Mitosis in Drosophila - Home | Journal of Cell Sciencejcs.biologists.org/content/joces/92/2/137.full.pdf · Mitosis in Drosophila ... as model fos r th stude y of th mitotie c cell

REVIEW

Mitosis in Drosophila

DAVID M. GLOVER

Cancer Research Campaign, Eukaryotic Molecular Genetics Research Group, Department of Biochemistry, Imperial College of Science,Technology and Medicine, London S\V7 2AZ, UK

Summary

Drosophila is an attractive organism in which tostudy both the rapid rounds of mitosis typical ofembryonic development in many species, and thelonger cell cycles of diploid tissues later in develop-ment. Mutations in genes essential for mitosis inDrosophila may result in lethality in late embry-onic, larval or pupal stages of development. Inaddition, mutations in many genes required for thenuclear divisions of early embryogenesis have been

found in screens for female sterility. The mitoticmutations have phenotypes indicative of lesions at avariety of mitotic stages. A combined molecularand genetic analysis of these genes has the potentialto unravel the complex set of protein-protein inter-actions that occur in this dynamic process.

Key words: mitosis, Drosophila, cell cycle.

Introduction

Mitotic cycles in early embryosThe early embryos of many organisms, including insects,echinoderms, molluscs and amphibians, have been usedas models for the study of the mitotic cell cycle. Indeedthe mitotic divisions in the early embryos of suchorganisms consist of rapid successions of M and S phaseswith no discernible Gi or G2 phases as found at laterstages of development. Studies of the early cell divisioncycles in sea-urchins or in Xenopus embryos have led tothe idea of an underlying master oscillator. The cytoplas-mic origins of this oscillator have been suggested by thedemonstration of continuing cell surface contractions inenucleated Xenopus (Hara et al. 1980) and sea-urchinembryos (Sluder et al. 1986) with a similar frequency tothose normally observed. More recently it has beenshown that cycles of protein phosphorylation, histonekinase activity, and M phase or maturation promotingfactor (MPF) activity also continue in Xenopus embryosin the absence of any nuclear components (Dabauvalle etal. 1988). It has been postulated that MPF plays a keyrole in the cell cycle. It was first described as a factor thatinduces G2-arrested Xenopus oocytes to mature by com-pleting the second meiotic division. Subsequently, MPFactivity was shown to oscillate in mitotic divisions,peaking in each M phase (Wasserman & Smith, 1978;Gerharte/o/. 1984). When partially pure MPF is injectedinto interphase arrested eggs or cell-free extracts, itinduces nuclear envelope breakdown, chromosome con-densation, and spindle formation (Miake-Lye et al.

Journal of Cell Science 92, 137-146 (1989)Printed in Great Britain © The Company of Biologists Limited 1989

1983). As MPF activity decays, these processes undergoreversal and S-phase begins as indicated by the onset ofDNA synthesis. MPF relieves G2 arrest in the presenceof protein synthesis inhibitors, and has been postulated tobe a kinase critical for initiating a cascade of mitoticevents. Protein synthesis is required, however, for matu-ration and cleavage and for the cyclical appearance ofMPF activity (Lohka & Mailer, 1985; Gerhart et al.1984). Just as MPF activity oscillates during these earlymitotic divisions, another class of proteins (the cyclins)have been described, which undergo periodic synthesisand degradation during each cell cycle in the earlyembryos of several organisms (Rosenthal et al. 1982;Evans et al. 1983; Swenson et al. 1986; Standart et al.1987). It seems likely that the cyclins play a significantrole in regulating these early mitotic cycles, although atpresent the relationships between MPF and the cyclinsare not clear.

Genetic studies of the cell cycleThe sequential pathway of mitotic events that hasemerged from the formalized conclusions of geneticstudies with yeasts seems in contrast to the idea of anoscillatory mechanism that has emerged from the study ofembryonic systems. Most of the genetic studies on thecell cycle carried out to date have been with either bakers'yeast, Saccharomyces cerevisiae, or fission yeast, Schi-zosaccharomyces pombe. In their pioneering studies onthe cell cycle in S. cerevisiae, Hartwell & Pringle isolatedconditional lethal 'cell division cycle (cdc)' mutants thatarrest the development of cells at characteristic stages of

137

the cycle (Pringle & Hartwell, 1981; Hayles & Nurse,1986). Studies of the phenotypes of these mutants eithersingly, or as double mutant combinations, have enabledfunctional relationships between various mutants to bedetermined. The interrelationships between these genesare formally expressed as pathways of sequential activi-ties.

Several observations point towards a unique step in theyeast cell cycle that has been termed 'start', which has tobe completed in order to initiate DNA synthesis andsubsequent mitotic events. Start is the point at which cellgrowth is coordinated with division and it also marks thepoint at which the mating pheromones arrest the cycleprior to conjugation. The presence of this conditionalblock in the cell cycle of the yeasts is a major differencefrom the oscillatory cycles of early embryos. This reflectsthe disparity of the experimental systems under study.The early embryos of the organisms mentioned above areessentially closed systems in which there is no de novogene activity within the developmental time frame understudy and all 'life-support systems' have been provided bythe mother. The yeast cell on the other hand is a wholeorganism in its own right. It has to utilize externalnutrients and can undertake the process of the divisiononly when conditions are appropriate.

Underlying principlesIt would not be surprising to find aspects of the control ofcell division unique to one or other of these systems,although it is clear that many steps must be held incommon. Commonality has recently been shown for theproduct of the cdc2 gene of S. pombe, which is required,not only for 'start', but also for the Gi to M transition.The cdc2 gene product has been conserved throughouteukaryotes, as is most strikingly demonstrated by theability of the cDNA encoding the homologous humanprotein to complement cdc2 mutations of 5. pombe (Lee& Nurse, 1987). More recently, Gautier et al. (1988)have demonstrated that an antibody against the highlyconserved region of the 34K (K=10 3 M r ) proteinencoded by cdc2, will recognize a protein of similarmolecular mass in a preparation of MPF purified fromXenopus. Furthermore, the 13K product of the sucl genethat interacts with cdc2 kinase in yeast cells will inhibitMPF activity, and can be used as an affinity reagent topurify the Xenopus cdc2 homologue, a 32K protein withan associated 45K protein (Dunphy et al. 1988). It isclear, therefore, that the differences between the cellcycles of yeasts and early embryos have been overempha-sized by the experimental approaches used to analysethem; the Xenopus embryo lends itself to biochemicalstudies, the yeasts lend themselves to genetic analysis.

