As a step toward a fuller understanding of the role that Delta-Notch signaling plays in patterning the zebrafish neural tube, we compared expression of several delta and notch genes to the distribution of neural precursor cells and differentiating neurons. We use the term, neural precursor cell, to indicate proliferative neuroepithelial cells that did not express Hu proteins [40]; we identified proliferative cells by their incorporation of BrdU. Conversely, we categorized those cells that expressed Hu proteins as post-mitotic, differentiating neurons. We examined the RNA expression patterns of delta and notch genes using a sensitive fluorescent detection method, choosing as our timepoint 24 h (hours of development at 28.5°C; [41]). At this stage, primary neurons have differentiated and some secondary neurons are postmitotic and have differentiated while others remain in the cell cycle [42]. The assays described here do not discriminate between primary and secondary neurons.

We show here expression of three notch genes: notch1a (n1a), notch1b (n1b) and notch5 (n5) [43, 44]. Cells throughout the dorsoventral extent of the trunk neural tube expressed each of these genes. Both Hu-negative and Hu-positive cells expressed n1a (Figure 1A,B,C) whereas n1b and n5 expression appeared to be complementary to Hu localization (Figure 1D,E,F,G,H,I), indicating that neural precursors expressed n1a and n1b but differentiating neurons did not. Particularly intense staining for each gene was evident apically, suggesting that transcripts were concentrated where the neural tube lumen will form.

We showed previously that post-mitotic primary neurons express dlA transiently at the 4 somite stage [28]. Many neural plate cells in excess of known post-mitotic neurons expressed dlA and dlD, suggesting that these cells include neural precursors [26,27,28]. By contrast, dlB expression appeared to be limited mostly to post-mitotic neurons [27]. At 24 h, cells throughout the trunk neural tube, including Hu-positive cells, expressed dlA (Figure 1J,K,L). Both Hu-positive and Hu-negative cells also expressed dlB (Figure 1M,N,O), however, apparently fewer of each class expressed dlB at high level than dlA. Some Hu-positive and many Hu-negative cells expressed dlD (Figure 1P,Q,R). Together, these observations suggest that a subset of neural precursors and cells specified for neuronal development express dlA, dlB and dlD.

To examine directly whether proliferative cells of the trunk neural tube express delta and notch genes, we exposed embryos to a brief pulse of BrdU at 24 h. BrdU-positive cells expressed each of the notch (data not shown) and delta genes we tested (Figure 2A,B,C). However, only a fraction of BrdU-positive cells expressed delta genes:11.2% expressed dlB (169 cells counted), 19.2% expressed dlA (228 cells counted) and 20.1% expressed dlD (186 cells counted). Together with the above data, we conclude that at least a portion of proliferative, neural precursor cells express each of these genes and that post-mitotic neurons express n1a, dlA, dlB and dlD. We speculate that the subset of BrdU-positive cells that express delta genes are those fated to exit the cell cycle following the next mitosis and we interpret our failure to detect n1b, n5, dlB and dlD RNA in all or some presumptive neurons as evidence that transcription of these genes is not maintained after cell cycle exit.

Widespread neural expression of delta and notch genes is consistent with the possiblity that cell-cell signaling contributes to pattern formation within the zebrafish nervous system. To examine this directly, we performed experiments designed to test whether lateral inhibition patterns primary motor neurons. Zebrafish primary motor neurons are distributed with a striking periodicity [45, 46]. In previous studies we found that islet1 (isl1) expression, which marks primary motor neurons, was modified over time. Initially, contiguous cells in a row close to floorplate expressed isl1, whereas later, isl1-expressing cells were interspersed with cells that did not express isl1 [46]. Currently, we cannot distinguish between the possibilities that alternating cells downregulate isl1 expression or that cells that do not express isl1 move between those that do. Nevertheless, to maintain this periodicity, isl1-negative cells must not transcribe isl1, despite being equidistant with isl1-positive cells from notochord and floorplate, presumed sources of signals that induce motor neuron development [47]. To test whether newly born primary motor neurons inhibit neighboring cells from adopting similar fates, we ablated identified CaP and VaP motor neurons [48] and asked whether these cells were replaced. Normally, CaP develops in every hemisegment, midway between adjacent somite boundaries. VaP develops in about one half the hemisegments; when present it is next to CaP. Both cells express isl2 whereas other primary motor neurons do not [46]. Previous experiments showed that when CaP was removed from 13 somite stage embryos, it was never replaced [49]. For this experiment, we extirpated CaP and, when present, VaP from the right side of spinal hemisegment 7 (R7) in 11 somite stage embryos, 1 hour earlier in developmental time than in previous experiments [49]. We let these embryos develop until the 19 somite stage and probed for expression of isl2. Of a total of 13 embryos in which these ablations were performed, 11 developed a CaP or CaP and VaP in the ablated position, indicating that the ablated cells were replaced. The remaining two embryos showed no replacement. Although we do not know the normal fates of the cells that replaced ablated primary motor neurons, these observations are consistent with the possibility that, during normal development, primary motor neurons inhibit neighboring cells from adopting a primary motor neuron fate.

