The Drosophila melanogaster gene brain tumor negatively regulates cell growth and ribosomal RNA synthesis

The overall size of a metazoan is controlled at the cellular level by the coordinate regulation of cell division and cell growth. Although it has long been established that inappropriate cell division can lead to cancer, it is becoming increasingly clear that cell growth, or increase in cell mass, is of equal importance. For example, the oncogenes Myc, Ras and Cyclin D (Prober and Edgar, 2001) and the tumor suppressors retinoblastoma (White, 1997) and Pten (Gao et al., 2000; Goberdhan et al., 1999; Huang et al., 1999) have all been shown to regulate cell growth. Although the factors that regulate cell division have been extensively studied (Sherr and Roberts, 1999), the processes that control cell growth are just beginning to be elucidated (Stocker and Hafen, 2000).

Given the dependence of cell growth on protein synthesis, regulation of translation is likely to play an important role in growth control. In fact, recent studies have shown that one mechanism of cell growth regulation is achieved through an insulin receptor signaling pathway, one of the most downstream targets of which is the ribosomal protein S6 (Weinkove and Leevers, 2000). Phosphorylation of this protein appears to lead to stimulation of translation of mRNAs that contain 5′-terminal oligopyrimidine tracts. It has been proposed that the coordination between cell growth and division might be achieved through a balance of translation of messages that contain this sequence and those that lack it. Translation of the former, which include translation factors and ribosomal proteins, would favor growth, while translation of the latter, which include genes involved in cell cycle progression, would favor division (Thomas, 2000).

An additional way in which cells can regulate protein synthesis, and therefore growth, is through control of ribosome synthesis. For example, Myc overexpression in mice can induce the transcription of multiple genes involved in ribosome synthesis (Kim et al., 2000). The fruitfly D. melanogaster genes minifly (Giordano et al., 1999) and pitchoune (Zaffran et al., 1998) have been found to be required for ribosomal RNA (rRNA) processing and are also important for organism growth. Notably, pitchoune appears to be a target of Myc in flies (Zaffran et al., 1998). In contrast to genes that are required for ribosome synthesis, Rb appears to function to negatively regulate ribosome synthesis through its ability to inhibit both RNA polymerase I (Cavanaugh et al., 1995) and III (White et al., 1996) transcription.

The only other gene that has so far been shown to negatively regulate RNA polymerase I and III transcription is ncl-1 (for abnormal nucleolus) from the soil nematode Caenorhabditis elegans. Previously, we have demonstrated that ncl-1 functions not only to negatively regulate rRNA synthesis, but also to inhibit cell growth (Frank and Roth, 1998). ncl-1 mutant worms are larger than wild-type worms and have larger cells. Furthermore, they have enlarged nucleoli in almost all of their cells (Hedgecock and Herman, 1995), which is indicative of a higher rate of rRNA synthesis, that results in a higher steady state level of rRNA. Additionally, they have a higher rate of 5S RNA synthesis and probably contain more ribosomes as they contain more protein than do wild-type worms (Frank and Roth, 1998). NCL-1 protein is predominantly cytoplasmic and its levels of expression in cells of the embryo, gonad (Frank and Roth, 1998) and adult somatic tissue (D. J. Frank, PhD thesis, University of Washington: Seattle, 2000) are inversely related to the size of nucleoli: cells with small nucleoli have high level expression of NCL-1, whereas cells with large nucleoli have low level expression.

The gene most similar to ncl-1 is brat (brain tumor) from D. melanogaster (Arama et al., 2000) (G. R. Hankins, PhD thesis, University of Virginia: Charlottesville, 1991). Both genes contain two B-box zinc fingers, a coiled-coil domain and multiple NHL (NCL-1, HT2A and LIN-41) (Slack and Ruvkun, 1998) repeats. The NCL-1 and BRAT proteins are 38% identical overall and 80% identical in the most C- terminal 280 amino acids, the region that contains the NHL repeats (Arama et al., 2000). Homozygous brat mutants die before eclosion and have greatly enlarged brains, up to eight times the normal size (G. R. Hankins, PhD thesis, University of Virginia: Charlottesville, 1991). The brain tumor phenotype of brat mutants is primarily due to expansion of the optic neuroblasts (Kurzik-Dumke et al., 1992). Imaginal discs from third instar brat mutant larvae, although appearing normal in situ, are able to metastasize and form secondary tumors when injected into the abdomen of a wild-type host fly (Woodhouse et al., 1998), thus indicating that brat also has a function in the imaginal discs. Superficially, brat and ncl-1 mutants do not appear to have analogous phenotypes at the organismal level. We now address whether or not they have similar phenotypes at the cellular level. We show that, in addition to being structurally related, ncl-1 and brat are functionally homologous. Therefore, this mechanism for controlling growth through repression of rRNA synthesis is conserved between worms and flies. We propose that excess ribosome synthesis and cell growth may be important aspects of the tumorous phenotype of brat mutants.