Studies on cell division in Drosophila should be able tocombine the advantages of both systems. Not only doesthe fruit fly have extensively studied genetics, the virtuesof which have been extolled on many previous occasions,but also its developmental biology is highly suitable forstudies of the mitotic cycle. The early embryo has asuccession of rapid mitoses and, furthermore, in thesubsequent development of the organism, two differenttypes of tissue arise, having either proliferating diploid

cells or cells that do not divide but become highlypolyploid. As we shall see below, the co-existence of thesetwo types of tissue at certain developmental stagespermits several approaches for selecting mutations ingenes essential for mitotic cell division.

The isolation of mitotic mutants of Drosophila

Mutations with embryonic phenotypesThe Drosophila embryo is a syncytium for the first twohours of its development, in which time there occur 13rapid rounds of nuclear division. The first nine rounds ofmitosis occur within the embryo and then at telophase ofnuclear cycle nine the majority of the nuclei migrate tothe cortex. Once at the surface, the nuclei undergo afurther four cycles before cellularization occurs at inter-phase of cycle 14 (Zalokar & Erk, 1976; Foe & Alberts,1983). The organization of the cytoskeleton during thisperiod of rapid nuclear divisions has been carefullydocumented in both fixed and living embryos (Karr &Alberts, 1986; Warn et al. 1987; Kellogg et al. 1988).The cell cycle lengthens following cellularization andthere is a distinct interphase period, enabling transcrip-tion to occur. Until this stage, there has been little or nozygotic gene expression and so the components requiredfor the early mitoses must have been provided matern-ally. One might therefore expect to find a class ofmaternal-effect lethal mutations that disrupt the func-tions of the genes encoding these components. A femalehomozygous for such a mutation would be expected toproduce embryos in which the early divisions could notbe completed successfully. How do these homozygousmutant females themselves survive to adulthood if theyhave a defective gene essential for mitosis? This happensbecause, for many genes, the wild-type gene productsupplied by their heterozygous mothers persists through-out embryogenesis to permit the remaining three to fourrounds of cell division that occur following cellulariz-ation. Most of subsequent larval development involvescell growth with the endoreduplication of DNA in theabsence of mitosis. Nevertheless, the imaginal cells,destined to form the adult organism and not themselvesnecessary for the survival of the larva, continue to dividethroughout larval development, as do cells of the centralnervous system. Thus there is a requirement at thisdevelopmental stage for zygotic activity of genes essentialfor mitotic cell division. Some of the mitotic genesrequired for early embryogenesis do not appear to berequired for these later cell divisions. Either they areproducing a gene product that is specific for the earlyembryo or they are are members of families of genes, eachhaving different developmentally regulated expressionpatterns. In many cases, however, maternal effect mu-tations do show some additional zygotic effect upon cellproliferation in the diploid imaginal and neuroblast cellsof the larvae. The cytological examination of diploidtissues in these larvae reveals mitotic abnormalities insome cells. It would appear that in these cases themutation does not result in a complete arrest of mitosis inall of these relatively slowly dividing cells. This is in

138 D. M. Clover

contrast to early embryogenesis where the imposition of13 cycles of nuclear division within a 2-h period isevidently too great, and the gene product provided by thehomozygous mutant mother is either insufficient orinadequate for this task.

A second class of mitotic mutations can be recognizedin which homozygous mutant zygotes can survive earlyembryogenesis utilizing the wild-type gene product sup-plied by their heterozygous mothers, but which do notcomplete the division cycles that follow cellularization.These divisions occur in complex 'mitotic domains',which develop following a specific temporal programme(Hartenstein & Campos-Ortega, 1985; V. Foe, personalcommunication). The gene string is one example of sucha gene that is required for the progression of cells from G2into the fourteenth mitosis (B. Edgar & P. O'Farrell,personal communication). Other mutations appear tolead to the arrest of cell division in some lineages ofdividing cells in preference to others in later embryogen-esis. One example is a recessive lethal mutation thataffects the divisions of certain neuroblasts in the lateembryo (E. Giniger, H. Vaessin & Y. N. Jan, personalcommunication). The corresponding wild-type gene hasbeen cloned and found to be homologous to the cyclins(see above). This same gene that encodes cyclin A hasbeen independently cloned both by Lehner & O'Farrell(personal communication) and in our own laboratory(Whitfield & Glover, unpublished). In addition, we havealso found a Drosophila homologue of the cyclin B gene.Thus the door is opening for a genetic analysis of thecyclin genes in a multicellular eukaryote.

Mutations with late larval or pupal phenotypesA large group of homozygous mitotic mutants can surviveby utilizing maternally supplied proteins until late larvaldevelopment. In such cases, the imaginal cells of thehomozygous mutant larvae cannot proliferate and conse-quently death ensues during the larval or early pupalstages. This larval lethal phenotype of mitotic mutantswas first recognized in the analysis of DNA repair-defective mutants that produce elevated frequencies ofchromosome breakage (Baker et al. 1982), and in sub-sequent analyses of collections of late larval lethal mu-tations from three laboratories, Gatti & Baker (1988)were able to identify many mitotic mutants. The mitoticphenotype of repair-defective mutations is just oneexample of how genes required for progress through thecell cycle can have additional roles in related biologicalprocesses. Baker and his colleagues have demonstrated acell-cycle requirement for certain loci originally identifiedfrom mutants showing increased mutagen sensitivity.The assay used in these studies was the spontaneousproduction of genetically mosaic somatic tissue in fliesheterozygous for a recessive cell marker. The appearanceof clones of cells having the phenotype of the recessivegene shows that cells have become hemizygous or homo-zygous for the recessive marker. This can be shown to beindicative either of a problem with the transmission ofchromosomes to daughter cells, chromosome breakage,or mitotic recombination occurring during the cell cycle(Baker & Smith, 1979; Baker et al. 1982; Smith et al.

