MICROSCOPY RESEARCH AND TECHNIQUE 35:294-306 (1996)

Electron Microscopic Analysis of Drosophila Midline Glia During Embryogenesis and Larval Development Using p-Galactosidase Expression as Endogenous Cell Marker ANGELIKA STOLLEWERK, CHRISTIAN KLAMBT, AND RAFAEL CANTERA Institut fur Entwicklungsbiologie, Universitat zu Koln, 0-50923 Koln, Germany (AS., C.K.); Department of Zoology, University of Stockholm, S-106 91 Stockholm (R.C.)

KEY WORDS

ABSTRACT To thoroughly study developmental problems it is often desirable to identify spe- cific cells at the resolution of the electron microscope (TEM). Specific antibodies, and immunogold and other antibody labelling techniques can be successfully used with the TEM. But for these techniques to be successful there must be substantial adjustments for each antibody and tissue analyzed. To develop a more generally applicable labelling method we took advantage of the enhancer trap technique in Drosophila. Enhancer trap fly strains show cell- and/or tissue-specific P-galactosidase expression which can be visualized by a simple X-gal staining procedure. To com- bine the power of the enhancer trap approach with electron microscopy, we have improved the fixation and staining conditions, which allow detection of X-gal crystals (by TEM) and thus provide precise information on ultrastriictural morphology. We have tested our technique using the well- known midline glial cells and examined these cells between late embryonic and pupal develop- mental stages. The four embryonic midline glial cells found in each neuromere reside ventrally and dorsally to the midline of the neuropile and are closely associated with unpaired neurons, major commissures, and other types of glial cells. During larval and pupal life dramatic cell growth and endomitotic nuclear replication occur in midline glial cells. By the end of larval life, the giant midline glial cells fragment to give rise to a variable number of small midline glial cells. Here we show that the combination of transmission electron microscopy with cytochemical detection of p-galactosidase expression represents a promising and valuable tool for the study of the morphology and development of specific cell types.

Drosophila melanogaster, Nervous system, Glial cells, Enhancer trap

o 1996 Wiley-Liss, Inc.

INTRODUCTION The neuropile of the ventral nerve cord of embryos of

the fruit fly Drosophila melanogaster is organized into a scaffold-like form resembling a ladder. Each of the three thoracic neuromeres as well as the first eight abdominal neuromeres is characterized by two commis- sures crossing the ventral midline of the nervous sys- tem. Furthermore, all neuromeres of the embryo are interconnected by axonal tracts extending along the lateral longitudinal connectives. The formation of this precise and stereotyped axon pattern has a close rela- tionship with a number of specific glial cells (Jacobs and Goodman, 1989a,b). The importance of glia in neu- ronal development is not unique to invertebrates. Ex- tensive studies have shown the role of glial and other nonneuronal cells in the development of the vertebrate nervous system, both as a permissive and sometimes as an active substrate for migrating neurons and extend- ing growth cones (e.g., Rakic, 1971; Silver et al., 1982; Singer et al., 1979).

In Drosophila the diversity of glial cells has been disclosed with help of the enhancer trap methodology (Ito et al., in press; Klambt and Goodman, 1991). The introduction of this new lineage marker technique in Drosophila (Bellen et al., 1990; Bier et al., 1990; O'Kane et al., 1987) has lead to the isolation of a num-

ber of fly strains which specifically express P-galactosi- dase in a limited number of cells. These lines can sub- sequently serve as tools to specifically mark individual glial cells, which will make it possible to follow the development of these cells in wild-type as well as in mutant backgrounds (Choi and Benzer, 1994; Gi- angrande et al., 1993; Jacobs, 1993; Klaes et al., 1994; Klambt and Goodman, 1991; Klambt et al., 1991; Win- berg et al., 1993).

For some of these glial cells functional aspects are being clarified and detailed knowledge is accumulating about their development. One of the best studied ex- amples is the midline glial cells found in the ventral midline of the embryonic nerve cord of Drosophila. The midline of the developing nerve cord of the Drosophila embryo consists of only about 26 cells per neuromere. Among these are two pairs of midline glial cells and the ventral unpaired median (VUM) neurons, which are thought to have important functions during commis- sure formation (Bossing and Technau, 1994; Jacobs and Goodman, 1989a; Klambt et al., 1991). Once the two segmental commissures have been established, two

Received February 1, 1995; accepted in revised form March 1, 1995. Address reprint requests to Christian Klambt, Institut fur Entwicklungsbio-

logie, Universitat zu Koln. D-50923 Koln, Germany.

0 1996 WILEY-LISS, INC.

