Abstract

We developed a method to highly purify germline stem cells (GSCs) from the Drosophila ovary, one of the best understood types of adult stem cell. GSCs express variant isoforms of general transcriptional components, translation initiation factors, and several variant ribosomal proteins, including RpL22, a protein enriched in several mammalian stem cells. These novel isoforms may help regulate stem cell gene expression because a reversion assay indicated that at least four were specific for GSCs. By comparative analysis, we identify additional genes enriched in GSCs, including Psc, the Drosophila homolog of the Bmi-1 Polycomb group gene, as well as genes that may delay cytokinesis in pre-meiotic germ cells. By comparing GSCs arrested by BMP over-expression and bam mutation, we hypothesize that mRNA utilization is modulated in differentiating GSC daughters. Our findings suggest that Drosophila and mammalian stem cells utilize at least two regulatory mechanisms in common.

Keywords: Germ cell; Stem cell; Expression profile; Psc; Drosophila; RpL22

Abbreviations: GSC, germline stem cell; PGC, primordial germ cell; BMP, bone morphogenetic protein; FACS, fluorescence-activated cell sorting; PI, propidium diiodide; Prc1, Polycomb repressive complex 1; Tbps, TATA-binding proteins; TAFs, TATA-associated factors; hTR, human telomerase RNA

Article Outline

Introduction

Materials and methods

Culture of Drosophila strains and cells

Immunostaining and fluorescence microcopy

Purification of expanded GSCs

Purification of GSCs by FACS sorting

Induced stem cell differentiation and reversion

Microarrays

Annotation

Results

Purification of GSCs from genotypes with an expanded GSC population

Expanded GSCs delay cytokinesis like normal GSCs and PGCs

Purified GSCs express a representative gene profile

GSCs express few differentiation markers

GSCs express novel mRNAs encoding general transcription and translation factors

GSCs express high levels of Psc and other Polycomb group genes

Energy metabolism may be altered in GSCs

Candidate genes that control the GSC cell cycle

Candidate genes for cystoblast differentiation

Identifying GSC-specific gene candidates by induced differentiation and reversion

Discussion

GSCs may be maintained by mechanisms acting at multiple levels

Comparing cell structure and physiology using arrays

Modifications of transcription machinery in GSCs

Translational control of GSC maintenance and differentiation

Comparison to genes associated with mammalian GSCs

Multiple stem cells may utilize the Polycomb group gene Psc/Bmi-1

RpL22 may also be important in invertebrate and vertebrate stem cells

Acknowledgements

References

Introduction

Stem cells maintain and replenish the tissues of multicellular organisms, a function that when compromised may lead to deficiency, premature aging, or cancer. Despite their importance, however, the rarity of stem cells and their location within complex tissues continues to limit our knowledge (reviewed by Fuchs et al., 2004 and Ohlstein et al., 2004). A potentially powerful way to gather information about stem cells is to determine their gene expression profiles (reviewed by Kawasaki, 2004). Such studies identify genes and pathways whose function in stem cell biology can subsequently be tested. Recently, based on such comparisons, stem cells have been proposed to share a suite of common characteristics that contribute to their ability to serve as long-term cellular progenitors (Ramalho-Santos et al., 2002 and Ivanova et al., 2002). The evolutionary conservation of stem cell mechanisms can also be probed by comparing the gene expression patterns of similar stem cells from diverse species.

Among of the most accessible stem cell types for functional and molecular studies are the germline stem cells (GSCs) of the Drosophila ovary (reviewed in Lin, 2002 and Ohlstein et al., 2004). 2–3 GSCs normally reside at the tip of each ovariole within the ovary where under optimal nutritional conditions they divide asymmetrically every 20 h to produce a stem cell and a daughter cystoblast (Fig. 1A). Cystoblasts turn on the differentiation gene bag-of-marbles (bam) and divide synchronously to generate 16-cell germ cell cysts interconnected by ring canals that become surrounded by somatic cells to form a new follicle. Cystoblasts also initiate a complex process of cytoskeletal and cytoplasmic differentiation that is manifest by changes in the germ cell structure known as the fusome (de Cuevas and Spradling, 1998). Non-dividing somatic cells known as cap cells support a niche by adhering to GSCs and by activating BMP signaling that is essential for GSCs to repress bam transcription and remain stem cells (Ohlstein and McKearin, 1997, Xie and Spradling, 1998, Xie and Spradling, 2000, Song et al., 2002, Chen and McKearin, 2003 and Casanueva and Ferguson, 2004). Recently, many of the same mechanisms that regulate adult ovarian GSCs have been shown to maintain the late embryonic and larval populations of primordial germ cells (PGCs) (Zhu and Xie, 2003, Gilboa and Lehmann, 2004 and Kai and Spradling, 2004).

