Abstract
The Drosophila gene giant (gt) is a segmentation gene that affects anterior head structures and abdominal segments A5–A7. Immunolocalization of the gt product shows that it is a nuclear protein whose expression is initially activated in an anterior and a posterior domain. Activation of the anterior domain is dependent on the maternal bicoid gradient while activation of the posterior domain requires maternal nanos gene product. Initial expression is not abolished by mutations in any of the zygotic gap genes. By cellular blastoderm, the initial pattern of expression has evolved into one posterior and three anterior stripes of expression. The evolution, position and width of these stripes are dependent on interactions between gt and the other gap genes. In turn, gt activity in these domains affects the expression of the other gap genes. These interactions, typical of the crossregulation previously observed among gap genes, confirm that gt is a member of the gap gene class whose function is necessary to establish the overall pattern of gap gene expression. After cellular blastoderm, gt protein continues to be expressed in the head region in parts of the maxillary and mandibular segments as well as in the labrum. Expression is never detected in the labial or thoracic segment primordia but persists in certain head structures, including the ring gland, until the end of embryonic development.
Introduction
The development of the anteroposterior pattern of the Drosophila embryo takes place in stages, progressing from simpler, block-like domains to units spanning double segments and then to individual segments. This strategy is revealed by the existence of zygotic genes whose localized expression is required for the differentiation of the corresponding pattern elements (Nüsslein-Volhard and Wieschaus, 1980). The first stage in the formation of the anteroposterior pattern of the embryo is controlled by the zygotic gap genes. These are expressed in broad, partially overlapping domains and, in combination, are responsible for activating the expression of the next genes in the segmentation hierarchy, the pair-rule genes. Lack of gap gene activity causes the loss of several contiguous segments from that part of the body plan that corresponds roughly to the domain of expression of the gene.
The initial domains of expression of the segmentation gap genes are specified by maternal cues laid down in the egg. The maternal cues are produced by three organizing systems responsible for the anterior, postenor and terminal pattern elements, respectively (Nüsslein-Volhard et al. 1987). The anterior system, acting through a gradient of the bicoid (bcd) gene product, has been shown to activate at least one zygotic gap gene, hunchback (hb), in the part of the embryo in which bcd concentration is above a certain threshold (Driever et al. 1989). Genetic and molecular evidence (Hülskamp el al. 1989; Irish et al. 1989; Struhl, 1989) indicates that the posterior system, through the nanos (nos) product, acts negatively by preventing the accumulation of maternal hb product in a graded way from the posterior of the embryo. This activates posterior development because hb acts negatively on other gap genes. The terminal system, acting through the torso and l(1)polehole genes, activates the zygotic gap genes tailless (tll) and huckebein (hkb) in the terminal regions (Jürgens et al. 1984; Strecker et al. 1986; Pignoni et al. 1990; Weigel et al. 1990). By cellular blastoderm the initial domains of expression of the gap genes become sharper and their pattern is refined. Mutual interactions among the gap genes account for much of this elaboration of the pattern (Jackie et al. 1986).
The giant (gt) gene is required for normal development of the head and of the abdominal segments. Strong gt loss-of-function alleles or deficiencies of the gene were classified as segmentation gap mutations because they cause defects in the segmental pattern of two regions. One, in the posterior half of the larva, is a loss of contiguous abdominal segments A5, 6, 7 and sometimes 8, a type of pattern deletion characteristic of segmentation gap mutations. The other region is the head where defects have been interpreted as the loss of labral and labial structures resulting in failure to complete head involution and loss of parts of the cephalopharyngeal skeleton (Petschek et al. 1987; Mohler et al. 1989) and fusion of the labial and prothoracic segments (Petschek and Mahowald, 1990). In contrast to a classical gap mutation, lack of gt function causes less extensive gaps in the segmentation pattern and does not completely delete the abdominal, segments that it affects but only their anterior parts, leaving a considerable expanse of naked cuticle (Petschek et al. 1987; Mohler et al. 1989; Petschek and Mahowald, 1990). Earlier studies also showed that gt mutations had only minor effects on the pattern of expression of the pair-rule gene fushi tarazu (Carroll and Scott, 1985). More recent reports show that gt mutations cause alterations in the expression of even skipped (Frasch and Levine, 1987), a drastic weakening of stripes 5 and 6 of ftz expression and the concomitant fusion of engrailed stripes 10-13 (Petschek and Mahowald, 1990). These results together with the pattern of defects, the early requirement for gt activity and the early onset of expression all indicate that gt functions very much like a gap gene. In this work we have studied the interactions of gt with the three maternal systems that organize the anteroposterior pattern and with the gap genes. We find that the initial expression of gt is dependent only on maternal factors but that gt affects the expression of other gap genes and is in turn influenced by them. Our results show that gt acts like a typical gap gene and is required for the establishment of the anteroposterior pattern of the embryo.
Materials and methods
Preparation of giant antibody
We inserted a 522 bp SmaI –PvuII fragment that encoded amino acids 263 –436 of the cDNA open reading frame with no opa sequences (M. Capovilla, E. Eldon and V. Pirrotta, in preparation) into the pEX-1 expression vector of Stanley and Luzio (1984). High level expression of the giont-β-galactosi-dase fusion protein was induced in NF-1 cells by growth at 42 °C. Pelleted cells were lysed in 7 M guanidinium HCl and diluted with 6 volumes of Tris –EDTA, pH 8. Insoluble material was pelleted and resuspended in urea sample buffer (USB) containing 8M urea, 100mM Tris (pH7.6), 2% SDS and 5% beta mercaptoethanol, sonicated and applied to 2 mm preparative 7.5% polyacrylamide gels. The region of . the gel containing the hybrid protein was excised, homogenized and injected into rabbits as described by Benson and Pirrotta (1987). Serum obtained from the rabbits was purified first through CM Affi-Gel Blue (BioRad), to isolate an enriched immunoglobulin fraction. This was then precipitated with ammonium sulfate, resuspended and applied to affinity columns essentially as described by Pirrotta et al. (1988). The final product was greatly enriched for anti-giant activity and retained only very low levels of anti-beta galactosidase activity. Antibodies were used at a concentration of 1-2 μgml-1 when used with horseradish peroxidase (HRP)-conjugated secondary antibodies (Cappel) or 0.1 –0.6 μgmU1 when used with the Vectastain Elite kit (Vector Laboratories).