1985). The cytological analysis of these mutations hasconfirmed such genetic inferences and in many cases haspointed towards high levels of spontaneous chromosomebreakage (Gatti, 1979).

Mutations affecting both meiosis and mitosisAnother route towards mitotic mutants is through thefurther characterization of the meiotic mutants of Dros-ophila. Although meiotic and mitotic divisions havefundamental differences, they share obvious similarities,and utilize some common gene products. The firstmeiotic division differs from mitotic divisions in severalrespects, the most notable being that the centromeres donot split and homologous chromosomes are segregated tothe spindle poles. One might therefore expect a separateset of functions to be needed during this division. Thesecond equational division, on the other hand, is compar-able to mitotic divisions requiring splitting of the centro-meres. Of over 40 mutants that have been described asaffecting meiosis in Drosophila (Baker & Hall, 1976;Lindsley & Sandier, 1977), the majority affect the firstmeiotic division. However, Baker et al. (1978) haveshown, using the somatic cell clonal analysis describedabove, that of these mutants, six recombination-defectiveloci and four loci required for correct meiotic segregationalso show some effect on mitosis. Mutants that preferen-tially affect the second meiotic division are comparativelyrare. In their 1976 review, Baker & Hall suggest that thismight be because these genes would be expected to play acrucial role in mitosis and so would not have beenrecovered in the mutant selection schemes employed tosearch for meiotic mutants. Meiosis has now been exam-ined in males carrying several mutations selected by theirmitotic phenotypes. A number of meiotic effects areapparent. Males homozygous for the mitotic mutantsabnormal spindle (asp) or polo, show chromosome non-disjunction in meiosis (Ripoll et al. 1985; Sunkel &Glover, 1988). The mitotic mutation merry-go-round(mgr) is a recessive lethal, but cytological observationsshow that both meiotic divisions in this mutant are alsoabnormal, resulting in the formation of \n rather thanhaploid spermatid nuclei (Gonzalez et al. 1988).

Maternal effect loci essential for mitosis

gnu, a mutation specifically affecting earlyembryogenesisThe mutation gnu identifies a gene whose product isneeded for nuclear division during early development.Females homozygous for gnu lay eggs that develop giantnuclei as a result of continued DNA replication, in theabsence of chromosome segregation and nuclear division(Freeman et al. 1986). Fertilization of GNU eggs is notrequired for the giant nuclei to develop, contrasting withwild-type eggs in which fertilization is required beforeany further development can take place (Freeman &Glover, 1987). Whether or not the GNU egg is fertilized,any of the four products of female meiosis, the three polarbodies or female pronucleus, can participate in DNAsynthesis to give giant nuclei. By marking the paternal

Mitosis in Drosophila 139

DNA microtubules centrosomes

Wild-type

GNU

Fig. 1. The upper panels show a field of nuclei in the cortex of a blastoderm Divsophila embryo in prometaphase viewed byfluorescence microscopy. The chromosomes, stained with Hoechst, have undergone condensation and are becoming aligned overthe centre of the spindle microtubules stained with an anti-tubulin antibody. The centrosomes at the spindle poles are stainedwith an antibody against the centrosomal associated antigen, Bx63 (Frasch et al. 1985). The lower panel shows a field from aGNU embryo (Freeman et al. 1986; Freeman & Glover, 1987). Neither nuclear division nor the chromosome condensation-decondensation cycle occurs, although centrosomes replicate and continue to nucleate asters of microtubules.

genome with a bacterial gene, it has been shown that ifthe GNU egg is fertilized, then the DNA derived fromthe male pronucleus also undergoes DNA replication infertilized GNU embryos. This same approach has alsobeen used to demonstrate that DNA of the paternalgenome fails to replicate in fertilized eggs of mothershomozygous for the maternal haploid mutation, mh.Fertilization is required to trigger the development ofeggs laid by females homozygous for mh, but syngamydoes not occur and the female pronucleus undergoesmultiple rounds of haploid nuclear division. However, iffemales are homozygous for both gnu and mh, then bothmaternal and paternal genomes are replicated and, fur-thermore, the embryos develop giant nuclei (Freeman &Glover, 1987). It seems that somehow, the GNU cyto-plasm is lifting the repression of DNA synthesis thatnormally occurs following the completion of meiosis untilthe fusion of the male and female pronuclei has takenplace. The gene therefore appears to play a role in thecorrect establishment of coordinated DNA replicationand mitosis in zygotic development.

Uncoupling ofmitotic cycles from DNA replication inthe early embryoOne of the striking features of GNU embryos is thatalthough nuclear division does not take place, centro-somes continue to replicate (Fig. 1). Normally, eachwild-type interphase nucleus is associated with a singlecentrosome that replicates, giving rise to two daughtercentrosomes. These migrate to opposite sides of thenucleus in prometaphase to nucleate the microtubules ofa new spindle. The centrosomes of GNU embryos aredissociated from nuclei and do not function in theformation of mitotic spindles per se, but are capable ofnucleating asters of microtubles. They increase in num-ber and migrate to the cortex of the developing GNUembryo as they would in the wild-type, indicating thatalthough the nuclear division cycle has been disrupted,the centrosome cycle continues independently. Spindle-like structures can be seen when the embryos becomenecrotic and the giant nuclei break down to yield frag-ments of chromatin, which can organize microtubules.This would seem to indicate that there is nothing

140 D. M. Glover

tf

Wild-type polo

Fig. 2. Mitotic figures in larval neuroblasts. A wild-type anaphase is shown together with a circular initotic figure from larvaehomozygous for the mutation polo (Sunkel & Glover, 1988). The circular figures have certain characteristics of monopolarspindles.

inherently wrong with the mitotic apparatus of GNUembryos per se (Freeman et al. 1986). The GNUphenotype indicates that there is a centrosomal com-ponent of the cell cycle that is capable of runningindependently of the nuclear division cycle.

