Prospero distinguishes sibling cell fate without asymmetric localization in the Drosophila adult external sense organ lineage

The development of the external sensory organs in the adult Drosophila notum begins with the formation of sensory organ precursors (SOP) in the wing disc epithelium. The SOP divides along the anteroposterior (AP) axis to produce the IIa and IIb secondary precursors; the posterior IIa cell generates the external socket (tormogen) and hair (trichogen) cells, while the anterior IIb cell gives rise to the internal neuron and sheath (thecogen) cells (Gho et al., 1996; Gho and Schweisguth, 1998; Hartenstein and Posakony, 1989; Wang et al., 1997). Macrochaete SOPs begin dividing around 0-1 hours after pupal formation (APF) and the microchaete SOPs divide at 14-16 hours APF (Hartenstein and Posakony, 1989). A careful orchestration of intrinsic and extrinsic cues is required for the specification of cell fate in the SOP lineage (Campos-Ortega, 1996; Jan and Jan, 1995; Posakony, 1994). Extrinsic signaling is mediated by the Notch pathway, in which Delta or Serrate ligand activates the transmembrane Notch receptor, resulting in nuclear translocation of an intracellular domain of Notch together with the Suppressor of Hairless (SuH) transcription factor (reviewed in Bray, 1998). Loss or reduction of Notch signaling can transform IIa into IIb, socket into hair or sheath cell into neuron; an activated Notch receptor can produce the opposite cell fate transformations for each of these cell types (reviewed in Campos-Ortega, 1996). Thus, regulating Notch activity is essential for establishing distinct sibling cell fates at each division in the SOP lineage.

Asymmetric localization of the membrane-associated Numb protein is a primary mechanism for restricting Notch signaling activity to just one of two sibling cells. Numb is segregated into the IIb cell during mitosis (Rhyu et al., 1994) and numb mutants have a transformation of IIb into IIa cell fate that is dependent on Notch function (Guo et al., 1995; Uemura et al., 1989). These results suggest that Numb antagonizes Notch signaling to confer IIb cell fate. A similar relationship of Numb inhibition of Notch signaling also controls hair/socket and neuron/sheath sibling fates at the next step in the SOP lineage (Jan and Jan, 1995; Posakony, 1994) and many or all sibling neuron fates in the CNS (Skeath and Doe, 1998; Spana and Doe, 1996; Spana et al., 1995).

The prospero gene encodes a divergent homeodomain transcription factor that is asymmetrically localized in CNS, PNS and non-neural lineages in the embryo (Hirata et al., 1995; Knoblich et al., 1995; Spana and Doe, 1995). In the embryonic CNS, neuroblasts divide asymmetrically to produce smaller ganglion mother cells (GMCs), with each GMC generating a pair of neurons and/or glia (Goodman and Doe, 1993). Prospero protein is asymmetrically localized into the daughter GMC (Hirata et al., 1995; Knoblich et al., 1995; Spana and Doe, 1995), and prospero mutants fail to establish GMC-specific gene expression and show defects in neuron axon outgrowth (Doe et al., 1991; Vaessin et al., 1991). In the embryonic SOP lineage, Prospero protein is asymmetrically localized in the dividing SOP and partitioned into the IIb cell. Following IIb division, Prospero is transiently detected in the neuron and persists in the sheath cell throughout embryogenesis (Knoblich et al., 1995; Spana and Doe, 1995). prospero mutants show no cell fate defects in the embryonic SOP lineage, but there are clear defects in axon outgrowth (Doe et al., 1991; Vaessin et al., 1991). prospero is also expressed in the R7 photoreceptor neuron, where it is required for proper axon connectivity in the medulla (Kauffmann et al., 1996).

The Drosophila adult external SOP lineage is a model system for the study of asymmetric cell division (Campos-Ortega, 1996; Jan and Jan, 1995; Posakony, 1994), yet the expression and function of prospero in this lineage has not been determined. Here we investigate the role of prospero in the adult external sense organ lineage.

