marbles mutants: uncoupling cell determination and nuclear migration in the developing Drosophila eye

Morphogenesis may be thought of as two distinct yet coordinated processes. Cells become specialized to express particular gene products and not others, and also adopt particular shapes and positions. Much progress has been made in under-standing how cells are determined to express particular genes. Far less is known about how cell movements and shape changes are coordinated with cell specification during development.

Cell shape in the developing Drosophila eye is governed by the position of the nucleus in the cell (Tomlinson, 1985). The eye develops in an epithelial monolayer called the eye imaginal disc, in which cells are suspended between apical and basal membranes. The nuclei of undifferentiated cells are randomly positioned within the monolayer and as the cells undergo pattern formation, their nuclei migrate in a characteristic manner (Tomlinson, 1985). The beginning of eye morphogenesis at the posterior of the disc is marked by the basal migration of nuclei and cell bodies into the morphogenetic furrow. The furrow moves anteriorly through a pool of unpatterned cells and in its wake, cells begin to assemble into rows of unit eyes, also called facets or ommatidia (Ready et al., 1976). Each facet is composed of 22 cells, which assemble in a particular sequence, guided by a series of cell interactions (Tomlinson, 1988; Ready, 1989; Banerjee and Zipursky, 1990; Rubin, 1991). First, in the larval eye disc, the eight photoreceptors assemble in a particular order, followed by the four cone cells (Ready et al, 1976; Tomlinson, 1985; Tomlinson and Ready, 1987a). The accessory (pigment and bristle) cells assemble later in the pupal eye disc (Cagan and Ready, 1989). As the photoreceptor and cone cells join the facets, their nuclei and cell bodies rise apically (Tomlinson, 1985).

We have identified a Drosophila gene called marbles (marb), which is required for many of the nuclear migrations that accompany cell determination in the developing eye. We show that cell determination and differentiation in marb mutant eyes occurs for the most part normally despite the failure of the nuclei to rise. Thus, although nuclear migration and cell determination are temporally coordinated during normal eye development, the two processes are separable. The marb mutant phenotype indicates that apical localization of nuclei and cell bodies is not required for cell determination.

The ‘enhancer trap’ lines used (see Table 1) were gifts of the laboratory of G. M. Rubin. The transgenic line used to express Rh4-lacZ is p[Rh4.1900 lacZ]11, which is described in Fortini and Rubin (1990). Other mutant markers are described in Lindsley and Grell (1968). Flies were kept on standard food at 25°C.

Four marb alleles were isolated in an extensive screen for viable recessive mutations on the third chromosome with abnormal eye morphology (J. A. F.-V., R. W. Carthew and G. M. R., unpublished data; see Fischer-Vize et al., 1992a). Four marb mutants (marb^CD4, marb^FN1, marb^FQ19, marb^FU2) were recognized on the basis of their weakly rough eyes and irregular reduced corneal pseudopupils, indicative of retinal defects. (The reduced corneal pseudopupil is a red image visible with incident polarized light in white-eyed flies (Francheschini and Kirschfeld, 1971).) 20 additional marb alleles were induced with X-rays as follows. st males were exposed to X-rays (4000 rads) and then crossed to bw; marb^CD4st cu sr e gro^BFP2faf^FO8/TM6B virgins. (gro^BFP2 (Fischer-Vize et al., 1992a) and faf^FO8 (Fischer-Vize et al., 1992b) are two other rough-eye mutations.) Approximately 30,000 bw; *st/marb^CD4st… male progeny were screened for rough eyes. One additional marb allele was isolated in an attempt to generate a P-element-induced allele of marb by hybrid dysgenesis, as follows. Birm2; st males were crossed to Δ(2-3)99B virgins at 16-18°C. (The Birm2 and Δ(2-3)99B chromosomes are described in Robertson et al., 1988). The male progeny (Birm2/+; Δ(2-3)/st) were crossed to bw; marb^CD4st cu sr e gro^BFP2faf^FO8/TM6B virgins at 22°C, and approximately 100,000 *st/ marb^CD4st… male progeny were screened for rough eyes.

marb^CD4 was mapped meiotically to the tip of chromosome 3L in several steps. First, bw; st marb^CD4/th st cu sr e ca virgins were crossed with bw; th st cu sr e ca males. Many male progeny of all genotypic classes were individually mated with bw; st marb^CD4/TM6B virgins to determine whether the recombinant chromosome contained marb^CD4. By this process, marb^CD4 was positioned near the tip of 3L. Next, marb^CD4 was found to be distal to three P-element insertions marked with white+ (P[w+]) in polytene chromosome regions 66C, 64B and 63D (obtained from the laboratory of G. M. Rubin) by crossing w; marb^CD4st/ P[w+] virgins to w; marb^CD4st males and scoring many progeny for the w+ and marb eye phenotypes.