Digoxigenin-labeled (Boehringer Mannheim) brat RNA probes were made by transcribing plasmid LD28374, linearized with EcoRI or BamHI, with SP6 polymerase (antisense probe) or with T7 polymerase (sense probe), respectively. Tissue fixation and hybridization were performed as described previously (Kozopas et al., 1998).

For anti-Nop60B staining, dissected larvae were fixed for 30 minutes in 4% paraformaldehyde in PBS containing 0.1% Tween-20. Antibody was used at a 1:500 dilution and detected using a rhodamine-labeled anti-rabbit secondary antibody (Jackson Laboratories). Images of 0.2 μm optical sections were acquired using a Delta Vision microscope and processed using deconvolution software (Applied Precision). Projections of 20 or 30 sections were generated and Adobe Photoshop was used to measure nucleolar areas.

The construct KS+/ncl5′&3′/ncl, which consists of the ncl-1 cDNA flanked by ncl-1 promoter region (8 kb of genomic DNA upstream of the ncl-1 transcription start site) and 1.4 kb of genomic sequence downstream of the ncl-1 polyA site, was able to rescue the Ncl phenotype of ncl-1(e1942) mutant worms (data not shown). The ncl-1 cDNA sequence was replaced by the coding region of brat cDNA LD28374 to create the plasmid KS+/ncl5′&3′/brat. This plasmid was injected at a concentration of 80 ng/μl with the dominant marker Rol (Mello et al., 1991) at 80 ng/μl into the syncytial gonad of ncl-1(e1942) worms. Adult Rol progeny were picked and their progeny were assayed by Nomarski microscopy for any effect on nucleolar size. Five transgenic rescued lines were identified.

The FLP/FRT system (Xu and Rubin, 1993) was used to generate control and brat¹¹ mutant clones. The brat¹¹ allele has a nonsense mutation that creates a truncated protein (Arama et al., 2000). Although it is not known if brat¹¹ is a null allele, it is one of the strongest alleles of brat (G. R. Hankins, PhD thesis, University of Virginia: Charlottesville, 1991) (Woodhouse et al., 1998). Flow cytometry analysis of approximately 20 dissociated wing discs was performed as described (Neufeld et al., 1998). Similar results were obtained from five independent experiments in which mitotic recombination was induced by 37°C heat shock for 1 hour at 48 hours after egg deposition (AED) or for 1.5-2 hours at 72 hours AED. In all cases, discs were analyzed at 115 hours AED. For clone area measurements, a 45 minute heat shock was applied at 48 hours AED and discs were dissected and fixed at 115 hours AED. Areas were measured using Adobe Photoshop and data was analyzed using Microsoft Excel. The chromosome Ub-GFP FRT40A is from C. Martín-Castellanos and B. Edgar (unpublished).

UAS-brat lines were generated by P element-mediated transformation using the pUAST vector (Brand and Perrimon, 1993). The flip-out technique (Neufeld et al., 1998; Pignoni and Zipursky, 1997; Struhl and Basler, 1993) was used to generate clones overexpressing Gal4 in HS-FLP¹²²; UAS-P35; Act>CD2>Gal4, UAS-GFP_NLSS65T (± additional UAS transgenes) animals. Larvae were heat shocked for 45 minutes at 37°C at 72 hours AED. Similar results were obtained in two experiments without P35 using line UAS-brat(5) and seven experiments with P35 using lines UAS-brat(1A) and UAS-brat(5).

Induction of brat overexpressing clones was achieved with 18 to 25 minute heat shocks at 72 hours AED to generate approximately 5-10 clones per disc. Wing discs were dissected from wandering third instar larvae and fixed in 4% paraformaldehyde/phosphate-buffered saline. The number of GFP⁺ cells per clone was counted on a Zeiss Axioplan microscope. Cell doubling times were determined using the formula (log2/logN)hr, where N=median number of cells/clone and hr=time between heat shock and disc fixation.

