Drosophila cellularization is a large-scale cytokinetic event in which thousands of syncytial nuclei are simultaneously packaged into individual cells. Cleavage furrows extend from the embryonic surface between neighboring nuclei, rapidly generating a polarized epithelial sheet (reviewed in Knoblich, 2000 ; Schejter and Wieschaus, 1993 ). The apical, lateral, and basal domains of these cells are established during cleavage furrow extension by the targeted delivery of membrane components from the Golgi to the cell surface (Lecuit and Wieschaus, 2000 ). Domain borders are marked by adherens type junctions: the basal junction separating the basal and lateral domains during early cellularization and the apical spot junctions separating the apical and basolateral domains during late cellularization (Müller and Wieschaus, 1996 ; Hunter and Wieschaus, 2000 ). Although cellularization is a specialized process, many of the components and mechanisms of conventional cytokinesis and polarity establishment are conserved, making it a powerful system for in vivo studies of cytokinesis, cell polarity, and the establishment of cell–cell junctions.

The patterns of synchronous nuclear division and migration during the 13 division cycles preceding cellularization lead to the formation of a cortical monolayer of nuclei just beneath the embryonic surface. A bulge of plasma membrane, or somatic bud, forms above each nucleus within the cortical array (Foe and Alberts, 1983 ). During cellularization, regions of somatic bud contact give rise to two structures: the furrow canal and the basal junction (Lecuit and Wieschaus, 2000 ). The furrow canal is the basal tip of the nascent cleavage furrow and contains the presumptive basal membrane, along with cytokinetic proteins including actin, myosin, anillin, and septins (Warn and Robert-Nicoud, 1990 ; Young et al., 1991 ; Fares et al., 1995 ; Field and Alberts, 1995 ). Just above this structure is the basal junction, a region of tight membrane association containing the adherens junction proteins E-cadherin, α-catenin, and the β-catenin homologue Armadillo (Arm; Oda et al., 1993 , 1998 ; Hunter and Wieschaus, 2000 ). During the first phase of cellularization, the cleavage furrows move inward slowly as new membrane is delivered to the apical surface of the cell. Once the cleavage furrows pass the base of the nucleus, they advance rapidly as new membrane is added to the lateral cell surface (Lecuit and Wieschaus, 2000 ). Throughout cellularization, the furrow canal is maintained as a distinct domain, separated from the newly generated apicolateral membrane by the basal junction. At the completion of cleavage furrow invagination, the membrane of the cleavage furrow expands laterally to form the basal surface of the cell, the basal junction is degraded, and junctional components in the lateral membrane coalesce to form permanent adherens junctions at the apical-basolateral boundary. Thus, cellularization is a complex pattern of membrane addition and junction formation that allows the coupling of cytokinesis and the establishment of cell polarity.

nullo is one of three zygotic genes specifically required for organization of the cellularization front in the Drosophila embryo (Merrill et al., 1988 ; Wieschaus and Sweeton, 1988 ). The Nullo protein localizes to the furrow canal and basal junction during the early stages of cellularization (Postner and Wieschaus, 1994 ) and is required for the proper formation of the basal junction (Hunter and Wieschaus, 2000 ). In nullo mutant embryos, Arm properly accumulates at sites of somatic bud contact, but instead of forming concentrated basal junctions it expands along the lateral membrane. This leads to the failure of cleavage furrow invagination and the formation of multinucleate cells. Ectopically expressed Nullo protein can block the lateral movement of catenin–cadherin complexes that is required for coalescence of apical spot junctions. These results suggest that Nullo prevents the movement of adherens junction components within the membrane, acting to stabilize the initial accumulation of cadherins and catenins while a stable basal junction is formed.