Other antisera
Antisera directed against other segmentation gene products were kindly provided by other laboratories. In all cases they were preabsorbed by incubation with an excess of wild-type embryos prior to being used at the recommended dilutions. Rabbit anti-ftz antibodies (from H. Krause; Krause et al. 1988) were used at a 1/1000 dilution. Mouse anti-hb and rabbit anti-Kr antibodies (from R. Kraut and M. Levine) were used at a 1/300 and 1/100 dilution, respectively. Rabbit anti-en antibodies (from T. Kornberg; DiNardo et al. 1985) were used at a 1/400-1/1000 dilution. Rat anti-kni antibodies (obtained from K. Howard) were used at a 1/500-1/1000 dilution. Biotinylated secondary antibodies were obtained from Vector Laboratories (anti-mouse IgG and anti-rabbit IgG) or Jackson Immuno Research Laboratories (anti-rat IgG). These antibodies were also preabsorbed at a 1/10 dilution with an excess of Canton S embryos and used at a final dilution of 1/500.
Embryo fixation and staining
Flies of the appropriate genotype were allowed to lay eggs on grape juice agar plates for specified periods of time. The embryos were aged, if necessary, at room temperature (approx. 22°C), then collected, dechorionated, fixed and devitellinized essentially as described by Mitchison and Sedat (1983) as modified by Karr and Alberts (1986) and DiNardo and O’Farrell (1987). Embryos to be stained with peroxidase-conjugated antibodies were pretreated with 0.3 % H2O2 prior to rehydration in phosphate buffered saline (PBS, pH7.4) containing 0.1 % Tween-20 and 0.1 % bovine serum albumin (BBT). Blocking was done in BBT containing 2% serum (BBS). Primary antibody incubations were carried out in BBS at 4°C overnight. The embryos were washed extensively in BBT, reblocked in BBS and incubated with the appropriate preabsorbed secondary antibody for two hours at room temperature. They were then washed extensively in PBS containing 0.1 % Tween-20 (PBT) and incubated for one hour at room temperature with ABC reagent (Elite Kit, Vector Laboratories) at concentrations recommended by the manufacturer. Following extensive washes in PBT, the embryos were treated with the appropriate substrate. For horseradish peroxidase staining, we used diaminobenzidine (0.5 mg ml-1) in the presence of 0.001-0.003% hydrogen peroxide. In the exu embryo shown in Fig. 4B, 2 fit of a solution of 2 % nickel ammonium sulfate, 2% cobalt chloride was added to 1ml of substrate solution (DeBlas and Cherwinski, 1983). For alkaline phosphatase staining, we used 0.34 mg ml-1 nitroblue tetrazohum salt and 0.175 mg ml-1 of 5-bromo-4-chloro-3-indolyl phosphate, toluidinium salt, in buffer containing lOOmM NaCl, 50mM MgCl2, lOOmM Tris, pH9.5, 0.1% Tween-20, 1 mM Levamisol. After staining, the embryos were washed in PBT, dehydrated, cleared briefly in methyl salicylate and mounted in GMM (Lawrence et al. 1986).
Mutant strains
Wild type embryos were obtained from Canton S parents. The giant alleles gtx11, gtYA82 and Df(1)62gl8 were balanced over an FM7 chromosome carrying a P-element insertion expressing β-galactosidase under the control of the ftz promoter (Kania et al. 1990). We used embryos from stocks carrying the following mutations: Krüppel, Kr1 (Gloor, 1950); knirps, kniIIE72 (Tearle and Nüsslein-Volhard, 1987); hunchback, hb6tul (Lehmann and Nusslein-Volhard, 1987); bicoid, bcdE1 (Frohnhôfer and Nüsslein-Volhard, 1986); exuperantia, exuQV (Schüpbach and Wieschaus, 1986); nanos, nos1-7 (Nüsslein-Volhard et al. 1987); oskar, osk166 (Lehmann and Nüsslein-Volhard, 1986); torso, tor4021 (Klingler et al. 1988) and tor17 7111 (Schiipback and Wieschaus, 1986); tailless, tlP (Pignoni et al. 1990); empty spiracles, Df(3R)red3L (Dalton et al. 1989); buttonhead, btdc’ (Jürgens et al. 1984); orthodenticle, Df(l)KA14 (Wieschaus et al. 1984); tailless-huckebein, tlft hkb2 (Weigel et al. 1990).
Data analysis
Embryos were examined on a Zeiss Axiophot with differential interference contrast illumination. Measurements of the domains of protein expression were made midlaterally using a Zeiss E7 eyepiece micrometer and converted to egg length percentages. Ages of blastoderm embryos were estimated by measuring the length of the nuclei and invaginating cell membranes at the cell periphery as a percent of egg diameter at the broadest part of the embryo (at approximately 50 % EL). After the onset of gastrulation, morphogenic criteria were used to stage embryos according to Campos-Ortega and Hartenstein (1985). Because the gt pattern of expression changes quite rapidly, care was taken to compare only those embryos at the same developmental stage. For the mutant studies, this was the brief period between completion of cellularization (when the cellular layer was at least 10 % of the diameter of the embryo) and the earliest signs of gastrulation. At least 10 embryos of the appropriate stage and orientation were measured for embryos from maternal effect collections. At least 25 embryos of the appropriate stage and orientation were measured from zygotic lethal collections. Mean and standard deviations were calculated for the position of each stripe boundary. In most cases, determining which embryos were mutant was straightforward: the pattern of protein expression was obviously altered in the appropriate percentage of embryos. In a few cases of zygotic mutants, however, changes in the pattern were subtle. In these cases generally a standard deviation at one or more domain boundaries was greater than 2%, suggesting a large degree of scatter in the data points. The values were then plotted to see if two peaks could be resolved, consistent with a reproducible shift in approximately 25 % of the embryos. As a wild-type standard, we took the pattern obtained with Canton-S embryos but, in measuring the zygotic mutants, the giant domains of wild-type and heterozygous embryos were measured as well and compared to the domains in the 25 % of embryos that were mutant. Shifts of less than 2% EL were considered insignificant. To confirm the identification of mutant embryos, whenever possible we used double-stained embryos with a second antibody against ftz or en protein to reveal changes in the embryonic fate map corresponding to the mutant phenotype.