Scanning electron microscopy was performed as described in Kimmel et al. (1990). Unstained adult eyes were fixed, embedded in plastic and sectioned as previously described (Tomlinson and Ready, 1987b). Adult eyes were stained with monoclonal antibodies to β-galactosidase (Promega) or rabbit polyclonal antibodies to Rh1 (a gift of C. Zuker) and prepared for sectioning as follows (U. Gaul, personal communication). Heads were bisected, dipped in 70% ethanol and then fixed 1 hour in 4% paraformaldehyde in PEM buffer (0.1M Pipes pH 7.0, 2 mM MgSO⁴, 1 mM EGTA) on ice. Retinas were dissected out of the cuticle during fixation. Following four washes in PBST (PBS+0.1% Triton X-100), retinas were incubated overnight at 4°C in primary antibody diluted 1:250 in PBSDT (PBST+0.3% deoxycholate), washed 4 times with PBSDT, then incubated for 4 hours at room temperature with HRP-conjugated secondary antibodies (Bio-Rad or Vector) diluted 1:200 in PBSDT. Following four washes in PBST, retinas were incubated for 5 minutes in developing solution (0.5 mg/ml DAB, 0.03% NiCl₂+CoCl₂, 1× PBS), then incubated in developing solution +0.003% H₂O₂ for 10-20 minutes until black. Stained retinas were washed four times in PBS, then dehydrated, embedded and sectioned exactly as the unstained retinas were.

For whole-mount preparations, third instar larval eye discs were dissected, fixed and stained with monoclonal antibody (mAb)22C10 (a gift of S. Benzer), mAb42G11 to the sevenless protein (a gift of M. Simon), or a monoclonal antibody to β-galactosidase (Promega) using PEM+paraformaldehyde fix, PBST washes, biotinylated secondary antibodies, Vectastain (Vector) and DAB,Co+Ni developing as previously described (Fischer-Vize et al., 1992b). Prior to embedding for sectioning, discs were stained with monoclonal antibodies against elav (a gift of G. M. Rubin; Bier et al., 1988), rough (Kimmel et al., 1990), or β-galactosidase (Promega) using the procedure described above, except that discs were left attached to brains, peripodial membranes were not removed and discs were overstained until black (10 –20 minutes). After staining, the discs were treated exactly as the stained retinas were (above) and dissected from other tissues prior to embedding.

Pupal eye discs were stained with cobalt sulfide (Melamed and Trujillo-Cenoz, 1975) as described by Wolff and Ready (1991).

Four EMS-induced mutant alleles of the marb locus with similar phenotypes (marbCD4, marbFN1, marbFQ19 and marbFU2) were isolated in a screen for viable recessive mutants with abnormal eye morphology. Three of the four alleles display no defects outside the eye (see Table 2 and its legend). In order to determine whether the eye phenotype represents the complete loss of function (the null phenotype) of the marb gene, we wanted to look at the phenotype of the four alleles in trans to a deficiency for the locus, but none was available. Thus, 20 independent marb alleles were induced with X-rays and isolated in simple F₁ screens in an effort to produce deletions and chromosomal rearrangements that might completely destroy the marb locus (Table 2 and Materials and methods). One additional marb allele was induced by hybrid dysgenesis (Table 2 and Materials and methods).

Of the 21 additional alleles induced, 13 were viable. The lethality of the other eight alleles is likely due to breakpoints outside the marb locus as all of the lethal alleles are viable in trans to each other. With one exception (marbBX18; see Table 2 and below), all of the viable alleles and all heterozygous combinations of the lethal and viable alleles display eye phenotypes similar to the original four alleles (data not shown). None of the alleles is a chromosomal deficiency for the entire marb locus, but four of the viable alleles (marbBX1, marbBX3, marbBX9, marbBP) contain chromosomal rearrangements with a common breakpoint in 61C4-7 (Table 2 and legend). This location is consistent with meiotic mapping data (Materials and methods) and therefore the four X-ray alleles above are likely to break within the marb gene.