Cloning of the nullo gene revealed little about potential mechanisms for its involvement in basal junction formation: nullo encodes a small, highly basic protein that lacks homology to any previously characterized proteins (Rose and Wieschaus, 1992 ). In this article we identify a homologue of Nullo from D. virilis and use interspecies rescue and deletion analysis to identify conserved regions of Nullo that are required for basal junction formation. We then show that Nullo is a myristoylprotein and demonstrate in vivo that N-terminal myristoylation and an N-terminal basic cluster both contribute to the membrane targeting of Nullo, although the relative contributions of these motifs changes as cellularization progresses.

Rose and Wieschaus, 1992

Transgenes were assayed for adult rescue activity by crossing w; P[nullo]/P[nullo] males to Df(1)6F1-2/FM7 females and scoring the number of surviving Df(1)6F1-2/Y; p[nullo ]/+ offspring as a percentage of the Df(1)6F1-2/+; p[nullo ]/+ offspring. These crosses were done at 18°C, 25°C, and 29°C using multiple lines carrying each transgene. Rescue at 18°C is reported relative to the background survival of Df(1)6F1-2/FM7 females crossed to males lacking a transgene.

To measure transgene rescue of the nullo phenotype, embryos from C(1)DX y w; P[nullo]/P[nullo] lines raised at 25°C were stained to visualize cell outlines and nuclei (see below). The percentage of embryos displaying a mutant phenotype was determined and a rescue efficiency was calculated based on the expected 25% mutant embryos produced by the C(1)DX y w stock.

To assay the activity of UAS-nullo deletion transgenes, males carrying the UAS transgene over a balancer were crossed to mat67.15 driver females, and the progeny were raised at 18°C. The percentage lethality is given as 1 − (surviving transgene flies/surviving balancer flies).

For transgenes that rescued the nullo phenotype, subcellular localization was examined in Df(1)6F1-2/Df(1)LVII9; p[nullo]/p[nullo] lines. Both Df(1)6F1-2 and Df(1)LVII9 contain additional lethal mutations, so the transheterozygote was produced to make a stock that would be viable in the presence of the nullo transgene. Nonrescuing transgenes were examined in nullo-X embryos produced by the C(1)DX y w; P[nullo]/P[nullo] stock.

Rose and Wieschaus, 1992

O'Neil and Belote (1992)

Tautz and Pfeifle, 1989

To visualize Armadillo or Nullo proteins, embryos were heat-methanol fixed (Müller and Wieschaus, 1996 ) and stained by standard methods using mouse anti-Armadillo 7A1(Riggleman et al., 1990 ) or mouse anti-Nullo 5C3-12 or 2F8-18 (Postner and Wieschaus, 1994 ). HA-tagged Nullo protein was detected using anti-HA antibodies (Covance Laboratories, Madison, WI).

To label Golgi vesicles, embryos were fixed in 18.5% formaldehyde saturated with heptane, methanol-popped, and stained using mouse anti–β-COP antibodies (gift of V. Malhotra; Ripoche et al., 1994 ).

For actin staining, embryos were fixed in 18.5% formaldehyde saturated with heptane, manually devitellinized, and treated with Alexa 488 phalloidin.

Alexa 568–labeled antibodies (Molecular Probes, Eugene, OR) were used for single stainings, and a combination of Alexa 546– and Alexa 488–labeled secondaries (Molecular Probes) were used for double labeling. All embryos were stained with Hoechst, mounted in Aquapolymount (Polysciences, Warrington, PA), and imaged using a Zeiss LSM-510 confocal microscope (Thornwood, NY).

The construct DMDV was produced by PCR amplification of the DV nullo gene using primers containing NdeI and BglII sites. The resulting DV nullo ORF cassette was then placed in the 2.2-kb DM nullo genomic fragment.