Results
The gt RNA is first detected by in situ hybridization during nuclear cycle (NC) 12 (Mohler et al. 1989). Our antiserum begins to detect gt protein at the end of NC 12, localized in nuclei in two distinct domains that parallel the pattern observed by in situ hybridization: a broad stripe at 62 – 80 % EL and a posterior domain at 2 – 33% EL, excluding the pole cells (Fig. 1A). Expression is weak initially but gradually increases in strength through NC 13 and the pattern begins to sharpen. Embryos carrying strong giant mutations such as gri11 or giYA82 still express the protein and show a very similar evolution of the pattern as the wild-type but the intensity of the antibody staining does not increase, suggesting that functional gt protein may stimulate its own expression. Embryos hemizygous for Df(1)62g18, a deletion that removes the gt gene entirely, show no staining at all under our conditions.
By the middle of NC 13, the posterior domain has retracted and sharpened forming a stripe whose posterior boundary is at 15 – 20% EL (Fig. 1B). During NC 14 the pattern changes rapidly. As membranes progress inward from the cortex to form cells, the anterior domain begins to resolve into two stripes, a stronger one at 63 – 73 % and a somewhat weaker one at 73-84% EL (Fig. 1C). By the time cellularization is complete, a new band of expression appears anteriorly at position 88-95 % EL (Fig. 1D). This domain, stripe 1, extends only laterally and dorsolaterally. Meanwhile, the separation of stripes 2 and 3 has progressed so that they now occupy positions 74 – 81 % and 62 – 70 % EL, respectively. The posterior stripe gradually shifts forward, reaching a final position at 24 – 35 % by the end of cellularization. The complete pattern of expression at cellular blastoderm consists then of four stripes.
The pattern of expression in post-blastoderm development is complex, with some regions becoming strong and others fading throughout germ band extension. We cannot determine at this point whether all the cells expressing giant at later stages are descendants of the cells that constitute stripes 1 – 3 at blastoderm. When we speak of elongation or changes in the three head stripes of expression, we refer only to patterns that show some local continuity and apparent evolution through time. The rapid disappearance of giant expression from certain regions indicates that the protein as well as the mRNA must be rapidly turning over. In fact the pattern is continuously evolving throughout the process of cellularization. The posterior band, stripe 4, begins to fade towards the end of NC14. The ventral part of stripe 2 also fades at this time and is replaced by a distinct domain of expression that develops as a short ventral stripe slightly more anterior and more intense than the part of stripe 2 that faded away. As gastrulation begins, the anterior end of the ventral furrow reaches the ventral stripe and the giant-expressing cells line the interior and the edges of the resulting anterior midgut invagination (Fig. 1E,F). These cells become rapidly internalized forming a mass that stains most intensely bilaterally (Fig. 2A). These cells are not part of the anterior midgut primordia, which also differentiate at this time in the same region and are displaced posteriorly by the stomodeal invagination. Instead, the gt-expressing cells appear to move laterally and dorsally during germ band extension, forming an arch that meets anteriorly and medially with a group of strongly expressing cells inside the clypeolabrum (Fig. 2B,C).
Initially, stripe 3 straddles the cephalic furrow, staining cells on both sides of it and with its posterior border overlapping with the thin maxillary stripe of engrailed-expressing cells (Fig. 2A). At no time from now on is there appreciable giant expression more posterior than this engrailed stripe, hence no expression in the primordia of the labial lobe. As the germ band begins to extend, stripe 3 decreases in intensity dorsally and laterally while it expands ventrally as cells enter the deepening cephalic furrow. Expression is seen internally anterior of the furrow in a region between the maxillary stripe of engrailed and the anterior midgut invagination (Fig. 2A). On the ventral side, the cells of stripe 3 fold into the ventral furrow and give rise to a thin layer of strongly expressing cells that extends to the posterior border of the maxillary segment, directly under the ectodermal layer (Fig. 2D). Most of these cells will gradually move forward as the stomodeal invagination deepens and will cease expressing gt as they fold into the stomodeum (Fig. 2D,E,F). Intriguing, but difficult to explain, is the appearance of staining in the germ cell precursors that are found in the amnio-proctodeal invagination (arrowhead in Fig. 2B). This staining, observed in many embryos during germ band elongation and again during germ band retraction (Fig. 3A), is not detectable in the pole cells at earlier stages. Whether it is a transient stage of expression or an odd artefact, we cannot tell.
The staining region that corresponds to stripe 2 continues fading on the sides of the embryo but remains strong on the dorsal side where it moves posteriorly towards and then into the cephalic furrow (Fig. 2B,C). Strong expression persists in a group of dorsolateral cells while the rest fade away as the germ band reaches its maximal extension. The anteriormost stripe in the labral region, stripe 1, broadens laterally and advances anteriorly and then ventrally as the clypeolabrum begins to form and fold ventrally into the stomodeal invagination. A group of cells at the tip and inside the infolding clypeum express strongly and continue to do so until the beginning of germ band retraction. The intensity and extent of gt expression in the head peaks shortly before the germ band reaches its maximum extent. The number of cells expressing gt continues to decrease during germ band retraction (Fig. 3A) but some staining continues to be detectable even after the cuticle is secreted, nearly up to the end of embryonic development. It is interesting to note that among the last cells in which expression persists is a cluster located dorsally just behind the dorsal sac and the pharyngeal musculature but above the brain and corresponding in shape and position to the ring gland (Fig. 3B). Petschek et al. (1987) have observed that in strong gt mutants the ring gland appears to be missing or defective and have speculated that very mild defects in the ring gland of the viable gt1 allele might be responsible for the insufficiency of ecdysteroid that is responsible for the eponymous giant larva phenotype in this mutant.