Germline stem cells have also been widely studied in male gonads. In the Drosophila testis, GSCs reside in an apical niche adjacent to non-dividing hub cells. The regulation of male and female GSCs has much in common, including niche cell adhesion (Yamashita et al., 2003), bam repression by BMPs (Shivdasani and Ingham, 2003, Kawase et al., 2004, Schulz et al., 2004, Song et al., 2004 and Bunt and Hime, 2004), and ring canal/fusome morphogenesis (Lin et al., 1994, Hime et al., 1996 and de Cuevas and Spradling, 1998), although details differ between the sexes. In both sexes, germline cystocytes were recently shown to be capable of reverting into functional GSCs in vivo (Kai and Spradling, 2004 and Brawley and Matunis, 2004). However, unlike the situation in female GSCs, JAK-STAT signaling functions as the major male GSC regulatory signal (Tulina and Matunis, 2001 and Kiger et al., 2001).

Mouse testes also contain niches that support large numbers of GSCs along the basal surface of the seminiferous tubules (reviewed in Zhao and Garbers, 2002). Mouse GSCs give rise to cells that divide asymmetrically to form clusters of interconnected A-type spermatogonia that are analogous to cystocytes. Mouse GSCs can be cultured with other testicular cells in vitro on feeder cells and retain the ability to function when re-introduced into a host animal. Recently, conditions have been found that either maintain enriched mouse or rat GSC populations, or cause them to differentiate into clusters of spermatogonia-like cells. This has allowed a profile to be determined of genes associated with GSC maintenance in culture (Hamra et al., 2004).

Here we report a method to expand Drosophila ovarian GSCs in vivo and to enrich them by fluorescence-activated cell sorting (FACS) to essentially 100% purity. We describe several novel genes expressed by GSCs and their possible implications for the mechanisms of stem cell maintenance and function. We take advantage of our ability to cause GSCs to begin differentiation and then to revert back to the stem cell state to associate several of these genes with GSCs. Finally, our study allows us to compare some of the genes expressed by Drosophila and mouse stem cells.

Materials and methods

Culture of Drosophila strains and cells

Drosophila were cultured on standard food. Genotypes used are described in Kai and Spradling (2004) and in Flybase (http://www.flybase.bio.indiana.edu). The Psc-lacZ strain was isolated in a large enhancer trap screen (Karpen and Spradling, 1992). Michael Buszczak constructed the Arginine kinase–GFP fusion strain by inserting a P element encoding a GFP exon within the endogenous gene according to the method of Morin et al. (2001). c587-Gal4;vas-GFP;UAS-Dpp or vas-GFP;bam^Δ86 females were fed additional yeast for several days and aged 7–14 days prior to the harvest of GSCs to allow the number present to increase. Drosophila Kc cells were grown in serum-free HyQ-CCM3 medium (HyClone Laboratories, Inc.) at room temperature.

Immunostaining and fluorescence microcopy

Ovaries were stained as described in Kai and Spradling (2003). Isolated cells were stained on poly-lysine-coated slides. Briefly, cells were allowed to settle for 45–60 min in a humid chamber, fixed, and stained as for ovaries. After fixation, the cells were washed with PBS and pre-absorbed with PBS containing 0.2% Triton X-100 and 5% normal goat serum (PBT-NGS) for 30 min or longer. Cells were subsequently washed 3 times with PBS, and incubated with primary antibody diluted in PBT-NGS for 1 h. After 3 additional washes with PBS, the cells were stained with secondary antibody diluted in PBT-NGS for 1 h. Cells were washed 3 more times with PBS and mounted in PBS containing 50% glycerol and 1 μg/ml of DAPI. Images were taken using a Leica TCS-NT, Leica TCS SP2, or Zeiss Axiovert S100 microscope.

