Jagged2a-Notch Signaling Mediates Cell Fate Choice in the Zebrafish Pronephric Duct

Pronephros, a developmental model for adult mammalian kidneys (metanephros) and a functional kidney in early teleosts, consists of glomerulus, tubule, and duct. These structural and functional elements are responsible for different kidney functions, e.g., blood filtration, waste extraction, salt recovery, and water balance. During pronephros organogenesis, cell differentiation is a key step in generating different cell types in specific locations to accomplish designated functions. However, it is poorly understood what molecules regulate the differentiation of different cell types in different parts of the kidney. Two types of epithelial cells, multi-cilia cells and principal cells, are found in the epithelia of the zebrafish distal pronephric duct. While the former is characterized by at least 15 apically localized cilia and expresses centrin2 and rfx2, the latter is characterized by a single primary cilium and sodium pumps. Multi-cilia cells and principal cells differentiate from 17.5 hours post-fertilization onwards in a mosaic pattern. Jagged2a-Notch1a/Notch3-Her9 is responsible for specification and patterning of these two cell types through a lateral inhibition mechanism. Furthermore, multi-cilia cell hyperplasia was observed in mind bomb mutants and Mind bomb was shown to interact with Jagged2a and facilitate its internalization. Taken together, our findings add a new paradigm of Notch signaling in kidney development, namely, that Jagged2a-Notch signaling modulates cell fate choice in a nephric segment, the distal pronephric duct.


Introduction
In vertebrates, development of the excretory system is characterized by the successive formation of three distinct kidneys with increased complexity: pronephros, mesonephros, and metanephros. The pronephros is found in all vertebrates, but in mammals it is a nonfunctional transitory structure that is replaced by the mesonephros and then the metanephros. In the early life of fish and amphibians, however, the pronephros is a functional filtration organ that develops very similarly to metanephros, and has been used as a model for kidney development. Nephrons, the fundamental functional units of the kidney, possess several segments, which regulate fluid balance, osmolarity, and the disposal of metabolic waste products [1]. While pronephroi in amphibians and fish contain two functional nephrons, the metanephroi of mammals have millions of nephrons [1][2][3]. The zebrafish pronephros consists of paired glomeruli coalescing at the midline ventral to the dorsal aorta, and two pronephric tubules that project bilaterally from the glomeruli to the pronephric (Wolffian) ducts that run caudally and fuse just before their contact with the exterior at the cloaca [2]. The glomerulus is the site of blood filtration. Epithelia of the tubules are the primary site of selective reabsorption and secretion, while the duct carries the modified urine to the outside world [1]. Though quite uniform in appearance, the tubule and duct epithelia are further subdivided into distinct segments, recognized by the expression of specific membrane transporters [4]. This is a general feature of vertebrate kidneys, where osmoregulatory function depends on an organized disposition of different transporters operating sequentially along the nephron [5][6][7].
Morphogenesis and cell fate determination of different nephric segments have attracted much attention recently. Multiple transcription factors and signaling pathways have been shown to be involved in these processes in different model organisms. Wnt4 is essential for tubulogenesis in mouse metanephroi and Xenopus pronephroi [8,9]. Brn1 is required for the development of Henle's loop, the distal convoluted tubule, and the macula densa in mice at the primitive loop stage [10], and so is pax2a for the differentiation of proximal tubule and duct epithelial cells and cloaca morphogenesis in zebrafish [11]. Some segments of the nephron comprise only one cell type, while others include two or more cell types. The mammalian collecting duct contains two major cell types: principal cells (for salt and water absorption) and intercalated cells (for acid/base transport) [12]. It was reported that Foxi1 plays a crucial role in the specification of intercalated cells [13].
Notch signaling is an evolutionarily conserved pathway that multicellular animals use in regulating pattern formation and cell fate determination through local cell interactions [14,15]. One of the well-known mechanisms of Notch signaling is lateral inhibition during neurogenesis: initially equivalent cells differentiate into a ''salt and pepper'' pattern of cells with different fates via a regulatory loop [14]. Notch is a transmembrane receptor that interacts with Delta and Serrate/Jagged ligands. Ligand-activated intramembrane proteolysis, which is partly through the c-secretase activity of Presenilin, is required to release the Notch intracellular domain (Notch icd ), which is then translocated to the nucleus, where Notch icd and CSL (CBF1/RBPjj, Su(H), and Lag-1) proteins bind and activate downstream target genes, such as Hairy/Enhancer of split related (Hes/her) homologs [16]. Ubiquitylation is a multistep process that results in the conjugation of Ubiquitin to a substrate protein. Recent studies have identified the roles of Neuralized and Mind bomb (Mib) in ligand ubiquitylation and endocytosis, which is essential for activating Notch [17][18][19][20]. Similarly, Jagged2 is ubiquitylated by a Mib paralog, Skeletrophin [21].
Notch signaling is required for the development of different kidney segments. By manipulating Notch activity, Notch signaling in the tubule was shown to inhibit duct fate in the dorsoanterior Xenopus pronephric anlage and to control subsequent tubule patterning [22]. Homozygous Notch2 del1 , a hypomorphic allele, and transheterozygous Notch2 del1 /Jag1 dDSL mice exhibit a similar glomerular defect: lack of capillary tufts and mesangial cells [23]. Presenilin is indispensable for the formation of mouse proximal tubules and glomerular podocytes [24,25]. Zebrafish jagged1b/jagged2a double morphants have small glomeruli or segments of glomerulus replaced by dilated blood vessels [26].