Maternal effects
Three classes of maternally acting genes have been identified, whose function during oogenesis is to lay down the cues that result in the establishment of the anteroposterior axis of the embryo. Their domains of activity are independently established, largely nonoverlapping and affect a) the anterior segments, b) the posterior segments and c) the terminal structures at both ends of the embryo (Nüsslein-Volhard et al. 1987). giant expression is found in all three of these domains. It is not surprising therefore, to find that loss of activity of any of these three pattern-forming systems has profound effects on the distribution of giant protein in the early embryo.
The anterior domain
The key maternal anterior gene, bicoid (bcd), is distributed in the early embryo in a concentration gradient with a high point near the anterior tip (Driever and Nüsslein-Volhard, 1988a,b). The bcd gradient can activate the hb gene only in the anterior part of the embryo, where bcd concentration is high enough to bind to its response elements in the hb regulatory region (Driever et al. 1989; Struhl, 1989). By manipulating the number and affinity of the bicoid binding sites, the domain of activation of the hb gene can be shifted more anteriorly or more posteriorly, demonstrating how this maternal morphogen can generate a pattern of zygotic gene expression. The activation of hb by bcd is insufficient to account for the complexity of the anterior gt pattern since the specification of head structures requires bcd levels higher than those required to activate hb expression and mutations in hb do not account for the totality of the phenotypes produced by lack of maternal bcd. Driever et al. (1989) hypothesized a gene X that would be activated by higher concentrations of bicoid and might therefore serve to specify the more anterior head segments. Our results show that giant is such a gene X, whose anterior domain of expression is most likely directly controlled by bcd.
In embryos produced by bcd mothers, the anterior pattern of giant expression is never observed. Stripes 1 – 3 fail to appear and stripe 4 develops as a broader domain shifted to a more anterior position (28-47% EL instead of 24 – 35 % EL; Fig. 4A). This indicates that the anterior expression of giant requires bcd activity either directly or indirectly but that stripe 4 is established independently of bcd. The anterior activation of giant might be mediated by some other zygotic gene dependent on bicoid, either hb or some other bcddependent gene, but several lines of reasoning suggest that the bcd controls gt directly: (1) gt anterior expression begins very early, before appreciable expression of other zygotic genes; (2) it is independent of hb, the major known mediator of bcd’, (3) it is not abolished by mutations in the other maternal factors or in the other known zygotic genes.
Furthermore, the anterior pattern of gt expression depends on the gradient of bcd concentration. This is shown by the distribution of gt expression in embryos derived from mothers mutant for exuperantia (exu), a gene required for the proper localization of bcd mRNA in the oocyte. Phenotypically such mutant embryos have greatly reduced anterior structures and expanded gnathal and thoracic structures (Schüpbach and Wieschaus, 1986). In exu eggs, the bcd RNA fails to be preferentially localized anteriorly and as Driever and Nüsslein-Volhard (1988b) have shown, the bcd protein gradient in these embryos is much shallower than normal, with the highest concentration at the anterior end corresponding to that normally found at about 65%. In exu embryos, gt expression at cellular blastoderm is found in a broad domain from 70-92 % EL anteriorly and 15-30% EL posteriorly (Fig. 4B). We interpret this as a broadened and anteriorly shifted stripe 3, in agreement with the general anterior shift of thoracic and gnathal structures. We suppose that bcd concentrations are insufficient to specify stripes 1 and 2 but that levels adequate to activate stripe 3 are now found over a good part of the head region. The fact that the domain of giant expression does not extend to the anterior tip of the embryo suggests that factors that prevent giant expression are present in the terminal region. The posterior shift of stripe 4 is probably an indirect effect mediated through other gap genes (see below). These results argue in favor of a direct dependence on the bcd concentration at least for the initial gt anterior domain. The separation of stripe 2 from stripe 3 and the appearance of stripe 1 are probably due to interactions with zygotic gene products.
The posterior domain
Embryos derived from mothers mutant for genes of the posterior group fail to differentiate abdominal structures (Nüsslein-Volhard et al. 1987). Unlike the anterior and terminal systems, the posterior system controls the embryonic pattern by negative regulation. It has been clearly established that nanos (nos), the key gene of the posterior class, acts solely by preventing the expression of the maternal hb protein in the posterior half of the embryo and is not required in embryos lacking maternal hb product (Hülskamp et al. 1989; Irish et al. 1989; Struhl, 1989). The maternal hb protein is uniformly distributed in a nos mutant but forms a decreasing gradient in the posterior half of nos+ embryos. Similarly, the distribution of maternal hb remains uniformly high along the entire anteroposterior axis of embryos derived from mothers mutant for oskar (osk) (Tautz, 1988), another member of the posterior group of genes which is required for correct positioning of nos product in the embryo. These results show conclusively that it is the abnormal persistence of the maternal hb product that suppresses abdominal development in these mutants.
In embryos derived from mothers mutant for nos or osk, the posterior stripe of gt fails to appear while the anterior pattern appears to be normal (Fig. 4C). This suggests that maternal hb product inhibits the expression of gt in the posterior domain but has no effect on the anterior expression. However, we have not generated hb- germ line clones to demonstrate this directly. This distinction between the anterior gt expression, which is bcd dependent and not repressed by maternal hb, and posterior expression, which is independent of bcd and repressed by hb, strongly suggests that they are controlled by two separate regulatory elements in the gt gene that act independently and respond to different regulatory proteins.
The terminal domains
The nonsegmental termini, anteriorly the acron and posteriorly the telson, require the action of the maternal genes of the terminal class. These are thought to act through a signalling pathway that includes genes such as torsolike, trunk etc. and results in the activation of a membrane-bound tyrosine kinase, the product of the torso gene, and of the D-raf l(1)polehole ser/thr kinase (Sprenger et al. 1989; Ambrosio et al. 1989; Stevens et al. 1990). In contrast to bicoid, these gene products are not localized in the embryo but are instead activated only at the poles by a localized signal and in turn activate zygotic genes such as tailless and hucke-bein in the terminal domains (Klingler et al. 1988; Strecker et al. 1989a; Weigel et al. 1990). We studied gt expression in embryos produced by mothers mutant for two members of the terminal class, torso (tor) and torsolike (tsl). As expected from the signal transduction model, their effects on gt expression are virtually identical. The posterior domain of expression appears normally at early syncytial blastoderm but fails to mature and to withdraw from the posterior region. At cellular blastoderm it still extends up to, but not including, the pole cells. The anterior domain of expression is also affected. Stripe 1 never forms. Stripe 2 is shifted anteriorly by 4 – 7 % EL but stripe 3 remains in its normal position (Fig. 4D). This indicates that the terminal genes have contrasting effects on giant: a repressive effect posteriorly and an inductive effect anteriorly. The fact that both of these effects take place after the initial stages of giant expression suggests that they are mediated by one or more zygotic effectors.