The ΔM mutation was created using site-directed mutagenesis to change the start of the DM nullo protein from MGSTHS to MASTHS. ΔE was created by amplifying a fragment of DM nullo corresponding to amino acids 169–213 using standard PCR. The remaining constructs were created using PCR gene SOEing (Vallejo et al., 1995 ). ΔP to changes amino acids 35–60 of DM Nullo from KIQRLVLRKLSISARKQKRLNKRSKH to NIQNLVLNNLGVGGGGFQNGCSANTA. These changes neutralize the net charge of the unconserved N-terminal region and were based on the final 17 amino acids of the unconserved sequence from DV Nullo. ΔMP incorporates the changes from both ΔM and ΔP. ΔA removes amino acids 61–82, ΔB removes amino acids 89–102, ΔC removes amino acids 109–135, and ΔD removes amino acids 145–164.

All of the PCR products were subcloned into the NdeI and BglII sites of the cassette vector using NdeI and BglII sites on the external PCR primers. The 2.2-kb PstI nullo genomic fragment was then subcloned into the CaSpeR-4 transformation vector (Thummel and Pirotta, 1991 ).

The UAS full-length nullo construct was created by placing a PstI fragment from nullo-HA (Hunter and Wieschaus, 2000 ) into pBS KS− (Stratagene). A NdeI-SfiI fragment was then removed from the nullo ORF and replaced with the corresponding fragment from each of the deletion constructs ΔM, ΔP, and ΔMP. The resulting HA-tagged deletions were PCR amplified with primers that introduced a 5′ EcoRI site and a 3′ XhoI site, and these fragments were subcloned into pUAST (Brand and Perrimon, 1993 ) to produce the UAS-N (full-length nullo), UAS-ΔM, UAS-ΔP, and UAS-ΔMP constructs.

Germline transformation was carried out by standard methods (Spradling, 1986 ).

To quantitate UAS-Nullo expression, Western blot analysis was carried out as above, except that detection was performed using the ECL Plus kit (Amersham). The membrane was exposed to Hyperfilm ECL (Amersham) and then scanned and analyzed using the Storm imaging system (Molecular Dynamics, Sunnyvale, CA).

Duronio et al., 1990

Simpson and Wieschaus, 1990

Rose and Wieschaus, 1992

Raising the LII27.32R5 stock at 18°C increased the proportion of viable nullo males. To determine if temperature would also affect the viability of nullo null mutations we examined a line transheterozygous for Df(1)6F1-2 and Df(1)LVII9, two small deficiencies that delete the nullo gene. Although this line produces no viable adults at 25°C, we could recover up to 10% of the nullo mutant class as viable adults at 18°C. To quantitate the effect of temperature on basal junction formation we examined Df(1) 6F1-2 embryos at 29°C, 25°C, or 18°C and determined the percentage of nuclei in multinucleate cells (Figure 1A). Wild-type embryos do not produce multinucleate cells at any of the temperatures tested. In contrast, at 29°C the majority of nullo mutant embryos had >80% of their nuclei in multinucleate cells (Figure 1, A and D). At 25°C, nullo mutant embryos had 50–80% of their nuclei in multinucleate cells. When the temperature was lowered to 18°C, there was a further reduction in the number of defective cleavage furrows: most embryos had <30% of their nuclei in multinucleate cells, and these cells were generally bi- or trinucleate (Figure 1, A and C). In addition, only 17% of the embryos had a nullo phenotype, as determined by anti-Arm staining of basal junctions, suggesting that the remaining 8% (of an expected 25%) of nullo mutant embryos were phenotypically wild type. This class may correspond to the viable mutant adults recovered at 18°C. Thus, reducing the temperature can compensate for the absence of Nullo protein and increase the number of functional basal junctions.

Figure 1 Decreased temperatures can partially compensate for the loss of Nullo protein. (A) Df(1) 6F1-2 embryos were raised at 18°C, 25°C, or 29°C. At each temperature, the phenotypic severity of 20 embryos was determined by counting (more ...)

In wild-type embryos, Nullo is required to maintain the initial accumulation of basal junction components, allowing for the rapid construction of a stable junction before the onset of cleavage furrow invagination. This function may be more important when the nascent complexes are destabilized by high temperatures. In contrast, embryos raised at low temperatures may have an inherent stability of the immature junctions, which allows them to form in the absence of Nullo.