All Drosophila strains were raised on standard cornmeal, yeast and agar food. For identifying cells of the SOP lineage, we used A101 (an enhancer trap insert into the neuralized gene that expresses lacZ in all cells of the SOP lineage) and A1-2nd-29 (an enhancer trap line expressing lacZ in the hair and socket cell; Hartenstein and Jan, 1992). For generating prospero loss-of-function clones, we crossed hs-FLP/+; FRT82B pros¹⁷/+ males (kindly provided by V. Reddy and V. Rodrigues) to y w, P[ry+ hs-neo FRT] 82B, P[w+ hs-πM] 87E, Sb, P[ry+ y+] 96E / TM6B, Hu, Tb females and the progeny were heat shocked at 39°C for 60 minutes at second or third larval instar to induce FLP-mediated recombination. For generating misexpression of prospero in all cells of the SOP lineage, we used two methods: (1) we crossed y w; 109-68GAL4 / CyO (an enhancer trap insertion in the scabrous gene that expresses GAL4 in all cells of the SOP lineage; kindly provided by Y. N. Jan) to w; w+ UAS-prospero 17K-2 (kindly provided by Fumio Matsuzaki); when grown at 30°C the progeny from this cross show strong prospero expression in the SOP lineage at all stages assayed. (2) Larvae from the w; w+, UAS-hsp70-prospero 24H2 line (also kindly provided by Fumio Matsuzaki) were heat shocked at 39°C for 60 minutes at 14 hours APF to transiently induce prospero expression in all cells.

Dissections of 14-18 hours APF pupae were done in PBT (1× PBS, 1% BSA, 0.1% Triton), and the wing discs were either fixed for 40 minutes in fresh 4% formaldehyde diluted in PEM (100 mM Pipes, 2 mM EGTA, 1 mM MgSO₄) or treated for 2 hours in 5 mg/ml Colcemid (Sigma) in Schneider’s medium and then fixed as described above. Primary antibodies were applied for 2 hours at room temperature and the tissue was rinsed with PBT. Primary antibodies used were: fluorescein-conjugated anti-HRP (1:200; Jackson ImmunoRes.), rat anti-Elav 7E8A10 (1:10; Developmental Studies Hybridoma Bank at the University of Iowa), rabbit anti-β-galactosidase (1:3000; Cappel), rabbit anti-phosphohistone H3 (1:500; Upstate Biotechnology), mouse anti-Prospero MR2A (1:4; Spana and Doe, 1995) and rat anti-Supressor of Hairless (1:1000; kindly provided by F. Schweisguth). DTAF-, LRSC-or Cy5-conjugated secondary antibodies (1:300; Jackson ImmunoRes.) were applied for 2 hours at room temperature; the tissue was rinsed in PBT, mounted in 95% glycerol with 1% n-propylgallate (Sigma) and viewed on a BioRad MRC-1024 confocal microscope.

Adults were mounted in Hoyers medium and baked in a 55°C oven overnight and photographed on a Zeiss Axioplan. Scanning electron micrographs were taken of frozen specimens on a Hitachi microscope (kindly acquired by Dr John Bozzola at the Southern Illinois University electron microscope facility).

We used the A101 enhancer trap line to label all cells of the SOP lineage; A101 encodes a nuclear β-galactosidase (β-gal) protein (Huang et al., 1991; Usui and Kimura, 1993). To identify specific cells within the SOP lineage, we used the A1-2nd-29 enhancer trap line to mark the hair and socket cells (Hartenstein and Jan, 1992), anti-Elav to mark the pre-divisional IIb cell and the neuron (Bier et al., 1988), anti-BarH1 to mark the sheath cell (Higashijima et al., 1992; Leviten and Posakony, 1996), anti-phosphohistone H3 or condensed DNA to mark cells in mitosis (Ajiro et al., 1996; Broadus and Doe, 1997; Hendzel et al., 1997; Mahadevan et al., 1991), and external morphology to identify hair and socket cells; in addition, we use the dA10 enhancer trap line to mark hair and socket cells of the SOP lineage at the 4-cell stage (Gho and Schweisguth, 1998), and we use the stereotyped position of cells along the anteroposterior axis to confirm cell identity assignments (Gho et al., 1996; Gho and Schweisguth, 1998; Wang et al., 1997).