A confirmation of these conclusions was given by the pattern of gt expression in embryos produced by mothers carrying a dominant allele of torso, tor4021 (not shown). This mutation is thought to result in a constitutively active form of the torso product which would therefore activate terminal-specific zygotic gene functions throughout the embryo (Klingler et al. 1988; Strecker et al. 1989a). The phenotypic result is the expansion of the terminal, nonsegmental domains of the embryo at the expense of the segmental primordia. The distribution of giant product in these embryos is revealing: the posterior domain is completely absent and only two stripes are seen anteriorly. We interpret these results as showing that stripes 3 and 4, which normally correspond to segmented regions of the blastoderm fate map, are suppressed by the ectopic expression of terminal genes. Two anterior stripes remain. We suppose that these are stripes 1 and 2, which normally correspond to the anterior nonsegmented portion of the map and are shifted more posteriorly than normal by the expansion of the terminal domains. In approximately 10% of the embryos, the second stripe (63 – 73 % EL) succeeds in resolving into two stripes (69 – 77% and 59 – 64% EL). This incomplete suppression may reflect the fact that segmental primordia, though greatly reduced, are not altogether absent in these embryos.
The results with the three maternal classes of pattern forming genes do not tell us conclusively whether giant is directly affected by maternal cues. However, the fact that the anterior (bcd) and posterior (nos) systems affect the earliest expression of gt argues for a direct activating effect of bcd on stripe 3 and of a repressing activity of maternal hb on stripe 4 expression. The effects of the terminal system on the later-appearing stripe 1 and on the retraction of stripe 4 are instead most likely mediated by zygotic gene products.
Gap gene interactions with giant
The earliest zygotic genes that affect the anteroposterior pattern belong to the segmentation gap class. The gap genes begin to be expressed very early at the syncytial blastoderm stage and are thought to respond directly to the maternal cues. It has been shown in addition that the gap genes interact both positively and negatively with one another so that the final pattern of expression is a complex resultant of the maternal and zygotic influences (reviewed by Gaul and Jâckle, 1990). To determine how other gap genes influence gt expression, we examined the gt protein for Krüppel (Kr), knirps (kni), hb or tll mutations. In addition we looked for the effects of gt mutations on the expression pattern of hb, Kr and kni. To aid in identifying mutant embryos, we frequently double-stained using antibodies to ftz or en. As a rule, however, we measured the positions of the staining domains in a large number of unselected embryos and plotted each in a histogram. A statistically significant separate peak accounting for approximately 25 % of the embryos was taken to represent a domain of expression affected by the mutation (see the methods section). The results obtained with each mutation are summarized diagrammatically in Fig. 5.
hb – gt interactions
As expected from the analysis of nos mutations, zygotic hb, like maternal hb, has a negative effect on gt expression in the posterior domain. In hb mutant embryos, stripe 4 expands posteriorly into the region normally occupied by a late-appearing band of zygotic hb expression (10 – 20% EL; Fig. 6A). In addition, the anterior border is shifted 4% EL anteriorly. This confirms the negative effect of hb on the expression of stripe 4 and indicates that the withdrawal of this domain from the posterior end of the embryo is at least in part caused by the synthesis of zygotic hb in this region. However, in hb mutants giant expression does not continue all the way to the posterior pole, indicating that other negative interactions are involved, most likely with the zygotic interpreters of maternal terminal information. The shift of the anterior border is harder to account for. Loss of zygotic hb function is necessarily accompanied by the reduction of maternal hb to only one dose of the gene. This might lower the maternal hb concentration in the posterior half of the embryo and permit the activation of stripe 4 up to a more anterior position. Hülskamp et al. (1990) invoke a similar explanation for the anterior expansion of the abdominal band of knirps expression. This broadening of the kni band in response to a reduction in maternal and zygotic hb may have additional direct or indirect effects on the domain of gt expression. We do not know what causes stripe 4 to fade at the end of cellular blastoderm but it cannot be simply inhibition by increasing levels of hunchback posterior expression since the fading occurs also in hb- embryos.
The only significant effect of the absence of zygotic hb on the anterior pattern of gt is a slight anterior shift of stripe 3. This indicates that hb does not mediate the activating effect of bicoid on gt anterior expression. Given the complete overlap between the anterior domains of gt and of hb, this minor effect is not likely to be directly caused by hb. A more plausible explanation is that it is mediated through Kr. It has been shown that high levels of hb repress Kr and set the anterior edge of the Kr central domain. In the absence of zygotic hb, the Kr domain expands anteriorly by 10 % EL (Hülskamp et al. 1989) and could affect gt anterior expression.
There are no obvious changes in the domains of zygotic hb expression in embryos deficient for gt. A very subtle effect may be observed in the anterior of the embryo where, at cellular blastoderm, the. previously continuous expression of hb becomes modulated into three stripes that appear to be complementary to the three anterior gt stripes. If this is the case, it would suggest a negative effect of gt on the anterior expression of hb. However, the anterior modulation of hb is difficult to ascertain under our staining conditions and we cannot be certain that it fails in gt-deficient embryos.