Rose and Wieschaus, 1992

Rose and Wieschaus, 1992

Ibnsouda et al., 1995

Rose and Wieschaus, 1992

Figure 2 The D. virilis Nullo homologue. (A) In situ hybridization to D. virilis embryos shows no detectable nullo expression in early embryos. (B–D) Transcript levels increase just before cellularization and peak at the start of cycle 14. (E) During (more ...)

Like DM nullo, DV nullo is an intronless gene that encodes a small basic protein with a predicted size of 233 amino acids and a calculated pI of 10.8. The amino acid sequences of DM Nullo and DV Nullo are 47% identical, and the similarity of the proteins rises to 58% when conservative amino acid changes are included. Most of the conserved residues fall into clusters (Figure 2, A–E) in the C-terminal two thirds of the protein (Figure 2G). The N-terminal sequence of the proteins is comparatively unconserved, although it contains an N-terminal myristoylation site (M) and a cluster of positively charged residues (P) that are present in both proteins (Figure 2G).

We were interested in determining if the conservation between the two proteins was sufficient to allow rescue of D. melanogaster nullo mutant embryos using the DV nullo gene. We created transgenic lines that contained the D. virilis nullo ORF under the control of the D. melanogaster nullo promoter (DMDV). To assay the rescue of mild, moderate, and severe cellularization defects, we tested the ability of the transgene to restore adult viability to nullo mutants raised at 18°C, 25°C, or 29°C. A single copy of the DMDV transgene was sufficient to rescue mutant flies at all three temperatures (Figure 3A). To assess rescue of the cellularization defects directly, we examined embryos from a nullo mutant line that carried two copies of the DMDV transgene. At 25°C none of the embryos from this line had a detectable nullo phenotype, suggesting the transgene completely restores basal junction formation (Figure 3B).

Figure 3 Rescue activity of the nullo transgenes. (A) The viability of Df(1)6F1-2 adult males carrying each transgene was determined at 18°C, 25°C, and 29°C. A + indicates survival above background at each temperature. (B) Embryos (more ...)

Transgenes containing a D. virilis genomic fragment are also capable of rescuing D. melanogaster nullo mutant embryos, although the rescue activity is lower than constructs containing the D. melanogaster promoter region and the D. virilis protein coding region (our unpublished results). This would suggest that there is at least partial conservation of the regulatory regions that drive blastoderm-specific expression. These regions are likely to be contained within a small upstream region, because the D. virilis fragment contains only a limited region of upstream DNA. We have also observed that a D. melanogaster nullo transgene containing only 400 base pairs of upstream DNA is capable of reproducing the short burst of nullo expression, although the anterior-posterior striping seen during nullo degradation is less pronounced (our unpublished results). Interestingly, this small upstream region includes a regulatory motif shown to be conserved between DM nullo and sry-α, another zygotic cellularization gene with a similar pattern of expression (Ibnsouda et al., 1993 ).

The conservation of sequence and expression pattern, along with the ability to perform interspecies rescue, confirm that the D. virilis gene identified is a structural and functional homologue of D. melanogaster nullo. The rescue experiments also identify a set of conserved features that are sufficient for Nullo activity during cellularization.

Mutations ΔA-ΔD did not affect the level of Nullo protein expression: extracts from the transgenic embryos contained truncated proteins of the predicted size at levels comparable to wild type (Figure 4A). Surprisingly, deletion of these conserved regions had no effect on the subcellular localization of the Nullo protein. These deletion proteins accumulated at the cellularization front and showed normal punctate staining in the cortical cytoplasm (Figure 4B). ΔE deleted the epitope for the 5C3-12 anti-Nullo antibody, and although we were able to use an alternative mAb to demonstrate that this protein is produced at reduced levels, we were only able to detect faint staining in the embryo (Figure 4, A and B).

