Perlecan participates in proliferation activation of quiescent Drosophila neuroblasts


  • Aaron Voigt,

    1. Abteilung Molekulare Entwicklungsbiologie, Max-Planck-Institut für biophysikalische Chemie, Am Fassberg, Göttingen, Germany
    Search for more papers by this author
    • Drs. Voigt and Pflanz contributed equally in this work.

  • Ralf Pflanz,

    1. Abteilung Molekulare Entwicklungsbiologie, Max-Planck-Institut für biophysikalische Chemie, Am Fassberg, Göttingen, Germany
    Search for more papers by this author
    • Drs. Voigt and Pflanz contributed equally in this work.

  • Ulrich Schäfer,

    1. Abteilung Molekulare Entwicklungsbiologie, Max-Planck-Institut für biophysikalische Chemie, Am Fassberg, Göttingen, Germany
    Search for more papers by this author
  • Herbert Jäckle

    Corresponding author
    1. Abteilung Molekulare Entwicklungsbiologie, Max-Planck-Institut für biophysikalische Chemie, Am Fassberg, Göttingen, Germany
    • Abteilung Molekulare Entwicklungsbiologie, Max-Planck-Institut für biophysikalische Chemie, Am Fassberg, 37077 Göttingen, Germany
    Search for more papers by this author


Drosophila neuroblasts act as stem cells. Their proliferation is controlled through cell cycle arrest and activation in a spatiotemporal pattern. Several genes have been identified that control the pattern of neuroblast quiescence and proliferation in the central nervous system (CNS), including anachronism (ana), even skipped (eve) and terribly reduced optic lobes (trol). eve acts in a non–cell-autonomous manner to produce a transacting factor in the larval body that stimulates cell division in the population of quiescent optic lobe neuroblasts. ana encodes a secreted glial glycoprotein proposed to repress premature proliferation of optic lobe and thoracic neuroblasts. trol was shown to act downstream of ana to activate proliferation of quiescent neuroblasts either by inactivating or bypassing ana-dependent repression. Here, we show that trol codes for Drosophila Perlecan, a large multidomain heparan sulfate proteoglycan originally identified in extracellular matrix structures of mammals. The results suggest that trol acts in the extracellular matrix and binds, stores, and sequesters external signals and, thereby, participates in the stage- and region-specific control of neuroblast proliferation. © 2002 Wiley-Liss, Inc.


Control of cell proliferation is essential for the successful execution of developmental programs in multicellular organisms. These programs define in space and time whether cells are going to differentiate, proliferate, or are set aside to function as pluripotent stem cells. Stem cells are quiescent or cell cycle-arrested cell populations, which can be activated to differentiate or proliferate in response to developmental or environmental cues. They have been identified, for example, in the hematopoietic system, the germ line, and in the nervous systems (Wieschaus and Szabad, 1979; Truman and Bate, 1988; Spangrude and Johnson, 1990; Srour et al., 1991; Ito and Hotta, 1992; Ogata et al., 1992; Margolis and Spradling, 1995). An experimental system in which activation of quiescent stem cells can be studied both on a genetic and molecular level are the neuroblasts of the Drosophila central nervous system (CNS) (Datta and Kankel, 1992).

Neuroblasts are born during early and midstages of Drosophila embryogenesis, and they give rise to all parts of the embryonic brain and CNS (Truman and Bate, 1988; Ito and Hotta, 1992). However, a subpopulation of these neuroblasts remains quiescent until larval stages (White and Kankel, 1978; Truman and Bate, 1988; Ito and Hotta, 1992). In the larval brain, for example, three populations of neuroblasts with distinct proliferation profiles have been identified. One population divides continuously, whereas the other two populations of quiescent neuroblasts exhibit a distinct temporal profile of proliferation in the prospective optic lobe and the thoracic portion of the CNS during first larval instar stage (White and Kankel, 1978; Truman and Bate, 1988; Prokop and Technau, 1991; Ito and Hotta, 1992; Green et al., 1993; Datta, 1995).