POSTEMBRYONIC MIDLINE DEVELOPMENT 295

of the four midline glial cells start to migrate along cell processes of the VUM neurons to separate anterior and posterior commissures. To date a number of mutations are known which specifically affect the differentiation of the midline glial cells (Bier et al., 1990; Klaes et al., 1994; Klambt, 1993, Klambt et al., 1991; Kolodkin et al., 1994; Raz and Shilo, 1992; Rutledge et al., 1992).

To efficiently exploit the advantages of Drosophila in terms of its sophisticated genetics, it is important to have an appreciable knowledge of the morphology and fine structure of the diverse types of glia and neurons during all stages of development. Here we present a method which exploits the use of the growing number of enhancer trap lines which show a tissue- or, in many instances, even cell-specific expression of the p-galac- tosidase reporter protein. An improved staining pro- tocol is presented here which allows for the cellular detection of p-galactosidase and yields good ultrastruc- tural preservation. This method can be applied to all enhancer trap marker strains available. Here we dem- onstrate the usefulness of this approach in the analysis of Drosophila midline glial cells during embryonic and postembryonic stages. The present study reveals a pre- viously unknown separation of both segmental com- missures into three major compartments during em- bryonic and larval stages. In addition, we describe endomitotic DNA replication in the midline glial cells during larval development, which appears to precede a nuclear division at the end of larval development.

MATERIALS AND METHODS Fly Strains

The enhancer trap lines, 3.920 and 3.400, direct ex- pression of a tau-p-galactosidase fusion protein in the midline glial cells (Callahan and Thomas, 1994). p-ga- lactosidase activity in embryos and larval stages of these enhancer trap lines is found in the entire cell cytoplasm.

X-Gal Staining of Embryonic Stages Embryos were collected overnight on apple juice

plates. The chorion was removed chemically with 5% commercial bleach. After prefixation in equal volumes of 18.5% glutaraldehyde in 0.1 M phosphate buffer, pH 7.0, and n-heptane for 20 min, the vitelline membrane was removed mechanically. The embryos were then transferred into the staining solution [ lo mM phos- phate buffer, pH 7.2; 150 mM NaCl; 1 mM MgC1,; 3.1 mM K4(FeII(CN)6); 3.1 mM K3(FeIII(CN)6)3 and were incubated overnight a t 14°C. Following the staining reaction embryos were washed in 0.1 M phosphate buffer, pH 7.0, two times, 5 min each, and staged ac- cording to Campos-Ortega and Hartenstein (1985). Af- ter staging, the embryos were incubated in a solution of 2% acrolein and 4% glutaraldehyde (Henrikson and Vaughn, 1974) in 0.1 M phosphate buffer, pH 7.0, a t room temperature. The acrolein treatment has to fol- low the P-galactosidase activity since it destroys the enzymatic activities. Following simultaneous fixation with glutaraldehyde and Os04 (Franke et al., 19691, the embryos were stained en bloc with 2% uranyl ace- tate, dehydrated through an ascending ethanol series in 5 min steps, and transferred in Araldite over propy-

lene oxide. Polymerization was done at 60°C for 48 h. Best contrast for electron microscopy (EM) was at- tained by staining the sections with 2% uranyl acetate and saturated lead hydroxide at room temperature for 15 min.

X-Gal Staining of Larval, Pupal, and Adult Tissues

The nerve cord was dissected in phosphate-buffered saline (0.1 M, pH 7.2) and thereafter fixed for 30 min in ice-cold 1.542% glutaraldehyde in the same buffer. The histochemical reaction was performed as described above for embryonic nerve cords, except that a more acidic (pH 6.5) staining solution was employed. After the histochemical reaction, the tissue was postfixed for several hours in 0.1 M phosphate buffer containing 2.5% glutaraldehyde and 4% paraformaldehyde, fol- lowed by 1 h in 1% osmium tetroxide in the same buffer. Dehydration was performed along an ascending ethanol series in short (5-8 min) steps. The tissue was transferred to AGAR 100 over trichlore ethane. Poly- merization was done at 60°C for 48 h.

Sectioning Semithin (0.5-2 pm) sections were stained with tolu-

idine blue and mounted in Permount or AGAR 100. Strong stainings with toluidine blue gave better reso- lution than light stainings since the strong ones were found to enhance the contrast between the colors given by the toluidine blue and the characteristic blue color of the X-gal reaction. Sections from some specimens were also mounted without counterstaining. Ultrathin sections were mounted on copper grids, stained with uranyl acetate and lead citrate, and studied on a Zeiss EM900 or a Jeol 100 electron microscope. No differ- ences were observed in the resolution of the reaction product at the EM level after omission of either uranyl acetate or lead citrate.

