus ventral markers by ISH around the onset of gastrulation (wnt8a, vent-2b) or at mid/late gastrula for bmp4, as this marker is not well detected by this method at earlier stages. Injection of nicd1 mRNA, which encodes the constitutively transcriptionally active intracellular domain of Notch1 (Chitnis et al., 1995), increased or expanded the expression domains of wnt8a (81%; Fig. 4A-D,K) and vent-2b (87%; Fig. 5A-D,K). Bmp4 was upregulated in some cases (44%; Fig. 6A,B, green asterisk; Fig. 6M) it was downregulated in others (48%; Fig. 6A,D, red asterisk; Fig. 6M) and it was unaffected in a small proportion of injected embryos (8%; Fig. 6A,C, yellow asterisk). Overexpression of the full-length notch1 mRNA (notch1 FL) upregulated the three ventral markers (91% for wnt8a, Fig. 4E,F,K; 91% for vent-2b, Fig. 5E,F,K; 82% for bmp4, Fig. 6E,F,M). In contrast, Notch1 Mo consistently decreased their expression (93% for wnt8a, Fig. 4G,H,K; 88% for vent-2b, Fig. 5G,H,K; 100% for bmp4, Fig. 6G,H,M). In addition, we injected su(H)1^DBM mRNA to impair CSL-dependent, canonical Notch activity (Wettstein et al., 1997). This resulted in a consistent downregulation of wnt8a (93%, Fig. 4I-K) and vent-2b (77%, Fig. 5I-K), but the effect on bmp4 expression was more variable (27% unaffected, 50% downregulated, 23% upregulated on the injected side; Fig. 6I-M).

Three additional independent experiments were carried out using RT-qPCR, to verify the effect of Notch manipulation on ventral markers. Data are shown in Fig. 4L, 5L, 6N and Fig. S6F-J. Consistent with the ISH results for wnt8a and vent-2b, both Notch1 Mo and su(H)1^DBM significantly downregulated all the ventral markers analyzed at early gastrula stage (Fig. 4L and 5L), including bmp4 (Fig. 6N), which was not analyzed by ISH at this stage because it is not well detected by this method at early gastrula. Notch1 Mo kept these markers downregulated at mid gastrula, whereas nicd1 mRNA was able to upregulate them (Fig. 4L and Fig. S6F for wnt8a; Fig. 5L for vent-2b and ventx2.2; Fig. 6N for bmp4).

For one of the RT-qPCR experiments, we also co-injected Notch1 Mo with nicd1 mRNA and assessed the effect on wnt8a at s11, when the Notch target gene hes9.1 (which we used as a control of Notch responsiveness) is well expressed. Nicd1 mRNA rescued the downregulation of wnt8a and hes9.1 produced by Notch1 Mo (Fig. S6F and G, respectively), indicating that the effects of Notch1 Mo were specific. This rescue was confirmed by wnt8a ISH in another two independent experiments (Fig. S6A-E). In addition, Notch1 Mo downregulated other genes important for the ventral program, such as trim29 (Hayes et al., 2007; Cibois et al., 2015; Popov et al., 2017) (Fig. S6H), xk81a1 (Yan et al., 2009; Scerbo et al., 2017) (Fig. S6I) and tfap2a (Luo et al., 2002) (Fig. S6J). Nicd1 mRNA also rescued the effect of Notch1 Mo on these genes (Fig. S6H-J), further confirming the specificity of Notch1 Mo, and that the ventral program requires an intact notch1 function. In conclusion, our results indicate that Notch1 is required for the normal expression of genes with essential roles in the ventral program.

In the present work, we show for the first time that Notch1 protein is enriched in the ventral region from the beginning of embryogenesis in both Xenopus and zebrafish (Fig. 7). Our results suggest that a mechanism that controls notch1 mRNA localization underlies this asymmetry of Notch1 protein.