Several genes have been identified that affect neuroblast proliferation in Drosophila (Lipshitz and Kankel, 1985; Datta and Kankel, 1992; Ebens et al., 1993; Prokop and Technau, 1994; Park et al., 1998; Guan et al., 2000). These genes include anachroism (ana), terribly reduced optic lobes (trol), and even-skipped (eve). Of these genes, eve is a key player in many cell-fate decisions during embryonic development. It exerts pair-rule gene activity required for the establishment of the segment pattern and determination of neuronal identity during early neurogenesis (reviewed by Akam, 1987). In both processes, the eve-encoded homeodomain protein is part of a cell-autonomous cascade of transcription factors. Furthermore, eve was shown to act as a more general transcription factor necessary for the control of enzymatic activities such as alcohol dehydrogenase and rosy (Liang and Biggin, 1998). In addition, recent evidence by Park et al. (2001) showed eve requirement for the production of a transacting signal that regulates the activation of neuroblast proliferation in the larvae. It thereby mimics the function of the hormone ecdysone in coordinating developmental cell divisions. Importantly, however, the distribution of eve protein within the larval CNS and neuroblast reactivation in a heterogenetic extract assay system show that eve activity is neither produced in the regulated neuroblasts nor necessary in the larval CNS proper. It is likely to control or induce a long-range diffusible signal that impacts the reactivation of proliferation from mitotic quiescent neuroblasts (Park et al., 2001).

Another factor implicated in the developmental coordination of neuroblast proliferation is ana. ana encodes a secreted glial glycoprotein proposed to repress premature proliferation of quiescent optic lobe and thoracic neuroblasts (Ebens et al., 1993). Interestingly, however, when proliferation is activated in quiescent neuroblasts, expression of ana and the levels of the protein appear to remain constant. Because the amount of ana repressor remained unchanged when neuroblasts undergo the transition from quiescence to activation of proliferation, a second factor has been proposed to be needed to override ana-dependent repression (Ebens et al., 1993). ana-mediated proliferation repression of neuroblasts is antagonized by the activity of trol (Datta, 1995). However, the quiescent optic lobe and thoracic neuroblasts that are affected by the trol mutation are able to proliferate in the absence of ana activity, showing that trol acts downstream of ana either by inactivating or bypassing the repressor activity. trol activity may regulate the reactivation of neuroblast proliferation by up-regulating CyclinE expression to thereby stimulate the G1/S transition of the proliferation-activated formerly quiescent neuroblasts (Caldwell and Datta, 1998). eve activity ties in to this pathway, because eve mutations enhance the trol mutant proliferation phenotype, although eve expression is not detectable in regulated neuroblasts or in neurosecretory cells. Instead, eve activity is required for the production of an organismal-level transacting signal that is required to stimulate larval neuroblast proliferation in a trol-dependent manner (Park et al., 2001). Collectively, these observations suggest that trol does not encode a cell type-specific mitotic component and that trol mutations do neither cause cell death of neuroblasts nor do they affect cell fitness to the extent that neuroblasts were incapable of initiating cell divisions. Rather, the available evidence suggests that trol interferes with ana- and eve-dependent external signals, suggesting that its unknown product is either a component of or interferes with a cell–cell signal transduction mechanism.

Here, we show that trol codes for the Drosophila homolog of vertebrate Perlecan. Perlecan has been originally identified as a large, multidomain heparan sulfate proteoglycan of basal membranes that interacts with other extracellular matrix proteins, growth factors and receptors, and thereby influences cellular signaling (Noonan et al., 1991; Aviezer et al., 1994; Handler et al., 1997). The molecular nature of the trol gene product is, therefore, consistent with a function in regulating signaling activity either through the binding, storage or sequestering of specific ligands such as the secreted ana protein or external signals that are produced at the organismic level in response to eve activity.


Generation of New P-Insertion trol Alleles for Gene Cloning

The trol locus of Drosophila is localized in the chromosomal band 3A4 on the X-chromosome and is characterized by 134 mutant alleles (Datta and Kankel, 1992; Flybase, 1999). Several of the in-depth analysed trol-mutant alleles such as trolsd and trol15 show a severe size reduction of the larval optical lobe area, attributed to the loss of reactivation of neuroblast proliferation from mitotic quiescence (Datta, 1995). To molecularly identify the trol transcription unit, we screened a X-chromosomal collection of lethal P-element insertion lines. We found three P-element insertions, l(1)G0023, l(1)G0271, and l(1)G0374, which failed to complement the trol alleles trol13 and trol15 (Judd et al., 1972), suggesting that the corresponding P-element insertions have generated trol alleles. To show that the P-element insertion is indeed the cause of the trol mutation, we performed remobilization experiments. Each of the insertion lines could be reverted to viability, indicating that each of the three P-element insertions had hit the trol locus.