RESULTS X-Gal Staining Procedure

The combination of p-galactosidase histochemistry and electron microscopy (EM) presents difficulties sim- ilar to those encountered by the combination of EM with immunohistochemistry. The main problem is that these two methods have partially contradictory re- quirements. In general, performing the enzymatic staining reaction at optimal conditions results in tissue with poor ultrastructural preservation. Several limits are imposed by the fact that the histochemical detec- tion of p-galactosidase expression is based on the dem- onstration of its enzymatic activity. At the light micro- scopic level, five main parameters define the quality of the staining: fixation, temperature, pH, concentration of the substrate, and length of the reaction. Three of these parameters (fixation, temperature, and time), however, impose some restrictions on EM preservation.

The intensity of X-gal staining decreases with the intensity of the prefixation, whereas weaker prefix- ations result in poor ultrastructural preservation. This problem is more accentuated in the preparation of lar-

296 A. STOLLEWERK ET AL.

Fig. 1. Rostrocaudal series of semithin sections (1 pm) through a complete abdominal neuromere in a stage 16 embryo showing the location of midline glial cells (arrows) and their relations with important structures: connectives (Co), nerve roots (NR), and anterior (AC) and posterior (PC) commissures.

Val, pupal, and adult nerve cords, due to the increased size of the tissue and the growth of the neural sheath (perineurium and neural lamella). An additional prob- lem is that the formation of X-gal crystals can mechan- ically disrupt cytoplasmic organization. An illustration of these problems is presented in Figure 1. In this spec- imen, cells lacking P-galactosidase expression were rel- atively well preserved at the EM level, whereas the cytoplasm in well-stained cells was almost completely transformed into dense aggregations of X-gal crystals. Several concentrations of paraformaldehyde and glut- araldehyde in 0.1 M phosphate buffer were tested alone or mixed for both embryonic and larval tissues. For embryos, prefixation in 18.5% glutaraldehyde in 0.1 M phophate buffer, pH 7.0, gave the best results and ren- dered the P-galactosidase enzyme active and allows good ultrastructural preservation (25%, 23%, 20%, 15%, and 10% glutaraldehyde have been tested). For larval tissues, a 10 min fixation in ice-cold 0.5% glut- araldehyde gave acceptable results for light microscopy

(LM) but was insufficient for EM (data not shown). Increasing the fixation time orland the concentration of the fixative resulted in better EM preservation but se- riously impaired the histochemical reaction. A 30 min fixation with ice-cold 1.5-2% glutaraldehyde proved to be a good compromise between EM preservation and histochemical detection.

The optimal reaction temperature for the bacterial P-galactosidase enzyme is 37"C, in which the length of the staining procedure varies between a few minutes to several hours, depending on the transgenic lineltissues being studied. Performing the reaction at 37°C allows short staining times but results in poor preservation of the tissue even for LM. Setting back the reaction tem- perature to 14°C greatly improved the overall morphol- ogy. The reduced reaction temperature requires a pro- longed exposure to the staining solution. However, at this temperature even reaction periods of up to 24 h did not destroy morphological details.

In order to obtain shorter staining times for the

POSTEMBRY ONIC MIDLINE DEVELOPMENT 297

larger postembryonic tissues, a lower pH was chosen for the staining solution than that used for embryonic material. Incubation of the tissue in the staining solu- tion at pH 6.5 was found to increase the reaction rate considerably without affecting stain specificity. Fur- ther improvement of ultrastructural preservation, es- pecially for the presentation of membranes, was ob- tained after incubation of the stained embryos in a solution of 2% acrolein and 4% glutaraldehyde in 0.1 M phosphate buffer, pH 7.0. The acrolein treatment has to follow the P-galactosidase activity since it destroys the enzymatic activities.

Durcupan (Buchs, Switzerland) was used first as em- bedding resin, but it often resulted in the complete dis- appearance of the blue X-gal reaction product during the polymerization. Thus, different resins (i.e., Ar- aldite and AGAR 100) were adopted since they did not interfere with the reaction product.