Kr – gt interactions
Kr has a distinct negative effect on gt expression. In Kr mutant embryos stripe 4 is greatly expanded anteriorly, reaching 48 % EL and invading therefore both the kni and Kr domains of expression (Fig. 6B). This effect could be direct or mediated through kni. Pankratz et al. (1989) have shown that the full expression of kni requires Kr activity. In Kr mutants, the posterior expression of kni never reaches its normal intensity, though its position is not shifted. The effect of Kr on gt stripe 4 could therefore be a direct negative effect of Kr on gt or an indirect effect caused by Kr enhancement of Arnzrp.v-dependent repression of gt. We conclude that it is most likely a direct effect because kni mutations have no broadening effect on gt stripe 4 (see below) while, in Kr mutants, stripe 4 expands nearly to 50 % EL. What prevents this gt domain from expanding further is probably the steeply rising concentration of hb whose anterior domain extends to 48% EL. In the anterior region, Kr mutations have only slight effects on gt, consisting of a slight broadening of stripe 3 and stripe 2. This could also be explained as a repressive effect caused by Kr expression in the central domain and in the Kr anterior domain (a stripe around 82 % EL).
The effects of gt mutations on Kr expression are very mild. In agreement with Gaul and Jackie (1987), we could not observe a significant broadening of the Kr central domain. However, several lines of evidence suggest that gt has a negative effect on Kr: (1) bcd embryos, in which gt stripe 4 expands forward, show a contraction of the Kr posterior border; (2) conversely, in cases in which gt stripe 4 is suppressed, Kr posterior border expands posteriorly, e.g. in nos embryos; (3) more directly, we have shown that ectopic expression of gt under control of the heat shock promoter causes suppression of Kr expression in the central domain and generates cuticular phenotypes similar to those of Kr mutant embryos while gt protein binds in vitro to specific sites in the Kr gene in the regulatory region responsible for expression in the central domain (M. Capovilla, E. Eldon and V. Pirrotta, unpublished data). Similar observations with ectopic expression of gt have been made by Kraut and Levine (1991).
kni – gt interactions
Mutations in the knirps gene affect the development of the first seven abdominal segments, thus including most of the abdominal region affected by giant. In kni mutants, stripe 4 of gt expression is affected only very slightly, if at all (Fig. 6F). It is possible that the posterior border becomes less well defined and that the intensity of posterior expression, which normally begins to fade at the end of cellular blastoderm, fades somewhat more rapidly in kni embryos. However, it is clearly still detectable at the beginning of germ band extension and we hesitate to attach much significance to these slight effects. The lack of significant changes in the gt stripe 4 domain in these mutants indicates that neither the hb nor the Kr effect on gt are to be explained by a mediation through knirps and are therefore best interpreted as direct interactions. In the head region, it is interesting to note that the knirps anterior domain of expression is roughly complementary with that of giant (Fig. 6E). At syncytial blastoderm, knirps forms a cap at the anterior pole that extends ventrally until about 75 % EL. At cellular blastoderm, a thin ring of expression appears at this position that corresponds to the interval between stripe 2 and 3. In spite of this pleasing complementarity, knirps mutants show the normal separation between gt stripe 2 and stripe 3 which cannot therefore be accounted for by a repressive action of kni on gt anterior expression.
Of the gap genes, knirps is the one whose pattern of expression is most strongly altered in embryos deficient for gt. In these embryos, the posterior domain of kni, which is normally at 34 – 47 % EL, expands posteriorly to 28 % EL. The strong negative effect of gt on kni that is implied by this result may in fact help to explain the previously reported interactions of kni with Kr. Pankratz et al. (1989) found that in Kr mutant embryos the abdominal kni stripe is present at the normal position but its intensity is greatly decreased. They concluded from this that Kr enhances kni expression. This apparent enhancement may be better explained as an indirect effect: in Kr mutants, gt stripe 4 expands anteriorly up to the middle of the embryo, fully overlapping and inhibiting expression of kni in its posterior domain.
tll—gt interactions
Lack of tll function affects the structures derived from both anterior and posterior terminal domains. In tll mutants abdominal segments 8-10 are missing and a decrease in the extent of the procephalon is accompanied by an expansion of the remaining body segments (Strecker et al. 1986). While tll is not the only zygotic target of maternal terminal class genes, it is clearly one of the mediators of these maternal signals (Klingler et al. 1988; Strecker et al. 1989b). In the absence of tll activity, the posterior domain of gt expression never matures properly. At the end of cellularization, it does not complete its withdrawal from the posterior terminal region and forms an expanded band whose posterior border is shifted 10 % EL posteriorly (Fig. 6C). This position is in good agreement with the shift in the fate map seen in tll embryos by Mahoney and Lengyel (1987) and with the distribution of tll mRNA (0-15% EL at cellular blastoderm, Pignoni et al. 1990). Surprisingly, lack of tll has little effect on the anterior domains of gt expression. The apparent posterior shift of stripe three was not considered significant because the internal wild-type controls in this set of embryos showed a similar range of positions for this stripe.
More significant shifts are seen with tailless-huckebein double mutant embryos (Fig. 6D). According to Weigel et al. (1990), the zygotic gene huckebein (hkb) is required to specify parts of the fate map corresponding to the posterior midgut and to anterior terminal structures. They suggest that hkb acts like a gap gene and, together with tll, mediates maternal terminal cues from the tor gene product. In tll-hkb double mutants, the posterior border of gt stripe 4 retracts even less than in tll mutants but at cellular blastoderm it does not extend completely to the tip of the embryo as in tor or tsl embryos. This might suggest the existence of additional components that mediate the effect of tor but a more likely explanation is that the hkb mutation used does not cause complete loss of function (Weigel et al. 1990). While tll alone has little effect on gt in the head region, the sum of the two mutations results in a gt pattern approaching the greatly shifted pattern seen in the absence of tor function. Stripe 1 is shifted to the anterior pole and is present at barely detectable levels while stripes 2 and 3 are displaced forward as in tor embryos. This indicates that, at the anterior end, torso effects on gt are mediated primarily through hkb and not through tll.