Using these cell-type-specific markers, we first detect Prospero in the nucleus of the IIb cell; it is cytoplasmic during IIb mitosis and then distributed to both neuron and sheath cell progeny. It is maintained at high level in the sheath cell, but only transiently detected in the neuron. Prospero is never detected in the SOP, IIa, socket and hair cells (Figs 1, 7A). In addition, we detect the IIb cell dividing before the IIa cell (Figs 1, 2). These results are surprising for two reasons. Previous reports claim that IIa divides before IIb (Gho et al., 1996; Gho and Schweisguth, 1998; Hartenstein and Posakony, 1989; Wang et al., 1997), but our genotypes show IIb dividing ahead of IIa (see Discussion). Also, the embryonic SOP localizes Prospero into the IIb cell (Knoblich et al., 1995; Spana and Doe, 1995), but the adult SOP does not localize Prospero protein, although Prospero is ultimately detected in IIb but not IIa.

To confirm the lack of Prospero asymmetric localization in the adult SOP lineage, we treated A101 imaginal discs with Colcemid to arrest SOP or IIb cells in mitosis. Colcemid treatment in embryos or in vitro primary embryonic cell cultures results in the accumulation of Prospero protein in asymmetric cortical crescents in mitotically arrested neuroblasts or SOPs (Broadus and Doe, 1997; Knoblich et al., 1995; Spana and Doe, 1995). We found that mitotically arrested SOPs did not have detectable Prospero protein (Fig. 2A). We observed mitotically arrested IIb cells in which the more posterior IIa cell was not yet in mitosis (Fig. 2B), and mitotically arrested IIb and IIa cells (data not shown); in both cases, the IIb cell showed cytoplasmic Prospero localization, consistent with our observations in the wild-type SOP lineage. These results strongly support our conclusion that the IIb cell divides before the IIa cell, and also confirm our finding that Prospero protein is not asymmetrically localized in the mitotic SOP or IIb cells.

To remove prospero function in the adult SOP lineage, we used the FLP/FRT system (Golic and Lindquist, 1989) to create prospero mutant clones in the wing and eye imaginal discs (see Materials and Methods); sense organs developing in the center of these mutant clones will be termed ‘prospero⁻ sense organs’. We used morphological criteria to score the fate of the external bristle and socket cell types in prospero⁻ sense organs. In females, prospero⁻ clones in the notum contain 86% single bristles and 14% double bristles (two bristle shafts emerging from a single or fused double socket) (Table 1; Fig. 3); similar phenotypes are seen in males but with lower frequency (Table 1). A more penetrant phenotype is seen in eye clones, where there are 54% double bristles and 46% single bristles. Double bristle sense organs in notum and eye are usually composed of one morphologically normal bristle and one stunted bristle (Table 1; Fig. 3). We were unable to accurately quantitate whether there are one or two socket cells in the prospero⁻ sense organs because two socket cells can fuse to form a single socket (Shiomi et al., 1994); however, double sockets are occasionally observed in prospero⁻ sense organs (Reddy and Rodrigues, 1999). Our results are consistent with the normal differentiation of the IIa cell (into bristle and socket) and a partial transformation of the IIb cell into a IIa cell (in the sense organs with double bristles).

To score the fate of the IIb cell in adult prospero⁻ sense organs, we assayed neuronal identity. Wild-type sense organs have a single associated neuron that has an axonal process projecting into the CNS and a dendrite that attaches to the base of the bristle/socket structure (Fig. 4A,B). In prospero⁻ single bristle sense organs, a neuron is detected just internal to the bristle/socket cells, but the axon is stunted and often has a circular trajectory, whereas the dendrite is extremely stunted and usually fails to connect with the base of the bristle/socket cells (Fig. 4C,D). Our results suggest that loss of prospero results in abnormal neuronal differentiation in prospero⁻ single bristle sense organs. In prospero⁻ double bristle sense organs, there is a complete loss of neuronal staining (Reddy and Rodrigues, 1999). These results are consistent with a transformation of IIb towards a IIa cell fate in the double bristle prospero⁻ sense organs.