Figure 4 Conserved regions required for Nullo function. (A) Western blot showing the expression of the C-terminal deletions ΔA–ΔE. Expression of the endogenous Nullo protein is shown in the first lane. The extracts from ΔA–ΔD (more ...)

Adult rescue assays and the subsequent examination of nullo mutant embryos carrying two copies of each transgene revealed that ΔC and ΔD were capable of rescuing the nullo phenotype at 18°C, 25°C, and 29°C, whereas ΔA, ΔB, and ΔE failed to rescue at any temperature (Figure 3, A and B). The fact that ΔA and ΔB produce wild-type levels of Nullo protein, which is enriched at the cellularization front yet fail to rescue the nullo phenotype, suggests that these conserved regions are specifically required for Nullo to stabilize the accumulation of basal junction components in the nascent cleavage furrows.

Resh, 1999

To determine if DM Nullo is a substrate for N-terminal myristoylation we expressed the nullo gene in a bacterial strain carrying the gene for yeast N-terminal myristoyltransferase (NMT) and grew the strains in the presence of ³H-myristate (Duronio et al., 1990 ). The resulting extracts contain Nullo protein that is labeled with ³H-myristate in an NMT-dependent manner (Figure 5A). We further confirmed the specificity of this labeling by creating a glycine to alanine mutation in position two of the consensus site. This alteration has been shown in other systems to block N-myristoylation by eliminating the site of myristoyl attachment (reviewed in Resh, 1999 ). The Gly2 to Ala change in the Nullo protein completely blocked labeling with ³H-myristate (Figure 5A).

Figure 5 The N-terminal region is required for Nullo localization. (A) A fluorograph of [3H]myristate–labeled extracts from bacteria expressing Nullo, Nullo containing the G2A mutation, yeast N-terminal myristoyltransferase (NMT), Nullo (more ...)

These data indicate that the Nullo protein contains a functional N-terminal myristoylation site. Because the mechanism for N-terminal myristoylation appears to be conserved in Drosophila (Deichaite et al., 1988 ; Ntwasa et al., 1997 ; Benting et al., 2000 ), it is likely that Nullo exists as a myristoylprotein in vivo.

Although expressed at normal levels, each of these mutations caused specific disruptions in the subcellular localization of the Nullo protein. The wild-type Nullo protein begins to accumulate during nuclear division cycle 12 and localizes to the metaphase furrows during mitosis of cycle 13. At the start of cycle 14, Nullo protein localizes to the cell surface and then becomes enriched at the cellularization front and in punctate cytoplasmic structures (Figure 5C; Postner and Wieschaus, 1994 ). The identity of the punctate structures is still unknown, although they do not colocalize with actin granules or with β-COP in cis-Golgi–derived vesicles (our unpublished results).

Examination of ΔM revealed that, in the absence of the myristoylation site, Nullo has a normal localization to the pseudocleavage furrows during mitosis of cycle 13. However, at the onset of cellularization, Nullo adopts a nuclear localization (Figure 5C) that persists until midcellularization, when levels of localized protein decline abruptly. ΔP greatly reduces level of punctate cytoplasmic staining but has only a mild effect on membrane localization (Figure 5C). This suggests that association with the punctate cytoplasmic structures is not a prerequisite for membrane localization. The loss of both the myristoylation site and positive charge completely blocks Nullo protein localization: ΔMP is distributed diffusely throughout the cortex of the embryo (Figure 5C). These results suggest that the combination of conserved N-terminal motifs is essential for targeting Nullo protein to the cellularization front.

We then examined the ability of these mutant transgenes to rescue the nullo phenotype. We found that a single copy of either ΔM or ΔP was sufficient to rescue nullo flies to adult viability at 18°C, 25°C, or 29°C (Figure 3A) and that embryos with two copies of ΔM or ΔP have a normal hexagonal array at 25°C (Figure 3B). A single copy of ΔMP, however, partially restored viability only at 18°C. Two copies of ΔMP were capable of restoring a normal hexagonal array to 49–83% of embryos at 25°C, depending on the transgene line tested. Although ΔMP is not enriched at the plasma membrane, some portion of this protein may still reach the cellularization front. As with the hypomorph LII27.32R5, this low level of Nullo protein may increase the likelihood of basal junction formation.