gt expression in the head primordia
Three genes have been recently described as having a gap-like role in mediating the development of head segments. These are orthodenticle (otd), empty spiracles (ems) and buttonhead (btd) (Cohen and Jürgens, 1990). The genes for otd and ems have been cloned by Finkelstein and Perrimon (1990) and Dalton et al. (1989), respectively, who found that they both encode homeodomain proteins. Since, after blastoderm, gt protein is found almost exclusively in the head primordia, we looked at the gt expression pattern in embryos mutant for each of these three genes. Both ems and otd affect gt expression in the anterior domain. In ems embryos, stripe 1 is normal but stripes 2 and 3 fail to sharpen, and separate at cellular blastoderm (Fig. 6H). In otd mutants, the posterior border of stripe 1 and all of stripe 2 are shifted posteriorly (not shown). In both cases, the part of the gt pattern most affected lies in the region of expression of the corresponding gene and involves the part of the fate map affected by the corresponding mutations. Both otd and ems are positively regulated by bcd and it has been proposed that they are direct targets of the bcd protein (Cohen and Jürgens, 1990; Finkelstein and Perrimon, 1990; Dalton et al. 1989). However, they cannot be the mediators of bcd effect on gt since their effect on gt anterior expression is negative rather than positive. Instead, the shifts in the gt stripes of expression that result from their loss of function are similar to those caused by other gap gene mutations. This is consistent with the interpretation that otd, ems and btd act as segmentation gap genes in the head region, where a greater complexity may require a greater number of components to specify the pattern. However, not all of these genes interact with giant. In btd embryos the changes in the gt anterior pattern are so slight as to be most likely insignificant. Although molecular probes for btd product are not available, the structures affected by btd mutations imply that its domain of expression should include the mandibular, intercalary and antennal segments. Of these, at least the mandibular region should overlap with sites of gt expression at cellular blastoderm.
Discussion
Our results show that the giant gene behaves like a typical segmentation gap gene. Like other gap genes, gt produces a nuclear factor very likely involved in regulating the expression of other genes. The defects exhibited by giant loss-of-function mutants fall into two regions. One, in the posterior half of the larva, is a loss of contiguous abdominal segments A5, 6, 7 and sometimes 8, a type of pattern deletion characteristic of segmentation gap mutations. The posterior domain of giant expression corresponds well to this phenotype and the alterations of the gt pattern in various mutant backgrounds are indicative of interactions typical of segmentation gap genes. While the posterior domain of gt expression fades off soon after cellular blastoderm, the in situ hybridization results (Mohler et al. 1989) and the localization of the protein show that the head is the site of most intense, complex and prolonged gt expression. This later expression in the head region represents a second phase of giant activity that takes place after its function as a gap gene is completed.
The anterior defects caused by gt mutations are more difficult to characterize because of the complex events and the structural rearrangements that normally result from head involution but that are perturbed or fail to take place in gt mutants. The reported lack of labral structures, epi- and hypostomal sclerites, H-piece and dorsal bridge (Mohler et al. 1989) are difficult to interpret as segmentation gaps. Rather, the defects within the head appear restricted to individual structures or elements rather than whole segments. The segmental attribution of these defects does not always correspond well with the pattern of expression that we observe in the head region. The anteriormost stripe and the later strong expression in the clypeolabrum fit well with the labral defects and the failure of head involution but there is a discrepancy between the position of stripes 2 and 3 and the sites of the other head defects. Petschek et al. (1987) and Petschek and Mahowald (1990) noted the disappearance of the labial lobe and the transient fusion of T1 and T2 in gt embryos at the germ band retraction stage, accompanied by defects in the expression of Antennapedia and Ser genes. Carroll and Scott (1985) and Petschek and Mahowald (1990) have also shown that stripes 1 and 2 of ftz are significantly broadened in gt mutants. Our results show that the posterior boundary of gt expression is within the posterior compartment of the maxillary segment and coincides with stripe 2 (maxillary) of engrailed, just in front of the labial lobe. This discrepancy can be accounted for in part by supposing that some of the mature head defects might have been incorrectly assigned to labial primordia and in part by assuming that the effects on the labial and thoracic segments are indirect, not primarily due to the lack of gt product but to the effect that it produces on other gap genes, most likely Kr. The spread of pattern defects to regions broader than the domains of strong expression is a phenomenon common to the other gap genes (hb, Kr, kni, tll) and is most likely explained through combinatorial and indirect effects produced by the cross-interactions between these genes. The fact that in gt mutants no specific defects have been attributed to maxillary and mandibular derivatives may also be due in part to the early internalization of most of the cells that express gt during the complex in folding of the initial blastoderm sheet of cells in the cephalic furrow. It is possible that, as a result, the gt-expressing cells contribute principally to internal structures, most of which are not preserved in cuticle preparations and whose ontogeny is more difficult to follow.
Two different mechanisms of gt activation
An account of the gt pattern of expression must begin with the realization that the anterior expression is controlled by a different mechanism from that which activates the posterior. Anterior expression is bcddependent, the posterior is not. Posterior expression is Ai>-sensitive, the anterior is not. Molecular studies of several other segmentation genes have shown that their regulatory regions are complex and constituted by multiple elements that can act independently of one another (Harding et al. 1989; Pick et al. 1990; Hoch et al. 1990). A preliminary analysis of the gt regulatory region shows in fact that the control elements responsible for expression in the posterior domain are distinct, independent and widely spaced from those that control anterior expression (P. Bagnaresi and V. Pirrotta, unpublished results). The results presented in this paper show that bcd is responsible, most likely directly, for activating the initial anterior expression of gt. Vie may suppose that this occurs through a mechanism similar to that demonstrated for hb and anticipate that molecular analysis will demonstrate the presence of bcd binding sites in the regulatory domain responsible for anterior expression of gt. The failure of this bcd dependent activation to extend as far as the anterior tip even at the earliest stages suggests that maternal factors inhibit gt expression in the terminal region. If these were part of the terminal system, .we would expect that in tor or tsl embryos this repression would be lifted and expression would extend fully to 100 % EL. Since this is not the case, we suppose that the highest bcd concentrations have a repressive effect on gt expression. Although such repressive activities of bcd have not yet been demonstrated at the molecular level they have been postulated by Pignoni et al. (1990) to explain anterior inhibition of tll by bcd and by Hülskamp et al. (1990) to set the anterior border of Kr expression.