The above data suggest that the eye phenotype displayed by the original marb alleles is likely to represent the null phenotype of the locus. However, additional information leads to the conclusion that this may not be the case. Six alleles (marbBX4, marbBX5, marbBX7, marbBX9, marbBX19, marbBX21) show an additional phenotype when homozygous (if viable) and/or in heterozygous combination with any other allele: notal bristles point in unusual directions (data not shown). The most likely interpretation of this observation is that the other 19 alleles must produce some marb protein that functions at least in the notum. This interpretation leads to two conclusions. First, it remains possible that the six marb alleles described above are null alleles. Second, three of the viable alleles with chromosomal abnormalities (marbBX1, marbBX3, marbBP) are not null alleles. This conclusion is more easily understood in the light of preliminary molecular analysis of the marb locus which indicates that marbBX3 and marbBP both have break-points near the 3′ of end of what appears to be a very large and complex gene (J. A. F.-V. and A. Cadavid, unpublished).

In summary, marb appears to function in at least two places, the notum and the eye. We do not know whether any of our alleles are nulls, but we suspect that many of them are not, at least in the notum. Although 25 alleles have been obtained that show a similar eye phenotype, it is not certain that this is the null phenotype of the marb locus in the eye because none of the alleles are deficient for the entire locus.

All of the experiments presented below were performed with marbCD4. We chose to analyze marbCD4 first, although marbBX18 has a rougher eye, because marbBX18 has a different eye phenotype from all of the other 24 alleles and we were concerned that it may be neomorphic.

The external surface of a wild-type eye reveals a hexagonal array of about 800 facets with a bristle at each of three corners (Fig. 1A,C). The hexagonal shape is formed by a lattice of pigment cells surrounding each facet. marb mutants were identified by the slightly rough appearance of their eyes (Fig. 1B), due to the irregular shape of some facets and the absence or misplacement of some bristles (Fig. 1D) as well as by their irregular reduced corneal pseudopupils, indicative of internal retinal defects (Materials and methods). Internally, wild-type facets contain eight photoreceptor cells (R1-8), each identifiable by its unique position within a trapezoid (Fig. 1E). Projecting from each photoreceptor cell into the center of the ommatidium are light-gathering organelles, called rhab-domeres (Fig. 1E). marb mutant eyes show a normal complement of eight photoreceptor cells occupying their usual positions, however most of their rhabdomeres are malformed (Fig. 1F).

In order to investigate the earliest developmental defects that lead to malformed rhabdomeres in the adult eyes, the progression of facet assembly in developing larval marb eyes was visualized using a number of different antibodies. marb eye discs were first stained with two antibodies that reveal the apical tips of assembling cells in the monolayer and examined in whole-mount preparation. mAb22C10, a monoclonal antibody that recognizes a cytoplasmic neural antigen and thus stains photoreceptor cells (Fujita et al., 1982), reveals the sequence of photoreceptor assembly, as the R-cells stain in the order R8, R2/5, R3/4, R1/6 and R7 (Tomlinson and Ready, 1987a). Facets at all stages of photoreceptor assembly are seen in a single larval eye disc, the most developed facets at the posterior. In marb mutant eye discs, the photoreceptor cell assembly sequence as revealed by mAb22C10 appears normal (Fig. 2A). marb mutant eye discs also showed a wild-type staining pattern with antibodies to the sevenless protein, which is expressed in a dynamic pattern of specific photoreceptor and cone cell subtypes (Fig. 2B; Tomlinson et al., 1987). Thus, both photoreceptors and cone cells appear to be properly determined and assembled in the larval eye disc.