Postner and Wieschaus, 1994

Hunter and Wieschaus, 2000

Brand and Perrimon, 1993

Hunter and Wieschaus, 2000

Prolonging the expression of full-length Nullo at levels characteristic of early cycle 14 is sufficient to produce the ectopic phenotype. To ensure that the potential absence of an ectopic phenotype in embryos from UAS-ΔM, UAS-ΔP, or UAS-ΔMP lines was not due to subnormal protein levels, the transgene insertions were first evaluated for protein expression levels. For each construct, several transgene insertions were analyzed by Western blot, Nullo expression levels were quantitated, and transgene insertions were assigned relative expression values. We then examined the subcellular localization of transgene insertions with similar expression levels.

When expressed during late cellularization, Nullo protein accumulates along the entire plasma membrane and in punctate structures in the basal region of the newly formed epithelial cells (Figure 6; Hunter and Wieschaus, 2000 ). Although ΔM under the endogenous promoter localizes to the membrane before cellularization and later becomes nuclear, UAS-ΔM expressed during late cellularization shows no membrane association and is enriched in the nucleus (Figure 6). Ectopic UAS-ΔP shows weak membrane localization, along with a diffuse cytoplasmic staining but no visible punctate staining (Figure 6), as seen with the endogenous ΔP protein. UAS-ΔMP shows no membrane staining and no observed enrichment in the nucleus versus the cytoplasm (Figure 6).

Figure 6 The N-terminal motifs are required for ectopic Nullo localization and function. UAS-N, UAS-ΔM, UAS-ΔP, and UAS-ΔMP embryos were stained with anti-HA antibody and with anti-Arm to visualize the localization of the ectopic Nullo (more ...)

The activity of a given mutant transgene was assayed by measuring the degree of lethality caused by its ectopic expression. Using the full-length transgene to elevate Nullo protein levels during late cellularization to 69% of that observed at the beginning of cycle 14 caused lethality in 87% of embryos and blocked the formation of apical junctions (Figure 6). UAS-ΔP at similar levels caused lethality in only 43% of embryos and only resulted in a partial blockage of apical junction formation (Figure 6). UAS-ΔP therefore shows a reduction in activity but can still interfere with apical junction formation. The ectopic expression of UAS-ΔM and UAS-ΔMP, on the other hand, has no effect on viability and does not block apical junction formation (Figure 6). The failure of these mutant proteins to localize to the cell surface suggests that membrane localization is absolutely required for Nullo's interference with apical junction formation.

Nicklas and Cline, 1983

Wieschaus et al., 1984

Eberl and Hilliker, 1988

Like DM nullo, DV nullo appears to require a relatively small region of upstream DNA for proper expression. The DV nullo gene also lacks introns, a characteristic of DM nullo and the other zygotic cellularization genes bottleneck and serendipity-α (Vincent et al., 1985 ; Schejter and Wieschaus, 1993 ). This may reflect the need to produce large amounts of transcript in the short time between the onset of zygotic transcription and cellularization. The DV nullo gene is expressed in the same temporal pattern as DM nullo, with zygotic transcription beginning just before cellularization and degradation of the transcript taking place during late cellularization. The spatial pattern is also similar, with uniform expression until midcellularization and the presence of several stripes of expression during late cellularization. As with sry-α, the positions of the stripes differ between D. melanogaster and D. virilis (Ibnsouda et al., 1995 ). However, previous studies suggest that exact pattern of nullo expression along the anterior-poterior axis is not functionally relevant (Rose and Wieschaus, 1992 ).