Our analysis of zebrafish pronephric ducts revealed that the distal ducts are composed of two types of epithelial cells: multi-cilia cells and principal cells. We showed that multicilia cells interpolate principal cells and that their differentiation is mediated by Jagged2a-Notch1a/Notch3-Her9 signaling. We also demonstrated that this differentiation process requires Mib, an E3 ligase that facilitates Jagged2a endocytosis, and, hence, activates Notch signaling. This is the first time, to our knowledge, that Jagged2-Notch signaling has been shown to mediate cell fate determination within a kidney segment, but not between segments, via a lateral inhibition mechanism.

Multi-Cilia Cells Interpolate Principal Cells in the Zebrafish Pronephric Duct
Acetylated tubulin staining revealed that stumpy single primary cilia are present in the pronephric duct as early as the 20 somite stage (ss) (unpublished data). Cilia tufts or multi-cilia appeared later and were fully formed by 36 h postfertilization (hpf). These cilia tufts were located along the distal pronephric duct between the proximal pronephric duct and cloaca, corresponding to somites 8-14 (Figures 1A, S1, and S3) [11,27,28]. Pericentriolar material 1 (Pcm1) staining of 36-hpf embryos revealed that multiple basal bodies are associated with each cilia tuft ( Figure 1B-1D) [29,30] and Pcm1 is colocalized with c-tubulin at the apical site of epithelial cells ( Figure S2). To determine if each cilia tuft is generated from a single cell, we used antibodies against a membrane marker, wheat germ agglutinin (WGA), and a tight junction marker, Zonula occludens-1 (Zo-1) [11,31]. Triple labeling of Pcm1, acetylated tubulin, and WGA demonstrated that cilia tufts are in the lumen of the duct and that multiple basal bodies are within one cell. Individual Pcm1 staining was also found in the neighboring cells corresponding to individual basal bodies of the primary cilia ( Figure 1E). Immunostaining of Zo-1 and Pcm1 confirmed that multiple Pcm1-staining basal bodies are localized to the apical side of one cell in the pronephric duct ( Figure 1F).
Cilia identity was further confirmed by transmission electron microscope imaging of a transverse section of the distal duct of 36-hpf embryos. We found two types of cells: cells with a cilia tuft of at least 15 cilia, and neighboring cells with a single primary cilium. All cilia tufts and primary cilia projected along the axis of the duct lumen and were of the typical 9 þ 2 structure ( Figure 1B and 1G), suggesting that they are motile [32]. Indeed, it was demonstrated that cilia in the zebrafish duct are motile, generating a corkscrew-like wave pattern in the duct lumen directed toward the cloaca [28].
In mammals, collecting ducts are composed of two major cell types: principal cells and intercalated cells [12]. Na þ , K þ ATPases transport numerous solutes and water across epithelia [33] and are only expressed in the principal cells [34,35]. The zebrafish counterparts are expressed in the pronephric duct [36]. However, a meticulous examination of Na þ , K þ ATPase a1a2 and Na þ , K þ ATPase b1a expression using in situ hybridization revealed that these genes are not expressed in all the duct cells ( Figure 1H-1K). Some sodium pump-negative cells interpolated principal cells. To investigate their identity, we cloned the zebrafish homologs and examined the expression patterns of pendrin1, pendrin2 [37,38], rhcg [39], and vacuolar-type ATPase B [40], all of which are marker genes in mammalian intercalated cells (see Materials and Methods). Although they were expressed in other tissues, none of these genes were expressed in the duct up to 72 hpf, suggesting that the sodium pump-negative cells are not intercalated cells (unpublished data). To determine whether they are multi-cilia cells, we cloned ciliogenic genes and analyzed their expression in the duct. Zebrafish rfx2 is the homolog of Caenorhabditis elegans daf-19, which controls cilium formation in sensory neurons [41]. Zebrafish centrin2 is the homolog of mouse Centrin2, which associates with centrosome-related structures of the basal bodies of the ciliated cells [42,43]. In addition to the ciliated tissues, including

Author Summary
The kidney is a complex organ that regulates blood homeostasis through the maintenance of fluid and ion balance and disposal of metabolic waste. We used zebrafish pronephros, a primordial vertebrate kidney, to address how a kidney tissue acquires its cell types and pattern. Two types of epithelial cells were found in the pronephric duct: multi-cilia cells and principal cells, which could be distinguished based on morphology and expression of different marker genes. In the pronephric duct, the multi-cilia cells and principal cells form a ''salt and pepper,'' or mosaic, pattern. Using existing zebrafish mutants and a knockdown technique, we demonstrated that the mosaic pattern and differentiation of these two cell types are controlled through a Notch-dependent lateral inhibition mechanism. Notch signaling has been shown to be essential for other aspects of kidney development, such as formation of the glomerulus and the tubule. Here, to our knowledge for the first time, we show that the same signaling pathway is required for the differentiation of two different epithelial cells in a kidney segment known as the distal pronephric duct. The same mechanism is very likely to be employed by other similar developmental processes in the same context to generate distinct cell types in a tissue.
