INTRODUCTION

The differentiation of sperm from the relatively undifferenti-ated gonial cells is one of the most complex and elaborate ofdevelopmental processes. This process encompasses two typesof cell divisions: the mitotic divisions needed for the multipli-cation of gonial cells, and the meiotic divisions leading to theformation of haploid spermatids. In addition, both premeioticand postmeiotic cells undergo a series of dramatic changes incellular morphology that culminate in the differentiation of thesperm tail. All these events, collectively defined as spermato-genesis, are characterized by a remarkable degree of evolu-tionary conservation, so that insects and mammals exhibit onlya few differences in the basic morphogenetic features involvedin sperm development. Thus, an understanding of spermato-genesis in a model system should provide a conceptualframework of wide utility.

A model organism particularly suitable for the elucidationof complex developmental processes is

Drosophilamelanogaster. Mutational dissection of Drosophila embryo-genesis, followed by the molecular characterization of therelevant genes, has led to an increasing understanding of thefundamental mechanisms underlying embryonic development.Genetic analysis has also been applied to Drosophila sper-matogenesis and many mutants have been isolated that disruptspecific aspects of this process (see for example Lifschytz andHareven, 1977; Lifschytz and Meyer,1987; Hackstein, 1991).Although these studies have clearly shown that the genetic dis-section of spermatogenesis is feasible, very few genes involvedin sperm development have been characterized at the molecular

level, and our current conception of the mechanisms thatcontrol the formation of male gametes is still very poor.Recently, however, many new genes involved in Drosophilaspermatogenesis have been identified by single P elementmutagenesis (Castrillon et al., 1993; Fuller, 1993; unpublishedwork of our laboratory). The molecular analysis of some ofthese genes is thus expected to be completed in the next fewyears, providing substantial insights into the problem of spermdevelopment.

A fundamental step in the mutational dissection of sper-matogenesis is the precise definition of the phenotypic conse-quences of specific mutations at the cellular level. Theaccuracy of such definition must rely on a careful descriptionof the cellular transformations that occur during spermatogen-esis in wild-type males. D. melanogaster spermatogenesis hasbeen extensively studied at the ultrastructural level, andbeautiful descriptions of the premeiotic (Tates, 1971), meiotic(Tates, 1971; Lin et al., 1981; Church and Lin, 1982, 1985)and postmeiotic (for review see Lindsley and Tokuyasu, 1980)stages are currently available. Light microscope analyses,though more widely accessible, are unfortunately not asdetailed and complete. The classic cytological procedure forlight microscope analysis of Drosophila spermatogenesis is theobservation of living preparations in phase contrast (see forexample Lifschytz and Hareven, 1977; Castrillon et al., 1993;Fuller, 1993). The morphological differentiation of living cellsis remarkably good and most cell types and stages, includingthe meiotic divisions, can be readily recognized. However, thefact that the structures observed in living cells cannot be dis-criminated by immuno- and histochemical staining techniques,

3521

Larval and pupal testes of

Drosophila melanogaster werefixed with a methanol/acetone fixation procedure thatresults in good preservation of cell morphology; fixed cellsviewed by phase-contrast optics exhibit most of the struc-tural details that can be seen in live material. Fixed testispreparations were treated with anti-tubulin antibodies andHoechst 33258 to selectively stain microtubules and DNA.The combined analysis of cell morphology, chromatin andmicrotubule organization allowed a fine cytological dissec-tion of gonial cell multiplication, spermatocyte develop-

ment, meiosis and the early stages of spermatid differen-tiation. We placed special emphasis on the spermatocytegrowth phase and the meiotic divisions, providing adescription of these processes that is much more detailedthan those previously reported. In addition, by means ofbromo-deoxyuridine incorporation experiments, we wereable to demonstrate that premeiotic DNA synthesis occursvery early during spermatocyte growth.

Key words: Drosophila, spermatogenesis, meiosis

SUMMARY

Chromatin and microtubule organization during premeiotic, meiotic and early

postmeiotic stages of

Drosophila melanogaster spermatogenesis*

Giovanni Cenci1, Silvia Bonaccorsi1,2, Claudio Pisano1,2, Fiammetta Verni1 and Maurizio Gatti1,2

1Istituto Pasteur, Fondazione Cenci Bolognetti, and 2Centro di Genetica Evoluzionistica del CNR, Dipartimento di Genetica eBiologia Molecolare, Universita’ di Roma ‘La Sapienza’, P. A. Moro, 5. 00185 Roma, Italy

*This paper is dedicated to the memory of Fritz Sobels, in recognition of his studies on

Drosophila spermatogenesis

Journal of Cell Science 107, 3521-3534 (1994)Printed in Great Britain © The Company of Biologists Limited 1994

3522

limits the utility of this material. On the other hand, classicalfixation techniques, such as those described by Cooper (1965),allow a clear visualization of chromosomal material but not ofother cellular structures, such as the cytoskeleton and thespindle.

An ideal fixation technique would be one that preserves thecellular morphology observed in vivo yet simultaneouslyallowing staining with chemical and immunological reagents.We have developed such a technique during the course ofstudies on the Y chromosome loops (Pisano et al., 1993). Inthis report we have exploited this technique to describe thearrangement and the organization of microtubules, chromatinand mitochondria in the various cell types arising during D.melanogaster spermatogenesis.

MATERIALS AND METHODS

StocksAll the observations were made on larval or pupal testes of an Oregon-R stock that has been maintained in our laboratory for about 30 years.The flies were reared on standard Drosophila medium at 25±1°C; dis-sections and cytological procedures were performed at room temper-ature.

Fixation and staining proceduresPupal and larval testes were dissected and fixed according to Pisanoet al. (1993). Fixed preparations were washed three times (5 minuteseach) in PBS, before incubation with the primary antibody for indirectimmunofluorescence or Giemsa staining. Three different anti-tubulinantibodies were used for immunostaining: (1) an anti-

α-tubulin mon-oclonal antibody (Ab) raised against chick brain microtubules(Amersham; Blose et al., 1982) designated as A-Am; the ascites fluidcontaining this antibody was diluted 1:50 in PBS. (2) An anti-βtubulin monoclonal Ab also raised against chick brain microtubules(Amersham; Blose et al., 1982) designated as B-Am; the ascites fluidwith this antibody was diluted 1:30 in PBS. (3) The anti-α tubulinmonoclonal Ab 3A5 raised against Drosophila embryonic tubulin(Piperno and Fuller, 1985), which was diluted 1:2 in PBS. In everycase the slides were incubated for 45 minutes with 30 µl of dilutedprimary antibody in a humid chamber at room temperature. They werethen washed three times in PBS (5 minutes each) and incubated for45 minutes with the secondary antibody (sheep anti-mouse IgG,F(ab′)2 fragment, conjugated with 5(6)-carboxy-fluorescein-N-hydroxysuccinimide ester (FLUOS) from Boehringer, Mannheim, cat.no. 1214616) diluted 1:15 in PBS.