We next asked whether the newly identified trol alleles show the reduced optic lobe phenotype (Datta and Kankel, 1992). Hemizygous l(1)G0271 mutant larvae develop smaller optic lobes than their heterozygous siblings (Fig. 1A,B) as observed earlier with mutant trol alleles (Datta and Kankel, 1992). Other parts of the brain show overgrowth defects, which indicate that proliferation control in the brain is strongly impaired (Fig. 1B). Thus trol mutations do not only specifically interfere with the reinitiation of optic lobe neuroblast proliferation but have severe effects on larval brain growth in general. This conclusion is consistent with the finding that the imaginal discs of such larvae are extensively folded compared with wild-type discs at the corresponding developmental stages (Fig. 1C,D). In addition, we noted that mutant discs are either significantly enlarged or, in other instances, smaller or not detectable. These diverse and even opposing observations with wing discs suggest that trol participates not only in a cell-specific manner in the control of proliferation as shown in case of the optic lobe neuroblasts (Datta, 1995) but contributes also in a more general manner to larval development. This conclusion is also consistent with the finding that the development of hemizygous l(1)G0271 larvae is slowed down, meaning that it took the mutant larvae at least 1 day longer to reach the third larval stage where they eventually die.

Figure 1.

Reduced optic lobes and abnormal imaginal discs in l(1)G0271 larvae. A,B: Overall larval brain morphology as visualized by phalloidin-Alexa 546 (red) and DNA (green) stainings. A: Brain of a wandering third instar larvae of the genotype l(1)G0271/FM7i showing wild-type morphology. Positions of thoracic ganglia (TG) and the optic lobe area (OL) are marked. B: Brain of a hemizygous l(1)G0271 individual showing that the TG region is disrupted and the OLs of the brain are strongly reduced in size (compare A,B). In addition, variable regions of the brain, except for the OL, show an overproliferation phenotype (example of a mild overgrowth region is labeled by an arrow), indicating that trol action is not required for OL formation only. C,D: Wing disc morphology as revealed by phalloidin-Alexa 546 staining (red). C: Wing disc of a l(1)G0271/FM7i individual showing the wild-type morphology. D: Wing disc of a hemizygous l(1)G0271 individual showing that the disc is convoluted and reduced in size (compare C,D). Note that the wing pouch is rotated 90 degrees relative to the dorsoventral axis and that the dorsal body wall region is heavily reduced in size (arrowhead). The haltere disc still adheres to the wing disc and is heavily convoluted (arrow). Scale bars = 100 μm in A–D.

The strong hypomorphic trol15 allele causes polyphasic lethality during first and second larval instar (Datta and Kankel, 1992), whereas the three newly identified P-element insertion mutations survived until third instar larval stage. Only few escapers develop into pupae to die as pharate adults. These observations indicate that the P-element–associated trol mutations are generally weaker alleles than the strong trol15 hypomorphic allele.

Molecular Identification of the trol Transcription Unit

To correlate, by molecular means, the P-element insertion sites with a transcription unit, we performed “plasmid rescue” experiments to identify P-element–adjoining genomic DNA sequences within the Drosophila genome (Perucho et al., 1980; Adams et al., 2000). The three P-element insertions are located within a DNA segment of less than 1 kb (Adams et al., 2000), followed by several large and a multitude of small exons that could be conceptually combined to a single transcription unit spanning a genomic region of more than 40 kb (Fig. 2A). Sequences of the transcription unit are represented by several expressed sequence tags (ESTs), which cover approximately 55% of the coding region of a 13-kb mRNA (Fig. 2A). Conceptional translation of the open reading frame indicates that the transcript encodes the Drosophila homolog of mammalian Perlecan, parts of which had been described previously (Friedrich et al., 2000).

Figure 2.

Genomic structure of the trol transcription unit, size of the trolnull deletion, and schematic representation of mammalian and Drosophila Perlecan. A: Physical map of the trol locus, the P{lacW} insertion sites of l(1)G0023, l(1)G0271, and l(1)G0374 (gray triangle) between bases number 165,000 and 100,000 of the contig AE003424 in region 3A of the X chromosome (Adams et al., 2000). Note the position of the P-element insertions directly upstream of the putative transcription start site of trol and the EP(1)1619 insertion site, which is approximately 4 kb downstream of the trol open reading frame (grey triangles). Exons are represented by black boxes, 5′ and 3′ untranslated regions by thick lines, coding regions corresponding to domain II-V of Perlecan (Timpl, 1993) are shown below the transcript (see also B, text and Figs. 3–5 for details). The extent of the trolnull deletion (which was generated by combined P-element excision/recombination events involving the l(1)GO271 and EP(1)1619 insertions as described in the text and Experimental Procedures section) is indicated. B: Schematic representation of mouse and Drosophila Perlecan domain structure. Note that domain I is not conserved in Drosophila and domain II is extended. For a detailed comparison of domain II-IV and domain V, see Figures 3–5 (here) and Figure 1 in Friedrich et al. (2001). The black box represents the N-terminal signal peptide sequence. Note that the overall domain structure of mammalian Perlecans is conserved in Drosophila and that domains III to V have almost identical subdomain compositions (details in Figs. 3–5; see also Friedrichs et al., 2000).