Kupffer's vesicle, olfactory pits, hair cells of the otic vesicle, and the neural tube (unpublished data) [44], rfx2 and centrin2 were expressed in a mosaic pattern in the duct at 36 hpf ( Figure 1L-1N). Furthermore, fluorescent double in situ hybridization of rfx2 and Na þ , K þ ATPase b1a revealed a mutually exclusive pattern ( Figure 1O-1Q). This indicates that multi-cilia cells and principal cells are two distinct cell populations in the zebrafish distal pronephric duct. notch1a, notch3, jagged2a, and her9 are Expressed in the Duct Notch signaling has been shown to be required for differentiation of ciliated cells in Xenopus skin [45] and in sensory patches of the zebrafish inner ear [46]. The mosaic pattern of multi-cilia cells and principal cells prompted us to explore whether Notch signaling is required for their differentiation in the pronephric duct.
Among four known Notch receptors, notch1a and notch3 were found to be expressed in the intermediate mesoderm (IM) in early stages and later in the duct. notch1a was expressed in the IM from 1 ss and subsequently in the distal duct region at 18 ss (Figure 2A and 2B). notch1a expression, however, was not detected in the duct after 20 hpf. notch3 was expressed in the IM from 1 ss onward and in the entire duct region, with a higher level of expression in the distal part, at 24 hpf ( Figure 2C and 2D), where expression persisted until at least 48 hpf.
There are nine known zebrafish Notch ligands: deltaA [47], deltaB [48], deltaC [49], deltaD [50], dll4 (M. M. and Y.-J. J., unpublished data), jagged1a (also known as jagged1 or serrateC), jagged1b (also known as jagged3 or serrateA), jagged2a (also known as jagged2 or serrateB) [26,46,51], and jagged2b (M. M. and Y.-J. J., unpublished data). Of the ligands, deltaC is expressed in the anterior IM, presumably in the glomerulus, from 4 ss to 18 hpf [49], and jagged1b is expressed in the developing proximal tubule [51]. jagged2a expression in the IM appeared gradually from anterior to posterior, spanning from somite 3 to somite 13 at 15 ss ( Figure 2E-2G). In the posterior IM, jagged2a expression displayed a salt-and-pepperlike pattern (a mixture of high-and low-expressing cells) from 17 ss to 20 ss (spanning from approximately somite 9 to somite 13) ( Figure 2H and 2I). Beginning with 20 ss, jagged2a expression was limited to individual cells; this pattern persisted in the pronephric duct until at least 48 hpf (Figure 2J and 2K). jagged2a was also expressed in the proximal duct ( Figure 2E-2J), and a description of its function there will be published elsewhere. Here, we only explore the function of jagged2a in the distal pronephric duct.
We further examined the expression of published Notch downstream targets, hairy/enhancer of split related (her and hey) genes, by doing in situ hybridization to detect her1 [52], her2 [53], her3 [54], her4 [55], her6 [56], her7 [57], her8 [53], her9 [58], hey1, hey2, and heyL [59]; or by checking the deposited expression patterns in the ZFIN database (http://www.zfin. org) of her5, her12, and hes5 [60]. Only two of the her genes were expressed in the IM. her6 appeared between the tailbud stage and 10 ss but expression was not maintained in later stages. her9 expression was not detected in the duct domain before 15 ss (unpublished data). However, it was expressed in the distal duct from 17 ss. This correlates temporally with mosaic expression of jagged2a in the same region ( Figures 2L and S3). Spotted and uneven her9 expression persisted in the duct till about 21 hpf and disappeared afterwards ( Figure 2M).

Individual jagged2a-Positive Cells Are Multi-Cilia Cells
Since Jagged2a presumably starts signaling to the neighboring cells from 17 ss onwards, we investigated the onset of multi-cilia cell differentiation by examining the expression of rfx2 and centrin2 at earlier stages. Interestingly, rfx2 expression in the IM and duct was similar to that of jagged2a. rfx2 expression was uniform in all duct cells before 15 ss, which is consistent with the fact that all cells have cilia-either cilia tufts or a single cilium-in this kidney segment. Its expression was then limited to single cells in the distal duct from 17 ss until at least 36 hpf ( Figure 3A-3D). Similarly, centrin2 expression was limited to single cells from 20 ss onwards (unpublished data). These observations suggest that multicilia cells are jagged2a-expressing cells. Indeed, jagged2a and rfx2 transcripts were colocalized in individual cells from 17 ss to at least 36 hpf ( Figure 3E-3J). Furthermore, when we investigated whether her9 is expressed in the same distal duct  domain as rfx2, we found that although her9 was expressed in the same domain, it was expressed primarily in non-rfx2expressing cells ( Figure 3K-3M).