After three washes in PBS (5 minutes each) the immunostainedslides were stained with Hoechst 33258 according to Bonaccorsi etal. (1988). Giemsa staining was performed as previously describedwith 4% Giemsa (Bonaccorsi et al., 1988).

The slides were analyzed with a Zeiss III photomicroscopeequipped with a mercury light source for incident illumination.FLUOS and Hoechst 33258 fluorescence were detected using the 0.9(BP 450-490, FT 510, LP420) and the 0.1 (BP 365/11, FT 395, LP397) Zeiss filter combinations, respectively. Microphotographs weretaken with Kodak T-Max 100 film.

Bromo-deoxyuridine (BrdU) labelingDissected testes were incubated in 10 µg/ml BrdU (Sigma) dissolvedin PBS. They were then washed for at least 3 minutes in testis buffer(Kremer et al., 1986), squashed and fixed as described above. Testispreparations were incubated for 1 hour with the ready-to-useAmersham solution (code RPN 202) containing an anti-BrdU mono-clonal antibody, nuclease and 1% BSA. After washing in PBS for 5

minutes the slides were incubated with the secondary antibody (sheepanti-mouse IgG, F(ab′)2 fragment, conjugated with FLUOS fromBoehringer, Mannheim) diluted 1:15 in PBS, washed and mounted inPBS for observation.

RESULTS AND DISCUSSION

Our methanol-acetone fixation technique (Pisano et al., 1993;see Materials and Methods) results in good preservation of thecell morphology observed in live material by phase-contrastoptics. As shown in Fig. 1, most of the morphological detailsthat can be seen in living cells are also visible in fixed prepa-rations, suggesting that fixation artifacts are minimal. The onlyintracellular structures that are substantially affected byfixation are the Y chromosome loops. These filamentous struc-tures are usually faint and labile in living preparations butbecome clearly apparent after fixation (Bonaccorsi et al., 1988;Pisano et al., 1993).

We have immunostained fixed preparations from both larvaland pupal testes with three different anti-tubulin antibodies (A-Am, B-Am and 3A5; see Materials and Methods). It has beenshown that A-Am recognizes most of the Drosophila α-tubulinisoforms (Matthews et al., 1989), while B-Am reacts with boththe β1- and the testis-specific β2-tubulin (Kimble et al., 1989);the monoclonal antibody 3A5 reacts with α-tubulin (Pipernoand Fuller, 1985). These antibodies gave identical results withall the cell types of D. melanogaster spermatogenesis. Thus,we will not distinguish between them in describing the variousimmunostaining patterns.

Following tubulin immunostaining, testis preparations weretreated with Hoechst 33258 and viewed under a fluorescencemicroscope with filter combinations that detect either fluores-cein or Hoechst fluorescence. This enabled us to examinemicrotubule and chromatin organization in the same cell andto correlate the behavior of these cellular components with theother structures that can be seen under phase-contrastmicroscopy. Based on these observations we have proposed astage subdivision of spermatocyte growth, meiotic divisionsand early spermatid differentiation. Each stage is defined onlyby morphological criteria and corresponds to a facet of theprocess that can be unambiguously distinguished from both thepreceding and the successive one.

SpermatogoniaCytological analysis of the cells located at the apex of the testisis difficult because in squashed preparations these cells areusually very close to each other and very similar in morphol-ogy. Thus, we have been unable to discriminate between germ-line stem cells, primary spermatogonia and the progenitors ofthe cyst cells. However, cysts containing 2, 4, or 8 secondaryspermatogonia can be easily recognized and analyzed.Secondary spermatogonia of these cysts appear identical whenin the same phase of cell cycle, with nuclei almost entirelyoccupied by chromatin (Fig. 2). Hoechst staining permits aneasy discrimination between G1 nuclei and those that havecompleted DNA synthesis, which have roughly doubled thechromatin content (Fig. 2). In most cases, the cyst cell nucleican be readily identified because of their elongated appearancewith the chromatin often divided into a main larger mass anda smaller, more or less separated element. In addition, the

G. Cenci and others

3523Drosophila spermatogenesis

chromatin content of these nuclei is comparable to that of post-synthetic spermatogonia, indicating that cyst cells are in anextended G2 phase of the cell cycle (Fig. 2).

Interphase spermatogonia exhibit a dense cytoplasmicnetwork of microtubules that undergoes a dramatic reorgani-zation in preparation for mitosis, culminating in spindleformation. Spermatogonial mitoses are relatively rare, sug-gesting that mitotic division of these cells is very rapid. Themetaphase spindle consists of two relatively narrow andelongated hemispindles that shorten progressively duringanaphase and telophase. At telophase the hemispindles becomebroader as they flatten against the reforming nuclei; a midbodyconnecting the two daughter nuclei is often visible at this time.These structures progressively disappear as the next interphasebegins and a new cytoskeleton takes shape in the cytoplasmsurrounding each nucleus (Fig. 2).

Premeiotic DNA synthesis The four gonial divisions generate 16 primary spermatocytesthat are morphologically indistinguishable from the parentalsecondary spermatogonia in the G1 phase (we designate thisstage as S0; an early S0 stage is shown in Fig. 2). Theseprimary spermatocytes soon enter a program of growth accom-panied by a series of characteristic changes in nuclear mor-phology that result in a 25-fold increase in nuclear volume(Tates, 1971; Lindsley and Tokuyasu, 1980; Fuller, 1993).

An important question about spermatocyte growth concernsthe timing of premeiotic DNA synthesis (cf. Olivieri andOlivieri, 1965; Lindsley and Tokuyasu, 1980). Olivieri andOlivieri (1965) injected one-day-old males with tritiatedthymidine and examined their testes by autoradiography forthymidine incorporation at several post injection (p.i.) timeintervals. They detected DNA synthesis in spermatogonialcysts as early as 30 minutes p.i. In contrast, thymidine incor-poration into 16-cell primary spermatocyte cysts was firstobserved at 8 hours p.i. The same results were obtained bySato (quoted by Lindsley and Tokuyasu, 1980), using a higherspecific activity tritiated thymidine. Because only theyoungest spermatocytes were labelled at 8 hours p.i., thelabelling of these cells could result from thymidine incorpo-ration occurring during the S phase of the 8-cell-cyst sper-matogonia. Thus, these observations indicate either that pre-meiotic DNA synthesis occurs throughout the growth phaseof D. melanogaster spermatocytes, or that these cells do notincorporate tritiated thymidine for some unknown reason.

To examine the occurrence of the premeiotic DNAsynthesis dissected larval testes were incubated at 25°C inPBS containing bromo-deoxyuridine (BrdU). At differenttimes after the beginning of the incubation, testes were fixedwith our usual procedure, and treated with an anti-BrdUantibody to detect BrdU incorporation. The first clearlylabelled cysts were observed after 3 hours incubation.