The following anti-sera were used: rabbit anti-GFP (Torrey Pines Biolabs, Inc); rabbit anti-β Galactosidase (Cappel) (1:1000); mouse monoclonal antibody 1B1 (anti-Hts) from Developmental Studies Hybridoma Bank (1:50); rat anti-α-tubulin (Oxford Biotech); rabbit anti-Anillin was a gift from Dr. Christine Field (1:500); and secondary antibodies were goat anti-mouse, or goat anti-rabbit IgG conjugated to Alexa 488 or Alexa 568 (Molecular Probes) (1:400).

Purification of expanded GSCs

150–300 ovary pairs from flies with expanded GSCs were dissected in Schneider's media supplemented with 10% FBS, and rinsed 3 times with calcium-free PBS. Then ovaries were incubated in PBS with 0.5% of Trypsin (Invitrogen, Inc.) and 2.5 mg/ml collagenase (Invitrogen, Inc.) for 15 min at room temperature, with intermittent vigorous shaking. Cell suspensions were filtered twice through a 40-μm nylon mesh. Cells were collected by centrifugation at 1000 ×  g for 7 min and re-suspended in 0.5 ml of Schneider's media supplemented with 10% FBS and 10 μg/ml of propidium iodide, incubated for 30 min at room temperature, and immediately sorted.

Purification of GSCs by FACS sorting

Fluorescence-activated cell sorting (FACS) were performed with the Becton Deckinson FACSCalibur using CELLQUEST software. Alive germline cells expressing vas-GFP were sorted by gating green-positive and red-negative cells with the exclusion mode. Sorted cells were collected, kept at −80°C in the presence of 0.5% SDS and 0.5 μg/ml of proteinase K until RNA extraction.

Induced stem cell differentiation and reversion

Two-day-old adult females of the genoype w¹¹¹⁸c587-GAL4;P[w+, hsp70-bam]^11dUAS-dpp were heat shocked at 37°C for 2 h and cultured at room temperature. RNA was prepared from ovariole tips dissected from 50 to 100 flies either just before the heat shock (0 h), or 20 h or 50 h afterwards. Purification of RNA, probe preparation, and analysis on microarrays were subsequently carried out as with purified GSCs.

Microarrays

RNAs were extracted from 7 to 8 × 10⁵ of the isolated GSCs and 5 × 10⁶ of Kc cells using the RNeasy kit (Qiagen) according to the manufacturer's protocol. cDNAs were synthesized from 5 to 15 μg of total RNA using the Superscript double-stranded cDNA synthesis kit (Invitrogen) followed by the Expression Analysis Technical Manual (Affymetrix). cRNA reactions were carried out using the BioArray High-Yield Transcript labeling kit (Enzo). 13–15 μg of cRNAs were fragmented in fragmentation buffer (40 mM Tris-acetate pH 8.1, 100 mM KOAc, 30 mM MgOAc) at 94°C for 35 min. Probes were labeled and reacted with Drosophila gene arrays (Affymetrix, Inc. version 1 that analyzes approximately 13,600 genes) at the MIT/HHMI Biopolymers lab. The 3′ to 5′ end labeling ratios of the probes used were in the range of 1.8–5.3, ruling out significant RNA degradation prior to reverse transcription. The data were initially interpreted using the Affymetrix analysis software package MAS 5.0 (Liu et al., 2003). Presence/absence calls are based on a comparison of the signal levels observed with perfectly matched oligonucleotide probes specific for the gene in question, to the signal from similar probes containing single nucleotide mismatches using the default parameters (see Affymetrix, 2004). Values from three replicate arrays were averaged. All values were rounded to 3 significant figures for presentation in the Tables. One of the 9 replicates had a correlation coefficient of r = 0.92 with its two duplicates, instead of r > 0.96 as for the rest of the data. We determined that excluding or including this replicate had no effect on any conclusion in the paper. In addition, a few rare values were discarded (1% of genes) where a signal was clearly anomalous on one replicate (for example, absent in the case of a strongly expressed gene). Including these values in the averages would not have affected any of the conclusions in the paper.