To address if any functional homology exists between ncl-1 and brat, we first asked whether the two genes have analogous expression patterns. Because NCL-1 protein is most highly expressed in cells with low rates of rRNA and protein synthesis, such as cells of the early embryo (Frank and Roth, 1998) and neurons in the adult (D. J. Frank, PhD thesis, University of Washington: Seattle, 2000), we predicted that brat expression would be highest in cells with low levels of biosynthetic activity. We examined the brat expression pattern in D. melanogaster larvae using RNA in situ hybridization. Using a brat RNA probe, we observed high level brat expression in brains from wild-type third instar larvae (Fig. 1A). This expression was quite uniform throughout the entirety of the brain hemispheres, including the optic lobe. Weaker but fairly uniform expression was also seen in virtually all cells of the imaginal discs. Examples of eye-antennal, wing and leg discs from third instar larvae are shown in Fig. 1B-D, respectively. In the eye disc, higher brat expression levels were observed in small clusters of cells along the morphogenetic furrow (Fig. 1B). These are likely the neuronal preclusters, the first cells in the eye disc to exit the cell cycle and differentiate into relatively metabolically inactive cells. This expression pattern is consistent with a hypothesis that brat, like ncl-1, functions as an inhibitor of cell growth.

To determine whether brat and ncl-1 are functionally homologous, we asked whether the brat gene could rescue the large nucleolus phenotype of ncl-1 mutant worms. The brat-coding region was inserted into a vector such that it was flanked by the ncl-1 promoter region and downstream genomic sequence. This construct was injected into ncl-1 (e1942) worms and nucleolar sizes of transformants were observed using Nomarski optics. While ncl-1 mutants have large nucleoli, ncl-1 worms expressing the brat transgene have small nucleoli that are indistinguishable from those seen in wild type (Fig. 2). Therefore, the brat gene is able to functionally replace the ncl-1 gene in C. elegans, indicating that these two genes not only have similar sequence and are expressed in similar types of cells, but are indeed functionally homologous.

As brat is able to functionally replace ncl-1 in C. elegans, we wanted to learn whether brat mutants show the same cellular phenotypes as ncl-1 mutant worms, such as enlarged cells. To analyze brat mutant and wild-type cells within the same tissue, we used FLP/FRT-mediated mitotic recombination (Xu and Rubin, 1993) in larvae heterozygous for brat¹¹ to generate paired clonal populations of cells: those that are homozygous brat¹¹ and those that are homozygous wild type. In this system, the wild-type chromosome is marked with GFP. Thus, brat¹¹/brat¹¹ cells can be differentiated by the fact that they do not express GFP. Unfortunately, this system does not allow us to separate unrecombined brat¹¹/+ cells from +/+ cells (Fig. 3A). Thus, any cell size difference we observe could be an underestimate of the actual effect if brat¹¹/+ cells are larger than +/+ cells. We expect that cells might be sensitive to brat dose, as we previously found that ncl-1 heterozygous worms have larger nucleoli than wild-type worms (though smaller than ncl-1 homozygotes) (D. J. Frank and M. B. Roth, unpublished). Nevertheless, flow cytometric analysis revealed that brat¹¹/brat¹¹ cells from third instar wing imaginal discs were consistently larger than the internal control brat¹¹/+ and +/+ cells (Fig. 3B,C). Furthermore, area measurements showed that clones of brat¹¹ cells were larger than their corresponding sister clones in wing imaginal discs (Fig. 3D). The increased size of brat¹¹/brat¹¹ cells and clones suggests that, similar to ncl-1, loss of function mutations in brat lead to excess cell growth.

To further characterize any functional relationship between brat and ncl-1, we next asked whether brat affects nucleolar size. We used mitotic recombination as described above to generate brat¹¹/brat¹¹ mutant clones and +/+ control clones in wing imaginal discs. These were stained with an anti-Nop60B antibody (Phillips et al., 1998) to visualize nucleoli. We found that brat mutant cells have nucleoli that are 18-33% larger than wild-type cells (Fig. 4A). Because the size of the nucleolus is indicative of the level of rRNA synthesis (Altmann and Leblond, 1982; Kurata et al., 1978; Moss and Stefanovsky, 1995), this result suggests that brat mutant cells may have a higher level of rRNA synthesis than wild-type cells. To test this hypothesis, we compared rRNA levels in brains and wing imaginal discs from homozygous brat/brat mutants and control brat/+ heterozygotes. We found that while there was no effect on rRNA levels in brains, homozygous brat/brat wing disc cells contained 1.6 times more rRNA than control cells (Fig. 4B). Because rRNA is very stable (Liebhaber et al., 1978) reported a rRNA half-life of at least 700 hours in primary human fibroblasts), this increase in steady state rRNA level is probably due to an increased level of rRNA synthesis in brat mutant cells.