Jagged2a and Mib Regulate Cell Fate Differentiation through Lateral Inhibition
The above finding suggests that Jagged2a regulates differentiation of multi-cilia cells and principal cells. We designed morpholino antisense oligonucleotides (MOs) to knock down the function of Jagged2a, and used the mib ta52b mutant to study the function of the Jagged2a-Notch pathway in differentiation. In addition to two MOs, jagged2a-atg and jagged2autr, designed to be antisense to the jagged2a translation start site and the 59 UTR, respectively, one MO (jagged2a-sp) was designed to block RNA splicing between exon 1 and intron 1. The jagged2a-sp MO effectively blocked splicing until at least 48 hpf ( Figure 4A) and the jagged2a-utr MO was specific in a sequence-dependent manner ( Figure S4A-S4F). jagged2a MOs did not affect duct development ( Figure 4B). jagged2a-atg morphants displayed uniform rfx2 (100%, n ¼ 242) and centrin2 (89%, n ¼ 224) expression in almost all of the duct cells, in contrast to a mosaic pattern found in wild-type (wt) embryos ( Figure 4C-4F). Similar results were found in jagged2a-utr morphants ( Figure S4F; Table 1). Furthermore, we observed that Pcm1 and acetylated tubulin were significantly increased in jagged2a-sp morphants (100%, n ¼ 7; Figure  S5A-S5F). In contrast, Na þ , K þ ATPase b1a expression was highly reduced in the duct of jagged2a-atg morphants at 24 hpf (100%, n ¼ 22; Figure 4G and 4H) and 36 hpf (100%, n ¼ 40; unpublished data). Similarly, we observed multi-cilia cell hyperplasia in mib ta52b mutants as evidenced by rfx2 expression ( Figure 4I and 4J) and immunostaining of acetylated tubulin and Pcm1 ( Figure S5G-S5I). Statistically, mib ta52b mutants generated greater than 2-fold more multi-cilia cells than wt embryos (Table 1). Consistently, principal cells were (C-H) Multi-cilia cell number is increased in (D and F) jagged2a-atg morphants compared to (C and E) wt embryos as shown by (C and D) rfx2 and (E and F) centrin2 expression at 24 hpf, but principal cell number is decreased in (H) jagged2a-atg morphants compared to (G) wt embryos as revealed by Na þ , K þ ATPase b1a expression at 24 hpf. (I-L) Multi-cilia cell number is increased in (J) mib ta52b embryos compared to (I) wt embryos as shown by rfx2 expression at 24 hpf, but principal cell number is decreased in (L) mib ta52b embryos compared to (K) wt embryos as revealed by Na þ , K þ ATPase a1a2 expression at 24 hpf. Panels M-R focus on the duct around somite 11 to 13. (M-O) Fluorescent double in situ hybridization of rfx2 (green) and Na þ , K þ ATPase b1a (red) in 36-hpf (M) wt embryos, (N) jagged2a-sp morphants, and (O) mib ta52b mutants shows multi-cilia cell hyperplasia in jagged2a morphants and mib ta52b mutants. Arrows point to the rfx2-expressing cells in the duct of (M) wt embryos; arrowheads point to the Na þ , K þ ATPase b1a-expressing cells in the pronephric duct of (N) jagged2a-sp morphants.
While the number of rfx2-expressing cells was dramatically increased in jagged2a-sp morphants and mib ta52b mutants (Table 1), only three to five Na þ , K þ ATPase b1a-expressing cells were found in the jagged2a-sp morphants (89%, n ¼ 19), and a dramatically decreased number of Na þ , K þ ATPase b1aexpressing cells were found in mib ta52b mutants ( Figure 4M-4O). Similarly, double immunostaining with a6F (raised against the chick a1 subunit of the Na þ /K þ ATPase [61]) and Pcm1 showed that most of the duct cells adopted a multi-cilia cell fate and expressed Pcm1; only two to three cells were positive for the principal cell maker a6F in the distal duct of jagged2a-sp morphants (93%, n ¼ 15), and principal cells were dramatically decreased in mib ta52b mutants ( Figure 4P-4R). This observation suggests that the multi-cilia cell hyperplasia in jagged2a morphants and mib ta52b mutants is at the expense of the principal cells through lateral inhibition (see below). It is unlikely to be due to an inhibitory activity of Jagged2a on proliferation of multi-cilia cells, since there is no difference in cell proliferation between wt embryos and jagged2a morphants (unpublished data). The phenotypic severity of mib ta52b was not as strong as that of jagged2a morphants ( Table  1), indicating that some residual Notch activity remains in mib ta52b mutants, as reported previously [20].
Mib Binds and Internalizes Jagged2a in Cells mib ta52b mutants display a global compromise in Notch activation, and mib was identified to encode an E3 ligase that activates Notch signaling by ubiquitylating and endocytosing Delta [20].
The phenotypic analysis of multi-cilia cells and principal cells in the duct suggests that mib genetically interacts with jagged2a. Since Delta has been shown to be a substrate of Mib and endocytosed after ubiquitylation [20], we asked if Mib physically interacts with Jagged2a, as shown for a human Mib paralog, Skeletrophin [21]. We checked the in vivo interaction of Jagged2a and Mib by immunoprecipitation and cotransfection experiments. Indeed, Mib bound to full-length Jagged2a and Jagged2a icd ( Figure 4S). Moreover, Myc-Jagged2 was localized to the cell surface (membrane) and cytoplasm when transfected alone ( Figure 4T) and to the perinuclear granules when cotransfected with Flag-Mib ( Figure 4U). The fact that Mib binds Jagged2a and facilitates its internalization suggests that Mib regulates Jagged2a in a way similar to Delta.