Not only can the processes of centrosome replicationand nuclear division be uncoupled, but furthermorecentrosomes can proceed through multiple rounds ofdivision in the absence of DNA replication. This hasbeen shown by microinjecting aphidicolin, a specificinhibitor of DNA polymerase a, into syncytial wild-typeDrosophila embryos. The rounds of centrosome repli-cation correlate with cortical budding cycles that, as withuntreated embryos, spread in waves from both poles.When the buds are present at the surface of aphidicolin-injected embryos, the nuclei have decondensed chroma-tin surrounded by nuclear membranes as judged bybright annular staining with an anti-lamin antibody. Asthe buds recede, the unreplicated chromatin condensesand lamin staining becomes weak, diffuse and cytoplas-mic (Raff & Glover, 1988). There seems therefore to beno absolute requirement for the correct completion of Sphase in order for both nuclear and cytoplasmic events ofM phase to take place. This is not to say that some criticalaspect of S-phase is not completed and if, indeed,aphidicolin has its only primary effects on DNA polym-erase a, this could well be possible. Nevertheless, DNAsynthesis is dramatically inhibited and chromosome repli-cation, a major objective of the cell cycle, does not occur.These observations add further support to the hypothesisthat there are fundamental cell-cycle oscillators in manyearly embryonic systems.

Mitotic mutations with maternal and zygoticphenotypesMutations in many genes essential for mitosis showmaternal effect phenotypes, and yet the genes are alsorequired at developmental stages other than just early

embryogenesis. Females homozygous for either of themutations polo (Sunkel & Glover, 1988), or aurora(Leibowitz & Glover, unpublished data), for example,survive to adulthood to lay eggs that die during embryo-genesis, but nevertheless show some mitotic abnormali-ties in neuroblast cells during the larval stages of theirdevelopment. Immunocytological studies on POLO em-bryos reveal highly branched mitotic spindles with broadirregular poles that do not have distinct centrosomes.The centrosome-associated antigen, Bx63, is present asparticulate matter that gradually coalesces throughout theabnormal development of the embryo. Embryos fromhomozygous aurora females have normal mitotic spindlesfor the early cleavage divisions. However, in later cyclesthere is a characteristic change in the pattern of centro-some staining in the progression from anaphase totelophase. The aurora anaphase spindles are focusedupon well-defined 'dot-like' centrosomes, which developinto broad, telophase-like spindles that appear to benucleated from points around the nuclear envelope andshow weak, indistinct centrosome staining. In larvalneuroblast cells, polo alleles display a high proportion ofcircular mitotic figures, many of which are polyploid(Fig. 2). aurora also displays this phenotype when madeheterozygous with a chromosome deficient for the locus(Leibowitz & Glover, unpublished data). Nevertheless,the larvae do mature to adulthood. This is not the case forlarvae homozygous for the mutation meny-go-round(mgr), a late larval lethal mutation that also demonstratesthis phenotype (Gonzalez et al. 1988). The integrity ofmicrotubules appears to be required for these circularfigures to form as they are no longer seen if the cells aretreated with colchicine. This is supported by obser-vations of Gonzalez et al. (1988) on the phenotype oflarvae homozygous for both mgr and asp, the latter genebeing required for the integrity of the spindle (Ripoll etal. 1985). Circular figures are no longer seen in thisdouble mutant combination. One possible interpretation

Mitosis in Drosophila 141

of these mitotic figures is that they represent chromo-somes arranged around a monopolar spindle as couldoccur if centrosome division were not occurring cor-rectly. Taken together, the phenotypes of these mutationssuggest lesions affecting the centrosome. It is prudent toremain cautious as to the precise nature of the primarydefects, since aberrations in one step of the mitotic cyclecould have epistatic effects. Attempts to clone these genesare in progress and, ultimately, a molecular analysis willaid our understanding of these genes and their products.

Larval neuroblast phenotypes of mitotic mutants

In total, some 70 genes have been described that playsome role in the Drosophila cell cycle and these are listedin Table 1. The Table attempts to correlate the observedphenotypes with stages of the mitotic cycle, but it willinevitably be necessary to re-assess these crude groupingsas more alleles are analysed and in greater depth. I haveclassified the mitotic mutations into three main cat-egories: those affecting chromosome integrity; thoseaffecting chromosome condensation; and mutations thatappear to affect metaphase or anaphase and which inmany cases lead to the formation of polyploid cells.

Mutations affecting chromosome integrityGatti (1979) has carried out a cytological analysis ofchromosome integrity in a number of recombination-defective meiotic mutants and mutagen-sensitive mu-tants, and the reader is referred to his paper for a fulldescription of these phenotypes. A list of the mutationsthat show phenotypes of this general type is given inTable 1. The phenotypes of some double combinations ofmutations representative of different alleles within thisgroup have also been analysed. Synergistic sensitivity toradiation has been observed with simultaneous hemi- orhomozygosity for the DNA repair mutants mei-9 andmei-41 indicative of different, competitive pathways forDNA repair in somatic cells (Baker et al. 1978). musJOSand muslO9 each produce distinctive patterns of chromo-some breakage, and yet the combination of the twomutants suggests that musJ09 can in part substitute formitslOS (Baker et al. 1982). This approach of examiningthe phenotypes of double mutant combinations has in thepast proved invaluable for analysing interactions betweencdc mutants in yeast, but is still in its infancy in theDrosophila field. It will no doubt prove equally valuablein future studies of the interactions between mitoticmutants in Drosophila.