Notch1 establishes an opposite distribution in relation to dorsal nuclear β-catenin at blastula stages, consistent with our previous hypothesis of a Notch1-mediated destabilization of transcriptionally active β-catenin in the ventral side (Fig. 7, route 1). Further work is needed to understand how this mechanism operates at the cellular level. β-Catenin is involved in complex intracellular dynamics, is distributed in different pools, and is regulated by phosphorylation and interactions with cadherins, which all complicate the analysis (McCrea et al., 2015; Muñoz-Descalzo et al., 2015). We found that, whereas Notch1 is consistently enriched in the ventral region from the beginning of embryogenesis in all stages analyzed, the global distribution of total β-catenin was more variable in Xenopus until s7. However, in some s7 embryos we noticed an opposing distribution of Notch1 and β-catenin proteins in ventral-animal cells. Notch1 was enriched in the apical region, but β-catenin levels increased towards the basal region. Interestingly, it has been proposed that, in the absence of Wnt signaling, Notch sequesters a cell surface-located pool of transcriptionally active Armadillo/β-catenin and induces its degradation in Drosophila. In this way, Notch functions as a buffer against fluctuations in the levels of transcriptionally active Armadillo/β-catenin that leaks from the Gsk-3β-dependent destruction complex (Sanders et al., 2009). The variability that we found in the DV distribution of total β-catenin at early stages might be indicative of such fluctuations during the process of Notch buffering.

We also found that Notch1 is required for the normal expression of genes essential for a fully functional ventral program. Notch1 is able to positively control ventral genes in a CSL-dependent fashion around the onset of gastrulation (Fig. 7, route 2), although we do not know whether it does it in a direct or indirect way (Fig. 7, dashed lines). Intriguingly, ISH analysis at mid/late gastrula revealed that constitutively active nicd1 and blockade of canonical CSL-dependent activity produced variable results for bmp4. This might be related to time-dependent changes, as it is known that Notch signaling can elicit opposite effects depending on time and context (Glavic et al., 2004; Contakos et al., 2005; Revinski et al., 2010). Notch1 might also positively control bmp4 indirectly, by contributing to the destabilization of β-catenin in the ventral side (Fig. 7, route 1), thus preventing the expression of BMP antagonists in this region, which would otherwise interrupt the known positive feedback loop by which bmp4 increases its own expression (De Robertis, 2009) (Fig. 7). The situation could be more complex because bmp4 is deeply involved in an intricate network of biochemical interactions between the dorsal and ventral centers that self-regulates the DV field. bmp4 is able to indirectly decrease its own expression by upregulating sizzled, an inhibitor of the Tolloid metalloproteinase that degrades Chordin. If BMP levels increase, sizzled expression is enhanced, thus leading to an increase in Chordin levels, which results in inhibition of BMP signaling by blocking BMP function, decreasing its levels by interrupting the BMP positive-feedback loop (De Robertis, 2009) (Fig. 7). Thus, it is conceivable that, in our experimental conditions, if nicd1 tends to upregulate bmp4, and BMP4 reaches a threshold, the self-regulatory network might operate, attempting to downregulate bmp4. This might, in part, underlie the mixed results that we obtained by ISH for bmp4 expression at mid/late gastrula. Notch-dependent β-catenin degradation has been shown to be primarily elicited by membrane-tethered uncleaved Notch in other systems (Hayward et al., 2005; Kwon et al., 2009). If notch1 FL is stronger than nicd1 in promoting β-catenin degradation, this effect would prevent the operation of the self-regulatory network that balances bmp4 expression, as it depends on β-catenin signaling. If this is the case, the upregulation of bmp4 by notch1 FL would prevail. This might explain the differences in the response of bmp4 to nicd1 and notch1 FL. These kinds of mechanisms are difficult to show using RT-qPCR, as it measures the mean levels of mRNA in a population, thus masking the variabilities of response that occur within the population, which can be analyzed by ISH in individual phenotypes.

The downregulation of ventral markers by su(H)1^DBM mRNA around the onset of gastrulation, and the higher CSL-dependent transcriptional activity in the ventral side at early gastrula, revealed by the gene reporter assay that we show here, support the hypothesis of a spatial regulation of canonical Notch activity in the DV axis. In addition, we have previously shown that notch1 restricts the expression of chordin and Xnr3 in the BCNE in a CSL-independent fashion by promoting β-catenin degradation (Acosta et al., 2011). We propose that Notch1 canonical and non-canonical activity is spatially regulated in the animal hemisphere, with lower levels at the dorsal side and higher levels at the ventral side. An unknown mechanism controlling Notch1 mRNA and protein distribution from the beginning of embryogenesis underlies this process (Fig. 7).