Fig. 1. Comparison between living and fixedcells of Drosophila spermatogenesis. (a,b)Live (a) and fixed (b) mature primaryspermatocytes; note the Y chromosome loops.(c,d) Live (c) and fixed (d) primaryspermatocytes in late meiotic prophase I; notethe intranuclear granules produced by thedisintegration of the Y loops. (e,f) Live (e) andfixed (f) prometaphases, metaphases or earlyanaphases I; observation of unstained materialdoes not allow unambiguous identification ofthe meiotic stage. (g,h) Live (g) and fixed (h)telophases I. (i,j) Live (i) and fixed (j)secondary spermatocytes. (k,l) Live (k) andfixed (l) prometaphases, metaphases or earlyanaphases II; also in this case examination ofunstained material does not allow preciseidentification of the meiotic stage. (m,n) Live(m) and fixed (n) telophases II. (o,p) Live (o)and fixed (p) onion-stage spermatids; note theclear nuclei (n) and the phase-densenebenkerns (nk). Bar, 10 µm.

3524

Labelled nuclei were observed in 2-, 4- and 8-cell spermato-gonial cysts, as well as in 16-cell cysts containing very youngspermatocytes morphologically indistinguishable fromsecondary spermatogonia in the G2 stage (Fig. 3). Cysts con-

taining spermatocytes even slightly larger than gonial cellswere invariably unlabelled.

The spermatocyte labeling in 16-cell cysts cannot be due toan immunostaining artifact because the cyst cell nuclei, whichare in an extended G2 phase, do not react with the anti BrdUantibody (Fig. 3). Moreover, the fact that three hours are theminimum time required for labeling of both spermatogonia andspermatocytes strongly suggests that spermatocyte labeling isthe direct consequence of BrdU incorporation into these cells,and not the result of BrdU uptake during the last spermatogo-nial S phase. Thus, although we cannot explain why thetritiated thymidine experiments failed to detect premeiotic Sphase, we conclude that premeiotic DNA synthesis occurs invery young primary spermatocytes. The early occurrence ofthis synthesis may facilitate the elevated transcriptionalactivity of primary spermatocytes, which have embarked onthe production of most, if not all, mRNA needed for spermio-genesis (for review see Fuller, 1993).

Spermatocyte growthPrimary spermatocytes that have just completed DNA synthesis(stage S1, Fig. 4a) are similar in size and morphology to the G2secondary spermatogonia. As a primary spermatocyte begins itsgrowth, the nucleus progressively assumes an eccentric positionwithin the cytoplasm (stage S2, Fig. 4b,c). This so-called polarspermatocyte stage (Tates, 1971) is characterized by an asym-metrical distribution of the mitochondria within the cytoplasm,with these organelles clustered at the pole of the cell oppositeto the nucleus (Cooper, 1965; Tates, 1971). In our cytologicalpreparations polar spermatocytes exhibit a dense network ofcytoskeletal microtubules and a cloud of Hoechst-fluorescent,dotted material clustered in a cytoplasmic region opposite to thenucleus (Fig. 4b,c). These fluorescent dots most likely corre-spond to the Hoechst-stained mitochondrial DNA, thus identi-fying the cytoplasmic location of these organelles.

During the polar spermatocyte stage there is a substantialincrease in nuclear diameter accompanied by changes inchromatin organization within the nucleus. In young polarspermatocytes (stage S2a, Fig. 4b), the chromatin appears as acompact mass occupying the center of the nucleus. As the polarspermatocytes grow, chromatin subdivides into three masses

G. Cenci and others

Fig. 2. Spermatogonial cysts in different phases of the cell cycle. (a) Phase-contrast; (b) tubulin immunostaining; (c) Hoechst 33258staining. A, an 8-cell cyst in the G1 phase. B, a 4-cell cyst atmetaphase; note the metaphase spindles in b. C, a 16-cell cyst inearly G1. Tubulin immunostaining in b clearly shows that the cells ofthis cyst have just completed telophase; in one case a central spindleis still present between the two daughter nuclei. D, an 8-cell cyst inthe G2 phase. The phases of the cell cycle can be easily deducedtaking as a standard the metaphase nuclei of cyst B and the telophasenuclei of cyst C. It is evident that the nuclei of cyst A contain aboutone half the DNA of those of cyst B, whereas the nuclei of cyst Dcontain more or less the same DNA as cyst B nuclei. In cysts B andC the cyst cell nuclei (arrows) are clearly identifiable in that they arenot involved in cell division. The DNA content of these nucleiappears to be the same in both cysts, suggesting that the cyst cells arein an extended G2 phase. Only one of the two cyst cell nuclei of cystD is recognizable (large arrow), the other could be any of the nineadditional fluorescent elements of this cyst. Cyst A contains only 9nuclei and it is probably lacking one of the cyst cell nuclei; a largearrow points to the identifiable cyst cell nucleus. Bar, 10 µm.

3525Drosophila spermatogenesis

or clumps that remain closely apposed to the inner nuclearenvelope (stage S2b, Fig. 4c; see also Cooper, 1965). Thespace between these clumps increases with nuclear growth, sothat in mature spermatocytes most of the karyoplasm is notoccupied by the chromatin. Two of these chromatin clumps arelarger and denser than the other and probably correspond to thesomatically-paired large metacentric autosomes (Cooper,1965). The third chromatin mass appears to be composed by3-5 fluorescent elements that are likely to correspond to the Xchromosome, the portions of the Y chromosome not involvedin loop formation, and the tiny fourth chromosomes (Cooper,1965).

As nuclear growth continues, the nucleus progressivelyresumes a central position within the cell and the mitochondriadisperse uniformly into the cytoplasm (stage S3). This stage isalso characterized by the first appearance within the kary-oplasm of the Y chromosome loops. In D. melanogaster thereare three lampbrush-like loops, formed by the kl-5, kl-3 and ks-1 fertility factors (Bonaccorsi et al., 1988). We have previouslyshown that these structures develop asynchronously duringspermatocyte growth, with the kl-5 and ks-1 loops appearingearlier than the kl-3 loop (Bonaccorsi et al., 1988). Theyoungest apolar spermatocytes (stage S3, Fig. 4d) have acentrally located nucleus surrounded by a symmetric networkof microtubules, and exhibit two phase-dark bodies that corre-spond to the primordia of the kl-5 and ks-1 loops (Bonaccorsiet al., 1988).