Notch1a and Notch3 Receptors Function Redundantly in Jagged2a-Mediated Lateral Inhibition
In the duct, we observed slight multi-cilia cell hyperplasia in notch1a/des (deadly seven) mutants and notch1a morphants ( Figure 4W; Table 1). notch3-utr MO and notch3-sp splicing MO against the exon 1-intron 1 boundary were designed; the former was specific in a sequence-dependent manner ( Figure  S4G-S4L) and the latter was able to induce splicing defects until at least 48 hpf ( Figure S4M). More multi-cilia cells were found in notch3 morphants ( Figure 4X; Table 1). The stronger cilia phenotype seen in notch3 morphants compared with that of notch1a mutants or morphants was consistent with the persistent notch3 expression in the pronephric duct and also demonstrated that Notch3 plays a more important role than Notch1a.
Loss of function of a single Notch receptor resulted in a phenotype that was less severe than that of jagged2a morphants. This suggests that Notch1a and Notch3 act redundantly. In fact, 91.1 6 12.2 multi-cilia cells were generated in notch3-sp MO-injected des th35b mutants in contrast to 42.9 6 4.7 and 55.7 6 12.2 multi-cilia cells in notch1asp and notch3-sp morphants, respectively (Figure 4V-4Y; Table  1). However, the multi-cilia cell phenotype in notch3-sp MOinjected des th35b mutants was not as severe as that of jagged2a morphants (Table 1). Consistently, Na þ , K þ ATPase b1a downregulation in notch3-sp MO-injected des th35b mutants was not as severe as that in jagged2a morphants (unpublished data). These data suggest that there may be a yet-unidentified Notch involved in this differentiation process. Her9 Acts Downstream in Jagged2a-Notch1a/Notch3-Mediated Lateral Inhibition The temporal and spatial expression of her9 in the distal pronephric duct suggests that it is one of the downstream target genes of Jagged2a-Notch1a/Notch3 signaling. We asked whether activation of her9 in the duct requires Jagged2a, Notch1a, and Notch3. her9 expression in the pronephric duct was reduced in jagged2a-sp morphants (77%, n ¼ 54; Figure 5A and 5B). While her9 expression was slightly reduced in des th35b / notch1a mutants ( Figure 5C and 5D) and in notch3-sp morphants (100%, n ¼ 45; Figure 5E), its expression was almost completely lost in notch3-sp MO-injected des th35b mutants (94%, n ¼ 36; Figure 5F). Similarly, its expression was almost completely lost in mib ta52b mutants ( Figure 5G and 5H). We next examined whether her9 is activated by Notch1a and Notch3. The constitutively active form, the intracellular domain (icd) of both Notch1a and Notch3, were used. her9 expression was activated by both Notch1a icd (22%, n ¼ 98; Figure 5I-5K) and Notch3 icd (17%, n ¼ 70; unpublished data). These experiments demonstrate that the activation of her9 expression in the pronephric duct requires Notch1a and Notch3, in addition to Jagged2a and Mib.
We further studied her9 function in the duct with her9-atg (effectiveness verified in [62]) and her9-utr MOs. her9 morphants exhibited multi-cilia cell hyperplasia, as demonstrated by rfx2 expression (Figure 5L and 5M; Table 1). The requirement of Jagged2a, Notch1a, and Notch3 for activation of her9 expression in the duct, and the multi-cilia cell hyperplasia in her9 morphants demonstrate that Her9 acts downstream of the Jagged2a-Notch1a/Notch3 pathway. However, multi-cilia cell hyperplasia in her9 morphants was not as severe as that in jagged2 morphants or notch3-sp MO-injected des th35b mutants (Table 1). One possibility is that Her9 is not completely knocked down by her9 MOs, because of the potential negative autoregulatory feedback on the transcription by its protein, similar to Hes7 [63]. Alternatively, there may be more effector(s) working in parallel with Her9. The latter explanation is particularly likely, since the her9 expression domain only partially overlaps with that of jagged2a ( Figure S3).

Multi-Cilia Cell Differentiation Requires Jagged2a from 17ss Onwards
rfx2 and jagged2a displayed mosaic patterns from 17 ss onwards ( Figure 3E and 3F); her9 was expressed in the distal duct domain from 17 ss onwards (Figures 2L and 3L). Moreover, her9-expressing cells were not multi-cilia cells ( Figure 3K-3M). The dynamic expression of these genes suggests that multi-cilia cells start to differentiate from 17 ss onwards. We next investigated whether Jagged2a-Notch signaling is required from as early as 17 ss. We found that rfx2 expression is uniform in the IM in wt embryos ( Figure  6A), mib ta52b mutants ( Figure 6B), and jagged2a-sp morphants (91%, n ¼ 33; Figure 6C) at 15 ss, while a neurogenic phenotype was obvious in mib ta52b mutants, indicating that multi-cilia cells do not start to differentiate before 15 ss. However, when rfx2 expression was limited to individual cells at 18 ss in wt embryos ( Figure 6D), rfx2-expressing cells were increased in the mib ta52b mutant ( Figure 6E) and jagged2a-sp morphants (90%, n ¼ 43; Figure 6F). Similarly, her9 and notch3 morphants exhibited multi-cilia cell hyperplasia from as early as 17 ss (unpublished data). These data indicate that Jagged2a-Mib-Notch3-Her9 is required for cell differentiation from as early as 17 ss.