Figure 3.

Comparison of domain II sequences between mouse and Drosophila Perlecan. Mouse Perlecan was used as the representative of the mammalian Perlecans. Whereas mouse Perlecan has only 4 LDL-R motifs (mLDL-R), Drosophila Perlecan has at least 22 such motifs (dLDL-Rs). Note that nearly all of the LDL-Rs are coded by individual small exons. For some exons in domain II of Drosophila Perlecan, however, no LDL-R motif was detected by the interpretation programs but could be identified as degenerated motifs by eye comparison; these were not included in the alignment. The order of the sequences in the alignment follows their order in the proteins and does not reflect grouped similarities. m-LDL-R 1 to 4 represent amino acids 210-231, 295-316, 335-356, and 379-400 of mouse Perlecan (NP_032331). The 22 LDL-R domains of Drosophila Perlecan represent amino acids 38-59, 78-99, 123-144, 204-225, 293-314, 402-423, 441-462, 490-511, 544-565, 584-605, 621-642, 659-680, 721-742, 769-790, 805-826, 845-866, 882-903, 922-943, 990-1011, 1176-1197, 1216-1237, and 1259-1280, respectively.

Figure 4.

Comparison of domain III between mouse and Drosophila Perlecan. Mouse and Drosophila Perlecan have three laminin B-type domains each (m lamB 1 to 3 and d lamB 1 to 3, respectively). The order of the sequences follows their order in the proteins and does not reflect grouped similarities. m lamB 1 to 3 represent amino acids 590-716, 985-1112, and 1393-1516; d lamB 1 to 3 represent amino acids 1475-1591, 1839-1964, and 2190-2312.

Figure 5.

Comparison of domain IV between mouse and Drosophila Perlecan. Mouse and Drosophila Perlecan have 14 immunoglobulin-type domains each (mIG 1 to 14 and dIG 1 to 14, respectively). Thirteen mIG and 12 dIG motifs are in domain IV, the remainder is between domains II and III. The order of the sequences follows the order in the proteins and does not reflect grouped similarities. mIG 1 to 14 represent amino acids 414-490, 1682-1762, 1778-1857, 1871-1950, 1961-2026, 2063-2136, 2156-2233, 2253-2331, 2442-2523, 2540-2617, 2626-2704, 2727-2805, 2828-2894, and 2902-2979. Drosophila immunoglobulins (dIG) represent amino acids 1041-1118, 1307-1375, 2416-2484, 2513-2592, 2609-2692, 2711-2788, 2801-2880, 2892-2971, 3014-3073, 3098-3175, 3192-3253, 3309-3383, 3395-3469, and 3500-3455.

Perlecan is a posttranslationally modified extracellular matrix protein that has heparan/chondroitin sulfate side chains (Hassell et al., 1980; Noonan et al., 1991; Danielson et al., 1992). It is subdivided into five distinct regions that harbour domains structurally related either to LDL receptors (domain II), the N-terminal region of both laminin A and B short arms (domain III), the N-CAMs (domain IV), or the globular C-terminus of the laminin A chain (domain V) (Noonan et al., 1991; Kallunki and Tryggvason, 1992; Noonan and Hassell, 1993). Sequence analysis of three overlapping cDNA clones (SD04592, GM03359, and GM01116), which cover approximately 7 kb, revealed coding regions corresponding to the regions IV and V, and a major part of domain III of mammalian Perlecan. In addition, we performed reverse transcriptase-polymerase chain reaction (RT-PCR) with primers specific for the annotated transcript corresponding to the coding region of domain II. A diagram summarizing the domain structure of Drosophila Perlecan compared with mammalian homologs is shown in Figure 2B.