Maturation of the apolar spermatocytes is characterized bya further increase in nuclear size and by the elaboration of theY loops. The kl-5 and ks-1 loops continue to expand and soonlose their initial compactness, revealing their underlying fila-mentous structure. At this stage (stage S4, Fig. 4e), a thirdcluster of filamentous material, the kl-3 loop, also becomesapparent. The three loops enlarge steadily and reach theirmaximum size in mature spermatocytes (stage S5, Fig. 4f).These cells are the largest cells produced during Drosophilaspermatogenesis. Their nuclei exhibit three fully grown Ychromosome loops that differ both in the type of thread theycontain and in their relative degree of compactness: the kl-5and ks-1 loops are more compact and contain a coarse thickthread, while the kl-3 loop consists of a thinner, less twistedand folded filament (Bonaccorsi et al., 1988). The Y loops ofmature spermatocytes occupy most of the nucleus and often

overlap, so that in many cases they cannot be distinguishedfrom each other. However, they can be easily identified byGiemsa staining at pH 10 or by immunostaining with variousantibodies that specifically decorate each of these structures(Bonaccorsi et al., 1988; Pisano et al., 1993).

The disintegration of the Y loops marks the end of thegrowth phase of primary spermatocytes and the beginning ofthe first meiotic division. When the loops begin to fall apartinto pieces (stage S6, Fig. 4g), the nuclei remain at more orless the same size as in the mature spermatocytes. However, inS6 nuclei the nucleolus has already disappeared; the chromatinhas begun to condense in preparation for meiosis but is stillclose to the inner nuclear envelope. The microtubular systemof these cells has also changed from that of mature spermato-cytes in that the microtubules are now more concentratedaround the nucleus than in the rest of the cytoplasm (Fig. 4g).

Traditionally, the spermatocyte growth phase is consideredas the meiotic prophase. However, as already observed byCooper (1965) in D. melanogaster and by Kremer andcoworkers (1986) in D. hydei, spermatocyte growth is notaccompanied by the typical progressive condensation andsynapsis of meiotic chromosomes that can be seen in themeiotic prophase of most organisms. After premeiotic DNAsynthesis the chromosomes are dispersed within the nucleusand become organized into three clumps as primary spermato-cytes enlarge. Although there is no direct evidence that eachclump contains a couple of paired homologs, this conclusionis generally accepted by students of Drosophila spermatogen-esis (for reviews see Cooper, 1965; Fuller, 1993). In addition,it has been suggested that homologous chromosome associa-tion in the clumps may be the consequence of the persistenceof ‘somatic pairing’ during gonial mitoses and throughout sper-matocyte development (Metz, 1926; Cooper, 1965; reviewedby Fuller, 1993).

Although each clump is likely to consist of a bivalent, noleptotene- or zygotene-like chromosomes can be discernedwithin these structures by aceto-orcein or Hoechst staining(Cooper, 1965; present results). Thus, the clumps represent adiffuse state of meiotic chromosomes, analogous to the stateof interphase chromatin. This decondensed state of spermato-cyte chromosomes persists until the M1 stage (see below),when chromatin condenses very rapidly, just prior to the onsetof the first meiotic metaphase. According to Cooper (1965),when the bivalents become visible by orcein staining, theyhave already attained a degree of condensation equivalent tothat of diakinetic chromosomes.

The absence of the leptotene-pachytene stages in primaryspermatocytes is probably related to the lack of synaptonemalcomplex (Meyer, 1964) and meiotic crossing over in D.melanogaster males. Interestingly, D. melanogaster females,that exhibit a normal synaptonemal complex and undergomeiotic recombination, also exhibit a canonical meioticprophase that lasts about 80 hours and comprises chromosomecondensation stages equivalent to leptotene/zygotene andpachytene (King, 1970; Carpenter, 1975; Davring and Sunner,1976, 1977). Thus, D. melanogaster possesses the geneticinformation necessary for progressive meiotic chromosomecondensation, but this information is exploited only duringfemale meiosis. Meiotic chromosome condensation in males isvery rapid and is brought about in a fashion similar to thatobserved in mitotic cells. This observation suggests the intrigu-

Fig. 3. Bromo-deoxyuridine labelling of 16-cell cysts of youngprimary spermatocytes. (a) Hoechst 33258 staining; the arrows pointto the cyst cells. (b) Bromo-deoxyuridine incorporation detected byantibody staining; note that the cyst cells are not labelled.

3526 G. Cenci and others

Fig. 4. Primary spermatocytegrowth. (Row a) A complete cyst ofyoung primary spermatocytes in theS1 stage; the arrows point to the cystcells. (Row b) Polar spermatocytesin the S2a stage; the arrows indicatethe asymmetric cap of mitochondria.(Row c) Polar spermatocytes in theS2b stage; the arrows point to theclusters of mitochondria. (Row d) Ayoung apolar spermatocyte in the S3stage; note the primordia of the kl-5(A) and ks-1 (C) Y chromosomeloops, and the three chromatinclumps. (Row e) A primaryspermatocyte in the S4 stage inwhich some filaments of the kl-3loop (B) have become apparent.(Row f) A mature primaryspermatocyte in the S5 stageshowing fully developed Ychromosome loops; note that thenucleolus (arrow) is surrounded bythe sex chromosome chromatin.(Row g) A primary spermatocyte inthe S6 stage with disintegrating Yloops. Bar, 10 µm.

Phase-contrast Tubulin Hoechst Stage

S1

S2a

S2b

S3

S4

S5

S6

3527Drosophila spermatogenesis

ing possibility that meiotic chromosome contraction in malesexploits the same genetic program that governs chromosomecondensation during mitosis.

Meiotic divisionsAs the Y loops disintegrate the nucleus shrinks and its diametereventually becomes about 20% shorter than that of maturespermatocytes (stage M1, Fig. 5a,b). A distinguishing featureof the M1 stage is the presence within the nucleus of a fewcompact granules derived from Y-loop disintegration. Themicrotubular system of these M1 cells has already completedits reorganization: the network of microtubules around thenucleus has disappeared and two prominent asters havebecome apparent. During the M1a stage (Fig. 5a), the asters,initially located on the same side of the nucleus, migrate toopposite poles, while remaining closely apposed to the outernuclear envelope. By the M1a stage, the chromatin has alreadyattained a high degree of condensation and the three majorbivalents are clearly visible. The fourth chromosome bivalentis not always clearly identifiable. However, we often observeda dotted element associated with the smallest bivalent, sug-gesting that the fourth chromosomes lie close to the sex chro-mosomes, retaining their apparent vicinity during interphase.In the M1a stage the bivalents tend to be peripherally locatednear the inner of the nuclear envelope, while in the subsequentM1b (Fig. 5b) stage they appear to move freely in the nucleo-plasm.

The M1a nuclei are round or ovoid and the cytoplasmaround them has a rather homogeneous appearance underphase-contrast microscopy. In the later M1b stage (Fig. 5b),the nuclei become irregularly shaped and are surrounded byphase-contrast dense, elongated elements. Most likely thesestructures correspond to the system of parafusorial and astralmembranes that surround the spindle and the poles during malemeiosis (Tates, 1971; Church and Lin, 1982). During the M1bstage the asters are more prominent than in the M1a stage, anda few spindle fibers radiating from the asters appear to reachthe bivalents, which occupy variable positions within thenucleoplasm. The demarcation between the nucleus and thecytoplasm progressively disappears as meiosis proceeds, whileadditional spindle fibers become connected to the bivalents,which can be seen at different distances from the spindle poles(stage M2; Fig. 5c). The M1b and M2 stages almost certainlycorrespond to the prometaphase of the first meiotic division, aphase in which the bivalents, connected to the spindle poles bykinetochore microtubules, undergo complex motions necessaryfor proper metaphase orientation (Church and Lin, 1982,1985).