To determine whether misexpression of prospero in the IIa cell, or its hair and socket daughter cells, is sufficient to alter their cell fate or differentiation, we selectively misexpressed prospero throughout the adult SOP lineage. We crossed 109-68GAL4 flies (which express GAL4 in all cells of the adult SOP lineage) with flies containing a UAS-prospero transgene (which allows GAL4-inducible expression of the full length Prospero protein; see Materials and Methods). The 109-68GAL4/UAS-prospero flies are lethal as pharate adults, and are virtually bald, showing less than 1% of the normal external sense organ bristles in the notum and eye (Fig. 5A-D; Table 1). These results show that prospero misexpression within the SOP lineage is incompatible with the normal development of the IIa cell and/or its bristle and socket progeny.

To determine whether the IIa cell is transformed into a IIb cell following prospero misexpression, we stained wild-type and 109-68GAL4/UAS-prospero pupal nota for cell-type-specific markers: neuron, Elav; sheath cell, BarH1; socket cell, SuH; there is no marker for the hair cell (Fig. 5E-H). In wild-type nota, each SOP lineage contains one Elav⁺ neuron, one BarH1⁺ sheath cell and one SuH⁺ socket cell (Fig. 5E,G). In 109-68GAL4/UAS-prospero nota, we observe clusters expressing BarH1 (sheath marker) and/or Elav (neuron marker) but never SuH (socket marker). They were predominantly either 4-cell clusters (29/54) or 2-cell clusters (21/54); a small fraction were 3-or 5-cell clusters (2 of each). Among the 4-cell clusters, the most common phenotype is one cell strongly BarH1⁺, one cell strongly Elav⁺ and two cells weakly expressing both BarH1 and Elav (15/29; Fig. 5F, arrowhead); other phenotypes are four cells weakly BarH1⁺ and one cell weakly Elav⁺ (5/29; Fig. 5F, arrow), and four cells weakly BarH1⁺ and two cells weakly Elav⁺ (4/29; data not shown). Among 2-cell clusters, the most common phenotype is that both cells have strong Elav and weak or no BarH1 (20/21; Fig. 5F, inset); the remaining cluster shows one Elav⁺ cell and one BarH1⁺ cell (data not shown). Within each cell, we observe either inverse levels of BarH1 and Elav, or mutually poor levels, suggesting that BarH1 and Elav negatively regulate each other. Our results show that misexpression of prospero blocks IIa cell differentiation, and can lead to duplications of IIb cell progeny. Taken together with the external hair/socket phenotype, we conclude that misexpression of prospero in the SOP lineage can produce a IIa- to-IIb cell fate transformation, resulting in a loss of IIa progeny (bristle/socket) and a duplication of IIb progeny (neuron/sheath cell). In addition, clusters with three or four BarH1⁺ sheath cells and only one Elav+ neuron indicate that misexpression of prospero may induce a low frequency neuron to sheath cell transformation.

In addition to the 109-68GAL4/UAS-prospero misexpression studies, we used a hsp70-prospero transgene to misexpress prospero at specific times during the SOP lineage as well as in all surrounding cells (see Materials and Methods). There is fairly severe balding in the notum, with 90% of the heat-shocked flies missing ∼20% of the microchaete (Fig. 6; Table 1). We also noticed an ‘edge effect’ along the perimeter of the bald areas in which double bristles are observed and a very low frequency of triple bristles (Fig. 6E). Internal cell fates under the bald patches of notum include a duplication of Elav⁺ neurons within each SOP lineage (Fig. 6F), consistent with a IIa- to-IIb cell fate transformation of SOP lineages in the bald regions of the notum.