Midline Glial Cells at the End of Embryogenesis To label the midline glial cells, we used the enhancer

trap lines 3,400 and 3,920 kindly provided by C. Cal- lahan and J. Thomas. Embryos derived from these en- hancer detector strains express strong p-galactosidase activity in the cytoplasm of all midline glial cells. Us- ing the method described above, we performed serial semithin cross-sections of abdominal neuromeres of stage 16 embryos. Figure 1 shows a panel of 12 light level micrographs representing one segmental unit. In- tense P-galactosidase expression can be seen in four midline glial cells and within both anterior and poste- rior commissures. The accumulation of X-gal crystals in the ventral portion of the anterior commissures can also be seen in Figure 2, which shows a cross-section through an anterior commissure of the first abdominal neuromere. Commissural axons originating from the longitudinal connectives exit these bundles at three levels, subdividing the commissure into three compart- ments in the dorsoventral axis. Between the individual compartments, X-gal staining reveals the presence of midline glial cell bodies and processes. In some seg- ments even glial nuclei can be detected between the commissural axons (not shown). In Figure 2b, which is a higher magnification of the same section shown in Figure 2a, midline glial cell processes contacting com- missural axons are clearly labelled by X-gal crystals. Membrane specializations were not observed in this contact areas.

In order to further analyze a possible subdivision of anterior and posterior commissures by the midline glia, we examined the organization of the commissures by serial ultrathin sections of nine abdominal neu- romeres of three different embryos. Figure 3 shows a cross-section through the nerve cord of a stage 16 em- bryo. Two midline glial cells located between the ante- rior and posterior commissure of an abdominal neu- romer are intensively labelled by X-gal crystals. In addition to the heavily labelled midline glial cells (Fig. 3, asterisks), flanking neuronal and glial cell bodies also exhibit X-gal crystals. Whether this is due to ex- pression of lacZ in these cells or diffusion of the X-gal product is presently unknown.

Midline Glial Cells in the Larval Nerve Cord In the first instar larva, the pattern of P-galactosi-

dase expression is identical to that described above for the last embryonic stage. The stained structures ap- pear as a continuous series of 13 metameres from the suboesophageal to the last abdominal neuromere. The most rostra1 and the most caudal metameres have a slightly different morphology. The midline glial cells are segregated into a dorsal and a ventral row. This can be more clearly observed in second instar larvae, when the midline glial cells are larger (Fig. 4a). Single neu- romeres contain P-galactosidase-expressing midline glia either in a ventral position, a dorsal position, or, most commonly, in both levels. Cells located in an in- termediate position were also found in some neu- romeres. This variable distribution was also observed throughout the third larval instar and during pupation (Fig. 4b,c). Transverse sections show that each midline glial cell is flanked by the medianmost glial cells that form the neuropile cover (interface glia) (Fig. 5b). As in the embryonic central nervous system (CNS), the ven- tral midline glial cells are located along the midline of the ventral neuropile, whereas their dorsal counter- parts are located along the dorsal midline of the neu- ropile. Horizontal sections along either the ventral or the dorsal row of midline glial cells show that in most cases the cells found at each level form a single string along the entire nerve cord (Fig. 5a). Within each neu- romere, the midline glial cells occupy a central posi- tion, between the anterior and posterior commissures. Interposed between the row of midline glial cells along the anterior-posterior axis there is a complementary series of dorsoventral channel glial cells (Fig. 5a) (Ito et al., 1995).

Transverse serial sections across entire thoracic neu- romeres in a mature third instar larva were made. Ex- amples are seen in Figure 6a, b. As in the embryo, both the anterior and posterior commissures are subdivided into three main compartments located dorsally, medi- ally, and ventrally in the neuropile (Figs. 5b, 6a). Each dorsal midline glial cell extends a thick prolongation ventrally, and its ventral counterpart extends a similar prolongation dorsally. Both prolongations seem to end on the median commissure.

In LM and EM sections (Figs. 7, 81, it was observed that the ventral midline glial cells are associated with the ascending neurites of ventral unpaired median neurons (VUMs), whereas the dorsal midline glial cells are associated with the descending neurites of dorsal unpaired median neurons (DUMs). Figure 8 shows an example of the results obtained with this method for the electron microscopic detection of P-galactosidase by X-gal staining. Here, relatively large X-gal crystals were formed in the cytoplasm of a large midline glial cell at the same time that ultrastructural details are relatively well preserved.

Cell Growth and Endomitotic DNA Replication of the Midline Glia During Larval Stages

During larval life, the midline glia grow in size and by the third instar become the largest cells of the nerve cord (Fig. 7a,b). The segregation of the midline glia

Fig. 2. a: Transverse sections through an abdominal neuromere in a stage 16 embryo a t the level of the anterior commissure. At this stage, the commissure is divided along its dorsoventral axis into three main compartments (dorsal (dAC), median (mAC), and ventral make close contacts (arrows) with commissural axons (Ax).