It will be important to elucidate how Notch1 controls the expression of ventral genes. Although the most studied Notch targets are the HES/HEY genes, which encode transcriptional repressors of the bHLH family, other targets are beginning to emerge in different animal models, including genes involved in proliferation (such as Myc, Cyclin D), apoptosis (reaper, hid), cell fates (Gata3, Pax2), signaling pathways (lip-1, ERBB2, EGFR, Notch), metabolism and cytoskeletal regulators (Bray and Bernard, 2010). Strikingly, GATA factors are required for wnt8a and ventx1.2 but not for bmp4 and ventx2.2 expression in Xenopus (Sykes et al., 1998). Interestingly, a link between notch and wnt8 has recently been revealed in posterior development in short-germ arthropods, in which posterior segments are added sequentially from a segment addition zone that requires a positive regulation of wnt8 by Delta-Notch signaling to develop (Schönauer et al., 2016). Comparative studies indicated that the common ancestor of arthropods employed delta-notch, wnt8 and caudal for posterior development and it was suggested that this gene regulatory network was already present in the common ancestor of arthropods and vertebrates (Urbilateria) (McGregor et al., 2009).

Although it is well accepted that relocation of dorsal determinants soon after fertilization activates the maternal Wnt/β-catenin pathway that drives the development of the dorsal center, it remained unclear whether an analogous event of asymmetric distribution of ventral determinants occurs in the earliest stages of development. Overall, there is no published evidence of early asymmetries of maternal components of the BMP signaling pathway before the mid-blastula transition. While zygotic components of the ventral cascade start to be transcribed at the mid-blastula stage, they are initially expressed in a uniform fashion and become restricted to the VC during gastrulation (Ventx genes), or they begin to be expressed ventrally at blastula stages (wnt8a) or at the mid-gastrula stage (bmp4) (see Introduction). To our knowledge, ventral enrichment of Notch1 mRNA and protein is the earliest localized sign of ventral development reported in vertebrates, preceding the appearance of the strong localized expression of wnt8a, bmp4 and Ventx genes in the VC, and the dorsal accumulation of transcriptionally active nuclear β-catenin in the opposite region. Together with our previous findings (Acosta et al., 2011), our results indicate that during early embryogenesis, ventrally located Notch1 participates in the control of the initial DV polarity by a dual mechanism: (1) by promoting the development of the VC and (2) by destabilizing transcriptionally active β-catenin (Fig. 7).

Albino and wild-type Xla embryos were obtained using standard methods (Franco et al., 1999) from adult animals obtained from Nasco, and staged according to Nieuwkoop and Faber (1994). Danio rerio (zebrafish) embryos were obtained as previously described (Kamm et al., 2013) and staged according to Kimmel et al. (1995). Xenopus laevis in S.L.L.'s laboratory and Danio rerio in L.F.F.'s laboratory were employed according to protocols approved by the Laboratory Animal Welfare and Research Committees (CICUAL) from Facultad de Medicina-UBA and INGEBI-CONICET, respectively. Experiments on Xenopus laevis in L.K.’s laboratory were approved by the ‘Direction départementale de la Protection des Populations, Pôle Alimentation, Santé Animale, Environnement, des Bouches du Rhône’ (agreement number F 13 055 21). All experiments were performed following Directive 2010/63/EU of the European Parliament and of the council of 22 September 2010 on the protection of animals used for scientific purposes.

Synthetic capped mRNAs for microinjections were obtained as follows. Xla notch1 full length in pSP72 (notch1 FL) (Coffman et al., 1990) was digested with XhoI and in vitro transcribed with the T7 Megascript transcription kit (Ambion, AM1334) with a 4:1 cap analog:GTP ratio, using m7G(5′)ppp(5′)G (Ambion, AM8050). Xla nicd1 pCS2+MT and Xla su(H)1^DBM pCS2+MT (Chitnis et al., 1995; Wettstein et al., 1997) were digested with NotI and in vitro transcribed with the mMESSAGE mMACHINE SP6 Transcription Kit (Ambion, AM1340). Capped mRNAs were purified with the RNeasy mini kit (Qiagen, 74104).