The adult Drosophila external SOP lineage is a model system for the study of asymmetric cell division, in which both intrinsic and extrinsic mechanisms interact to control binary cell fate (Campos-Ortega, 1996; Jan and Jan, 1995; Posakony, 1994). The Prospero homeodomain protein is one of the first identified and best understood intrinsic determinants, yet until now, the expression and function of prospero in the adult external sense organ lineage has not been determined. We show that Prospero is first detected in the IIb nucleus, and during IIb mitosis, it is distributed to both neuron and sheath daughter cells. Loss of prospero results in either a IIb- to-IIa transformation or the defective development of the IIb neuronal progeny (Fig. 7B); ectopic prospero produces IIa- to-IIb cell fate transformations (Fig. 7C). Our observations suggest that Prospero plays an important role in specifying sibling cell differences, even if it is not asymmetrically localized. These findings have implications for the function of prospero homologues in other organisms, where the Prospero ‘asymmetric localization domain’ (Hirata et al., 1995) does not appear to be conserved.

The adult external SOP produces the IIa and IIb cells; IIa generates the external socket and hair cells, while IIb generates the internal neuron and sheath cells. Previous reports claim that IIa divides ahead of IIb (Gho et al., 1996; Gho and Schweisguth, 1998; Hartenstein and Posakony, 1989; Wang et al., 1997). However, in this and the accompanying paper (Reddy and Rodrigues, 1999), we find that the IIb cell divides before the IIa cell (Fig. 1). Although this has never been observed previously, we feel that our lineage is accurate for the following reasons. (1) At the 2-cell stage, we are sure the anterior cell is the IIb cell. Previous studies conclude that the anterior cell is the IIb cell (Gho et al., 1996; Gho and Schweisguth, 1998; Wang et al., 1997), and we show that anterior cell expresses Elav and Prospero, well-characterized markers for the neuron and sheath cells, which are progeny of the IIb cell (Bier et al., 1988; Bodmer et al., 1987; Gho et al., 1996; Gho and Schweisguth, 1998; Higashijima et al., 1992; Nakamura et al., 1994; Wang et al., 1997). (2) At the 2-cell stage, we are sure the anterior cell divides first. Mitotic cells are unambiguously identified using established markers: anti-phosphohistone (Ajiro et al., 1996; Hendzel et al., 1997; Mahadevan et al., 1991), condensed DNA (Broadus and Doe, 1997), or anti-tubulin in the accompanying paper (Reddy and Rodrigues, 1999). (3) We detect the neuron and sheath cells (IIb progeny) prior to IIa mitosis. At the 3-cell stage, we observe Elav and Prospero in the two anteriormost cells, which are thus the neuron and glia, derived from the IIb cell. The posterior third cell is the undivided IIa cell. (4) Finally, Colcemid-treated discs show A101⁺ 2-cell stages in which only the anterior cell is in mitosis; this clearly shows that the anterior cell, IIb, enters mitosis before the posterior cell, IIa.

How can we explain the previous findings showing that IIa divides before IIb? The initial description of IIa/IIb division timing was done by BrdU labeling (Hartenstein and Posakony, 1989); BrdU labeling detects the time of DNA replication (S phase) but not the time of mitosis, and it is possible that IIa goes through S phase before IIb, even though IIb enters mitosis before IIa. A more recent analysis was performed using tubulin staining to mark mitotic cells; this study appears to show the posterior IIa cell dividing prior to the anterior IIb cell (Gho and Schweisguth, 1998). We can imagine three possible explanations to resolve the difference between our results and that of Gho and Schweisguth (1998). (1) There may be genetic background differences between the A101 lines used in each study, with IIa dividing first in their genotype and IIb dividing first in our genotype. However, we detect IIb dividing first in several genotypes (data not shown; Rodrigues and Reddy, 1999). (2) It may be that the IIb cell divides first to produce two previously unrecognized daughters: a ‘IIIb’ tertiary precursor cell, that will later generate the neuron and sheath cells, and the previously identified ‘fifth cell’, the origin of which is unknown (Hartenstein and Posakony, 1989; Usui and Kimura, 1993). This hypothesis requires that the ‘fifth cell’ is Prospero⁺ and rapidly migrates away. If so, the remaining IIa/IIIb might be mistaken for a IIa/IIb pair, with IIa dividing first to produce the socket/hair cells and the IIIb dividing later to generate the neuron/sheath cells. (3) The ‘2-cell stage’ with the anterior cell dividing first (observed by Gho and Schweisguth, 1998) may actually be a 3-cell stage in which the sheath cell was not identified. Although we think this is unlikely, the sheath cell does have a low level of A101 β-gal and is often positioned internal to the neuron and IIa cell; without a sheath cell marker this cell might be missed.