Given that brat mutant cells are larger than wild-type cells, we hypothesized that brat functions to inhibit cell growth, such that overexpression of brat would be expected to lead to a decrease in cell and organ size. Because ubiquitous overexpression of brat resulted in lethality (not shown), the Gal4-UAS system (Brand and Perrimon, 1993) was used to overexpress a wild-type brat cDNA specifically in the developing eye using the eyeless-Gal4 line (ey-Gal4). The eyeless enhancer directs expression in actively proliferating cells of the eye disc (Halder et al., 1998). Expression of brat in the developing eye, using two different UAS-brat lines, resulted in a dramatic decrease in organ size (Fig. 5).

We next used the Gal4-UAS system to overexpress brat in the developing wing using the decapentaplegic-Gal4 line (dpp-Gal4). In this line, Gal4 is expressed between wing veins LIII and LIV (Staehling-Hampton and Hoffmann, 1994). Overexpression of brat led to an obvious decrease in the size of this intervein region using two different UAS-brat lines (Fig. 5). To quantitate this growth inhibition, we measured the wing blade area bounded by veins LIII and LIV and compared it with the area bound by veins LII and LIII (Table 1) which served as an internal control as it was affected only slightly. We found that brat overexpression resulted in a 36% decrease in wing area relative to the control. The decrease in eye and wing size caused by brat overexpression is probably due to a combination of cell growth inhibition and cell death. To determine if there was an effect on cell size in the wing, we counted the number of bristles in a defined area. As each cell in the wing is associated with a single bristle, the inverse of the number of bristles in a region of a defined area gives a relative estimate of cell size. Surprisingly, the UAS-brat line appeared to have increased cell sizes in the wing (Table 1).

Because overexpression of brat in the wing appeared to cause an increase in cell size while inhibiting organ growth (Table 1; Fig. 5), we wanted to examine the effect of brat overexpression in clones of cells, thus allowing us to compare overexpressing and control cells directly in the same tissue. The flip-out technique (Struhl and Basler, 1993) was used to overexpress brat and GFP in clones of cells. We dissociated wing discs from staged larvae in which overexpression was induced and analyzed them by flow cytometry. We found that overexpression of brat resulted in a slight increase in cell size (Fig. 6A,B) with no effect on cell cycle phasing (not shown). Microscopic examination of clones revealed that overexpression of brat led to cell death as evidenced by pycnotic nuclei visualized by DAPI staining of clones (data not shown). To overcome this effect, we also co-expressed the cell death inhibitor P35 (Hay et al., 1994) with brat in the clones and observed an even larger increase in cell size (Fig. 6C,D). P35 expression appears to be somewhat deleterious to cells, as on its own it caused a small but reproducible decrease in cell size. The increased cell size in the presence of P35 is probably due to the fact that P35 expression inhibited the cell death caused by brat overexpression, thus allowing a greater proportion of the brat overexpressing cells to be analyzed.

Although brat overexpression results in enlarged cells, analysis of clone areas showed that brat overexpression actually inhibits total clone growth. We compared the areas of wing imaginal disc clones expressing brat, GFP and P35 with control clones expressing only GFP and P35, and found that brat overexpression led to a significant decrease in clone area (Fig. 6E).

Because brat overexpression inhibited clone growth yet resulted in enlarged cells, we hypothesized that brat might be causing a slowing of cell division. To address this possibility, we induced clones to express brat, P35 and GFP at 72 hours AED, and counted the number of cells per clone 43 hours later. Clones expressing brat had significantly fewer cells than control clones expressing only P35 and GFP (Fig. 7). We calculated that these cells had 50% longer doubling times than control cells (Fig. 7). Thus, overexpression of brat resulted in a slowing of cell division. As cell size is controlled by the rates of both cell growth and cell division, we interpret the fact that brat overexpressing cells are larger than control cells to mean that the inhibition of cell division rate is more severe than the inhibition of cell growth.