Mutations affecting chromosome condensationOf this second group, perhaps the best characterizedmutation is muslOl, originally identified as a mutagen-sensitive mutation. The striking feature of this mutationis that it results in abnormal condensation of hetero-chromatin but not euchromatin (Gatti et al. 19836). Theavailability of a temperature-sensitive mutant allele of thelocus permitted the onset of this abnormal chromosomecondensation to be followed after cells are shifted to therestrictive temperature. There appears to be no gross

effect of muslOl upon the replication of DNA in hetero-chromatin as judged by an autoradiographic study of[3H]thymidine incorporation, and it has been suggestedthat the effects of the mutation upon mutagen sensitivityand DNA repair are secondary consequences of theprimary effect on condensation of heterochromatin.Nevertheless, there are instances in which mutant allelesof this locus do affect DNA replication. One allele, K451,prevents the extra rounds of DNA replication that occurat the X and 3rd chromosome clusters of chorion genes,in follicle cells at a developmentally specific phase ofoogenesis (Orr et al. 1984; Snyder <?£«/. 1986). Normallythese extra rounds of DNA replication result in the 15- or60-fold amplification of these genes on the respectivechromosomes, enabling the follicle cells to undertake thesynthesis of large amounts of chorion protein for the shellof the developing egg. Whilst it is not inconceivable thateffects upon the organization of chromatin could havesecondary consequences upon DNA replication, it isprobably prudent to await further molecular characteriz-ation of this locus before drawing any conclusions aboutthe mode of action of the gene.

Gatti and co-workers have also described a number ofother mutations having a variety of effects upon chromo-some condensation (Table 1; Gatti et al. 1983a; Gatti &Baker, 1988). In some of these mutations the irregularchromosome condensation is accompanied by polyploidy,in some by chromosome breakage, whereas in othersthere appears to be no additional effect. The variety ofphenotypes in this group of mutations suggests a widerange of wild-type functions for these genes and there isas yet no indication of the primary lesion in any of themutations.

Mutations affecting metaphase or anaphaseThis remaining group covers a much broader set ofphenotypes. It is difficult to assign the effect of thesemutations to stages of the mitotic cycle, although it hasbeen suggested that three mutations lead to cells arrestingat metaphase (Gatti & Baker, 1988). The mitotic index ofcells from non-colchicine-treated cells from the mutantI(l)d.deg4, for example, is three to four times higher thancontrol cells; anaphases are rare; and 30% of the figuresare tetraploid or hyperploid. I(l)d.deg3 and l(l)d.deglOhave similar phenotypes to each other, their metaphasefigures displaying over-condensed chromosomes and splitchromatids. In both these mutants, chromosome frag-mentation is common and occurs primarily near thecentromere (Gatti & Baker, 1988). 1(3)7 m62 andl(l)d.degll result in highly polyploid nuclei. Thesemutations have normal mitotic indices, and show a highproportion of multipolar anaphases. Cells with spectacu-lar arrays of 500-1000 chromosomes can be readilyobserved, indicating that segregation can fail completelyfor six to seven successive cycles. The spindle poles ofthese structures appear to be undergoing replication inconcert with the increase in ploidy, suggesting that thetwo mutations might identify genes that function eitherdirectly in cytokinesis or in its coupling to other mitoticevents (Gatti & Baker, 1988). Cells affected by themutation 1(3)13 m281 also show increased ploidies, but in

142 D. M. Glover

Table 1. Cell cycle genes o/Drosophila

Mutation (or gene) Phenotype Reference

Interphase (small or no discs)l(l)disdess

Chromosome integritymei9mei4J; nuislO2miislOSmuslO9

l(3)MR109fs(3)820mit(I)2; mit(l)3; mit(l)7; mit(l)8; mit(l)9;

mit(l)U; mit(l)12; mit(l)13

Chromosome condensationmit(l)4; I(3)8ml2; l(3)Um254; I(3)12ml37;

1(3)1 lbl; l(3)IX-U; I(3)m45l(3)snapmuslOl1(3)1902; l(3)e20; 1(3)K43; I(3)IX-14;

mit(l)5; mit(l)14; I(3)15m25; I(3)7m75;l(3)g60A; l(3)13m230; 1(3)2004

1(3)2612; I(l)d.degl2l(l)d.deg9; I(l)d.het2; l(3)XH-10; 1(3)2004

bam

Metaphase-anaphasepoddotl(l)d.deglO; I(l)d.deg3

I(l)d.deg4aspmarmit(3)R2mit(3)R72mil(3)rl35l(l)zwlOrough dealaurora, thulepolomgr (merry go round)c204lodestarl(l)T\V-6cs;fs(3)2755braI(3)13m281l(l)d.degll; I(l)7m62

G2-Mstringcyclin Acyclin B

Required in embryogenesisgnu (giant nuclei)

No figures

Breaks without regional specificityBreaks and interchanges without regional specificityPrimarily euchromatic breaks and interchangesBreaks and interchanges preferentially located at

euchromatin/heterochromatin junctionsBreaks and interchanges without regional specificityBreaks and interchanges near the nucleolus organizers

Elevated frequency of chromosome breakage

Irregular chromosome condensation

Irregular chromosome condensationIrregular condensation of heterochromatin

Irregular chromosome condensation; chromosome breakage

Irregular chromosome condensation; polyploid cellsIrregular chromosome condensation; chromosome breakage;

and polyploid cellsIrregular condensation and breaks at metaphase

Extreme chromatin condensationOvercondensation of individual chromatidsExtremely condensed chromosomes with split chromatids;

chromosome breakage; polyploid cells; and no anaphasesPolyploid cells with few anaphasesPolyploid cells with few anaphasesMetaphase arrestMetaphase arrest, overcondensed chromatin, aneuploidyMetaphase arrest, overcondensed chromatin, polyploidyMetaphase arrest, overcondensation of chromatinAneuploid cellsAneuploid cellsBranched spindles and polyploidy (in embryos)Circular neuroblast metaphases, pole defect in embryosCircular neuroblast metaphasesNo sister chromatid apposition in heterochromatic regionsAnaphase bridges, branched spindles in embryosAnaphase bridgesChromosome breakage at anaphaseEndoreduplicatedGiant polyploid cells

Arrests in G2 after cellularizationAffects neuroblast divisions in embryogenesis

Uncontrolled DNA synthesis giving giant nuclei

f, j , k

"The four genes each for a- and ^-tubulin are not included in this table. The majority of the mutations listed in this table represent singlealleles of mutant loci. In cases where a locus is represented by several mutant alleles, either the name of the locus or the name of arepresentative allele is given.