In the embryonic SOP lineage, Prospero is asymmetrically partitioned into the IIb cell during the SOP division (Hirata et al., 1995; Knoblich et al., 1995; Spana and Doe, 1995), but we show that asymmetric localization of Prospero does not occur in the adult SOP lineage. Prospero is not detected in the adult SOP; it is first seen in the IIb cell but not in the IIa cell. It is unlikely that there is low level of Prospero protein present in the SOP and asymmetrically partitioned to the IIb cell, because prolonged mitotic arrest of the SOP fails to reveal an accumulation of cortical asymmetric Prospero protein (similar treatment leads to a robust cortical crescent of Prospero in embryonic SOPs and neuroblasts; Broadus and Doe, 1997). In addition, Prospero protein is present in the cytoplasm of the mitotic IIb cell, and is distributed to both the neuron and sheath daughter cells; it subsequently becomes nuclear in the sheath cell and disappears from the neuron. Finally, Miranda protein is required to anchor Prospero protein at the cortex of mitotic neuroblasts and SOPs in the embryo (Ikeshima-Kataoka et al., 1997; Shen et al., 1997), but there is no detectable Miranda protein in the SOP lineage (Fuerstenberg et al., 1998). Taken together, these observations show that accumulation of Prospero in the IIb cell is not the initial cause of IIb cell fate, but rather a secondary step in the pathway of IIb cell specification. Similarly, elimination of Prospero from the neuron is an intermediate step in neuronal differentiation, rather than a consequence of partitioning Prospero into the sheath cell and out of the neuron.

What mechanisms lead to prospero expression in the IIb cell but not in the IIa cell? Specification of IIa/IIb cell fates is determined by the relative activity of Notch signaling. Productive Notch signaling results in IIa cell fate; asymmetric localization of Numb protein into the IIb cell blocks Notch signaling and results in the IIb cell fate (reviewed in Campos-Ortega, 1996). We propose that productive Notch signaling prevents prospero expression in the IIa cell, whereas lack of Notch signaling is allows prospero expression in the IIb cell. Consistent with this model, SOP lineages with unregulated Notch signaling produce a pair of IIa cells that both fail to express prospero, while SOP lineages lacking Notch function produce two IIb cells that both express prospero (Reddy and Rodrigues, 1999). One effector of Notch signaling in the IIa cell is the zinc-finger transcriptional repressor Tramtrack, which may directly or indirectly repress prospero expression (Guo et al., 1995). Interestingly, prospero is expressed in the R7 neuron during eye development (Kauffmann et al., 1996; Spana and Doe, 1995) and tramtrack mutants have supernumerary R7 neurons while tramtrack misexpression reduces R7 differentiation (Li et al., 1997; Tang et al., 1997). Thus, a similar Notch-, tramtrack-dependent pathway may repress prospero expression in both the R7 photoreceptor neuron and the IIa cell. It should be noted that a somewhat different mechanism must be involved in repressing prospero in the neuron but not sheath cell; in this case, Notch signaling is required for sheath cell fate (Guo et al., 1996; Hartenstein and Posakony, 1990), the cell that maintains prospero expression. The lack of Notch-mediated repression of prospero expression in the sheath cell may reflect the fact that Notch signaling is SuH-dependent in the IIa cell, but SuH-independent in the sheath cell (Wang et al., 1997).