Annotation

Annotation of gene function was based on the Affymetrix software, Drosophila gene ontology (http://www.geneontology.org), and RNAi screen data (Boutros et al., 2004, Somma et al., 2002 and Eggert et al., 2004) followed by manual correction of the results based on primary literature. Additional data sets for comparison including those of Hamra et al. (2004) were obtained from the Entrez GEO site (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?CMD=search and DB=geo). Excel Pivot tables containing the raw expression values determined by MAS5.0 for all the experiments described in this paper are available from this site.

Results

Purification of GSCs from genotypes with an expanded GSC population

We succeeded in purifying GSCs by first circumventing the problem of their rarity. Wild-type ovarioles contain only 2–3 GSCs that constitute an insignificant fraction of the total germ cell mass. However, GSC number can be increased using two methods that simultaneously eliminate all other germ cell stages. When Dpp (a Drosophila BMP) is over-expressed in the somatic cells surrounding the stem cell niche, GSCs divide and accumulate (Xie and Spradling, 1998) because bam expression (and hence differentiation) cannot be induced. A virtually identical “stem cell tumor” forms in bam mutant animals (Fig. 1B). To facilitate the purification of stem cells from both genotypes we marked germ cells using the germline-specific vasa gene fused to GFP (Nakamura et al., 2001). We used Drosophila Kc tissue culture cells growing under conditions where they doubled approximately every 20 h as a comparison population.

GSCs were purified over a period of 2–3 h starting with groups of 200–400 ovaries containing expanded GSC populations (see Materials and methods). Individual cells were released by protease treatment and sorted on a FACs machine (Fig. 1C). Approximately 20–40% of the cells fell into a well-separated population of GFP^hi PI^low cells. Upon microscopic examination more than 99.9% of the cells recovered from this sub-population exhibited high level of Vasa-GFP expression and contained a normal-looking spectrosome as judged by morphology and anti-Hts labeling (Fig. 1D). Initially, each ovary contained 1000–2000 GSCs and we collected 150,000–400,000 sorted GSCs per experiment, representing an estimated 20–50% recovery. This protocol worked equally well when GSCs were expanded by Dpp mis-expression or bam mutation (Figs. 1E and F). In both cases, the expanded GSCs were collected, characterized, and the RNA extracted and stored for future use.

Expanded GSCs delay cytokinesis like normal GSCs and PGCs

The isolated GSCs appeared to be virtually identical to normal stem cells located at the tips of Drosophila ovarioles. True GSCs (de Cuevas and Spradling, 1998) and larval germ cells (Kai and Spradling, 2004) undergo a non-canonical cell cycle typified by delayed cytokinesis in the absence of cell cycle arrest. Consequently, daughters remain connected following each mitosis via an arrested cytokinesis furrow within which a new “plug” of fusome material accumulates. We observed that purified GSCs contain a high frequency of cell pairs bridged by fusome structures, strongly suggesting that they divide according to the normal stem cell cycle (Figs. 1E and F). Staining dpp-expanded GSCs and bam-expanded GSCs in intact ovaries with antibodies that recognize fusomes, spindles, or ring canals, confirmed that these cells undergo an asymmetric M phase (Figs. 2A and E), fusome plug growth in G1 (Figs. 2B and F), fusome fusion in S (Figs. 2C and G), and ring canal closure in late S/G2 (Figs. 2D and H). However, expanded GSCs more closely resemble primordial germ cells than GSCs in one character—the absence of associated cap cells (compare Figs. 2I and J). This is probably why the new fusome of expanded GSCs, like PGCs (Kai and Spradling, 2004), grows to the same size as the old fusome, and why these fusome pieces do not stretch out into the “exclamation point” configuration characteristic of GSCs in contact with cap cells (Figs. 2D and H). These differences have no known function in GSCs and bam-expanded GSCs can re-populate the germline of host embryos and produce functional gametes (Niki and Mahowald, 2004).