However, staining with mAb22C10 revealed that the morphology of many of the photoreceptor cells in the marb larval eye disc is unusual. In wild-type discs, the photoreceptor cell nuclei occupy characteristic positions at particular stages of ommatidial assembly (Tomlinson, 1985). As cells join facets, their nuclei migrate from basal positions in the disc to the apical surface where the contacts that induce cell determination are initiated. Subsequently, nuclei and cell bodies settle into characteristic positions as the facets mature (Tomlinson, 1985; Tomlinson and Ready, 1987a). The positions of the photoreceptor cell nuclei beneath the apical surface can be seen in mAb22C10-stained whole-mount eye discs as cytoplasmic stain surrounds them (Tomlinson and Ready, 1987a). In marb discs, many of the nuclei appear to absent from their normal positions (data not shown).

Photoreceptor cell nuclei were visualized directly using an enhancer trap line called BG9408 that expresses the E. coli lacZ gene in all photoreceptor cell nuclei (Table 1). In whole-mount preparations of BG9408 larval eye discs stained with an antibody to β-galactosidase, clusters of photoreceptor cell nuclei in rows of developing facets are seen in characteristic apical positions (Fig. 2D). In contrast, in a marb mutant back-ground, there are very few stained apical nuclei and a disordered array of stained nuclei, resembling a mass of ‘marbles’, is seen at the basal surface of the eye disc (Fig. 2E).

In order to examine the later stages of development of the cone cells and the assembly of the hexagonal pigment and bristle cell lattice, marb pupal eye discs were stained with cobalt sulfide, which reveals the apical outlines of cells (Fig. 2C). The cone and primary pigment cells appeared to develop normally. In addition, most of the secondary and tertiary pigment cells and bristle precursor cells that form the hexagonal lattice appeared to assemble normally when viewed apically. However, occasional patterning defects in the lattice were observed (Fig. 2C and legend).

In summary, the facet assembly sequence appears to occur normally in marb mutant eye discs, with the exception of occasional patterning defects in the cells of the hexagonal lattice surrounding the photoreceptors. However, a majority of the photoreceptor cell nuclei appear to reside basally instead of occupying their normal apical positions.

In order to demonstrate more clearly that the nuclei of most determined cells are basal in marb larval eye discs, and to determine the identities of these cells, wild-type and marb discs were stained with antibodies that label the nuclei of specific cell types, and then sectioned longitudinally so that the apical and basal planes are visible in one section (Fig. 3, Materials and methods). Enhancer trap lines that express lacZ in subsets of photoreceptor and cone cells (Table 1, 3) were crossed into a marb background and discs stained with anti-bodies to β-galactosidase. In addition, antibodies to elav and rough, two proteins expressed in specific cell types in larval eye discs, were used (Table 3).

Stained nuclei corresponding to determined photoreceptor or cone cells always appeared apical in wild-type discs (Fig. 3). In contrast, in marb discs, all stained nuclei were basal in those discs stained with markers specific for subsets of R8,2,5,3,4 and 7, only apical stained nuclei are seen with markers expressed in subsets of R1,6 and cone cells, and stained nuclei were observed both apically and basally with markers specific for cell types belonging to both of the above groups (Fig. 3, Table 3). We conclude that in marb larval eye discs only the nuclei of the R1,6 pair and the cone cell nuclei migrate apically as they join developing facets. In addition, the basal nuclei in marb discs appear to have fallen into the photoreceptor cell axons, beneath the fenestrated membrane that supports the photoreceptor cells (Fig. 3). Finally, although the abnormal positions of the nuclei make it impossible to identify their individual cell types on the basis of their positions in marb developing facets, we think it likely that the markers retain their cell type specificity in the mutant discs because a similar number of stained nuclei appear to be present in wild-type and marb eye discs. This observation is consistent with the idea that cell determination is unaffected in marb mutants.

In order to determine the locations of the nuclei of the various cell types in marb adult eyes, three different enhancer trap lines were used, each of which expresses lacZ specifically in the nuclei of R1-6, R7 or pigment cells (Table 1 and Fig. 4A,D,G). Adult eyes of wild-type and marb flies containing each enhancer trap were stained with anti-β-galactosidase anti-bodies and sectioned horizontally so that the apical and basal surfaces of the eye were both visible. All of the photoreceptor cell nuclei, except that of R8, and all of the pigment cell nuclei, are normally apical (Fig. 4B,E,H,J). In contrast, in marb eyes, most of the R1-6 cell nuclei are basal (Fig. 4C) as are most of the pigment cell nuclei (Fig. 4I). As in marb larval discs, the basal photoreceptor cell nuclei appear to be located in the axons of the adult eyes (Fig. 4C). This is most apparent by comparing the positions of the basal photoreceptor nuclei (Fig. 4C) and those of the pigment cells (Fig. 4I), which have no axons. Unexpectedly, the R7 nucleus although it was basal in the larval eye disc, appears to be apical in the marb adult eye (Fig. 4F). Tangential sections of anti-β-galactosidase-stained marb eyes, in which it is possible to identify each photoreceptor cell, were examined in order to investigate the possibility that the cell type specificity of the R7-specific enhancer trap changed in marb mutants. However, only R7 cell nuclei were stained (data not shown).