Prometaphase of the first meiotic division is completedwhen the bivalents congress to a metaphase plate (stage M3;Fig. 5d). During metaphase, the bivalents cluster in a singleHoechst-bright structure equidistant from the poles and areconnected to the spindle poles by two symmetrical sets ofmicrotubules. In addition to these pole-to-chromosome micro-tubules the metaphase spindle exhibits bundles of microtubulesthat do not encounter the chromosomes. These microtubulesappear to originate from the poles, run outside the congregatedbivalents and end either in the opposite pole or distally to themetaphase plate (Fig. 5d).

Nearly all the cysts with cells in the M3 stage also containearly anaphase I cells (stage M4a; Fig. 5e). Interestingly, in

most of these early anaphases the segregating bivalents havealready migrated a considerable distance towards the poles,suggesting that initial anaphase chromosome movement is veryrapid. The spindle organization of early anaphase I cells is verysimilar to that of metaphase I cells in the M3 stage, consistingof two closely apposed umbrella-shaped half spindles.However, as anaphase proceeds the spindle microtubules reor-ganize: the density of microtubules in the region between thesegregating sets of chromosomes (central spindle) progres-sively increases, while the number of microtubules located inthe proximity of the chromosomes decreases. By mid-anaphase(stage M4b, Fig. 5f) the microtubules emanating from eachpole overlap in the central region of the spindle, so that the twoopposed umbrella-shaped half spindles of early anaphase arenow fused in a unique oval structure. By late anaphase/earlytelophase (stage M4c, Fig. 5g,h) there is a further reduction inthe number of microtubules emanating from the centrosomes,accompanied by an apparent increase of the density of centralspindle microtubules. Concomitant with this spindle reorgani-zation, there is a further increase in the distance between thetwo daughter nuclei, suggesting that the M4a and M4b-c stagescorrespond to anaphase A and anaphase B, respectively.

The dense bundle of microtubules in the central spindleformed during stage M4b-c is progressively squeezed andeventually attains an hourglass shape (telophase I, stage M5,Fig. 5i). In many cases, and more frequently in broken cysts,these hourglass-shaped spindles are bent at the mid-body.During telophase I (stage M5, Fig. 5i), the two daughter nucleibecome clearly demarcated from the cytoplasm and oftenexhibit a dark dot, which stands out from the phase-contrastclear nucleoplasm. In favourable preparations these ‘roundbodies’ (Grond, 1984) can also be seen in spermatid nuclei, butthey disappear soon after the initiation of spermatid elongation(Lindsley and Tokuyasu, 1980). The nature and the biologicalrole of these intranuclear structures, also called the proteinbodies (Kremer et al., 1986) or the nucleolus-like bodies(Lindsley and Tokuyasu, 1980), are still largely unknown(Tates, 1971; Lindsley and Tokuyasu, 1980; Grond, 1984;Kremer et al., 1986).

As meiosis proceeds, telophase I nuclei enlarge and thesecond division asters become apparent (stages M6a and M6b;Fig. 6). Actually centrosome separation occurs earlier andsometimes two distinct centrosomes are already visible in lateanaphase I (stage M4c; see Fig. 5h). These centrosomes, whichcontain a single centriole (Tates, 1971), progressively nucleatethe aster microtubules and migrate to the cell poles. Intelophases with hourglass-shaped spindles, two close foci ofmicrotubule nucleation can often be seen. As the central spindleprogressively disappears, these initial asters become moreprominent and move to opposite poles, while remaining closelyapposed to the outer nuclear envelope (stage M6a, not shown).When the asters have completed their migration to the poles,secondary spermatocyte nuclei have considerably enlarged andtheir chromatin appears to be relatively decondensed,occupying the center of the nucleus (stage M6b, Fig. 6a). TheM6b cells are relatively rare, indicating that this stage, whichprobably corresponds to interphase II, has a short duration.

As the chromosomes condense, in preparation for meiosisII, three Hoechst-bright elements become apparent (stage M7,Fig. 6b). These elements (the two major autosomes and the Xor the Y chromosome) in many cases lie very close to each

3528

other due to the small size of secondary spermatocyte nuclei,so that only one or two Hoechst-bright spots can be observed.Soon after the chromosomes have condensed, the cytoplasmic

nuclear demarcation disappears and additional spindle fiberspenetrate the nucleus, while the chromosomes are found atvariable distances from the spindle poles (stage M8; Fig. 6b).

G. Cenci and others

Fig. 5(a-d). The first meiotic division in D. melanogaster males. (Row a) Late meiotic prophase nuclei in the M1a stage; note the prominentasters and the intranuclear granules resulting from disintegration of the Y loops. (Row b) Prometaphase nuclei in the M1b stage; note theparafusorial and astral membranes surrounding these nuclei. (Row c) A prometaphase nucleus in the M2 stage in which the nuclear-cytoplasmic demarcation is no longer visible. (Row d) A metaphase I cell with congregated bivalents (M3 stage).

Phase-contrast Tubulin Hoechst Stage

M1a

M1b

M2

M3

3529Drosophila spermatogenesis

These features suggest that during the M7 and M8 stages chro-mosomes undergo the prometaphase motions necessary forproper orientation and congression.

The subsequent stages of the second meiotic division areanalogous to those of the first meiotic division. Thus, chromo-

somes eventually congregate into a metaphase plate (stage M9,Fig. 6c), while the density of both kinetochore and polar micro-tubules increases. During early anaphase II (stage M10a, Fig.6d), the chromosomes move rapidly towards the poles and thespindle retains a metaphase-like configuration. Later in

Fig. 5(e-i). (Row e) A metaphase I cell in the M3 stage (top) plus an early anaphase (bottom) in the M4a stage. (Row f) A mid-anaphase in theM4b stage. (Row g) Two late anaphases in the M4c stage; note the prominent central spindle. (Row h) A late anaphase in the M4c stageshowing two distinct centrosomes at each pole. (Row i) A telophase in the M5 stage showing a bent central spindle and the forming secondaryspermatocyte asters. In all the panels, Hoechst staining detects both nuclear and mitochondrial DNA; note the characteristic arrangement ofmitochondria along the central spindle. Bar, 10 µm.