Purified GSCs express a representative gene profile

RNAs were extracted from 5 to 10 preparations of each cell type, pooled, probes prepared (without amplification), and Affymetrix oligonucleotide arrays representing more than 13,000 Drosophila genes were screened. Raw data were analyzed using Affymetrix MAS5.0 software, which determines from the signals associated with multiple gene-specific oligonucleotides whether RNA homologous to the gene is likely to be present or absent, as well as a quantitative expression value (Liu et al., 2003). Hereafter, when transcripts from a gene are said to be absent, this means the transcripts could not be experimentally detected by this criterion. Three separate arrays were used for each sample and the results found to be highly reproducible. For each of the probes 5200–5650 genes were called as present, and the correlation coefficients (r) of their log transformed expression levels were greater than 0.96 (see Materials and methods). The mean variances in gene signals between replicates were only 0.16–0.19.

We used multiple criteria to determine if the gene profiles obtained from purified GSCs are in fact representative of GSCs in vivo. First, to identify variables associated with cell and RNA probe purification, we compared the profiles obtained from dpp and bam-expanded GSC populations, which are expected to be highly similar. Since different Drosophila stocks were used, and the cell purification, RNA preparation, and analysis of these GSC populations were carried out separately, such a comparison should reveal variation due to uncontrolled technical parameters. Instead, we found that the levels of gene expression in the dpp- and bam-expanded GSCs are extremely similar (Fig. 3A). Among the 5964 genes that were scored as positive in either dpp- or bam-expanded cells, the mean variance was only 0.20 and the correlation coefficient r = 0.97, values close to those between replicates. In comparison, r = 0.79 and the mean variance was 0.54 when the 6290 genes scored as positive were compared between dpp-expanded GSCs and Kc tissue culture cells (Fig. 3B). These observations provide confidence that our measurements of GSC gene expression are reproducible and little-affected by technical factors.

To determine whether purified GSCs express the same genes as GSCs in vivo, we examined which known GSC genes were scored as present in our GSC profiles. To probe GSC specificity, we also calculated the ratio of each gene's expression in GSCs to that in Kc cells. All the genes known to function in GSCs and all genes GSCs have been reported to express based on RNA and protein studies were detected as present (see Table 1). These include the BMP reception pathway (but not its ligands), all known fusome and nuage components, the genes implicated in germline translational regulation, RNAi, adhesion to cap cells, as well as the few known germline transcription factors. Not only are transcripts of these genes present, but many are highly enriched in GSCs relative to Kc cells. Strikingly, GSCs also express 8 genes encoding tudor domain proteins, five of which had not previously been documented in ovarian germ cells: Pcl, CG9925, CG14303, CG15707, and CG7084. Tudor domain proteins are known to be enriched in Drosophila and mammalian germ cells (Bardsley et al., 1993 and Chuma et al., 2003). In contrast, bam and bgcn, two genes that are specifically required for cystoblast development, were under-represented in GSCs vs. Kc cells. As expected, bam transcripts are not detected in GSCs purified from bam mutants (Table 1). Taken together, these observations argue strongly that the purified GSCs express all the genes known to be preset in GSCs within functioning ovaries.

GSCs express few differentiation markers

The degree to which germline and tissue stem cells have become lineage restricted remains an important question. Two aspects of the GSC gene profile argue that these cells are less differentiated than many other adult cells (Web Table 1). First, purified GSCs do not express Hox genes at detectable levels and express very few of the many transcription factors required zygotically for embryonic development, or for tissue differentiation (Web Table 1). Of 59 relatively well-known such genes listed in Web Table 1, only 8, including hairy, Hairless, Rel, tramtrack, U-shaped, and emc were expressed at detectable levels in GSCs. In contrast, Kc cells express the Hox gene Deformed (Dfd), as well as high levels of several transcription factors associated with blood cell differentiation including serpent.