Because brat mutant wing imaginal disc cells contained more rRNA than control cells (Fig. 4B), we wished to determine if brat overexpression would inhibit rRNA accumulation. We used the flip-out method to generate clones overexpressing GFP and P35 as a control, or GFP, P35 and brat in larvae. We then isolated wing imaginal discs from third instar larvae, dissociated the cells and used fluorescence activated cell sorting (FACS) to isolate GFP-expressing and non-expressing cells. RNA was isolated from equivalent numbers of cells and the relative amount of rRNA was determined. While GFP + P35 expression led to an increase in total rRNA, cells overexpressing brat, GFP and P35 had approximately half as much rRNA as control cells (Table 2). This decrease in rRNA per cell occurs even though brat overexpressing cells are larger than control cells (Fig. 6D). Given that rRNA is very stable (Liebhaber et al., 1978), this decrease in the steady state level of rRNA in cells overexpressing brat probably represents a significant downregulation of rRNA synthesis. We conclude that brat can negatively regulate the level of cellular rRNA.

Our results demonstrate that brat functions to repress ribosomal RNA synthesis and cell growth. We found that brat mutant cells are larger than control cells, have enlarged nucleoli and contain excess rRNA. Furthermore, brat overexpression inhibits clone and organ growth, and leads to a decreased level of rRNA per cell. Excess cell growth may be a requisite precursor to the excess cell division that is observed in brat mutant brains (G. R. Hankins, PhD thesis, University of Virginia: Charlottesville, 1991). There is growing evidence to suggest that this model of hyperplasia being preceded by hypertrophy may be an important mechanism of tumor formation. For example, excess cell growth is seen before transformation in mice in which Myc is overexpressed (Iritani and Eisenman, 1999), and many of the transcriptional targets of Myc are genes involved in cell growth (Coller et al., 2000). Additionally, tumor promoting agents such as phorbol esters cause rapid increases in ribosomal RNA transcription (Allo et al., 1991; Garber et al., 1991; Vallett et al., 1993), suggesting that excess ribosome synthesis may also be an important early step in transformation. In future, it will be important to try to understand how it is that excess cell growth and ribosome synthesis can trigger excess cell division.

The brat gene has been shown to function in both the brain and imaginal discs of D. melanogaster larvae. brat mutants have enlarged brains, and imaginal discs from brat mutants can form tumors when transplanted into a wild-type host. Furthermore, we have shown that brat is expressed in imaginal discs and that mutant wing imaginal disc cells are larger than control cells and contain more ribosomal RNA. However, brains do not contain excess rRNA. So why do brat mutants get brain tumors but not tumorous discs? This may be due to the plasticity of the imaginal discs. For example, experimental manipulations that affect cell division rates are compensated for by changes in cell growth so that the disc always ends up the same size, regardless of its number of cells (Neufeld et al., 1998). Therefore, if brat mutant cells in wing discs are larger than wild-type cells, there may be compensation so that the wing does not overgrow. In the brain, such compensation might not exist; thus, excess cell growth might stimulate excess cell division and result in an overgrown brain containing normal sized cells with normal rRNA levels.

If brat and ncl-1 are functional homologs, then why do mutations in these two genes not result in the same phenotype? The answer to this probably lies in the fact that flies and worms have very different patterns of development. While worms have determinate lineages in which every cell division and cell fate decision is absolutely identical from one worm to the next (Sulston et al., 1983), this is not the case for flies. Studies of clonal populations of cells have shown that cell proliferation patterns in D. melanogaster imaginal discs differ from one fly to the next (Bryant, 1970; Bryant and Schneiderman, 1969). Furthermore, in D. melanogaster development, proliferation is often temporally separate from differentiation. For example, the wing disc starts as an embryonic primordium of about 50 cells that grow and proliferate during the four days of larval development to result in 50,000 cells that finally differentiate to form the adult wing (Cohen, 1993). By contrast, the only tissue type in C. elegans in which a stem cell population divides throughout life is the germline. In fact, the germline is the only tissue in worms in which tumorous phenotypes have been clearly observed (Schedl, 1997). This tissue does not become tumorous in ncl-1 mutants, most probably because ncl-1 does not function in the germline. In support of this, the germline nuclei have very large nucleoli that do not appear to enlarge in ncl-1 mutants, possibly indicating that these nuclei are synthesizing ribosomes at maximum capacity even in wild-type worms.

Given that brat mutant cells are larger than wild-type cells, it seems surprising that overexpression of brat should also result in larger cells. brat overexpression also resulted in a slowing down of cell division; the doubling time for brat overexpressing cells was 21 hours compared with 14 hours for wild type. A possible model to explain these results comes from recent work showing that mouse liver cells in which the 40S ribosomal protein S6 was conditionally knocked out were able to grow but not proliferate in the absence of nascent ribosome synthesis (Volarevic et al., 2000). These authors suggest that cells will not divide unless there is a sufficient level of nascent ribosome synthesis. Applying this model here, because overexpression of brat leads to a dramatic decrease in rRNA synthesis (and therefore decreased ribosome synthesis), cell division is slowed. Cell growth also is slowed, as evidenced by the small brat overexpressing clones. Growth is not completely inhibited, however, as cells are able to use ribosomes synthesized before the onset of brat overexpression. The end result, therefore, is large, slowly dividing cells.