The most striking difference to mammalian Perlecan is the absence of an equivalent to domain I, which consists mainly of one SEA domain (Henikoff and Henikoff, 1994), in Drosophila. Sequences corresponding to this domain have neither been found in cDNA clones nor in the genomic DNA sequences. It is also absent from C. elegans Perlecan (Noonan et al., 1991; Kallunki and Tryggvason, 1992; Rogalski et al., 2001). Another major difference is that domain II, which is characterized by LDL-R motifs, is extended in Drosophila Perlecan. Whereas mammalian homologs exhibit four such motifs (Noonan et al., 1991; Cohen et al., 1993), Drosophila Perlecan has a total of at least 22 (Fig. 3). Domains III and IV of Drosophila Perlecan and sequence comparison with mouse Perlecan are shown in Figures 4 and 5. For a detailed description of domain V see Friedrich et al. (2000).

trol Expression

The expression pattern of Drosophila Perlecan during embryogenesis was previously analysed with anti-dPCN antibodies directed against domain V of the protein and digoxigenin-labeled antisense RNA probes (Friedrich et al., 2000). We monitored the distribution of transcripts by in situ hybridisation of whole-mount embryos by using digoxigenin-labeled antisense RNA probes corresponding to the different regions of the 13-kb transcript. The different probes, including probes directed against the 5′ most untranslated sequences of trol corresponding to GM02428 cDNA, which are located close to the insertion sites of the three P-elements (see above) and probes corresponding to all four domains of the open reading frame revealed an identical expression pattern. This finding suggests that the cDNA clones represent different portions of a single, large transcription unit. This conclusion is consistent with the location of the P-element insertions in the 5′ region of the trol gene which, together with the 5′ untranslated region of transcription units, are the preferred sites for P-element integrations on the X chromosome (Peter et al., 2002).

The spatiotemporal in situ expression patterns of the various portions of the trol transcript (Fig. 6) are identical to those reported earlier by using probes directed against domain V (data not shown; for details see Friedrich et al., 2000). Our results add to the published patterns that the transcripts were strongly expressed during oogenesis (Fig. 6A). The finding of maternal transcription provides a molecular basis for earlier results showing that the trol15 allele causes maternal-effect embryonic lethality (Robbins, 1990). In addition to the maternal expression, we noted expression of transcripts in a subset of dorsal midline glial cells of the CNS (Fig. 6B). These cells were found to express Perlecan as detected by anti-Perlecan antibody staining at a later stage of development (Friedrich et al., 2000).

Figure 6.

trol expression during oogenesis and in embryonic and larval nervous systems. Whole-mount in situ hybridisations were performed by using a probe specific for the largest exon of trol, which codes for domain III (see Fig. 2A). A: During oogenesis, transcripts are detected at the tip of the germarium (left side). In the consecutive stages of follicle formation, no transcripts are found. However, transcription is reinitiated in region 2b of the germarium and transcripts remain up until stage 6 of oogenesis both in nurse and follicle cells. With the onset of vitellogenesis, transcription fades away and only the dorsal follicle cells overlying the oocyte nucleus show transcript accumulation. B: During embryonic development trol is expressed at high levels in a subset of the cells of the central nervous system (CNS). Ventral view of a stage 15 embryo. The arrow points to a cell of the midline glia. C,D: In the larval CNS, trol is expressed in few cells only. C: Dorsal view of the larval brain, the boxed region is enlarged in D. Two clusters of cells at the inner side of the brain hemispheres express higher levels of trol. Note that the optic lobes are virtually devoid of trol expression at this stage of development.

During larval stages when trol activity is required for the activation of proliferation of the quiescent optic lobe neuroblasts, Perlecan transcripts are expressed in a distinct subpattern of neural cells, which are located outside the optic lobe region (Fig. 6C,D). We were unable to identify the nature of these cells. By analogy to the earlier expression of Perlecan and the transcript in a subset of glial cells of the CNS (Friedrich et al., 2000; see above), we anticipate that the Perlecan-expressing cells represent a subset of brain glia. Irrespective of the identity of these cells, however, the remarkable result is that trol-encoded Perlecan is not prominently expressed in the larval optic lobe cells.

Deletion of the trol Transcription Unit by Recombination After P-Element Excision

To obtain a trol null mutation, we generated a deletion spanning the genomic region of the transcription unit. The deletion was the result of the simultaneous mobilization of the P-elements l(1)G0271 and EP(1)1619, which are located immediately 5′ of the putative transcription start site and approximately 4 kb 3′ to the transcription unit (see Fig. 2A; details in Experimental Procedures section), respectively. Excision of the two P-elements was monitored by the reappearance of the white-eye phenotype, indicating that both P-element–associated marker gene copies were lost. Heterozygous female individuals were collected and examined by PCR analysis by using primers directed to 5′ adjacent sequences of the l(1)G0271 insertion site and to 3′ adjacent sequences of the EP(1)1619 insertion site, respectively, followed by sequencing of the amplified DNA. The sequence data showed that the excision of the P-elements in combination with a recombination event has caused a 47,908,bp deficiency, removing the entire Perlecan transcription unit exclusively (Fig. 2A). The deficiency mutation causes lethality without resulting in an obvious and morphologically distinct larval cuticle phenotype as has been observed with the hypomorphic trol mutations. Questions concerning the maternal effect of this new trol allele, which represents a trolnull mutation, can now be addressed.