Hlitotic phenotypes described by Gatti et al. (\983a,b) and Gatti & Baker (1988); cGatti (1979); ''Smith et al. (1985); eMitotic mutantscharacterized in our laboratory at Imperial College; f Gatti et al. (1983fl,b); BMitotic mutants from the laboratory of Ripoll in Madrid;'' Perrimon et al. (1985) and Gatti (unpublished observations); ' Edgar & O'Farrell (personal communication); ' Giniger, Vaessin & Jan (personalcommunication); kLehner & O'Farrell (personal communication).

this case as a consequence of endoreduplication. This isbelieved to result from successive rounds of DNAreplication without mitotic division, resulting in bundlesof four, eight or 16 sister chromosomes (Gatti & Baker,1988).

It is often difficult to assess the real nature of the lesionin mutants from these terminal phenotypes, which resultfrom the gradual depletion of maternal gene products.The availability of conditional lethal mutations wouldgreatly assist this problem. This has been advantageous

Mitosis in Drosophila 143

in work on the yeast cdc genes, where temperature-conditional mutants are available. These allow the rapidinactivation of the gene product within a single cycle,thus getting a step closer to the primary defect. There aresome temperature-sensitive mitotic mutations of Dros-ophila: l(l)T\V6cs, for example, is a cold-sensitive lethalthat gives anaphase bridges in the cycle following the shiftto the restrictive temperature. Most of the genes listed inTable 1 are represented by single alleles. Progress inunderstanding the functions of these genes thereforeawaits the isolation of further alleles, both conditionaland non-conditional.

Relating phenotypes to function

It is important to study a number of alleles of any locusbefore one can assess the nature of a lesion. Theavailability of chromosomes deficient for the regioncontaining a locus can be used to accentuate the pheno-type and thereby help in the understanding of thefunction of the wild-type gene product. Larvae homo-zygous for asp, for example, show an elevated mitoticindex, aneuploid cells, and highly condensed chromatin.If asp is made heterozygous with a deficiency, theresulting larvae show an increased mitotic index, withsome overcondensation of chromosomes compared withwild-type larvae, but all the cells are diploid, as ifabruptly arrested in metaphase. As mentioned above, aspis thought to affect the mitotic spindle and biochemicalstudies have shown that microtubules are more stable inmutant than in wild-type cell extracts (Ripoll et al. 1985).A polypeptide has been identified by two-dimensionalelectrophoretic analysis, which varies in concentration asa function of gene dosage of the region containing the asplocus (F. Wandosell, personal communication, 1988). Itis proposed that this protein acts to modify a secondprotein involved in spindle dynamics. More work needsto be done to confirm this model, and it will be helped bya molecular analysis.

A genetic approach will be invaluable in understandingmitosis, but it has limitations that can be overcome by theconcerted application of molecular studies. It is only amatter of time before many of these genes are cloned,sequenced, and expressed in Escherichia coli in orderthat the gene product can be used as an immunogen.Antibodies raised in this way will be powerful tools inanalysing the functions of these gene products in Dros-ophila cells.

Perhaps it is not surprising that concerted biochemicaland genetic studies have progressed furthest with themajor components of the microtubules, the tubulinmolecules themselves. Cloned DNAs of the Drosophilamulti-gene family for tubulin genes were first isolated byvirtue of their cross-homology with a chicken tubulincDNA clone. Genetic analyses, on the other hand, havebeen carried out on the major cv-tubulin gene, tubA84B(Matthews & Kaufman, 1987), and the testis-specific j32-tubulingene, Bit (Kemphuesef al. 1982, 1983; Fuller^al. 1987). The latter studies have been facilitated by themale sterile phenotype of these mutants. The /32-tubulins

encoded by the first recessive alleles at this locus wereunable to form a-^-heterodimers, resulting in the failureof chromosome movement at meiosis, axoneme formationand spermatid elongation. A second class of mutationshave been isolated encoding partially functional /32-tubulin, which can still assemble into a-^-heterodimers.One such allele has recently been shown to direct thesynthesis of /32-tubulin that assembles into aberrantmicrotubules both in vivo and in vitro. Genetic screensfor additional B2t alleles have also yielded non-comple-menting mutations that map to several different locations(Raff & Fuller, 1984). In these cases, males heterozygousfor both a B2t mutation and second-site non-comple-menting mutation are sterile, even though they have onewild-type allele for each gene. This can be explained ifthe second-site mutation produces a defective productthat can still interact with the ^-tubulin from the onewild-type gene, and so reduce the amount of functionalcomplex to one quarter of the wild-type level. This screenhas yielded a mutation in the a'-tubulin gene at 84B, andalso a series of other mutations, which most probablyrepresent genes encoding other proteins that interact with/3-tubulin. The recent analysis of one of these second sitenon-complementing mutations, haywire™"1, suggests thatthe gene encodes a protein required for microtubulefunction in a variety of ways. Males homozygous for themutation are sterile and show defects in meiosis, flagellarelongation and nuclear shaping (Regan & Fuller, 1988).This general approach will prove invaluable in establish-ing interactions between gene products.

The progression from a biochemical towards a geneticanalysis is also being made in situations where less isknown about the protein under investigation. Goldsteinand co-workers (1986), for example, have purified amicrotubule-associated protein from Drosophila cells,raised antibodies against it and used these to screenexpression libraries. In this way they have cloned thesegment of Drosophila DNA encoding this protein andlocalized it to region 100EF by in situ hybridization tosalivary gland chromosomes. A similar approach has beenused by Whitfield et al. (1988) to clone the gene encodinga centrosome-associated antigen. The next step in thesestudies is to generate mutations at these loci and so takeadvantage of Drosophila genetics. These are early days inthe study of mitosis in Drosophila. As the field develops,we will no doubt see the success of a multidisciplinaryapproach in which genetics, cell and molecular biologyare brought to bear upon this fundamental process ofeukaryotic cells.