Duct Cells Adopt a Principal Cell Character at the Expense of Multi-Cilia Cells When Notch Is Constitutively Activated
Multi-cilia cell hyperplasia is found in the mutants and morphants defective in the Jagged2a-Notch1a/Notch3-Her9 pathway. The increase of multi-cilia cells is most likely at the expense of principal cells, since no cell proliferation and apoptosis were detected in the duct of either wt or mib ta52b embryos ( Figure 7A-7F; Videos S1 and S2). Thus, we demonstrated that Jagged2a-Notch1a/Notch3-Her9 is required for specification of multi-cilia cells and principal cells through a lateral inhibition mechanism. We next asked whether duct cells will adopt a principal cell fate if Notch is constitutively activated in the duct. We crossed transgenic lines hsp70:Gal4 and UAS:myc-notch1a-intra and induced the expression of constitutively active Notch (Notch1a icd ) by heat-shock from 6-8 ss. Fluorescent double in situ hybridization with rfx2 and Na þ , K þ ATPase b1a at 24 hpf revealed that multi-cilia cells (rfx2-expressing cells) interpolate principal cells (Na þ , K þ ATPase b1a-expressing cells) in control embryos (heat-shocked hsp70:Gal4 only) (100%, n ¼ 4; Figure  7G), while principal cells are uniformly present at the expense of multi-cilia cells in hsp70:Gal4/UAS:Notch1a icd embryos (100%, n ¼ 9; Figure 7H). This confirms that Notch signaling makes binary choices between multi-cilia cells and principal cells in the pronephric duct.

Discussion
In this paper, we have shown that there are two major epithelial cell types found in the zebrafish distal pronephric duct. The mosaic pattern of multi-cilia cells and principal cells is controlled by Jagged2a/Notch-mediated lateral inhibition. Using available mutants and morphants deficient in genes functioning in Notch signaling, we demonstrated that one ligand, Jagged2a; two receptors, Notch1a and Notch3; and one downstream effector, Her9, are required for the differentiation and patterning of these two cell types. In addition, we showed that Mib is essential for this process, since it interacts with Jagged2a and facilitates Jagged2a internalization. In summary, our findings indicate a new function of Notch signaling in cell fate choice within a zebrafish kidney segment.
Interestingly, such a function of Jagged2-Notch signaling has not to our knowledge been found in mammals, although Jagged2 is expressed in the postnatal murine kidney [64,65]. This may be due to the early lethality of Jagged2 knockouts, which prevents the detection of such a function in metanephric kidneys. There are two zebrafish Jagged2 homologs, jagged2a and jagged2b. Most likely, the subfunctionalization of these two genes makes it possible for us to identify the function of jagged2a in zebrafish pronephros. Our findings warrant further study of the role of Jagged2 in mammalian kidneys by conditional knockouts.
The physiological functions of multi-cilia cells and principal cells are apparently different. While motile cilia on the apical side of the multi-cilia cells propel urea along the lumen of the pronephric duct [28], principal cells, which account for the majority of the cells in the kidney, reabsorb ions and other molecules according to fluid balance requirements. A plausible physiological significance of the interpolating pattern of multi-cilia cells and principal cells may be to coordinate the movement of the fluid and the process of reabsorption of the ions and other small molecules. Pax2a staining in the neural tube (asterisk) indicates the neurogenic phenotype in (C) mib ta52b mutants compared to that of (A) wt embryos. Three wt embryos and four mib ta52b mutants were examined. In addition, three wt embryos and three mib ta52b mutants were sectioned, and all sectioned slices were examined. No proliferating cells were found in the duct domain (unpublished data). (E and F) Apoptosis assay with TUNEL method on (E) wt embryos and (F) mib ta52b mutants at 21 hpf. TUNEL staining was found in the somite and neural tube (arrowheads), while TUNEL staining was not found in the pronephric duct (arrows). The brown staining in the duct is background staining. Ten wt embryos and five mib ta52b mutants were examined. (G and H) Fluorescent double in situ hybridization of rfx2 (red) and Na þ , K þ ATPase b1a (green) in 24-hpf embryos demonstrated that multi-cilia cells interpolate principal cells in (G) heat-shocked hsp70:Gal4 control embryos, while in (H) heat-shocked hsp70:Gal4/UAS:myc-notch1a-intra embryos, Na þ , K þ ATPase b1a expression is robustly found in the duct cells but rfx2 is not. Arrows point to rfx2-expressing cells.

Jagged2a-Mediated Lateral Inhibition of Multi-Cilia Cell Differentiation
Notch signaling is used for binary cell fate specification in many developmental processes. Notch activation in the signal-receiving cells inhibits them from expressing a set of genes leading to one fate and diverts them to an alternative program of differentiation. Consistent with other recent expression studies in mice, chicks, and zebrafish [46,66,67], we found that the temporal and spatial expression patterns of notch1a, notch3, and jagged2a fulfill their predicted roles in a multi-cilia cell to principal cell inhibitory signaling process in the zebrafish distal pronephric duct (Figure 8). While the expression of Notch receptors is evident throughout the duct epithelium, jagged2a expression becomes restricted to developing multi-cilia cells ( Figure 3G and 3J). In addition, her9 is expressed unevenly within this domain and most her9expressing cells are not colocalized with the multi-cilia cells ( Figure 3M). Our observations support the notion that lateral inhibition regulates cell characters in the distal pronephric duct. In all of the mutants/morphants, Jagged2a-Notch signaling is thought to be blocked to different degrees, and many to almost all epithelial cells in the zebrafish pronephric duct become positive for rfx2 and Pcm1, implying that they have adopted a multi-cilia cell-rather than a principal cellcharacter.