Overexpression of trol

EP(1)1160 is inserted within the region of the trol gene where the three above-described P-elements are located. We asked, therefore, whether this line could be used to drive the expression of trol by the GAL4/UAS system (Brand and Perrimon, 1993; Rorth, 1996). We crossed the homozygous EP(1)1160 females with males homozygous for the engrailed-GAL4 driver (Tabata et al., 1992). In situ hybridisation on embryos with trol anti-sense RNA probes and anti-Perlecan domain V-specific antibodies revealed an engrailed-like expression pattern, indicating that endogenous trol can be activated in response to transgene-dependent GAL4 activity (data not shown).

We next asked whether ectopic trol expression interferes with normal embryonic and larval development. GAL4 driver lines were crossed with EP(1)1160 to overexpress trol in response to (1) the maternal and ubiquitous V3-GAL4 driver (Häcker and Perrimon, 1998), (2) the sca-GAL4 driver in neurons (Budnik et al., 1996), (3) the en-GAL4 driver in a series of stripes along the longitudinal axis (Tabata et al., 1992), and (4) the Actin5C-GAL4 driver to obtain constitutive ubiquitous expression (Ito et al., 1997). Lethal and phenotypic consequences of ectopic trol expression were examined in each case. Much to our surprise, only minor effects of the ectopic trol overexpression could be observed in low penetrance. They include defects in the arrangement of macrochaete in response to neurospecific trol expression and extra wing veins in response to ectopic trol activity. These observations indicate that ectopic expression of trol under the experimental procedures described has only subtle effects on development as compared to, for example, the overexpression of heparan sulfate proteoglycans encoded by dally-like and dally (Jackson et al., 1997; Tsuda et al., 1999; Strigini and Cohen, 2000; Baeg et al., 2001).


The results shown here provide evidence that the gene trol encodes the Drosophila homolog of Perlecan. Perlecan has been originally identified in the basement membrane (Hassell et al., 1980; Morriss-Kay and Tucket, 1991; Brown et al., 1997; Handler et al., 1997). It is a heparan sulfate proteoglycan (HSPG), proteins that recently have become the focus of interest as modulators of intercellular signals in development and morphogenesis. Growth factors can bind to the glycosaminoglycan side chains of HSPGs (Reichsman et al., 1996; Plotnikov et al., 1999) and mediate receptor-ligand interactions (Perrimon and Bernfield, 2000). Furthermore, genetic studies have implicated HSPGs in fibroblast growth factor (FGF), Wnt/Wingless, Hedgehog, and Decapentaplegic/bone marrow protein function in both vertebrates and invertebrates (reviewed in Perrimon and Bernfield, 2000; Selleck, 2001). Moreover, the Drosophila glypicans Dally and Dally-like have been shown to bind and stabilize Wingless at the cell surface (Strigini and Cohen, 2000; Baeg et al., 2001) and specific glypicans have been implicated in Wnt signaling of zebrafish (Tsuda et al., 1999; Topczewski et al., 2001).

Mammalian HSPGs were shown to promote growth factor receptor binding and influence the mitogenic activity of cells (Aviezer et al., 1994). In particular, Perlecan mediates FGF-receptor binding and thereby stimulates mitogenesis and angiogenesis (Aviezer et al., 1994). In addition, studies on Perlecan knock-out mice have indicated that the protein plays an important role in matrix structure and affects the proliferation of chondrocytes in mice likely by the binding, storage, and sequestering of FGF at the site of expression to regulate appropriate FGF signal activity for proper cartilage and bone development (Arikawa-Hirasawa et al., 1999). The key result of our analysis, i.e., that the trol gene codes for Perlecan, indicates that the homologous HSPG protein plays an important role in the control of proliferation activation in a subset of quiescent neuroblasts of the Drosophila CNS. The results suggest that trol functions by a mode of action similar to mammalian Perlecans by either conferring a ligand and ligand receptor binding specificity or by binding and storing corresponding signals at the site of expression necessary for the proper regulation of proliferation.