To further test the role of Delta-Notch mediated lateral inhibition in zebrafish neurogenesis, we examined embryos homozygous for a missense mutation of dlA (dlA^dx2). We previously described the effect of this mutation on development of floorplate, notochord and hypochord [37] and dorsal spinal sensory neurons and neural crest [38]. To gauge the extent to which Notch signaling is disrupted in dlA^dx2 mutant embryos we examined expression of ngn1, a proneural gene negatively regulated by Notch activity [50]. Zebrafish neural plate stage embryos express ngn1 in three longitudinal domains of the prospective trunk neural tube ([50]; Figure 3A), which correspond to regions that express elevated levels of dlA and in which the earliest neurons are born [28]. Cells within the longitudinal domains of wild-type embryos do not express ngn1 uniformly; rather, in situ RNA hybridization revealed higher levels of ngn1 transcripts in some cells than in others ([50]; Figure 3A). By contrast, cells within the longitudinal domains of dlA^dx2 mutant embryos expressed ngn1 at uniformly high levels (Figure 3B). Notably, the mediolateral extent of each domain appeared unchanged. We conclude that the dlA^dx2 allele disrupts Notch signaling within neurogenic regions of zebrafish embryos but does not influence the formation of neurogenic domains.

In addition to elevated levels of ngn1 expression, dlA^dx2 mutant embryos have excess Rohon-Beard spinal sensory neurons and reduced numbers of trunk neural crest precursors [37, 38]. To investigate the basis for these and other neural phenotypes in dlA^dx2 mutant embryos we compared the balance of proliferative neural precursor cells and post-mitotic neurons in wild-type and dlA^dx2 mutant embryos. When treated with a pulse of BrdU at 24 h, wild-type embryos revealed BrdU-positive cells and a distinct, non-overlapping set of basally located huC-positive neurons (Figure 4A). By contrast, 24 h dlA^dx2 mutant embryos had few BrdU-positive cells and a large excess of huC-positive cells (Figure 4B). We conclude that, in dlA^dx2 mutant embryos, most neural precursor cells cease dividing and develop as neurons. Consequently, the neural precursor pool is depleted in mutant embryos and replaced by post-mitotic neurons.

The above observations indicate that neural precursor cells prematurely differentiate in dlA^dx2 mutant embryos. We imagined two possible consequences for neural cell fate specification. First, all normally developing neural cell types might be specified but some might arise ahead of schedule. Second, the progenitors for cell types that normally arise late might instead develop identities appropriate for early-born cell types, resulting in an excess of some cell types and reduction of others. We previously showed that disruption of Delta-Notch signaling, by injection of RNA encoding a dominant negative form of frog Delta-1 protein into early cleavage stage zebrafish embryos, caused formation of excess early-specified primary motor neurons and reduced numbers of later-specified secondary motor neurons [28]. Our observations of dlA^dx2 mutant embryos confirm and extend those results. dlA^dx2 mutant embryos had excess presumptive primary motor neurons and Rohon-Beard neurons (Figure 5A,B; [37, 38]). By contrast, the number of secondary motor neurons in mutant embryos was greatly reduced (Figure 5C,D). Primary motor neurons can be distinguished by cell body position, axon projection, and gene expression [46, 51]. To learn how the dlA^dx2 mutation affected specification of primary motor neuron identity, we probed embryos for expression of isl2, which serves as a marker for CaP and VaP, two of four identified primary motor neurons [46].18 h mutant embryos had excess isl2-positive cells in the ventral trunk neural tube (Figure 5E,F). These cells appeared as clusters of about 3–5 cells, located midway between adjacent somite borders, consistent with the normal position of CaP and VaP motor neurons. This indicates that disruption of Delta-Notch signaling causes development of excess primary motor neurons, which have identities appropriate for their positions in the neural tube. Previously, we showed that primary motor neuron identity is dependent upon cell body position within the ventral spinal cord [46, 52]. We conclude that Delta-Notch signaling regulates specification of appropriate numbers of primary motor neurons but does not regulate specification of motor neuronal identity.