The studies shown here exemplify a very striking parallel between the role of Jagged2-Notch in the distal pronephric duct and the inner ear, and that of Delta-Notch signaling in neural tissue, the inner ear, and the intestine. In all of these cases, the obstruction of Notch signaling leads to a failure in lateral inhibition and to a great excess of one cell type at the expense of another. The supernumerary cell types are multicilia cells in the distal pronephric duct (this study), hair cells in the ear [46,67,68], neurons in the neural system ( [48,69,70] and reviewed in [14]), and secretory cells in the gut [71]. Similar to Delta-Notch signaling, the blockage of Jagged2a-Notch signaling results in an up-regulation of jagged2a expression, implying that expression of jagged2a itself is negatively regulated by Notch activity. If a cell in the wt organism expresses jagged2a, thereby activating Notch in neighboring cells, it will not only inhibit these neighbors from adopting the primary fate, but it will also down-regulate their expression of jagged2a. This generates a feedback loop that, over time, tends to amplify differences between adjacent cells so as to create a mixture of different cell types (Figure 8; [72]).

A Similar Mechanism for Other Similar Systems?
Multi-cilia cells are largely absent in mammalian kidneys, even though the primary cilium is present on principal cells of the tubule segment. We found interpolating multi-cilia cells and principal cells in the zebrafish distal pronephric duct. This mosaic cell pattern has been shown to be present in other anamniote vertebrates including marine teleosts [73], lampreys [74], and amphibians [75]. Notch signaling was required for the differentiation of speckled 4A6-positive cells in the posterior duct of Xenopus [22]. Our findings in zebrafish multi-cilia cells and the conserved pattern of cilia cells in amphibians [75] suggest that 4A6-positive cells are multi-cilia cells and that, in general, lateral inhibition may be involved in establishing the interpolating pattern of multi-cilia cells and principal cells in the ducts of anamniote vertebrates.
The renal collecting duct of mammalian kidneys comprises various kinds of intercalated cells (mediating acid and base transportation), principal cells (responsible for salt and water absorption), and inner medullary cells, which moderate all three types of transport. Inner medullary cells are ''hybrid'' cells-positive for both intercalated and principal cell markers [13,76,77]. Since Jagged1 expression [23] and a similar mosaic pattern of intercalated cells and principal cells [13] were observed in the collecting ducts of mouse kidneys, it is tempting to speculate that Notch signaling is involved in the differentiation and patterning of these different cell types in the mammalian collecting duct.
The efferent duct transports material from the rete testis to the epididymis by motile cilia [78]. Similarly, multi-cilia cells and principal cells are found exhibiting a mosaic pattern in the efferent duct of reptile (turtle, [79]) and mammal (rat, [80]). It would be interesting to see how these two cell types differentiate and whether Notch signaling is involved in this differentiation process.
MO and mRNA injection. To achieve maximal knockdown effect, To examine the knockdown specificity of jagged2a-utr MO and notch3-utr MO, we cloned the 59 UTR of both jagged2a (À130 to þ3) and notch3 (À226 to þ3) to the EcoRI and XbaI sites of pCS2-XLT-GFP vector [87]. Plasmids were linearized with NotI, and mRNA syntheses were carried out with mMessage mMachine Kit (Ambion, http://www. ambion.com). Then, 250 pg of mRNA, 250 pg of mRNA with utr-MO, and 250 pg of mRNA with mis-match-utr-MO were injected into eggs at the one-cell stage. Green fluorescent protein (GFP) was examined under a Leica dissecting microscope, and images were taken by the equipped Nikon Digital (DXM1200F; Nikon, http://www.nikon.com) at 24 hpf. The sequences of the five-mismatch MOs (in lower case, designed by Gene Tools) are as follows: 5mis-jagged2a-utr (À69 to À45) MO: ATcACgGGCGAgAG-GATCgTCCcTT and 5mis-notch3-utr (À76 to À51) MO: AgATgCTT-TAAcAAATcAATCGcCG.
We used mRNA of notch1a icd and notch3 icd to examine the effect of Notch activation on her9 expression. pCS2-myc-notch1a icd [55] and pCS2-myc-notch3 icd [20] were linearized with NotI, and mRNA synthesis was carried out with mMessage mMachine Kit (Ambion). Then, 100 pg of notch1a icd or 100 pg of notch3 icd mRNA was coinjected with 50 pg of GFP mRNA into one blastomere of two-cell-stage embryos. The mRNA-containing side was traced by following GFP expression, and mRNA functional expression was recognized by disrupted somite boundaries [88].
Fluorescent double in situ hybridization. The method was as previously described [91] except that substrates fluorescein-tyramide and Cy3-tyramide were respectively replaced with Alexa 488tyramide and Alexa 568-tyramide (Molecular Probes). The substrates are diluted in amplification buffer/0.0015% H 2 O 2 according to the product manual. Images were taken using a Zeiss Confocal LSM 510.
Heat-shock treatment. Embryos (6-8 ss) from hsp70:Gal4 (homozygous) and UAS:myc-notch1a-intra (heterozygous) crossings were transferred to petri dishes with prewarmed (39 8C) egg water and incubated at 39 8C incubator for 40 min. Embryos were transferred to Petri dishes with 28 8C egg water afterwards and incubated at 28 8C until 24 hpf. The embryos in which Notch was activated were recognized by their short body axis [83].