trol was originally identified in a genetic screen for abnormal larval brain morphology that was due to defective patterns of neuroblast proliferation in the larval brain (Datta and Kankel, 1992). The subset of trol-dependent neuroblasts is normally kept quiescent in response to the gene product of ana (Datta, 1995), a secreted glial glycoprotein that represses premature proliferation of optic lobe and thoracic neuroblasts (Ebens et al., 1993). Several studies on trol and ana have led to the proposal that trol may regulate the reactivation of neuroblast proliferation by suppressing or bypassing the repression by ana, thereby stimulating the G1/S transition through up-regulation of Cyclin E expression (Caldwell and Datta, 1998). Other factors required for the activation of mitotically quiescent neuroblasts include the hormone ecdysone and a transacting factor of unknown identity, which is produced in response to the activity of eve (Park et al., 2001). The homeodomain transcription factor Even-skipped acts in a cell-autonomous manner in areas outside the regulated neuroblasts and is, therefore, likely to control a diffusible signal that impacts neuroblast proliferation in a trol-dependent manner (Park et al., 2001). In contrast, the activating effect of ecdysone on neuroblast proliferation occurs in a trol-independent manner (Datta, 1999). On the basis of these results, a trol pathway had been proposed in which both ana and eve take part. Our finding that trol encodes Perlecan provides a molecular basis toward an understanding of how a trol-dependent pathway could function through the binding and sequestering of proliferative signals.

A surprising observation was that Perlecan is not expressed in quiescent neuroblasts of the optic lobe area but rather in only few brain cells outside the area. Although we have no definite proof for the identity of these cells, it appears likely that in analogy to the early expression pattern observed in the embryo, these cells represent a subset of glia cells. Glia cells were shown to produce the secreted ana protein that is necessary to prevent premature optic lobe neuroblast proliferation (Ebens et al., 1993). These observations suggest that, like in the case of mammalian Perlecan, the Drosophila homolog is capable to bind, store, and sequester proliferation-controlling signals that derive from few specialized glial cells to regulate appropriate signal activity to be received and processed by neighbouring neuroblasts (Arikawa-Hirasawa et al., 1999). It is conceivable that Perlecan participates in the control of both activating factors such as those generated in response to eve activity and repressing factors like the secreted ana glycoprotein.

The Perlecan-dependent control of cell proliferation by promoting ligand-receptor interaction for FGF signal, as proposed on the basis of in vitro studies (Aviezer et al., 1994), would not be consistent with Perlecan function in the fly, because the promotion of any kind of signal-receptor interaction as well as stabilization of the extracellular matrix would require the expression of Perlecan close to or within the target cells. It may, therefore, also not act in a manner analogous to the Drosophila glypicans Dally and Dally-like, recently shown to bind and stabilize the Wingless signal molecule at the cell surface of target cells (Strigini and Cohen, 2000; Baeg et al., 2001). The finding that Perlecan is not expressed in most of its functional target cells, the quiescent optic lobe neuroblasts, is more consistent with the conclusion drawn from the mouse knock-out mutant studies, suggesting that Perlecan binds, stores, and sequesters ligand molecules and thereby modulates the signal activity. This conclusion would also be in agreement with the observation that, in contrast to overexpression of the glypican Dally-like protein (Baeg et al., 2001), the overexpression of Perlecan has only subtle effects on development and morphogenesis. This result argues further that factors that generate and/or modify heparan sulfate chains of Perlecan, such as the products of the genes sugarless, sulfateless, and fringe connection (Selva et al., 2000), are spatially and temporally restricted and are necessary to contribute in a cell-specific manner to proper Perlecan activity. Alternatively, Perlecan expression in glia cells might reflect elevated expression only. It is, thus, possible that low uniform expression might be important, in a cell autonomous manner, for the role of Perlecan in neuroblasts.

The identification of Perlecan as the trol gene product is consistent with a model proposing the ana protein and the unknown eve-dependent factor as putative direct interactors, which are depleted from the interstitial fluid or become enriched by association with Perlecan. This scenario would suggest that the control that trol exerts on neuronal proliferation is dependent on its timely regulated ability to bind to the growth factors proper, rather than on its place of expression. In analogy to mammalian Perlecan, the Drosophila homolog may also interact with FGF and possibly also other ligand molecules that promote cell proliferation and/or patterning processes during fly development, a proposal that will be addressed by future experiments.