References

BAKER, B. S., CARPENTER, A. T. C. & RIPOLL, P. (1978). Theutilisation during mitotic cell division of loci controlling meioticrecombination and disjunction in Drosophila melaiiogaster.Genetics 90, 531-578.

BAKER, B. S. & HALL, J. C. (1976). Meiotic mutants: Geneticcontrol of meiotic recombination and chromosome segregation. InThe Genetics and Biology of Drosophila, vol. la (ed. M. Asbumer& E. Novitski), pp. 352-429. London: Academic Press.

BAKER, B. S. & SMITH, D. A. (1979). The effects of mutagensensitive mutants of Drosophila melaiiogaster in nonmutagenisedcells. Genetics 92, 833-847.

144 D. M. Glover

BAKER, B. S., SMITH, D. A. & GATTI, M. (1982). Region specificeffects on chromosome integrity of mutations at essential loci inDrosophila melanogaster. Proc. natn. Acad. Sci. U.S.A. 79,1205-1209.

DABAUVALLE, M. C , DOREE, M., BRAVO, R. & KARSENTI, E.

(1988). Role of nuclear material in the early cell cycle otXenopusembryos. Cell 52, 525-533.

DUNPHY, W. G., BRIZUELA, L., BEACH, D. & NEWPORT, J. (1988).

The Xenopus cdc2 protein is a component of MPF, a cytoplasmicregulator of mitosis. Cell 54, 423-431.

EVANS, T., ROSENTHAL, E. T., YOUNGBLOOM, J., DISTEL, D. &

HUNT, T. (1983). Cyclin: a protein specified by maternal mRNAin sea urchin eggs that is destroyed at each cell division. Cell 33,389-396.

FOE, V. & ALBERTS, B. M. (1983). Studies of nuclear andcytoplasmic behaviour during the five mitotic cycles that precedegastrulation in Drosophila embryogenesis. jf. Cell Sci. 61, 31-70.

FRASCH, M., GLOVER, D. M. & SAUMWEBER, H. (1985). Nuclear

antigens follow different pathways into daughter nuclei duringmitosis in Drosophila embryos. J. Cell Sci. 82, 115-172.

FREEMAN, M. & GLOVER, D. M. (1987). The gnu mutation ofDrosophila causes inappropriate DNA synthesis in unfertilised andfertilised eggs. Genes Dev. 1, 924-930.

FREEMAN, M., NUSSLEIN-VOLHARD, C. & GLOVER, D. M. (1986).

The dissociation of nuclear and centrosomal division in gnu, amutation causing giant nuclei in Drosophila. Cell 46, 457-468.

FULLER, M. T., CAULTON, J. H., HUTCHENS, J. A., KAUFMAN, T.

C. & RAFF, E. C. (1987). Genetic analysis of microtubulestructure: A (3-tubulin mutation causes the formation of aberrantmicrotubules in vivo and in vitro. Devi Biol. 104, 385-394.

GATTI, M. (1979). Genetic control of chromosome breakage andrejoining in Drosophila melanogaster. I. Spontaneous chromosomeaberrations in X-linked mutants defective in DNA metabolism.Pmc. natn. Acad. Sci. U.S.A. 76, 1377-1381.

GATTI, M., PIMPINELLI, S., BOVE, C , BAKER, B. S., SMITH, D. A.,

CARPENTER, A. T. C. & RIPOLL, P. (1983n)- Genetic control ofmitotic cell division in Drosophila melanogaster. In Proc. AT Int.Congr. Cen. Nezv Delhi, vol. 2, pp. 193-204. Oxford & IBHPublishing: New Delhi.

GATTI, M., SMITH, D. A. & BAKER, B. S. (19836). A gene

controlling the condensation of heterochromatin in Drosophilamelanogaster. Science 221, 83-85.

GAUTIER, J., NORBURY, C , LOHKA, M., NURSE, P. & MALLER, J.

(1988). Purified maturation promoting factor contains the productof a Xenopus homologue of the fission yeast cell cycle control genecdc2+. Cell 54, 433-439.

GERHART, J., WU, M. & KIRSCHNER, M. (1984). Cell cycle dynamicsof an M-phase specific cytoplasmic factor in Xenopus laevis oocytesand eggs. J . Cell Biol. 98, 1247-1255.

GOLDSTEIN, L. S. B., LAYMON, R. A. & MCINTOSH, J. R. (1986). A

microtubule associated protein in Drosophila melanogaster.Identification, characterisation, and isolation of coding sequences.J. Cell Biol. 102, 2076-2078.

GONZALEZ, C , CASAL, J. & RIPOLL, P. (1988). Functional

monopolar spindles caused by mutation in mgr, a cell division geneof Drosophila melanogaster. J. Cell Sci. 89, 39-47.

HARA, K., TYDMAN, P. & KIRSCHNER, M. (1980). A cytoplasmic

clock with the same period as the division cycle in Xenopus eggs.Proc. natn. Acad. Sci. U.S.A. 77, 462-466.

HARTENSTEIN, V. & CAMPOS-ORTEGA, J. A. (1985). Fate mapping inwild-type Drosophila melanogaster I. The pattern of the embryoniccell divisions. Wilhetm Roux' Arch, devl Biol. 194, 181-195.

HAYLES, J. & NURSE, P. (1986). Cell cycle regulation in yeast. J . CellSci. Suppl. 4, 155-170.

KARR, T. L. & ALBERTS, B. M. (1986). Organisation of thecytoskeleton in early Drosophila embryos. J. Cell Biol. 98, 156-162.