Previous studies have demonstrated that activation of the insulin receptor and its downstream targets affects the activity of ribosomes and ultimately regulates cell, organ and organism size in D. melanogaster (Weinkove and Leevers, 2000). For example, the phosphoinositide 3-kinase (PI3K) Dp110 is a positive regulator of growth; its overexpression leads to increased cell growth (Leevers et al., 1996; Weinkove et al., 1999) while Dp110^–/– cells are smaller than wild-type cells (Weinkove et al., 1999). Conversely, D. melanogaster Pten is a negative regulator of the insulin receptor/PI3K pathway. Pten^–/– cells are bigger than wild-type cells, while overexpression leads to decreased cell growth (Gao et al., 2000; Goberdhan et al., 1999; Huang et al., 1999). One of the final downstream targets of this pathway is ribosomal protein S6. Phosphorylation of ribosomal protein S6 results in increased translation of a set of mRNAs that contain a unique 5′ sequence; these include mRNAs for ribosomal proteins and translation factors (Thomas, 2000).

In contrast to the cell growth regulation pathway of the insulin receptor and its effectors, our results indicate that brat affects cell growth not through the activity of ribosomes, but rather through the regulation of their synthesis. We have shown that, similar to Pten, brat is a negative regulator of cell growth in that brat^–/– cells are bigger than wild-type cells. Unlike Pten, however, brat overexpression results in an increase in cell size, presumably because brat inhibits cell division yet allows some cell growth to continue. Furthermore, Pten is not known to inhibit nascent ribosome synthesis. We therefore propose that brat does not function in the insulin receptor/PI3K pathway, but instead functions in a unique pathway to regulate the synthesis of ribosomes.

The cellular phenotype caused by overexpression of brat is similar to the effect of other genetic perturbations that affect ribosome synthesis. Heterozygous mutations in any one of a large class of genes termed Minutes, most of which encode ribosomal proteins, result in slow growing flies (Lambertsson, 1998). Clones of heterozygous Minute cells are also slow growing (Neufeld et al., 1998) and, in the one case that has been examined, consist of cells that are larger than wild type (Martín-Castellanos and Edgar, 2002). The tumor suppressor Retinoblastoma (Rb) has also been shown to control ribosome synthesis through its ability to inhibit rRNA and 5S RNA synthesis (Cavanaugh et al., 1995; White, 1997). Similar to our results with brat, overexpression of the D. melanogaster Rb homolog Rbf has been shown to slow cell division by 50% while increasing cell size (Neufeld et al., 1998). It is not yet known whether Rbf functions to control ribosome synthesis in flies.

How might brat and ncl-1 affect ribosome synthesis? It has been proposed that in E. coli (Nomura et al., 1984) and in D. melanogaster (Yamamoto and Pellegrini, 1990) rRNA synthesis is regulated by the polysome to free ribosomal subunit ratio. When this ratio is high, rRNA synthesis is upregulated. Conversely, when translation, and therefore this ratio, are low, rRNA synthesis is inhibited. As NCL-1 (Frank and Roth, 1998) and BRAT (Sonoda and Wharton, 2001) are both cytoplasmic proteins, one possibility is that brat and ncl-1 serve as sensors of this ratio. Alternatively, they could directly affect this ratio by serving as translational repressors. Interestingly, recent work has shown that brat functions in the translational regulation of at least one mRNA (Sonoda and Wharton, 2001). Future work should provide insight into the specific mechanism of brat and ncl-1 action.

We thank C. Gee and A. Shearn for the gift of the D. melanogaster lines y w; brat¹¹/CyOY⁺ and y; brat¹⁴/CyOY⁺, T. Lukacsovich for brat cDNA LD28374, and S. J. Poole for the anti-Nop60B antibody. We are grateful to members of the Roth and Edgar laboratories for helpful discussions; to A. de la Cruz for technical advice; and to B. Buchwitz, D. Chalker, S. Parkhurst and D. Prober for critical reading of the manuscript. This work was supported by NIH grants GM48453 to M. B. R. and R0151186 to B. A. E.