[vACI), each formed by fascicles of axons leaving the connectives (Con) at different levels. b: At higher magnification, it is clearly ob- served that processes from midline glia labelled with X-gal crystals

Fig. 3. Transverse section of the nerve cord in a stage 16 embryo showing two midline glial cells located between anterior and posterior commissure of an abdominal neuromere, corresponding to a section shown in Fig. If. The midline glia (asterisks) are heavily labelled by

X-gal crystals. Sparse crystals are also seen in this case in glial and neuronal cell bodies (arrowheads) located close to the midline glia. Con, connectives.

300 A. STOLLEWERK ET AL.

Fig. 4. Whole mounts of CNS from second (a) and third (b) instar larvae, white pupa (c), and 24 (d,e) and 48 (0 h pupae processed for X-gal histochemical detection of p-galactosidase in midline glial cells. Anterior is to the right in the specimens shown in lateral view (a-e) and to the top in f, which is a slightly tilted frontal view. Throughout larval development (a-d), the large cell bodies of midline glia are located at different positions but often in similar distributions in groups of adjacent neuromeres. By the beginning of metamorphosis, the large nuclei are progressively replaced by clusters of smaller nu- clei clustered in the ventral and dorsal positions. In each metamere, the two clusters and the corresponding cytoplasmic processes extend-

along the dorsoventral axis, started during late embry- onic stages, becomes more pronounced as the larval neuropile grows in size (compare, for example, Fig. 4a and 44. The size of the nucleus and the number of chromosomes observed during the last larval instar (Fig. 7a-d) indicate that the midline glia undergo sev- eral rounds of endomitotic DNA replication during lar- val life. By the end of larval life the giant nuclei of the midline glia are found both ventrally and dorsally in the neuropile. During the transition to pupal stages (e.g., during metamorphosis), starting in the suboe- sophageal and thoracic neuromeres, the giant glia nu- clei disappear, and tight clusters of much smaller nu- clei can be detected instead (Figs. 4d,e, 6). Mixed clusters of large and small midline glia nuclei were also observed at this stages (Fig. 6e). Electron micros-

ing between both “poles” result in the formation of the characteristic barrel-shaped structures found in early pupae as shown in panel d. In a higher magnification (e), some of the nuclei are indicated by arrows. The barrel-shaped structures diverge according to a segmental pat- tern: the three thoracic metameres (Tl-T3) grow and become much larger than the abdominal ones (Al-A8), whereas the suboesophageal metameres ( S ) turn 90” as they move rostrally and become integrated into the growing brain, where they eventually form the ring-shaped structure (arrow) shown in panel f. mBr, midbrain; OL, optic lobes; Re, retina. Scale bars: 50 pm.

copy revealed that a t this stage the large nuclei of mid- line glia are replete with cytoplasmic invaginations (Fig. 9). This apparent cell fragmentation extends later to the abdominal neuromeres, where clusters of small midline glial cells can be found predominantly in the dorsal position during pupation. This process continues during the first 2 days of pupal life and yields cell clus- ters each comprising at least 20 small nuclei, found both dorsal and ventral in the neuropile. The pupal clusters of midline glia are arranged in a fashion that emphasizes their dorsoventral polarity, forming barrel- shaped structures in which the nuclei occupy opposite poles, from which several cytoplasmic prolongations extend into the neuropile. In old pupae P-galactosidase expression is no longer detectable at the midline. In the adult only rare small glial cells can be observed in the

POSTEMBRYONIC MIDLINE DEVELOPMENT 301

AC

PC

IG IG

b Fig. 5. Camera lucida reconstruction from horizontal (a) and

transverse (b) sections of the CNS of a third instar larva stained with X-gal. The asterisks in the lumen of the dorsoventral channels de- marcate approximate borders between consecutive neuromeres. Along the midline there is an alternate series of midline glia (MG) and dorsoventral channel glia (CG). The anterior and posterior com- missures (AC and PC) are shown only in one neuromere as a few axons depicted by dotted lines. Also, a fascicle of ventrodorsal axons (arrow) surrounded by a midline glia is shown. In transverse section

midline in toluidine blue-stained sections (data not shown).

Some Midline Glia Are Incorporated Into the Brain

Shortly after pupation, the most rostra1 midline glia comigrate with the suboesophageal neuromeres (Fig. 4d,f) and become incorporated in the developing adult brain. Midline glial cells were not observed within the cervical connectives.

DISCUSSION The amount of data concerning the development of

neurons and glial cells in the fruit fly is increasing with the aid of the enhancer trap method (Bellen et al., 1989; Bier et al., 1989; Ghysen and O'Kane, 1988; OKane and Ghering, 1987). This technique utilizes the bacterial gene encoding the enzyme p-galactosidase, which in different enhancer trap lines is expressed as a tissue- or cell-specific marker.