KELLOGG, D. R., MITCHISON, T. J. & ALBERTS, B. M. (1988).

Behaviour of microtubules and actin filaments in living Drosophilaembryos. Development 103, 675-686.

KEMPHUES, K. J., KAUFMAN, T. C , RAFF, R. A. & RAFF, E. C.

(1982). The testes specific /3-tubulin subunit in Drosophilamelanogaster has multiple functions in spermatogenesis. Cell 31,655-670.

KEMPHUES, K. J., RAFF, E. C. & KAUFMAN, T. C. (1983). Genetic

analysis of B2t, the structural gene for a testes specific )3-tubulinsubunit in Drosophila melanogaster. Genetics 105, 345-356.

LEE, M. G. & NURSE, P. (1987). Complementation used to clone ahuman homologue of the fission yeast cell cycle control genecdc2+. Nature, Land. 327, 31-35.

LINDSLEY, D. L. & SANDLER, L. (1977). The genetic analysis ofmeiosis in female Drosophila melanogaster. Phil. Trans. R. Soc. B277, 295-312.

LOHKA, M. J. & MALLER, J. L. (1985). Induction of nuclearenvelope breakdown, chromosome condensation and spindleformation in cell free extracts. J . Cell Biol. 101, 518-523.

MATTHEWS, K. A. & KAUFMAN, T. C. (1987). Developmentalconsequences of mutations in the 84B a'-tubulin gene of Divsophilamelanogaster. Devi Biol 119, 100-114.

MIAKE-LYE, R., NEWPORT, J. & KIRSCHNER, M. (1983). Maturation

promoting factor induces nuclear envelope breakdown incyclohexamide-arrested embryos of Xenopus laevis. J. Cell Biol.97, 81-91.

ORR, W., KOMITOPOULOU, K. & KAFATOS, F. (1984). Mutants

suppressing in trans chorion gene amplification in Drosophila.Pmc. natn. Acad. Sci. U.S.A. 81, 3773-3777.

PERRIMON, N., ENGSTROM, L. & MAHOWALD, A. P. (1985).

Developmental genetics of the 2C-D region of the Drosophila Xchromosome. Genetics 111, 23-41.

PRINGLE, J. & HARTWELL, L. (1981). The Sacchammyces cerevisiaecell cycle. In The Molecular Biology of the Yeast Saccharomyces(ed. S. Strathern, E. Jones, & J. Broach), pp. 97-142. New York:Cold Spring Harbor Laboratory Press.

RAFF, E. C. & FULLER, M. T. (1984). Genetic analysis ofmicrotubule function in Drosophila. In Molecular Biology of theCytoskeleton (ed. G. G. Borisy, D. W. Cleveland & D. B.Murphy), pp. 293-304. New York: Cold Spring HarborLaboratory Press.

RAFF, J. W. & GLOVER, D. M. (1988). Nuclear and cytoplasmiccycles continue in Drosophila embryos in which DNA synthesis isinhibited with aphidicolin. J. Cell Biol. (in press).

REGAN, C. L. & FULLER, M. T. (1988). Interacting genes that affectmicrotubule function: the nc2 allele of the haywire locus fails tocomplement mutations in the testes-specific /3-tubulin gene ofDrosophila. Genes Dev. 2, 82-92.

RIPOLL, P., PIMPINELLI, S., VALDIVIA, M. M. & AVILA, J. (1985). A

cell division mutant of Drosophila with a functionally abnormalspindle. Cell 907-912.

ROSENTHAL, E. T., HUNT, T. & RUDERMAN, J. V. (1980). Selective

translation of mRNA controls the pattern of protein synthesisduring early development of the surf clam, Spisula solidissima.Cell 20, 487-494.

SLUDER, G., MILLER, F. J. & REIDER, C. L. (1986). The

reproduction of centrosomes: nuclear versus cytoplasmic controls.J. Cell Biol. 103, 1873-1881.

SMITH, D. A., BAKER, B. S. & GATTI, M. (1985). Mutations in genescontrolling essential mitotic functions in Divsophila melanogaster.Genetics 110, 647-670.

SNYDER, P. B., GALANOPOULOS, V. K. & KAFATOS, F. C. (1986).

Trans acting amplification mutants and other eggshell mutants ofthe third chromosome of Divsophila melanogaster. Pmc. natn.Acad. Sci. U.S.A. 83, 3341-3345.

STANDART, N., PINES, J. N., MINSHULL, J. & HUNT, T. (1987).

Cyclin synthesis, modification and destruction during meioticmaturation of the starfish oocyte. Devi Biol. 124, 248-258.

SUNKEL, C. E. & GLOVER, D. M. (1988). polo, a mitotic mutant ofDrosophila displaying abnormal spindle poles. J. Celt Sci. 89,25-38.

SWENSON, K. I., FARRELL, K. M. & RUDERMAN, J. R. (1986). The

clam embryo protein cyclin A induces entry into M phase and theresumption of meiosis in Xenopus oocytes. Cell 47, 861-870.

WARN, R. M., FLEGG, L. & WARN, A. (1987). An investigation of

microtubule organisation and functions in living Divsophilaembryos by injection of a fluorescently labeled antibody againsttyrosinated alpha tubulin.J Cell Biol. 105, 1721-1730.

Mitosis in Drosophila 145

WASSERMAN, W. & SMITH, D. (1978). The cyclic behaviour of a 467-480.cytoplasmic factor controlling nuclear envelope breakdown. J'. Cell ZALOKAR, M. & ERK, 1. (1976). Division and migration of nucleiBiol. 78, R15-R22. during early embryogenesis of Dmsopltila inelanogaster.J.

WHITFIELD, W. G. F., MILLAR, S. E., SAUMWEBER, H., FRASCH, M. microbiol. Cell 25, 97-106.& GLOVER, D. M. (1988). Cloning of a gene encoding an antigenassociated with the centrosome in Dmsophila. J. Cell Sci. 89, {Received 17 August 1988 — Accepted 17 October 1988)

146 D. M. Glover