The posterior initial pattern, like much of the rest of abdominal development, appears to be negatively rather than positively controlled. Several groups have shown that the removal of maternal hb suffices for normal abdominal development even in the absence of nos function (Hülskamp et al. 1989; Irish et al. 1989; Struhl, 1989). The activation of gt in the posterior domain must therefore be constitutive, dependent on ubiquitous transcription factors and severely inhibited by even relatively low levels of hb. Maternal hb RNA is initially uniformly distributed but, under the influence of nos product, the accumulation of hb protein tapers off in the posterior half of the embryo (Tautz, 1988). We can explain the activation of gt in the posterior third of the embryo as the derepression that occurs when hb concentrations drop below a threshold. Initial expression of gt would then take place from the position at which this threshold is reached to the posterior tip. The posterior expression of kni could be similarly derepressed but at a higher threshold of hb concentration. Maternal hb product, in this interpretation, acts as a graded morphogen, exerting negative control on genes that differ in their sensitivity to hb inhibition. Additional control may be exerted by bcd on the posterior expression of gt. This is likely because if maternal hb is removed entirely, normal development can still ensue (Lehmann and Nüsslein-Volhard, 1987). Though it has not been shown directly, this implies that gt expression in these embryos is normal and not ubiquitous. Some other maternal factor, in addition to maternal hb, must therefore inhibit gt expression in the anterior two thirds of the embryo. A likely candidate for this inhibitor is the bcd product which would then have a positive effect on the anterior regulatory element of gt and a negative effect on the posterior regulatory element. That this may be the case is also suggested by the fact that a consequence of the lack of maternal bcd is the broadening and anterior shift of the entire posterior stripe of gt expression (Fig. 4A) as well as of kni and Kr expression (Hülskamp et al. 1990). The result of this contrasting effect on anterior and posterior gap gene expression is to cause both the loss of anterior structures and the anterior shift of the rest of the pattern.
The gradual withdrawal of gt stripe 4 from the posterior, which occurs at the end of nuclear cycle 13, is clearly dependent on the terminal system: it is abolished in tor or tsl embryos. The effect of tor is mediated in part through the zygotic gap genes tll and hkb, as soon as their products begin to accumulate. In part, however, the withdrawal of stripe 4 is caused by the appearance of the posterior stripe of zygotic hb, which exerts a repressive effect on posterior gt expression.
Interactions with other gap genes
While maternal mutations can abolish gt expression in the anterior (bcd) or posterior (nos) domains, none of the known gap genes is required for initiating gt expression. This strongly suggests that gt is a direct interpreter of maternal cues and confirms the status of gt as a gap gene. In typical gap gene fashion, giant enters into a network of cross-regulatory interactions with other members of the gap gene class. To understand the interactions of gt with other gap genes, it is once again important to consider the different stripes of gt expression as independently controlled. It is also important to bear in mind that, because of the complex cross-regulatory interactions between genes of this class, it has been difficult to distinguish direct effects of one gene product on the expression of another and indirect effects mediated by other members of the group. In many cases, molecular experiments will be necessary to establish the direct regulatory relationships. As already discussed, hb, both maternal and zygotic, plays a major role in inhibiting or sharpening the expression of stripe 4 but our results show that it has little effect on the anterior gt stripes. At the same time, gt has little effect on hb expression. Kr clearly has a strong negative effect on the gt posterior stripe and gt in turn has a negative effect on Kr. This negative effect of gt on Kr is not very visible under normal conditions, as found also by Gaul and Jackie (1987), not because it is not strong but probably because Kr is also repressed by kni at its posterior border and by hb at its anterior border. Hence, in gt mutants, the expansion of the Kr domain is immediately checked by the presence of kni and hb immediately flanking it. However, the strong repressing effect of gt on Kr is evident when gt is expressed ectopically under the control of the heat shock promoter (Kraut and Levine, 1991; M. Capovilla, E. Eldon and V. Pirrotta, unpublished data). The anterior border of Kr is therefore set by hb repression and also by gt anterior expression, as hypothesized by Hülskamp et al. (1990). This is yet another case in which expression boundaries are multiply determined.
We found very little effect of kni on gt expression in the posterior domain but strong repressive effects of gt on the kni posterior stripe. The interactions of gt and Kr and of gt and kni may help to explain the apparent activating effect of Kr on kni expression as caused indirectly by Kr repression of gt. This negative role of Kr is more consistent with the demonstration that Kr protein acts as a transcriptional repressor (Licht et al. 1990). However, Kr may also interact directly with kni as suggested by Pankratz et al. (1989) who have identified a binding site for Kr protein in the kni regulatory region. The terminal gap genes tll and hkb appear to have negative effects on the other gap genes, both at the anterior and at the posterior end, with the exception of hb whose late-appearing posterior stripe actually requires terminal gene function (Tautz, 1988). Although no direct data are yet available, it is possible that gt in turn represses tll since the initial expression of tll over a broader posterior domain recedes to a posterior cap by NC14 (Pignoni et al. 1990). Failure to reduce tll posterior expression in a gt mutant might perhaps account for the appearance of secondary filzkôrper in the abdominal region, noted by Petschek et al. (1987) and Mohler et al. (1989).
The domains of expression of the gap genes are set by a network of mutual interactions and their boundaries are often multiply determined by the maternal morphogens, the neighboring gap genes as well as the non-adjacent gap genes. These interactions are essential to account for the stability and robustness of the patterns of expression and probably for the broad and graded morphogenetic effects that result from gap mutations. Accordingly we can interpret the extensive defects seen in Kr mutations as due not only to absence of Kr but also to the resulting expansion of gt stripe 4, which invades and overruns the kni domain and most of the Kr domain. In contrast, the more restricted effects of gt mutations in the abdomen could be attributed to the fact that the posterior expansion of kni is limited by the hb and tll domains.
ACKNOWLEDGEMENTS
We are very grateful to Peter Gergen, Thom Kaufman, Henry Krause, Steve Cohen, Tom Kornberg, Ken Howard, Rachel Kraut and Mike Levine for gifts of mutant stocks and antibodies. We thank Sharon Bickel for help in antibody purification, Steve Cohen and Maria Capovilla for helpful suggestions and Rachel Kraut and Mike Levine for communicating their results to us prior to publication. E.D.E. was the recipient of an NIH postdoctoral training grant stipend and the research was supported by NIH grant GM 34630 to V.P.