Comparison of the DM Nullo and DV Nullo predicted proteins revealed several interesting features. The first of these is the conservation of the N-terminal myristoylation site and the presence of an adjacent unconserved, but highly basic, region. This suggested that the N-terminal myristoylation site was important for Nullo localization or function and that it likely acted as part of a “myristate plus basic” motif used in membrane localization (see below). The second feature was the presence of five distinct conserved domains, separated by short nonconserved stretches. This pattern of conservation is evident even among closely related Drosophila species. In the nullo genes of D. oreana, D. lutescens, and D. yakuba (Caccone et al., 1996 ) the N-terminus and C-terminal conserved blocks have a high degree of sequence identity, whereas the C-terminal “linker” regions have already begun to diverge.

By deleting single conserved regions we identified two regions that affect Nullo activity without affecting protein localization. These proteins contain all the domains required not only to reach the plasma membrane but also to become enriched at the cellularization front. However, they still fail to rescue the basal junction phenotype, suggesting that they may define regions of Nullo that interact with components of the junction itself. Although the novel sequence of these regions makes it difficult to draw conclusions about their specific role, they provide a basis for further analysis, including the identification of interacting proteins. Two other conserved regions can be removed without affecting Nullo localization or activity, suggesting that they are dispensable or that they provide a redundant function. Truncation of the C terminus reduced protein levels, suggesting this region cannot be removed without affecting the stability of the Nullo protein. In addition to identifying functionally important regions, the deletion analysis also shed light on the previous difficulties in isolating point mutations in nullo. Given the small size of the nullo gene, the lack of conservation in the N-terminal third of the protein and the ability to delete substantial regions of the C terminus without affecting function, the probability of isolating a lethal point-mutation in nullo would be quite small.

Resh, 1999

Taniguchi and Manenti, 1993

Sigal et al., 1994

Zhou et al., 1994

Nullo is one of ~35 sequences in the D. melanogaster genome that contains the MGXXXS/T consensus site for N-terminal myristoylation. Although the majority of these genes have been characterized genetically and developmentally, only a handful have actually been shown to be myristoylated (Neel and Young, 1994 ; Teng et al., 1994 ; Rossi et al., 1999 ), and the in vivo role of the myristoylation site has been studied only in the neural protein Numb (Knoblich et al., 1997 ). Using a yeast NMT system (Duronio et al., 1990 ), we demonstrated that Nullo is modified by the addition of a myristoyl group, and this is prevented by the substitution of alanine for glycine in position 2. We believe that this in vitro demonstration of myristoylation is relevant to the in vivo modification of Nullo for two reasons. First, the consensus target for NMT activity appears to be well conserved between Drosophila and other organisms. Both Drosophila embryonic extracts and Schnieder cells have been shown to have an NMT activity that modifies myristoyl proteins from other organisms in a specific manner, and the corresponding NMT gene has been shown to be expressed before cellularization (Deichaite et al., 1988 ; Ntwasa et al., 1997 ; Benting et al., 2000 ). Second, the glycine to alanine mutation which blocks myristoylation in vitro has a striking effect on the subcellular localization of Nullo in vivo, suggesting that the N-myristoylation site is important for the targeting of Nullo to the plasma membrane.

During metaphase of nuclear cycle 13, Nullo lacking a myristoylation site (ΔM) still shows normal localization to the plasma membrane of the metaphase furrows. However, at the onset of cycle 14, ΔM is lost from the cellularization front and begins to accumulate in the nucleus where it remains until midcellularization. UAS-ΔM expressed during late cellularization showed no membrane association, but was observed in the nucleus. The difference in ΔM localization before and after the onset of cellularization suggests that some aspect of Nullo targeting or membrane association changes at this point. The fact that ΔMP, which lacks both N-terminal domains, shows no early membrane association suggests that the initial association of Nullo with the plasma membrane requires the presence of a positive charge and that the positive charge is no longer sufficient for membrane localization after the onset of cellularization. The electrostatic interaction may be disrupted by changes in the phospholipid content of the apical membrane as new membrane is inserted during cellularization. In contrast, endogenous and ectopic expression of ΔP demonstrates that the myristoylation site alone is sufficient for partial membrane localization throughout cellularization.