We also investigated whether other cell types were affected in mutant embryos by examining expression of markers of various interneurons, including lim1, lim5 and Pax2. Spinal cord cells begin to express lim1 and pax2 by about 14–15 h [53, 54] and lim5 by 24 h [55] indicating that these cells are born relatively early in neurogenesis. Excess numbers of cells expressed each of these markers in dlA^dx2 mutant embryos (Figure 5G,H,I,J,K,L). Notably, the dorsoventral position of cells expressing each of these markers was similar in mutant and wild-type embryos. Finally, dlA^dx2 mutant embryos had greatly reduced numbers of presumptive radial glia cells, labeled by antibody zrf-1 (Figure 5M,N). Although we do not know the time or place that these cells originate in the neural tube, we do not detect expression with this marker until 30–36 h, well after the onset of primary neuron birth at about 11 h and secondary motor neuron birth at about 16 h. Together with the above BrdU results, these observations indicate that disruption of Delta-Notch signaling causes excess neural precursors to be specified for early-born neuronal fates at the expense of later-born neuronal and glial fates, without interfering with mechanisms that specify early-born neuronal identity along the dorsoventral axis.

Expression of activated Notch proteins can prevent neural cell differentiation [15, 16, 20, 24, 25, 29]. In vertebrate embryonic nervous systems, proliferative undifferentiated cells express Notch genes [13,14,15,16,17,18,19, 21, 24]; thus, Notch activity may maintain the neural precursor state. Notably, some differentiated neural cells, such as Müller glia and radial glia, express Notch genes [30, 32] raising the possibility that Notch signaling does more than simply block differentiation. By contrast, only post-mitotic neural cells have been shown to express Delta genes [18, 20, 23]. This pattern of expression is consistent with the idea that specified, post-mitotic cells use Delta signaling to prevent neighboring cells from differentiating. However, it differs from the fly embryonic central nervous system wherein all cells of the prospective neuroectoderm express Delta with the exception of delaminated neuroblasts [56]. This indicates that, in flies, precursors express Delta and that those specified for neuroblast development downregulate its expression.

Here, we show that some zebrafish neural cells in S phase of the cell cycle and post-mitotic differentiating neurons express delta genes. We interpret these data to mean that a subset of proliferative neural precursors upregulate delta gene expression. Upregulation might engage a feedback loop mechanism, in which Notch activation in neighboring cells maintains them in a proliferative, unspecified state whereas those cells that upregulate Delta expression, and have decreased Notch activity, are driven to exit the cell cycle and to differentiate.

Lateral inhibition was originally defined by functional tests in which ablation of a cell of primary fate allowed an adjacent cell to take that fate, but ablation of a cell of secondary fate had no effect [8]. Ablation experiments performed on grasshoppers showed that neuroblasts laterally inhibit neighboring cells from also developing as neuroblasts [5]. Genetic analyses of fly development then led to the view that lateral inhibition is mediated by interaction of Delta and Notch proteins expressed on neighboring cells [12]. Our ablation studies, coupled with our studies of Delta and Notch expression and with analysis of a dlA mutation, provide evidence consistent with the possibility that similar mechanisms operate during specification of individual neural fates in a vertebrate nervous system.

Mutant and gene overexpression phenotypes in various vertebrate model systems support the notion that Delta-Notch signaling mediates lateral inhibition in the central nervous system [16, 20, 24,25,26,27,28,29, 35]. However, to our knowledge, until now direct evidence for lateral inhibition in the vertebrate central nervous system has not been reported. Previously, when primary motor neurons were ablated in 13 somite stage embryos, approximately 6 hours after birth, they were not replaced [49]. Here, we ablated primary motor neurons 1 hour earlier in developmental time and found that, in nearly every case, they were replaced. Together, our ablation results reveal an important feature of vertebrate neuronal specification: that the mechanisms that specify fate must be exquisitely timed. We view the failure to replace primary motor neurons ablated at 13 somite stage as indicating that the time at which a cell can be specified for that fate has passed, either because neighboring precursors have been specified for other fates or because the signals that specify primary motor neurons are no longer present. Future experiments, aimed at learning the fates of cells neighboring primary motor neurons may distinguish between these possibilities.

Here we took advantage of the fact that dlA^dx2 mutant zebrafish embryos survive several days after onset of neuronal differentiation to investigate how Delta-Notch signaling influences neural cell fate. We show that mutant embryos have a significant reduction in proliferative neural cells and a concomitant increase in neurons. These neurons have fates characteristic of the earliest-born neurons in zebrafish. Strikingly, the molecular identities of the excess, early-born neurons in dlA^dx2 mutant embryos are appropriate for their positions in the trunk neural tube. This is evident across the dorsoventral axis of the neural tube as well as within localized areas of the ventral neural tube occupied by distinct motor neurons. In other words, early-born neurons are specified appropriately in dlA^dx2 mutant embryos; what these embryos lack is the ability to regulate the number of cells specified for any particular fate.