Drosophila Strains

l(1)G P-element insertion strains were generated and mapped as described (Peter et al., 2002). The stocks trol13 (Judd et al., 1972), Actin5C-GAL4 (Ito et al., 1997), and EP(1)1160 (Rorth, 1996) have been obtained from the Bloomington Stock Center. EP(1)1619 (Harvie et al., 1998) was obtained from the Szeged Drosophila Stock Center, trol15 (Judd et al., 1972) from the Umea Drosophila Stock Center. FM6/Y; TM2, ry P{ry+Δ2-3}(99B)/MKRS, Sb P{ry+Δ2-3}(99B) (Robertson et al., 1988) served as transposase source. V3-GAL4 (Häcker and Perrimon, 1998), sca-GAL4 (Budnik et al., 1996) and en-GAL4 (Tabata et al., 1992) were obtained from D. Brentrup. Targeted misexpression was accomplished by using the GAL4/UAS system (Brand and Perrimon, 1993).

l(1)G023, l(1)G0271, and l(1)G0374 were analyzed by inverse PCR. All carry a single P-element insertion within 500 bases upstream of a putative transcriptional start site of the trol transcription unit. Each mutation failed to complement the trol13 and trol15 alleles. The P-element of EP(1)1160 is inserted in the same DNA interval as the other insertion lines but is homozygous viable and fertile with no morphologically discernible phenotype. By transposase-induced P-element excision, alleles l(1)G023, l(1)G0271, and l(1)G0374 have been reverted to viability as described in Peter et al. (2002).

The trolnull deletion was generated by crossing the P{ry+Δ2-3}(99B) (Robertson et al., 1988) jumpstarter chromosome into the y l(1)G0271/EP(1)1619 B background to supply a source of transposase. Flies in which excision events had occurred were detected by scoring the progeny for the loss of the white+ eye color marker of the P-elements and recombination between the y and B markers. The corresponding progeny was balanced with FM6 to establish stocks, which were screened for lethality. Genomic DNA was prepared from individuals of the lethal stocks and used for PCR by using the primers 5′-GCATACTTTTGGGCGCGCTTCG-3′ and 5′-ACGAACCAAATGTGTGAAA-ATTGGC-3′ directed to sequences on either side of the two original P insertions. The resulting PCR product of 294 bp was sequenced to establish the exact genomic breakpoints of the deletion trolnull (see Fig. 2A).

Sequencing of cDNA Clones and Sequence Analysis

The cDNA clones GM03359, SD04592, GM01116, GM01493, and HL01107 (Rubin et al., 2000) were sequenced on both strands completely by using an ABI377 sequencing machine (Perkin Elmer) and Sequencher software (Gene Codes Corporation). Homologous sequences were identified with BlastN at the NCBI (URL: Protein motifs in the sequences were identified by using Motifs (GCG 10; Genetics Computer Group Wisconsin) and Blocks+ Search (URL:; Henikoff and Henikoff, 1994). Multiple alignments of the full-length Perlecans and the individual protein domains were done by using Pileup (GCG 10; Genetics Computer Group Wisconsin) and shaded by using MacBoxshade 2.15 (URL:

Immunohistology and RNA In Situ Hybridisation

Drosophila embryos were fixed and permeabilised as described (Tautz and Pfeifle, 1989). Imaginal discs and brains of larvae of various stages and ovaries of 3- to 4-day-old females were dissected in Grace's cell culture medium (GIBCO), fixed in 16.7 mM potassium phosphate, 75 mM potassium chloride, 25 mM sodium chloride, 3.3 mM magnesium chloride, and 5 % (w/v) paraformaldehyde for 30 min and then used for immunohistochemistry or RNA in situ hybridisations. In situ hybridisations were performed by using digoxigenin- (DIG) and fluorescin isothiocyanate (FITC) -labelled probes (Hauptmann and Gerster, 2000) with a hybridisation temperature of 55°C. Probes were detected by alkaline phosphatase-coupled anti-DIG and anti-FITC antibodies (Roche, Mannheim) and NBT/BCIP staining.

Primary antibodies were mouse mAb22c10 anti-futsch (1:50; DSHB, Iowa), mouse mAb1D4 anti-FasciclinII (1:5; DSHB, Iowa), rabbit anti-Perlecan (1:3,500; Friedrich et al., 2000) detected by Alexa 488 or Alexa 546 secondary antibodies (1:500; Molecular Probes, Leiden). Phalloidin-Alexa 546 and TOTO-3 were included before the final wash step for 20 min at 1 μg/ml and 0.1 μg/ml, respectively. Images were taken on a Zeiss Axiophot for brightfield, darkfield, and differential interference contrast microscopy. Fluorescent labels were detected by using a Zeiss LSM 410 laser scanning microscope. Pictures were processed by using Photoshop 5 (Adobe) and assembled by using Freehand 8 (Macromedia).


We thank our colleagues in the lab for critical support. Special thanks to Gordon Dowe for sequencing and S. Baumgartner for sharing unpublished results.