Finally, we wanted to determine whether the malformed rhabdomeres characteristic of marb eyes correlate with the unusual basal position of the photoreceptor cell nuclei. marb mutant eyes that express lacZ in the nuclei of R1-8 were generated (Materials and methods), and apical tangential sections of anti-β-galactosidase-stained eyes were examined. Many facets in marb eyes contain a few normally shaped rhab-domeres and many oddly shaped ones (Fig. 1F). As shown in Fig. 5, stained apical nuclei were observed only in those photoreceptor cells with normally shaped rhabdomeres. Thus, the photoreceptor cells with basal nuclei have malformed rhab-domeres, and those with apical nuclei have normal rhab-domeres. In addition, the nuclei of R1,6 and then R3,4 were apical most frequently.

Using subtype-specific markers, we have shown that the photoreceptor cells in developing marb eyes appear to be properly specified. In order to determine whether the photoreceptors are properly differentiated, we examined the expression of cell-type-specific rhodopsins in marb adult eyes. Outer photoreceptor cells (R1-6) express Rh1, and R7 cells express either Rh3 or Rh4 in their rhabdomeres (Mismer and Rubin, 1987; Pollack and Benzer, 1988; Fortini and Rubin, 1990). First, wild-type and marb adult eyes stained with an antibody to Rh1 were sectioned tangentially so that individual R-cells could be identified. Only the rhabdomeres of the outer R-cells stained in both wild-type and marb eyes (Fig. 6A,B). Second, a transgenic line carrying an Rh4-lacZ hybrid gene which expresses lacZ in the cytoplasm of R7 cells was used to examine R7 cell differentiation. In both wild-type and marb backgrounds, only the R7 cell cytoplasm stained with an anti-β-galactosidase antibody. Thus by the criterion of rhodopsin expression, the photoreceptor cells in marb adult eyes appear to be properly differentiated.

We have identified a Drosophila gene, marb, which is required for many of the apical nuclear migrations that accompany pattern formation in the developing eye. Remarkably, photoreceptor and cone cell assembly, determination and differentiation appear to proceed normally despite the failure of many nuclei to rise. Therefore, although the two processes are normally coordinated temporally, nuclear migration is not necessary for cell determination.

In light of the observation that cells in the developing eye initiate their contacts apically (Tomlinson and Ready, 1987b; Cagan and Ready, 1989), it is particularly surprising that cell determination can occur despite the severe malformations in the apical shape of marb mutant cells. The importance of apical contacts is particularly well-illustrated by the interaction between the sevenless tyrosine kinase receptor in R7 and its ligand, the bride-of-sevenless protein, expressed in R8 (Banerjee et al., 1987; Tomlinson et al., 1987; Kramer et al., 1991). The interaction between sevenless and bride-of sevenless, which is responsible for the recruitment of R7 into the facet, occurs apically in R7 and R8, and we have shown that sevenless is apically localized in marb eye discs. Apparently, the apical cell contacts between R7 and R8, and presumably the contacts between other cells in marb eyes, must be sufficiently normal for cell interactions to occur.

Although the photoreceptor and cone cells assemble normally in marb eye discs, formation of the hexagonal lattice consisting of secondary and tertiary pigment and bristle cells in pupal discs is sometimes disrupted, which accounts for the rough eyes of marb adults. It is thought that the determination of secondary and tertiary pigment cells relies on their contacts with primary pigment cells (Cagan and Ready, 1989). Why do the secondary and tertiary pigment cells sometimes develop abnormally in marb pupal discs when the primaries are normally positioned? The nuclear positions of the pigment cells and bristle precursors in the pupal eye disc are not likely to be directly responsible for the defects in the lattice. First, the primary pigment cells lose their basal footing, so their nuclei must be more-or-less apical. Second, secondary and tertiary pigment cell nuclei normally remain basal as they sort into the lattice and rise later. The bristle neuron nucleus also remains basal (Cagan and Ready, 1989). Thus, although our data suggest that the later apical migration of secondary and tertiary pigment cell nuclei does not occur normally in marb eyes, this should not affect their earlier determination. The secondary and tertiary pigment cells normally contact the R-cells basally. Perhaps these contacts are also required for proper assembly of the pigment cell lattice and are disrupted in marb eye discs.