The study of several important aspects of neuronal and glial cell development requires the high resolution offered by electron microscopy. In this paper we present an improved protocol for detection of p-galactosidase activity in embryos and larvae derived from enhancer trap lines using the electron microscope. We show that

(b), it is observed that, in addition to being contacted by channel glia, the midline glia are also contacted by the most median interface glia (IG). Here, three midline glia are shown aligned along the ventrodor- sal axis and associated with the ventral, median, and dorsal divisions of the posterior commissure (vPC, mPC, and dF'C, respectively). Mid- line glial cell bodies lying close to the interface between the cell body rind and the neuropil (depicted by a dotted line) come in contact with the two most median interface glia cells (IG) found at these positions. Scale bar: 25 km.

the present method is able to display specific expres- sion of p-galactosidase in cells with reasonably well- preserved ultrastructure. A further improvement of the technique could result from the use of Bluogal (Sigma, St. Louis, MO) which is reported to give better results for EM (Campbell and Peterson, 1993).

Midline Glial Cells During Development of the Fly

Using the present technique we have analyzed the differentiation of the midline glial cells from embry- onic to late pupal developmental stages. The early em- bryonic development of these cells has been recently described in some detail (Jacobs and Goodman, 1989a; Klambt et al., 1991). To facilitate our analysis of these glial cells, we choose enhancer trap lines which permit the labelling of the entire cytoplasm (Callahan and Thomas, 1994). We were able to show in the embryo a separation of the two segmental commissures found in each neuromere (the anterior and posterior commis- sures), a condition which was previously unknown. These tracts are divided into three further subcompart- ments located ventrally, medially, and dorsally in the neuropile. The separation of the commissures into three distinct fiber tracts is maintained throughout larval stages. During late embryogenesis about four

302 A. STOLLEWERK ET AL.

Fig. 6. Semithin sections (1 pm) through thoracic neuromeres of late third instar larva stained with X-gal. a,b: The sections shown, counterstained with toluidine blue, belong to the same neuromere and are displayed in rostrocaudal order. The median fascicle of the ante- rior commissure (AC) is shown in panel a. At this stage, either large nuclei, clusters of small nuclei, or a combination of both can be found in a single neuromere. In this example, a cluster of small nuclei (ar- rows in panel a) occupies the dorsal midline region, whereas a single cell with a large nucleus (asterisks in panels a,b) occupies the ventral

glial cells can be detected in each neuromere (Bossing and Technau, 1994).

In larval stages it is obvious that the nuclei of mid- line glial cells are arranged in an irregular pattern throughout the depth of the neuropile. The midline glial cells enwrap the two commissures and during lar- val life undergo several rounds of endomitotic DNA replication to become some of the largest cells of the nerve cord.

region. The different types of midline nuclei found a t this stage are more clearly seen at higher magnification in panels c-e (in sections that were not counterstained, to show the specificity of the X-gal staining). c: Single midline glial cell with large nucleus (asterisk). d Cluster of small nuclei (some are marked by arrows) extending from ventral to dorsal neuropil. e: Dorsal cluster of small nuclear fragments (arrow) close to a large nucleus (asterisk). Scale bars: 20 pm (a,b); 10 km (c-e).

Previous work documented the existence of endomi- totic DNA replication among several classes of glial cells in Drosophila (Prokop and Technau, 1994). Also, these midline glial cells were shown to incorporate bro- modeoxyuridine (BrdU) during larval stages, but it was suggested (Prokop and Technau, 1994) that these cells divide earlier than reported here. Our data rather suggest that, in the larva, DNA replication takes place first without any cell division and only at the onset of

POSTEMBRY ONIC MIDLINE DEVELOPMENT 303

Fig. 7. The large size attained by the midline glia (one of them labeled MG in panels a,b) by late third instar, the location of these cells withm the nerve cord, and their relationships with neuropil structures can be followed in sernithin sections made along the sag- ittal plane (conventional histology). A broken line (a) partially delin- eates the border between the two neuromeres shown here, as defined by dorsoventral channels. Axons oriented along the dorsoventral axis (arrowheads), emanating from ventral (VUM) or dorsal (DUM) un-

paired median neurons, are in contact with the midline glia. The dorsal (dAC), median (mAC), and ventral (vAC) subdivisions of the anterior commissure are marked in one neuromere (b). Endoreplicat- ing nuclei of midline glia are seen in the CNS of third instar larvae at dorsal (c) and ventral (d) positions. Numerous chromosomes (arrow- heads) indicate the level of polyploidy attained by these cells in late third instar larva. Anterior is to the right. Scale bars: 10 pm (a,b); 10 km (c,d).

metamorphosis do the giant midline glial cells begin to divide. Since a regular cell division is unlikely to occur in a highly polyploid cell, we suggest use of the term fragmentation for the process resulting in the replace- ment of a giant nucleus by a cluster of small nuclei. The resulting cell fragments are probably nonviable and should be cleared from the tissue within a short time. In agreement with this, we could rarely find mid- line glial cells with conventional histology in the adult ventral nerve cord (R.C., unpublished results).