We also showed that dlA^dx2 mutant embryos have reduced numbers of later-born cell types including secondary motor neurons and radial glia. One possible explanation for this effect is that the premature exit of neural precursor cells from the cell cycle depletes the precursor pool. In this view, Delta-Notch signaling diversifies neural cell fate by guiding the orderly, and prolonged, transition from neural precursor cells to specified cells. Specification of a particular cellular identity might depend entirely upon the influence of instructive signals that change with time. For example, zebrafish ventral neural precursors might be specified differently by different lengths of exposure to Hedgehog (Hh) signaling. Those competent to respond to brief exposure to Hh might be specified as primary motor neurons whereas those subject to prolonged exposure prior to differentiation might develop as secondary motor neurons [57]. Notch activity might be important for specification of secondary motor neurons by holding them in an unspecified state for long enough to receive sufficient exposure to Hedgehog signaling. Disruption of Notch signaling would then allow an excess of precursor cells to respond promptly to Hedgehog signaling, producing an excess of primary motor neurons and a concomitant deficit of secondary motor neurons, precisely the phenotype of dlA^dx2 mutant embryos we show here.

A second possibility, supported by recent publications, is that Notch activity specifies certain cell fates, in particular, glial fates. Notch activity promotes gliogenesis in forebrain, retina, and cultured neural crest [30,31,32, 34]. Our observation that dlA^dx2 mutant embryos lack radial glia of the trunk neural tube also is consistent with this possibility. In fact, Notch activity might both hold neural precursor cells in an unspecified state and subsequently promote a subset of them to develop as glia. Conditional misexpression of activated Notch in individual precursor cells of the trunk neural tube might provide a way to distinguish between these possibilities.

Embryos from the University of Oregon and Vanderbilt University laboratory colonies were raised at 28.5°C. Staging was according to Kimmel et al. (1995). The dlA^dx2 allele is incompletely penetrant and variably expressive [37,38,39]. Mutant embryos shown here are those exhibiting the most severe mutant phenotypes.

Individual CaP or CaP and VaP motor neurons were removed from the right side of segment 7 (R7) of 11 somite stage embryos by aspiration with a micropipette broken manually to a tip diameter of 8–12 μm, using methods described previously [52].

Manually dechorionated embryos were labeled with BrdU by incubating them for 20 minutes on ice in a solution of 10 mM BrdU and 15% DMSO in Embryo Medium [58]. The BrdU solution was replaced with Embryo Medium and the embryos were incubated for 20 minutes at 28.5°C. The embryos were then anesthetized using 3-aminobenzoic acid ethyl ester and fixed in 4% paraformaldehyde. in situ RNA hybridization on BrDU-labeled embryos was carried out as described previously [59] except that the probes were not hydrolyzed. Following color development, the embryos were treated with 2 N NaOH for 1 hr, washed with PBSTx (phosphate buffered saline with 0.01% TritonX-100), blocked 1 hr in PBDTx (PBS, 1% DMSO, 0.01% TritonX-100 plus 2% sheep serum and 2 mg/ml BSA, then incubated 2 hr with anti-BrdU (monoclonal G3G4, Developmental Studies Hybridoma Bank). Detection was carried out using a biotinylated goat anti-mouse secondary antibody followed by streptavidin conjugated to peroxidase (Jackson ImmunoResearch Laboratories, Inc.). A brown precipitate was formed using Fast Dab (Sigma).

For fluorescent detection of RNA hybridization and antibody labeling, embryos were first treated for RNA hybridization and then sectioned using a cryostat microtome. The sections were incubated with monoclonal 16A11 anti-Hu antibody (University of Oregon Monoclonal Antibody Facility) and anti-digoxygenin-AP antibody (Roche). Anti-Hu labeling was detected using goat anti-mouse Alexa 488 antibody (Molecular Probes) and anti-digoxygenin-AP labeling was detected using HNPP Fluorescent Detection (Roche). Images were collected using a Spot camera (Diagnostic Instruments, Inc.) mounted on an Olympus AX70 microscope. Double-label images were created using Photoshop to merge individual red and green images.

Additional in situ RNA hybridizations were performed as described previously [59]. Additional primary antibodies used were zn-8 to detect Dmgrasp and zrf-1 (Developmental Studies Hybridoma Bank) and polyclonal rabbit anti Pax-2 (Berkeley Antibody Co.).