Phase-contrast Tubulin Hoechst Stage

M3

M4a

M4b

M4c

M4c

M5

3530 G. Cenci and others

Fig. 6. The secondmeiotic division in D.melanogaster males.(Row a) A secondaryspermatocyte nucleus inthe M6b stage. (Row b)Two prometaphasesecondary spermatocytenuclei in the M7 stage(top) plus aprometaphase nucleus inthe M8 stage (bottom).(Row c) A metaphase IIin the M9 stage. (Row d)An early anaphase in theM10a stage. (Row e) Amid-anaphase in theM10b stage. (Row f) Alate anaphase in theM10c stage. (Row g)Two telophases in theM11 stage. (Row h) Atelophase in the M11stage with a bent spindle.Note that in c-h Hoechststaining reveals thearrangement ofmitochondria. Bar,10 µm.

Phase-contrast Tubulin Hoechst Stage

M7

M8

M6b

M10a

M9

M10c

M10b

M11

M11

3531Drosophila spermatogenesis

anaphase, the daughter nuclei continue to move apart, whilethe density of fibers in the central spindle progressivelyincreases (stages M10b, Fig. 6e; M10c, Fig. 6f). As for meiosisI, we propose that stages M10a and M10b-c correspond toanaphase A and B, respectively.

During telophase II (stage M11, Fig. 6g,h), the centralspindle is progressively squeezed, attaining an hourglassshape, while the two daughter nuclei become demarcated fromthe cytoplasm. Many telophase II spindles are bent at an anglejust as telophase I spindles. However, while most anaphasespindles of broken cysts are bent, spindle bending does notoccur, or is much less pronounced, in complete cysts.

Our observations have shown that meiotic nuclei undergocyclic changes in the nuclear-cytoplasmic demarcation. Lateprophase/early prometaphase spermatocytes in the M1 stageexhibit a clear demarcation between the nucleus and thecytoplasm. This demarcation disappears as the cells enter theM2 stage; the absence of a nuclear-cytoplasmic boundarypersists through metaphase and anaphase I (stages M3+M4)until telophase I (stage M5), when a sharply outlined nucleusbecomes visible again. Nuclear demarcation is maintainedduring interphase, prophase and early prometaphase II (stagesM6 and M7), and disappears again when secondary spermato-cytes enter late prophase II (stage M8), to reappear at telophaseII (stage M11). These cyclic changes in nuclear appearancecannot be due to the breakdown of the nuclear membrane,which remains largely intact during male meiosis (Tates,1971). Instead, they most likely reflect transformations of othercomponents of the nuclear envelope such as the nuclearlamina. This suggestion is supported by recent data of White-Cooper and coworkers (1993), who analyzed lamin immunos-taining during Drosophila spermatogenesis, and showed thatpremeiotic primary spermatocytes have a prominent nuclearlamina that breaks down in cells undergoing meiosis.

Figs 5 and 6 clearly show that tubulin immunostainingdetects a prominent fibrous structure that connects the twodaughter nuclei during both the first and the second meioticana-telophase. This structure is recognized by three differentanti-tubulin antibodies, which also react with the metaphasespindle and the cytoskeleton. Thus, this bundle of fibrousmaterial is undoubtely a bundle of spindle microtubules.However, in a recent cytological analysis of the first meioticdivision of Drosophila males, Casal and coworkers (1990)failed to detect a central spindle by tubulin immunostaining. Inthese studies dissected testes were treated with taxol and fixedwith formaldehyde, after an ethanol prefixation; the prepara-tions were then stained by indirect immunofluorescence witheither an anti-α tubulin antibody (Sera Lab, AccurateChemicals, Westbury/NY, YL1/2; Kilmartin et al., 1982) or thesame anti-β tubulin antibody employed here (Amersham;Blose et al., 1982; see Materials and Methods). Using thismethod they were able to detect metaphase and early anaphasespindles but not ana-telophase central spindles. The reason forthese results is currently unclear. Their fixation proceduremight have selectively disrupted the central spindle of ana-telophases. Alternatively, and more likely, the taxol treatmentmight have interfered with spindle reorganization duringanaphase, preventing the formation of the central spindle.

The central spindles of the first and the second meioticdivisions exhibit a remarkably similar behavior during ana-telophase. At mid-anaphase I (stage M4b) or II (stage M10b)

the central spindle shows an extended zone of microtubuleoverlap at its equator. However, while the anaphase spindleelongates, this apparent overlap zone does not shorten as seenin organisms such as diatoms, where spindle elongation occursby microtubule sliding in the overlap zone (for review seeCande, 1989). This observation suggests that anaphase Bspindle elongation in Drosophila male meiosis is broughtabout by mechanisms that do not involve microtubule slidingin the central spindle. Alternatively, if such a sliding doesoccur, it must be accompanied by microtubule growth, so thatthe extension of the overlap zone would not decrease as thespindle poles move apart.

An interesting feature of male meiosis is the precise partitionof mitochondria between the daughter cells at each meioticdivision. Electron microscopy studies have shown that atprometaphase mitochondria aggregate around the nucleus,lining up parallel to the spindle axis. This arrangement persiststhrough telophase until the occurrence of cytokinesis, thusmediating precise partition of these organelles (Tates, 1971;reviewed by Fuller, 1993). Due to their DNA content, mito-chondria can be also identified by light microscopy infavourable Hoechst-stained preparations (Figs 5 and 6). In latemeiotic prophase and early prometaphase I (stage MI) mito-chondria appear to be uniformly distributed within thecytoplasm. However, as cells enter late prometaphase I (stageM2) mitochondria aggregate around the equator of the nucleus.This arrangement persists through metaphase, early and midanaphase up until the formation of the central spindle. Thenmitochondria line up along this structure and remain associ-ated with it until the occurrence of cytokinesis, which dividesthem into two equal groups (Fig. 5). The same mitochondrialbehavior is observed during meiosis II (Fig. 6), so that the fourmeiotic products receive the same amount of mitochondria.

The timing of meiotic divisionPrevious cinematographic studies on live cells have shown thatabout 72 minutes elapse from the time that primary spermato-cytes enter the M1 stage to the time when anaphase begins(Church and Lin, 1985). M1 cells, selected on the basis of theircharacteristic morphology, were filmed for 1.5 hours, allowingprecise determination of the timing of the initial meiotic events(Church and Lin, 1985). Within 30 minutes from their identi-fication M1 cells undergo several morphological changes thatculminate in the appearance of chromosomes; this indicatesthat a maximum of 30 minutes elapses between the formationof M1 cells and the appearance of the bivalents.

From the time that chromosomes become visible to thebeginning of anaphase I there are about 42 minutes. During thefirst 20 minutes of this interval the bivalents undergo a seriesof complex saltatory prometaphase motions; in the subsequent22 minutes they exhibit a persistent bipolar orientation that ismaintained until the onset of anaphase (Church and Lin, 1985).