Apparently, cells in the developing eye use diverse and dynamic mechanisms for nuclear movement. Cells entering the morphogenetic furrow move their nuclei using an apparently marb-independent mechanism. In addition, R1,6 and the cone cells do not appear to require marb for apical localization of the nucleus in larval eye discs, although the other R-cells do. Moreover, it appears that the R7 nucleus and to some extent the other R-cell nuclei use a mechanism that does not involve marb for apical nuclear movement late in development, as R7 and sometimes other R-cells that were basal in the larval disc, most often the R3/4 pair, became apical in the adult eye. As all of these observations were made using a marb allele which may not be null in the eye, it is possible that marb is actually required for all of the nuclear movements in the eye, but to differing extents.

Our results do not rule out a general requirement for the marb protein in Drosophila development. Indeed, some of the mutant alleles reveal that notal bristle formation requires marb. All of the marb alleles were isolated as viable mutations, either when homozygous or in trans to marbCD4. Thus, we may have isolated a special class of marb mutants that affect only the eye and notal bristles.

Although microtubule and actin-based motor proteins (e.g. myosins and kinesins) are known to be involved in organelle transport (Vale and Goldstein, 1990), no specific protein has been implicated in the mechanism of nuclear movement in postmitotic cells. Preliminary molecular analysis suggests that marb is likely to encode some sort of cytoskeletal protein (J. A. F.-V. and A. Cadavid, unpublished). One exciting possibil- ity is that marb is a motor protein or a protein that interacts with one. An alternate hypothesis is suggested by the obser- vation that the nuclei of marb photoreceptors appear to fall into the axons. The role of marb in nuclear migration in the eye may be similar to that of the chickadee and singed proteins in cytoplasmic streaming in the ovary. chickadee encodes profilin and singed encodes fascin, both of which are required for the proper formation of cytoplasmic actin networks that hold nurse cell nuclei in place during the transfer of nurse cell cytoplasm to the oocyte (Cooley et al., 1992; Cant et al., 1994). In chickadee or singed mutant ovaries, the flow of cytoplasm is blocked by the nurse cell nuclei, which become lodged in the ring canals connecting the nurse cells and oocyte. Thus, the role of marb may be to shape the cell body, thereby prevent- ing the cell body cytoplasm, and the nucleus along with it, from flowing into the axons of the photoreceptors or into the basal regions of the pigment cells. If so, it is remarkable that as cells begin to differentiate behind the morphogenetic furrow and their nuclei and cell bodies rise, the mechanism for retaining cell shape and thus apical nuclear migration abruptly begins to require marb. The identification of marb mutants, which are viable and have an easily recognizable phenotype, opens an avenue for convenient genetic analysis of nuclear migration and cell shape changes during morphogenesis.

We are indebted to Todd Laverty for expert cytological analysis. Matthew Freeman, Marek Mlodzik, Zhi-Chun Lai and Gerald Rubin generously provided published and unpublished enhancer trap lines. We thank Richard Carthew for isolating marb^CD4, Bruce Kimmel for performing SEMs, Mike Simon for the negative for Fig. 1A, and Seymour Benzer,Ulrike Heberlein, Gerald Rubin, Mike Simon and Charles Zuker for antibodies. We are grateful to two anonymous reivewers of this and an earlier version of the manuscript for their helpful comments. We also thank Brian Haarer and Peter Vize for their helpful criticisms of the manuscript. The initial isolation and partial characterization of the marb mutation was carried out while J. A. F.-V. was a Helen Hay Whitney postdoctoral fellow in the laboratory of G. M. Rubin and she is grateful for their support and encour- agement. This work was supported by grant number R29HD30680 from the National Institutes of Health to J. A. F.-V.