We demonstrate a role for prospero in establishing different IIa/IIb cell fates based on both loss-of-function and misexpression experiments. A significant fraction of the SOP lineages lacking prospero function show a duplication of the external bristle (a progeny of the IIa cell; Fig. 3, Table 1) and a loss of the neuron (a progeny of the IIb cell) (Reddy and Rodrigues, 1999). We were unable to accurately score socket cell fate, because multiple socket cells can generate a single, fused socket structure. The simplest interpretation of the double bristle prospero⁻ sense organs is that the IIb cell has become partially or fully transformed into a IIa cell, resulting in duplicate hair/sockets and loss of neuron/sheath cell. We think it is unlikely, but cannot rule out, the possibility that the neuron is transformed into a duplicate hair cell and the sheath cell is unaffected. In both notum and eye, however, there are still many ‘single bristle’ sense organs that have an associated neuron and, in these sense organs, the IIb cell must have been specified relatively normally. Thus prospero is not strictly necessary for IIb cell specification, but its function is important for the high-fidelity specification of IIb cell fate.

While the presence of prospero in the IIb cell is important for reliable IIb cell specification, the absence of prospero from the IIa cell is absolutely essential for IIa cell specification. Misexpression of prospero in the IIa cell and its progeny results in a fully penetrant loss of a socket cell marker (SuH) as well as the morphological external socket and hair structures; there is a corresponding increase in the internal Elav⁺ neurons and BarH1⁺ sheath cells. Our misexpression experiments show that absence of Prospero in the IIa cell is required for normal IIa development, and that presence of Prospero in the IIa transforms it partially or fully to the IIb cell fate. Thus, differential expression of prospero between IIa and IIb siblings is essential for normal SOP development. Similar results were observed using transient heat-shock-induced misexpression of prospero although, in these experiments, we observed a very low frequency of double and triple bristle sense organs at the borders of the bald areas. The cell lineage of these rare sense organs is unknown.

It is interesting to consider the different mechanisms by which prospero acts to distinguish sibling cell fate. During embryonic neuroblast cell division, localization of Prospero into the daughter GMC is necessary for GMC development, but exclusion of Prospero from the neuroblast is relatively unimportant for neuroblast development (since neuroblast development is fairly normal in miranda mutants where Prospero remains in the neuroblast; C. Q. D., unpublished results). In contrast, during the adult SOP lineage, it appears equally important to remove Prospero from the IIa cell as well as provide it to the IIb cell. Another key difference between the adult SOP lineage and the embryonic SOP and neuroblast lineages is the timing of cell divisions. There are several hours between each cell division in the adult SOP lineage, considerably longer than the 40-60 minutes cell cycle of embryonic neuroblasts and SOPs (Bodmer et al., 1989; Hartenstein et al., 1987). The shorter cell cycles of the embryonic lineages may require asymmetric localization of Prospero for efficient specification of sibling cell fate, whereas the longer adult SOP cell cycles may provide time for the action of other regulatory mechanisms (e.g. Notch-mediated repression of prospero expression).

In single bristle prospero⁻ sense organs, we observe a single neuron with profound defects in neurite outgrowth. The defects in axon and dendrite outgrowth and connectivity could be due to lack of prospero function in the IIb cell, a non-autonomous effect due to lack of prospero function in the sheath cell, or the absence of prospero function in the neuron itself. We think the first possibility is unlikely because axon outgrowth defects can be observed in R7 neurons, which do not arise from a Prospero⁺ precursor cell. We think the second possibility is unlikely because lack of sheath cells (in glial cells missing embryos) does not generate similar axon outgrowth defects (Jones et al., 1995). We favor the third model, in which prospero has a direct function in the neuron, because many neurons with different origins (CNS, PNS, eye) transiently express prospero and all show a similar prospero mutant phenotype: stunted and misrouted axons (Doe et al., 1991; Vaessin et al., 1991; Kauffmann et al., 1996).

We thank V. Reddy and V. Rodrigues for kindly communicating unpublished results, F. Matsuzaki for providing numerous published and unpublished reagents, V. Reddy and V. Rodrigues, Y. N. Jan, J. Posakony, F. Schweisguth and the Bloomington stock center for antibodies and/or fly stocks, Dr John Bozzola at the Southern Illinois University electron microscope facility, and Tonia von Ohlen, Andrew Zelhof and Francois Schweisguth for comments on the manuscript. This work was supported by the NIH (Oregon) and the Howard Hughes Medical Institute (Illinois), where L. M. was a Research Technician II and C. Q. D. was an Associate Investigator.