We have determined the frequencies of the various meioticstages in fixed testis preparations. The frequency of each stageshould be proportional to its duration in live material. Thus, ifone ascertains the frequency and the duration in vivo of a givenmeiotic stage, the duration of all the other stages can be easilycalculated from their relative frequencies. Using these criteriawe have estimated the lengths of the meiotic phases (Table 1);calculations were made by grouping together stages that areeasily distinguishable from both the preceding and the

3532

following ones, and by assuming that stages M1-M3 lastapproximately 72 minutes (Church and Lin, 1985). Forexample, we have grouped stages M1a and M1b, which includelate prophase and early prometaphase I, and are characterizedby the presence of a clear nuclear-cytoplasmic demarcation;this demarcation is absent in the next group of stages (M2 andM3), that correspond to late prometaphase and metaphase I.Similarly, stages M6-M7, that comprise interphase II and earlyprometaphase II, exhibit a clear nuclear-cytoplasmic demarca-

tion that disappears in late prometaphase and metaphase II(stages M8 and M9). Since the relative frequency of stages M1-M3 is 27.1%, and their duration in vivo is 72 minutes, the totalduration of meiosis is expected to be 266 minutes (72×100/27).Thus, the lengths of stages M1a+b, M2+M3, M4+M5,M6+M7, M8+M9 and M10+M11 should be 44, 28, 61, 47, 26and 60 minutes, respectively (Table 1). Although these timesare only rough estimations, they clearly indicate that the firstand the second meiotic divisions have approximately the same

G. Cenci and others

Fig. 7. Spermatid differentiation in D. melanogaster. (Row a) Partial cyst with spermatids at various differentiation stages. 1, stage T1; 2, stageT2; 3, stage T3 with nuclei smaller than nebenkerns. (Row b) Onion stage spermatids (stage T4); note the characteristically dense cytoskeleton.Bar, 10 µm.

Phase-contrast Tubulin Hoechst

Table 1. The timing of meiotic cell division in D. melanogaster malesPhases and stages of meiosis

L. Prophase I Prometa. I Anaphase I Interphase II Prometa. II Anaphase II TotalE. Prometa. I Metaphase I Telophase I E. Prometa. II Metaphase II Telophase II

(M1a,b) ( M2; M3) (M4a,b,c; M5) (M6a,b; M7) (M8; M9) (M10a,b,c; M11)

Number of cells 202 128 280 214 117 274 1,215Relative frequencies 16.6 10.5 23.1 17.6 9.6 22.6 100Estimated duration 44 28 61 47 26 60 266(minutes)

The relative frequencies of the various stages were obtained by scoring 40 late larval/early pupal testes.E, early; L, late; prometa, prometaphase.

3533Drosophila spermatogenesis

durations (M1-M5=44+28+61=133minutes; M6-M11=47+26+60=133 minutes). The two meiotic divisions also appear tohave very similar durations in certain plant species (reviewedby Van’t Hof, 1974).

The timing of D. melanogaster male meiosis has been alsoestimated by time-lapse cinematography of living pupal testes(Cross and Shellenbarger, 1979). This analysis showed thatbetween the stage of ‘elongated nuclei’ during the first meioticdivision and the stage defined as ‘toward the end of the seconddivision’ there is a time interval of about 130 minutes. The‘elongated nuclei’ probably correspond to cells in mid-anaphase I (stage M4b-c), while the ‘end of the seconddivision’ is likely to correspond to telophase II (stage M11).We have calculated that the time elapsing between thebeginning of anaphase I (stage M4a) and the end of telophaseII (stage M11) is 194 minutes. Since the interval considered byCross and Shellenbarger is obviously more restricted than ours,the two estimations appear to be comparable.

Spermatid differentiationIn telophase II cells, the mitochondria associated with thecentral spindle move towards the nuclei, forming an irregularmass on one side of the nucleus (stage T1; Fig. 7a). This massquickly assumes a crescent-like shape (stage T2; Fig. 7a),which then results in a spherical shape (stage T3; Fig. 7a).Spermatids in the T3 stage consist of a spherical mitochondr-ial aggregate associated with a smaller, spherical nucleus. Asspermatid differentiation proceeds, the nucleus enlarges andmitochondria progressively fuse to form a complex organelle,called the nebenkern, containing multiple layers of interleavedmitochondrial membranes (Tates, 1971; Tokuyasu, 1975).Because cross sections of this mitochondrial derivativeresemble an onion, this stage of spermatid development hasbeen named the onion stage (Bowen, 1922; Tates, 1971;Tokuyasu, 1975). Light microscopy and Hoechst staining donot allow discrimination between T3 mitochondrial aggregatesand onion-stage nebenkerns. They both appear as phase-dense,round structures that exhibit a dotted Hoechst fluorescence, dueto the presence of mitochondrial DNA. However, onion stagespermatids (stage T4; Fig. 7b) can be easily distinguished fromthose in the T3 stage because they have a nucleus of the samesize or slightly larger than the nebenkern, and a densecytoskeleton with characteristic bundles of microtubules.

The onion-stage spermatid nuclei contain a very compactchromatin mass that can be easily visualized after both Hoechstand Giemsa staining (Figs 7b and 8a). When the elongationprocess begins, the nucleus remains spherical, while thenebenkern assumes an oval shape (stage T5, Fig. 8b). At stage

T5 the nuclear chromatin loses its previous compactness andbecomes more diffuse within the nucleoplasm. However, aselongation proceeds, chromatin condenses again, forming adense mass flattened against the fenestrated side of the nucleusunderlying the elongating axoneme (Fig. 8c). This extra cycleof chromatin decondensation and recondensation has beenalready described by light microscopy in D. hydei (Kremer etal., 1986). Although the precise biological meaning of this phe-nomenon is currently unclear, it has been correlated with themolecular reorganization of the chromatin that occurs inspermatid nuclei (Das et al., 1964; Hauschtek-Jungen andHartl, 1982).

We are grateful to Minx Fuller for her generous gift of the 3A5antibody, and to Mike Goldberg and Byron Williams for criticalreadings of the manuscript. This work has been in part supported bya grant from Progetto Finalizzato Ingegneria Genetica.

REFERENCES

Blose, S. H., Meltzer, D. I. and Feramisco, J. R. (1982). 10 nm filamentsinduced to collapse in cells microinjected with antibodies against tubulin. J.Cell Biol. 95, 229a.

Bonaccorsi, S., Pisano, C., Puoti, F. and Gatti, M. (1988). Y chromosomeloops in Drosophila melanogaster. Genetics 120, 1015-1034.

Bowen, R. H. (1922). Studies on insect spermatogenesis. III. On the structureof the nebenkern in the insect spermatid and the origin of the nebenkernpatterns. Biol. Bull. 42, 53-85.

Cande, W. Z. (1989). Mitosis in vitro. In Mitosis. Molecules and Mechanisms(ed. J. S. Hyams and B. R. Brinkley), pp. 303-326. Academic Press, SanDiego.

Carpenter, A. T. C. (1975). Electron microscopy of meiosis in Drosophilamelanogaster females. I. Structure, arrangement, and temporal change of thesynaptonemal complex in wild type. Chromosoma 51, 157-182.

Casal, J., Gonzalez, C. and Ripoll, P. (1990). Spindles and centrosomesduring male meiosis in Drosophila melanogaster. Eur. J. Cell Biol. 51, 38-44.