The notch1 antisense morpholino oligonucleotide (Notch1 Mo) with the base composition 5′-GCACAGCCAGCCCTATCCGATCCAT-3′ has previously been used and validated in works by our group and by other authors (López et al., 2003; Revinski et al., 2010; Acosta et al., 2011; Sakano et al., 2010). As control morpholino (control Mo), we used the standard control oligo or the random control oligo 25-N (Gene Tools). Each embryo was injected with 30-40 ng of the relevant Mo.

Embryos were injected in the animal hemisphere at the one-cell stage, or unilaterally, from early to late two-cell stage, as detailed in the figures. Injections included as tracer 20-40 ng of Dextran Oregon Green 488, MW 10000, anionic lysine fixable (DOG; Thermo Fisher Scientific, D7171).

For gain- and loss-of-function studies in Xenopus, sibling embryos were randomly allocated to control or experimental groups. Results are expressed as percentage of the total number (n) of embryos with the indicated phenotypes, which are described in the main text, in the figure legends or in the supplementary tables.

The preparation of digoxigenin-labeled antisense RNA probes and the whole-mount ISH were performed as previously described (Pizard et al., 2004), except that the proteinase K step was omitted. Probes for wnt8a, vent-2b, bmp4, wnt11 and chordin have been described previously by other authors (Christian and Moon, 1993; Rastegar et al., 1999; Fainsod et al., 1994; Freeman et al., 2008; Saka et al., 2000; Kuroda et al., 2004).

For immunodetection of endogenous Notch1 and β-catenin proteins or of NICD1 protein after injection of nicd1 mRNA, Xenopus embryos were fixed for 90 min with MEMPFA and then transferred to Dent's fixative (80% methanol, 20% DMSO) for 72 h at 4°C or −20°C. To orient the embryos in the DV axis, we cut left and right halves of regularly cleaving embryos from batches in which the paler pigmentation zone was bisected by the first cleavage furrow, as the paler region predicts the prospective dorsal side with ∼70% accuracy in Xla (Klein, 1987). Left and right halves, or animal halves, when required, were cut after rehydration. Afterwards, embryos were further processed as previously described, including a bleaching step to avoid quenching of the IF signal by the pigment (Acosta et al., 2011).

For preparing samples for cryosections, left and right embryo halves treated as previously described were impregnated in 15% sucrose in 1× PBS for 30 min at room temperature, and were then kept overnight in 30% sucrose in 1× PBS at 4°C. The following day, specimens were incubated in 15% gelatin (Sigma, G9391) in 1× PBS for 30 min at 37°C, in 2 ml tubes. Embedding was performed in the same solution, and specimens were oriented in such way that the internal surface of the embryo halves were laid on the bottom. In this way, serial sections were obtained, with the first section being the most internal. Embedded specimens were snap-frozen by immersion for 1 min in an isopropanol/dry ice bath at −40°C or below, and immediately transferred to −70°C overnight. Before sectioning, specimens were transferred to −20°C for 1 h. Sections of 20 µm thickness were obtained with a cryostat (Leica, CM1850) and collected on positively charged glass slides (Fisherbrand, Superfrost Plus). Sections were dried overnight at room temperature.

Immunodetection was performed in general as previously described for whole-mount specimens (Acosta et al., 2011), except that for sections, incubations were carried out with 0.2 ml of each solution per slide, the slides being covered with Parafilm. Antibody incubations were performed overnight at 4°C in a wet chamber, and the antibodies were washed four times for 30 min with 50 ml 1× PBS each. Slides were mounted with Vectashield containing DAPI (Vector Laboratories, H-1200).

Zebrafish embryos were fixed and processed for whole-mount IF in the same conditions as Xenopus embryos, except that, after rehydration, they were dechorionated with sharp forceps before proceeding to IF of whole embryos.

The commercially available Notch1 intra rabbit polyclonal antibody (N1 intra Ab) (Abcam, ab8387) was raised against a peptide from the human intracellular domain of NOTCH1, which has 93% identity with the protein encoded by Xla notch1. In human breast tumors, the antibody detects both nuclear and cytoplasmic NOTCH1 in tissue sections (Efstratiadis et al., 2007).