Plasmids, cell culture, transfection, immunoprecipitation, and Western blot analysis. pCS2-myc-Jagged2a and pCS2-myc-Jagged2a icd were cloned by in-frame fusion of jagged2a and jagged2a icd fragments to pCS2-myc vectors. The domain was predicted by the SMART program (http://smart.embl-heidelberg.de). pCS2-Flag-Mib was cloned by in-frame fusion of mib to the pCS2-Flag vector.
Immunocytochemistry. After a 24-h transfection, COS7 cells were fixed in methanol at 20 8C for 5 min and air-dried. Fixed cells were then incubated in blocking solution (10% goat serum in PBS) for 1 h, followed by staining with the appropriate primary antibodies (rabbit anti-Myc A14 [Santa Cruz Biotech] and mouse monoclonal anti-Flag M2 [Sigma-Aldrich], 1:1,000) in blocking solution for 1 h at room temperature. Subsequently, cells on coverslips were washed three times with PBS and incubated with Alexa 568-goat anti-rabbit antibody and Alexa 488-goat anti-mouse antibody (Molecular Probes). Coverslips were washed three times, mounted on glass slides, and observed under a Zeiss Confocal LSM 510. Figure S1. Yolk Extension Spans from Somite 8, the Location of the Anterior Limit of Cilia Tufts Nomarski pictures of 48-hpf zebrafish embryos revealed that the yolk extension spans from somite 8 (arrowhead). The arrow points to the pronephric tubule ventral to somite 3 [92]. The first three somites are not in a regular chevron shape, in contrast to the posterior somites. y, yolk; ye, yolk extension Found at doi:10.1371/journal.pgen.0030018.sg001 (859 KB TIF).   [93] revealed that notch1a is expressed in the pronephric duct spanning from somite 10 to 14 (arrows) at 18 ss. (C and D) Fluorescent double in situ hybridization of jagged2a and myoD revealed that mosaic jagged2a expression is found in the pronephric duct spanning from somite 8 to 14 (arrows) at 22 ss. (E and F) Fluorescent double in situ hybridization of jagged2a and myoD revealed that jagged2a-expressing single cells are found in the pronephric duct spanning from somite 8 to 14 (arrows) at 24 hpf. (G and H) Fluorescent double in situ hybridization of her9 and myoD revealed that her9 is expressed in the pronephric duct spanning from somite 10 to 12 (arrows) at 18 ss. (I and J) Fluorescent double in situ hybridization of jagged2a (green), slc4a2/ae2 (red, anterior), and ret1 (red, posterior) revealed that jagged2a-expressing single cells are found in the distal duct between the proximal duct (marked by slc4a2/ae2; [27]) and the cloaca (marked by ret1; [11]). Small arrows demarcate the jagged2a-expressing single cell domain, arrowheads demarcate the slc4a2/ae2-expressing domain, and big arrows demarcate the ret1-expressing domain. Found at doi:10.1371/journal.pgen.0030018.sg003 (3.6 MB TIF). Note that the number of multi-cilia cells was increased in jagged2a-utr morphants ( Table 1, 93%, n ¼ 231) but not in 5mis-match-jagged2a-utr morphants (97%, n ¼ 35). (G-I) Specificity of the notch3 morpholino. (G) Injection of notch3-utr-GFP mRNA at 250 pg produced green fluorescence, (H) coinjection of 0.38 pM notch3-utr-MO with 250 pg of notch3-utr-GFP inhibited GFP production, and (I) coinjection of 0.38 pM 5mis-match-notch3-utr-MO with 250 pg of notch3-utr-GFP did not inhibit its production. (J-L) Multi-cilia cells probed with rfx2 at 24 hpf in (J) wt embryos, (K) notch3-utr morphants, and (L) 5mis-match-notch3-utr morphants. Note that the number of multi-cilia cells was increased in notch3-utr morphants ( Table 1,   Video S1. Reconstruction (3-D) of pH3 and Pax2a Antibody Staining in the Distal Pronephric Duct of WT Embryos Reconstruction (3-D) of pH3 (green) and Pax2a (red) antibody staining in the distal pronephric duct of the embryo shown in Figure  7A reveals that pH3 nuclei are not localized to the pronephric duct domain of wt embryos at 18 ss. Note that the putative colocalized nuclei (yellow) turn partially to completely green at some rotating angles, suggesting that pH3-positive and Pax2a-positive cells are not colocalized, but in close vicinity to one another. Embryo is in lateral view, rotating around the dorsoventral axis. Found at doi:10.1371/journal.pgen.0030018.sv001 (747 KB AVI).

Supporting Information
Video S2. Reconstruction (3-D) of pH3 and Pax2a Antibody Staining in the Distal Pronephric Duct of mib ta52b Mutants Reconstruction (3-D) of pH3 (green) and Pax2a (red) antibody staining in the distal pronephric duct of the embryo shown in Figure  7C reveals that pH3 nuclei are not localized to the pronephric duct domain of mib ta52b mutants at 18 ss. Note that the putative colocalized nuclei (yellow) turn partially to completely green at some rotating angles, suggesting that pH3-positive and Pax2a-positive cells are not colocalized, but in close vicinity to one another. Embryo is in lateral view, rotating around the dorsoventral axis. Found at doi:10.1371/journal.pgen.0030018.sv002 (1.1 MB AVI).