Examination of the most abundant GSC transcripts provides further evidence of their undifferentiated state. GSCs contain 297 genes that constitute 0.05% or more of expressed RNA, while Kc cells contain 360 such genes. Most abundant transcripts in both cell types encode basic cellular machinery such as ribosomal proteins, translation factors, etc., and are expressed at similar levels to make up 48% (GSC) or 56% (Kc) of expressed RNA. However, Kc cells also abundantly express specific genes that are upregulated by activated ras signaling in hemocytes (Asha et al., 2003), as well as blood cell associated growth factors (Pvr, Idgf3), neuropeptide-like precursors (Nplp2), and other tissue-preferential genes. In contrast, the 14 highly transcribed genes expressed preferentially in GSCs (1.5% of expressed RNA) encode special components of the translation and RNAi machinery, or small heat shock proteins. Thus, Kc cells, but not GSCs, express abundant mRNAs characteristic of cellular differentiation.

GSCs express novel mRNAs encoding general transcription and translation factors

The observation that abundant, differentially expressed GSC mRNAs encode proteins involved in gene expression is consistent with the fact that many previously studied “germ cell specific” genes also encode such factors (Table 1). However, our experiments revealed additional previously undescribed GSC-enriched transcripts encoding other general gene expression proteins (Table 2). For example, the CG15398 protein (73-fold enriched relative to Kc) contains a TATA-binding motif and may interact within the TFIID general transcription machinery. Drosophila has three other better-known TBP-like proteins: canonical TBP, Trf, and Trf2. The later two TBP-like proteins are expressed predominantly during male gametogenesis (Crowley et al., 1993 and Aoyagi and Wassarman, 2000). CG15398 expression in female GSCs suggests that the basal transcription machinery may be modulated during oogenesis as well. In addition, TFIIA-S-2 (CG11639; 68-fold enriched) encodes a homolog of TFIIA-S, another general transcription factor component. Expression of TFIIA-S-2 has been shown previously to be enriched in the testis (Ozer et al., 2000).

Several other messages expressed differentially in GSCs encode proteins predicted to form part of the translational apparatus. Most of these proteins are predicted to act at an early step—the interaction of mRNAs with 40S ribosomal subunits. These factors include a new eIF-4E-like isoform (CG8023) and eIF-4E-binding protein (Phas1) that may compete with other components of the mRNA cap-binding eIF-4F complex. These interactions determine whether eIF-4E is available to interact with eIF-4G, and we also observed elevated levels of two genes, CG10192 and CG10990, that encode novel eIF-4G-like proteins. Both are expressed at about 10 times the level in GSCs as in Kc cells, and at higher levels than the endogenous eIF-4G protein in GSCs (Table 2). Other translational differences in GSCs may result from the relatively high expression of CG12413, the Drosophila homolog of IF-2M, the major mitochondrial protein translation factor, and of Efsec, a factor involved in the utilization of selenocysteine (Tujebajeva et al., 2000).

In addition to translation factors, we also observed that four putative ribosomal proteins are expressed preferentially in GSCs. Three of these are likely to be associated with the 40S subunit, and might interact with alternative eIF-4F members. RpS5b is expressed at about the same level as RpS5a, the normal S5, while RpS10b and RpS19b levels are only 7.5% or 33% of their normal counterparts, respectively. A fourth putative ribosomal protein, CG9871, resembles RpL22 but is expressed at only 6.6% the level of normal RpL22. These proteins may modify a subclass of GSC ribosomes and/or carry out novel functions.

GSCs express high levels of Psc and other Polycomb group genes

To better analyze the GSC gene profile, we developed a strategy that is outlined in Fig. 4. For most major aspects of cellular physiology, there likely exist genes whose expression varies with the magnitude to which that structure or process occurs in a cell. For example, Fig. 4A shows hypothetical Kc cells and GSCs with similar amounts of cytoplasm (and hence of ribosomes), but with different amounts of a component of interest such as mitochondria. Under these conditions, calculating the frequency distribution of the relative expression, log₁₀(GSC/Kc), of individual ribosomal protein mRNAs should yield a curve centered on 0. However, the distribution of relative expression of mitochondrial ribosomal protein mRNAs (green curves) is expected to be less than 0 for a cell such as GSC1 with fewer mitochondria per cell than Kc cells, or greater than 0 for a cell such as GSC2 with more mitochondria per cell than Kc cells.