After cycle 13, ΔM accumulates in the nucleus, which is surprising, because Nullo does not contain a predicted nuclear localization signal, and we have never observed wild-type Nullo protein in the nucleus. It is possible, however that the basic cluster confers a nuclear localization that is normally overridden by the membrane targeting of the N-terminal myristoyl group. In the absence of the myristoyl group, Nullo would associate with the nucleus via the polybasic region, and as we observed, the association with the nucleus would be diminished by the removal of the positively charged cluster (ΔMP). A similar observation has been made with the Moloney MuLV Gag protein, where mutation of the N-terminal myristoylation site roughly doubles the amount of protein found in the nucleus (Nash et al., 1993 ).

Although the ΔM, ΔP, and ΔMP proteins all retained some ability to rescue the nullo mutant phenotype, only ΔP was able to partially block apical junction formation when expressed ectopically during late cellularization. One explanation for the difference in activity may lie in the levels of functional protein required in each of these assays. As seen with the LII27.32R5 allele, the rescue of the nullo phenotype requires levels of Nullo protein much lower than those normally seen at the start of cycle 14. In contrast, the ectopic effects on apical junction formation require at least half the protein level seen at the onset of cellularization. In the case of ΔM the difference in activity may also result from the fact that this protein retains membrane localization during the cycle 13 to cycle 14 transition when basal junctions are formed but lacks any detectable membrane localization when ectopically expressed during late cellularization. The fact that both assays showed a decline in Nullo activity that correlated with the loss of Nullo from the plasma membrane suggests that the “myristate plus basic” motif controls membrane localization and that the membrane association of Nullo is essential for its role in junction formation.

The fact that both ΔM and ΔP retained some ability to associate with the plasma membrane is unusual, as removal of either the myristoyl group or the basic region is usually enough to prevent plasma membrane association in “myristate plus basic” proteins (reviewed in Resh, 1999 ). Interestingly, mutation of the myristoylation site of Numb also failed to affect membrane localization, although truncation of the first 40 amino acids of the N-terminus caused the protein to accumulate in the cytoplasm (Knoblich et al., 1997 ). It remains to be seen if Numb contains a bipartite localization motif similar to that seen in Nullo and whether the independent membrane targeting activity of myristoyl groups and basic clusters is a common occurrence in Drosophila.

Loss of Nullo has its primary effect on the formation of basal junctions from the small points of membrane contact that arise at the start of cellularization. The results presented in this article argue that Nullo is not an essential component of the junction itself but is required to stabilize nascent junctions, especially at higher temperatures where the membrane contacts may be more fluid. Ectopic expression of Nullo during late cellularization may also stabilize junctional components as they reach the plasma membrane, but in this case it would prevent the clustering required to form the apical adherens junctions (Hunter and Wieschaus, 2000 ). The effect of Nullo on adherens junctions is compatible with numerous direct and indirect modes of action. However, the structural studies presented here suggest that Nullo's activity requires its cell-surface localization and therefore may involve a physical interaction between Nullo and some component of the basal junction, or the accompanying cytoskeleton.

The authors thank members of the Wieschaus and Schupbach laboratories for helpful discussions; Jen Zallen, Jorg Grosshans, Jeff Thomas, and two anonymous reviewers for comments on the manuscript; Reba Samanta and Joe Goodhouse for assistance with histology and confocal microscopy; Romy Knittel for technical assistance in cloning and characterizing the D. virilis homologue; and Cynthia Hsuan-Hung for maintaining stocks. The authors are grateful to J. Gordon, V. Malhotra, P. Schedl, and D. St. Johnston for providing stocks and reagents. This work was supported by the Howard Hughes Medical Institute grant 5R37HD15587 from the National Institute of Child Health and Human Development and grant 95-00258/3 from the U.S.-Israel Binational Science Foundation.