Primary and secondary antibodies were diluted in blocking buffer as follows: β-catenin (mouse monoclonal IgG, which recognizes total β-catenin; Abcam, ab19448, 1/100), N1 intra Ab (1/200), anti-rabbit IgG-HRP (Santa Cruz Biotechnology, sc-2004, 1/1000), anti-rabbit IgG-Alexa 594 (Thermo Fisher Scientific, A-11012, 1/100), anti-mouse IgG F(ab′)2-Alexa Fluor 488 (Thermo Fisher Scientific, A-11017, 1/200). For double IF for Notch1 and β-catenin in whole-mount hemisections, we restricted the comparison of both patterns to those embryos in which Notch1 was enriched in the ventral region, according to the orientation predicted by pigmentation.

Whole embryos and cryosections were photographed in a MVX10 fluorescence microscope (Olympus) equipped with a DP72 camera (Olympus). Higher resolution IF and DAPI images for quantitative analysis at s8 were photographed in an inverted motorized fluorescence microscope IX83 (Olympus) equipped with a Hamamatsu C11440-22C ORCA FLASCH4.0U digital camera. For Notch1 IF combined with ISH of dorsal markers (wnt11 in cleavage stages, chordin in mid-blastula), Xenopus pigmented embryos were fixed with MEMPFA for 1 h and cut into right and left halves according to the pigmentation pattern as before. Then, right halves were transferred to Dent's solution, according to the immunodetection protocol, whereas left halves were submitted to the standard ISH protocol. Alternatively, we used albino Xenopus embryos to perform Notch1 IF and wnt11 ISH to correlate their patterns in the same embryo. For this purpose, embryos were fixed in MEMPFA overnight at 4°C and, the following day, they were subjected to the standard ISH protocol. After stringency washings, embryos were washed with TBSE, TBSET and TBSET+BSA as described by Acosta et al. (2011), and were incubated simultaneously with the ISH anti-digoxigenin-AP antibody (Roche, 11093274910, 1/2000) and the IF Notch1 antibody overnight at 4°C. To circumvent the possibility that quenching of IF with the ISH staining could obscure the interpretation of our results, we first proceeded to reveal Notch1 IF and image the embryos as in the standard IF protocol. Then, embryos were individually processed for revealing the wnt11 ISH in 96× multiwells, and the wnt11 pattern was correlated with the first image of the same embryo.

Pixel intensity was measured in images of cryosections in regions of interest (ROI) with ImageJ (NIH). For early cleavage stages, we designed a rectangular ROI on the animal hemisphere, and results were plotted as mean pixel intensity (y-axis) for each coordinate of the DV axis (x-axis) (Fig. 1A′-D′). In addition, two ROIs were selected for each embryo from s1 to s8, one in the dorsal-animal region (dROI) and the other one in the ventral-animal region (vROI). For blastula stages, as the blastocoel cavity precluded the analysis on a rectangular ROI, only comparisons of dROI and vROI were made. For analysis of nuclear IF levels at s8, dorsal and ventral ROIs containing nuclear areas (nROI) were selected, with the DAPI image as a source. Statistical analysis was carried out using Prism (GraphPad) (two-tailed Mann–Whitney U-test, P<0.05).

For WB analysis of Notch1, five whole Xenopus embryos (controls, control Mo- or Notch1 Mo-injected embryos) and ten dorsal or ten ventral halves were obtained at the indicated stages in the figures for each round of protein extraction. They were homogenized in 0.15 ml ice-cold PBS containing 1× protease inhibitors (Halt Protease and Phosphatase Inhibitor Cocktail, Thermo Fisher Scientific, 78440). The homogenates were centrifuged for 5 min at 700 g at 4°C, and the supernatant was precipitated with ten volumes of acetone for 20 min in ice. Samples were centrifuged for 5 min at 3000 g at 4°C, and the supernatant was discarded. Pellets were dried, then resuspended in 25 µl of 2× Laemmli buffer containing 10% of 2-mercaptoethanol and boiled for 5 min. Then 20 µl of protein samples were separated on an 8% SDS-PAGE gel and transferred to a polyvinylidenedifluoride membrane (Immobilon-P, Millipore, IPVH00010). Blots were blocked in 5% non-fat milk prepared in TTBS (0.05% Tween-20, 10 mM Tris-HCl pH 8, 150 mM NaCl) and then incubated overnight at 4°C with N1 intra Ab, 1/1000 diluted in blocking solution, and with GAPDH monoclonal antibody (6C5) (Thermo Fisher Scientific, AM4300, 1/10,000). After washing three times in TTBS for10 min each, blots were incubated for 2 h at room temperature with HRP-conjugated goat anti-rabbit IgG (H+L) (Jackson ImmunoResearch, 111-035-003, 1/10,000 diluted in blocking solution), and with HRP-conjugated goat anti-mouse IgG (H+L) (Bio-Rad, 1706516, 1/3000). After washing as before, blots were developed in SuperSignal West Pico Chemiluminescent Substrate solution (Thermo Fisher Scientific, 34580) and exposed to film (Agfa-Gevaert). To determine the actual levels of Notch1, the intensity of the bands corresponding to this protein was normalized to the corresponding GAPDH band used as internal control, in each assay.