In this respect it is interesting to note that during mammalian development midline glial cells are formed in the floor plate during embryogenesis but during later development (first postnatal week) appear to leave the midline (McKanna, 1993). There are three rows of primitive glial ependymal (PGE) cells which are already present when commissures form. The me- dian row stains for LC1, a marker for adult microglia,

and the two lateral rows stain for the SlOO-p marker of adult astrocytes. The nuclei of the PGE cells reside close to the ventricle and extend processes to the pia. During later development (first postnatal week), astro- cytes arise from the paramedian PGEs, whereas the median PGEs appear to give rise to microglial cells. Those events could be effected by a different mecha- nism still leading to the same result: that glial cells leave the midline in the nerve cord of Drosophila dur- ing adult stages as well as in the neural tube of the vertebrate CNS.

Using the specific enhancer trap lines and a combi- nation of X-gal staining and conventional histological methods, we could follow the fate of the midline glial cells throughout Drosophila development. Our findings indicate that the presented protocol serves as a valu- able tool in the study of differentiation in other cell types in Drosophila.

304 A. STOLLEWERK ET AL.

Fig. 8. p-galactosidase expression in larval tissue can also be de- tected at the EM level by the formation of very electron-dense crystals (arrowhead). As expected for this type of enhancer trap detection,

p-galactosidase is detected only in the cytoplasm of midline glial cells. Crystals are not formed in the interface glia (IG), neuropil (Ne), or associated dorsoventral axons (Ax) shown in this transverse section.

ACKNOWLEDGMENTS We are grateful to J. Thomas and C. Callahan for

making their enhancer trap lines available prior to publication. We thank B. Afzelius, J.A. Campos-Or-

tega, and F. Grawe, for their help and encouragement during many steps of this project. This work was sup- ported by the DFG through SFB243 and a Heisenberg fellowship to C.K. and a Graduierten fellowship to A.S.

POSTEMBRYONIC MIDLINE DEVELOPMENT 305

Fig. 9. Midline glial nuclei undergoing fragmentation. Cytoplasmic invaginations (arrows) penetrate deeply into the nuclei (asterisks) of two large midline glia located dorsally in a thoracic neuromere of a late third instar larva. Horizontal section. Scale bar: 1 pm.

REFERENCES Bellen, H.J., OKane, C.J., Wilson, C., Grossniklaus, U., Kurth-Pear-

son, R., and Gehring, W.J. (1989) P-element-mediated enhancer detection: A versatile method to study development in Drosophila. Genes Dev., 3:1288-1300.

Bier, E., Vassin, H., Shephard, S., Lee, K., McCall, K., Barbel, S., Ackerman, L., Carretto, R., Uemara, T., Grell, E.H., Jan, L.Y., and Jan, Y .N. (1989) Searching for pattern and mutations in the Dro- sophila genome with a P-lacZ vector. Genes Dev., 3:1273-1287.

Bier, E., Jan, L.Y., and Jan, Y.N. (1990) rhomboid, a gene required for dorso-ventral axis establishment and peripheral nervous system development in Drosophila melanogaster. Genes Dev., 4:190-203.

Bossing, T., and Technau, G.M. (1994) The fate of the CNS midline progenitors in Drosophila as revealed by a new method for single cell labelling. Development, 120:1895-1906.

Callahan, C.A., and Thomas, J.B. (1994) Tau-beta-galactosidase, an axon-targeted fusion protein. Proc. Natl. Acad. Sci. U. S. A., 91: 5972-5976.

Campbell, R.M., and Peterson, A.C., (1993) Expression of a lac2 trans- gene reveals floor plate cell morphology and macromolecular trans- fer to commissural axons. Development, 119:1217-1228.

Campos-Ortega, J.A., and Hartenstein, V. (1985) The Embryonic De-

velopment of Drosophila melanogaster. Springer-Verlag, Berlin, Heidelberg, New York.

Choi, K.-W., and Benzer, S. (1994) Migration of glia along photore- ceptor axons in the developing Drosophila eye. Neuron, 12:423- 431.