Plots of this type provide insight into the likely relative activity of several physiological pathways in GSCs compared to Kc cells (Figs. 4C–G). The center of the distribution of log₁₀(GSC/Kc) for cytoplasmic ribosomal protein mRNAs is close to 0, as expected (Fig. 4B). GSCs and Kc cells also express very similar levels of mitochondrial ribosomal protein mRNAs (Fig. 4C). However, many chromatin protein mRNAs are significantly elevated in GSCs compared to Kc cells (Fig. 4D). In particular, nearly all mRNAs encoding Polycomb group proteins (blue line) are elevated. The gene showing the greatest increase, more than 8-fold, is Psc, the Drosophila homolog of mammalian Bmi-1. To obtain further information about Psc expression during oogenesis, we identified a Psc-lacZ line (Fig. 5A). Psc is active in GSCs and remains present throughout the germarium. Psc-lacZ labeling subsequently declines and is lost about the time that nurse cell chromatin is remodeled in stage 5 (Dej and Spradling, 1999). These findings suggest that Psc and other Polycomb group proteins play an important role in GSCs, and in their progeny germ cells.

Some pathways are reduced in GSCs compared to Kc cells. mRNAs encoding extracellular matrix components and/or proteins involved in cell–matrix interactions are expressed at relatively low levels in GSCs (Fig. 4D). These include laminin A, laminin B2, the integrin ligand Tigrin, the ECM proteoglycan Papilin, and the cation-dependent collagen binding protein BM-40 (Martinek et al., 2002) mRNAs. GSCs, which normally contact cap cells and inner sheath cells, may depend less on interactions with extracellular matrices than many cells.

Energy metabolism may be altered in GSCs

GSCs express many genes involved in glycolysis or the Krebs cycle at lower levels than Kc cells (Fig. 4F). Transcripts of several such genes, including the glycolytic enzyme Aldolase, and the Krebs cycle enzymes CG1544 (oxoglutarate dehydrogenase), are below detectable levels. mRNA levels for Aconitase, another Kreb's cycle enzyme, lie scarcely above background in GSCs, representing only 15% of the level in Kc cells. These results suggest that GSCs acquire ATP from sources other than internal oxidative metabolism. One possibility is that they employ a phosphagen system such as arginine-phosphate and arginine kinase (Argk) (reviewed in Ellington, 2001) to store and utilize energy imported from neighboring cells. GSCs express Argk mRNA at high levels whereas it was not detected in Kc cells. Analysis of an ArgK-GFP fusion line revealed that the protein is expressed in GSCs and later germ cells (Fig. 5C). Only late region 2a and early region 2b cysts lacked robust expression of the fusion protein.

Candidate genes that control the GSC cell cycle

The process of cytokinesis (reviewed in Balasubramanian et al., 2004 and Strickland and Burgess, 2004) is greatly retarded in GSCs (Fig. 2). Consequently, a larger fraction of the GSC cell population should occupy cell cycle stages where cytokinesis genes (at least those that lie upstream from the arrest point) are highly expressed, elevating the levels of these transcripts. In contrast, positively acting genes responsible for the cleavage furrow arrest itself might show reduced expression, and inhibitory genes might be elevated, when averaged across the cell cycle. The relative expression plot of cytokinesis genes is shown in Fig. 4G. The distribution is broad and includes candidates for both types of genes. Most cytokinesis gene transcripts are found at somewhat higher levels in GSCs than in Kc cells, but two genes, Rho1 and Rop, are expressed at 3-fold lower levels, making them candidates for positively acting factors limiting ring canal closure. Rho1 activity is known to promote cleavage furrow closure (Prokopenko et al., 1999). Transcripts from just one of 7 septins, CG9699, is greatly elevated in GSCs relative to Kc cells, making it a candidate cytokinesis inhibitor.

Candidate genes for cystoblast differentiation

One of the first steps in GSC differentiation is a decrease in the level of BMP signaling activity as the stem cell daughter loses functional contact with cap cells and exits the GSC niche (Kai and Spradling, 2003 and Ohlstein et al., 2004</