For assessing endogenous levels of notch1 and wnt11 mRNAs, ventral and dorsal halves of s5 Xenopus embryos from two independent batches were cut, according to the pigmentation pattern. For functional experiments, two-cell-stage Xenopus embryos from three different batches were injected in both blastomeres with the mRNAs or morpholinos indicated in the figures. The following doses were injected per cell: 1 ng of nicd1 mRNA, 1 ng of su(H)1^DBM mRNA, 40 ng of control Mo, 40 ng of Notch1 Mo, or 40 ng of Notch1 Mo+1 ng of nicd1 mRNA. Embryos were snap frozen at the stages indicated in the figures and stored at −80°C. Ten embryos per biological replicate were used. Total RNAs were purified with an RNeasy kit (Qiagen). Primers are listed in Table S6. New primers were designed using Primer-Blast Software (NCBI). RT reactions were carried out using iScript Reverse Transcription Supermix for RT-qPCR (Bio-Rad). RT-qPCR reactions were performed in triplicate for each sample using SYBRGreen (Thermo Fisher Scientific, K0391) on a CFX qPCR cycler (Bio-Rad, CFX96 C1000 Thermal Cycler, 184-5384). RT-qPCR results were analyzed using Bio-Rad CFX Maestro Software and one-way ANOVA was used to determine statistical significance (P<0.05). The relative expression of each target was normalized to histone4 expression (H4).

Two-cell-stage embryos were injected in both blastomeres with 1 pg of circular plasmid pRL-TK (Renilla luciferase, Promega, E2231)+40 pg of circular Notch reporter plasmid (4×CSL-firefly luciferase, Addgene, 41726) (Saxena et al., 2001) or with 1 pg of circular plasmid pRL-TK+40 pg of circular Notch reporter plasmid+1 ng of nicd1 mRNA. Luciferase activity was quantified and normalized to Renilla activity in lysates from stage 10.25-10.5 embryos using the Dual-Luciferase Reporter assay system (Promega, E1910) in a TriStar LB 941 Modular Multimode Microplate Reader (Berthold Technologies). For dorsal and ventral luciferase activity quantification, embryos were sectioned in dorsal and ventral halves at stage 10.25-10.5. Five embryos were lysed for each measurement, and ten halves were lysed for each measurement. All assays were performed in duplicate. Embryos from the same fertilization batch were used for each experiment and repeated independently three times. To determine statistical significance, we applied one-way ANOVA analysis and Bonferroni's multiple comparisons test (t-test) (P<0.05). Statistical analyses were carried out using Prism 6.

Numbers of samples analyzed are indicated for each set of experiments in the figures and supplementary tables. Statistical tests applied are described for each set of experiments in the Materials and Methods sections, in the figures and supplementary tables. Differences were considered significant when P<0.05.

We acknowledge the following colleagues for providing us with the constructs for making synthetic mRNA and probes: Chris Kintner for notch1 FL, nicd1, su(H)1^DBM; Randy Moon for wnt8a; Walter Knöchel for vent-2b; and Eddy M. de Robertis for bmp4 and chordin. We are also grateful to Andrea Noemí Pecile, Manuel Ponce, Alejo Ramos and Ezequiel Yamus for animal husbandry; to Lucila Morono and Nerina Villalba for technical assistance; to Daniela Rojo and María Belén Favarolo for help with experiments; and to Ana María Adamo, María Cecilia Cirio, and Tomás Falzone for reagents. We are also indebted to Dr Marcelo Rubinstein for his constant support and advice.