Franke, W.W., Krien, S., and Brown, R.M. (1969) Simultaneous glu- taraldehyde-osmium tetroxide fixation with postosmication. His- tochemie, 19:162-164.

Ghysen, A., and OKane, C. (1988) Neural enhancer-like elements as specific cell markers in Drosophila. Development, 105:32-52.

Giangrande, A,, Murray, M.A., and Palka, J. (1993) Development and organization of glial cells in the peripheral nervous system of Dro- sophila. Development, 117:895-904.

Henrikson, C.K., and Vaughn, J.E. (1974) Fine structural relation- ships between neurites and radial glial processes in developing mouse spinal cord. J . Neurocytol., 3:659-675.

Ito, K., Urban, J., and Technau, G.M. (1995) Distribution, classifica- tion and development of Drosophila glial cells during late embryo- genesis. RQUX’S Arch. Dev. Biol. 204:284-307.

Jacobs, J.R., (1993) Perturbed glial scaffold formation precedes axon tract malformation in Drosophila mutants. J . Neurobiol., 24:611- 626.

306 A. STOLLEWERK ET AL.

Jacobs, J.R., and Goodman, C.S. (1989a) Embryonic development of axon pathways in the Drosophilu CNS. I. A glial scaffold appears before the first growth cones. J . Neurosci., 92402-2411.

Jacobs, J.R., and Goodman, C.S. (1989b) Embryonic development of axon pathways in the Drosophilu CNS. 11. Behavior of pioneer growth cones. J. Neurosci., 9:2412-2422.

Klaes, A,, Menne, T., Stollewerk, A,, Scholz, H., and Klambt, C. (1994) The ETS transcription factors encoded by the Drosophilu gene pointed direct glial cell differentiation in the embryonic CNS. Cell, 78:149-160.

Klambt, C. (1993) The Drosophilu gene pointed encodes two ETS-like proteins which are involved in the development of the midline glial cells. Development, 117:163-176.

Klambt, C., and Goodman, C.S. (1991) The diversity and pattern of glia during axon pathway formation in the Drosophila embryo. Glia, 4205-213.

Klambt, C., Jacobs J.R., and Goodman C.S. (1991) The midline of the Drosophila CNS: Model and genetic analysis of cell lineage, cell migration, and development of commissural axon pathways. Cell, 64:80 1-8 15.

Kolodkin, A.L., Pickup, A.T., Lin, D.M., Goodman, C.S. and Banejee, U. (1994) Characterization of Star and its interactions with seven- less and EGF receptor during photoreceptor development in Dro- sophilu, Development, 120:1731-1745.

McKanna, J. (1993) Primitive glial compartments in the floor plate of mammalian embryos: Distinct progenitors of adults astrocytes and microglia support the notoplate hypothesis. Perspect. Dev. Neuro- biol., 4:245-255.

OKane, C., and Gehring, W.J. (1987) Detection in situ of genomic regulatory elements in Drosophila. Proc. Natl. Acad. Sci. U. S. A,, 849123 -9127.

Prokop, A., and Technau, G.M. (1994) BrdU incorporation reveals DNA replication in non dividing glial cells in the larval abdominal CNS of Drosophila. Roux's Arch. Dev. Biol., 204:54-61.

Rakic, P. (1971) Neuron-glia relationship during granule cell migra- tion in developing cerebellar cortex. A golgi and electron micro- scopic study in Macacus rhesus. J. Comp. Neurol., 141:283-312.

Raz, E., and Shilo, B.-Z. (1992) Dissection of the faint little ball (flb) phenotype: Determination of the development of the Drosophilu central nervous system by early interactions in the ectoderm. De- velopment, 114:113-123.

Rutledge, B.J., Zhang, K., Bier, E., Jan, Y.N., and Perrimon, N. (1992) The Drosophila gene spitz encodes a putative EGF-like growth fac- tor involved in dorsal-ventral axis formation and neurogenesis. Genes Dev., 6:1503-1517.

Silver, J., Loren, S.E., Wahlstein, D., and Coughlin, J. (1982) Axonal guidance during development of the great cerebral commissures: Descriptive and experimental studies in vivo on the role of pre- formed glial pathways. J. Comp. Neurol., 21O:lO-29.

Singer, M., Nordlander, R.H., and Egar, M. (1979) Axonal guidance during embryogenesis and regeneration in the spinal cord of the newt. The blueprint hypothesis of neuronal pathway patterning. J . Comp. Neurol., 185:l-22.

Winberg, M.L., S.E., P., and Steller H. (1992) Generation and early differentiation of glial cells in the first optic ganglion of Drosophila melanogaster. Development 115:903-911.