Castrillon, D. H., Gonczy, P., Alexander, S., Rawson, R., Eberhart, C. G.,Viswanathan, S., DiNardo, S. and Wasserman, S. A. (1993). Toward amolecular genetic analysis of spermatogenesis in Drosophila melanogaster:characterization of male-sterile mutants generated by single P elementmutagenesis. Genetics 135, 489-505.

Church, K. and Lin, H. P. P. (1982). Meiosis in Drosophila melanogaster.The prometaphase-I kinetochore microtubule bundle and kinetochoreorientation in males. J. Cell Biol. 93, 365-373.

Church, K. and Lin, H. P. P. (1985). Kinetochore microtubules andchromosome movement during prometaphase in Drosophila melanogasterspermatocytes studied in life and with the electron microscope. Chromosoma92, 273-282.

Cooper, K. W. (1965). Normal spermatogenesis in Drosophila. In Biology ofDrosophila (ed. M. Demerec), pp. 1-61. Hafner Publishing, New York.

Cross, D. P. and Shellenbarger D. L. (1979). The dynamics of Drosophilamelanogaster spermatogenesis in in vitro cultures. J. Embryol. Exp. Morph.53, 345-351.

Fig. 8. Chromatin decondensation andrecondensation during early stages ofspermatid differentiation. Spermatidpreparations are stained with Giemsa atpH 10; note the intense staining of thechromatin and the pale staining of thenebenkerns. (a) Onion-stage spermatids(stage T4) showing a very compactchromatin mass. (b) Spermatids with ovalnebenkerns (stage T5) and relatively

diffuse chromatin. (c) Spermatids with elongated mitochondrial derivatives showing a dense chromatin mass flattened against the side of thenucleus underlying the growing axoneme. Bar, 10 µm.

3534

Das, C. C., Kaufman, B. P. and Gay, H. (1964). Histone-protein transition inDrosophila melanogaster. Exp. Cell Res. 35, 507-514.

Davring, L. and Sunner, M. (1976). Early prophase in female meiosis ofDrosophila melanogaster. Further studies. Hereditas 82, 129-131.

Davring, L. and Sunner, M. (1977). Late prophase and first metaphase infemale meiosis of Drosophila melanogaster. Hereditas 85, 25-32.

Fuller, M. (1993). Spermatogenesis. In Development of Drosophila (ed. A.Martinez-Arias and M. Bate), pp. 71-147. Cold Spring Harbor LaboratoryPress, New York.

Grond, C. J. (1984). Spermatogenesis in Drosophila. Ph. D. thesis. Universityof Nijmegen.

Hackstein, J. H. P. (1991). Spermatogenesis in Drosophila. A geneticapproach to cellular and subcellular differentiation. Eur. J. Cell Biol. 56, 151-169.

Hauschtek-Jungen, E. and Hartl, D. L. (1982). Defective histone transitionduring spermiogenesis in heterozygous segregation distorter males ofDrosophila melanogaster. Genetics 101, 57-69.

Kilmartin, J. V., Wright, B. and Milstein, C. (1982). Rat monoclonalantitubulin antibodies derived by using a new nonsecreting rat cell line. J.Cell Biol. 93, 576-582.

Kimble, H., Incardona, J. P. and Raff, E. C. (1989). A variant of β-tubulinisoform (β3) is expressed primarily in tissues of mesodermal origin inembryos and pupae, and is utilized in populations of transient microtubules.Dev. Biol. 131, 507-521.

King, R. C. (1970). Ovarian Development in Drosophila melanogaster.Academic Press, New York.

Kremer, H., Hennig, W. and Dijkhof, R. (1986). Chromatin organization inthe male germ-line of Drosophila hydei. Chromosoma 94, 147-161.

Lifschytz, E. and Hareven, D. (1977). Gene expressions and the control ofspermatid morphogenesis in Drosophila melanogaster. Dev. Biol. 58, 276-294.

Lifschytz, E. and Meyer, G. F. (1977). Characterization of male meiotic-sterile mutations in Drosophila melanogaster. The genetic control of meioticdivisions and gametogenesis. Chromosoma 64, 371-392.

Lin, H. P. P., Ault, J. G. and Church, K. (1981). Meiosis in Drosophilamelanogaster. I. Chromosome identification and kinetochore microtubulenumbers during the first and second meiotic divisions in males. Chromosoma83, 507-521.

Lindsley, D. L. and Tokuyasu, K. T. (1980). Spermatogenesis. In TheGenetics and Biology of Drosophila, vol. 2b (ed. M. Ashburner and T. R. F.Wright), pp. 225-294. London: Academic Press.

Matthews, K. A., Miller, F. B. D. and Kaufman, T. C. (1989).Developmental distribution of RNA and protein products of the Drosophilaα-tubulin gene family. Dev. Biol. 132, 45-61.

Metz, C. W. (1926). Observations on spermatogenesis in Drosophila. Z.Zellforsh. Mikroskop. Anat. 4, 1-28.

Meyer, G. F. (1964). A possible correlation between the submicroscopicstructure of meiotic chromosomes and crossing over. In Proc. 3rd EuropeanRegional Conference of Electron Microscopy. Pp. 461-462. CzechoslovakAcademy of Sciences, Prague.

Olivieri, G. and Olivieri, A. (1965). Autoradiographic study of nucleic acidsynthesis during spermatogenesis in Drosophila melanogaster. Mutat. Res.2, 366-380.

Piperno, G. and Fuller, M. T. (1985). Monoclonal antibodies specific for anacetylated form of α-tubulin recognize the antigen in cilia and flagella from avariety of organisms. J. Cell Biol. 101, 2085-2094.

Pisano, C., Bonaccorsi, S. and Gatti, M. (1993). The kl-3 loop of the Ychromosome of Drosophila melanogaster binds a tektin-like protein.Genetics 133, 569-579.

Tates, A. D. (1971). Cytodifferentiation during spermatogenesis in Drosophilamelanogaster: an electron microscope study. Ph.D. thesis. Rijksuniversiteit,Leiden. (Tates thesis has been extensively reviewed by Fuller, 1993).

Tokuyasu, K. T. (1975). Dynamics of spermiogenesis in Drosophilamelanogaster: VI. Significance of ‘onion’ nebenkern formation. J.Ultrastruct. Res. 58, 96-107.

Van’t Hof, J. (1974). The duration of chromosomal DNA synthesis, of themitotic cycle and of meiosis of higher plants. In Handbook of Genetics, vol. 2(ed. Robert C. King), pp. 363-377. Plenum Press, New York and London.

White-Cooper, H., Alphey, L. and Glover, D. M. (1993). The cdc25homologue twine is required for only some aspects of the entry into meiosisin Drosophila. J. Cell Sci. 106, 1035-1044.

(Received 4 July 1994 - Accepted 26 August 1994)

G. Cenci and others