The Drosophila Modifier of Variegationmodulo Gene Product Binds Specific RNA Sequences at the Nucleolus and Interacts with DNA and Chromatin in a Phosphorylation-dependent Manner*

modulo belongs to the modifier of Position Effect Variegation class of Drosophila genes, suggesting a role for its product in regulating chromatin structure. Genetics assigned a second function to the gene, in protein synthesis capacity. Bifunctionality is consistent with protein localization in two distinct subnuclear compartments, chromatin and nucleolus, and with its organization in modules potentially involved in DNA and RNA binding. In this study, we examine nucleic acid interactions established by Modulo at nucleolus and chromatin and the mechanism that controls the distribution and balances the function of the protein in the two compartments. Structure/function analysis and oligomer selection/amplification experiments indicate that, in vitro, two basic terminal domains independently contact DNA without sequence specificity, whereas a central RNA Recognition Motif (RRM)-containing domain allows recognition of a novel sequence-/motif-specific RNA class. Phosphorylation moreover is shown to down-regulate DNA binding. Evidence is provided that in vivo nucleolar Modulo is highly phosphorylated and belongs to a ribonucleoprotein particle, whereas chromatin-associated protein is not modified. A functional scheme is finally proposed in which modification by phosphorylation modulates Mod subnuclear distribution and balances its function at the nucleolus and chromatin.

Many eukaryotic gene products elicit remarkable functional diversity, even those involved in basic cellular processes. In particular, increasing evidence indicates that a number of ribosomal proteins fulfill a second function apart from ribosome and protein synthesis (1). Ribosome biogenesis takes place in the nucleolus where rDNA is transcribed in a large precursor RNA which is then processed by nucleotide modifications and specific cleavages into mature rRNAs that eventually assemble with ribosomal proteins into ribosomal subunits (2,3). This process requires a number of additional factors, such as small nucleolar RNAs (snoRNA) 1 and non-ribosomal nucleolar proteins. Our knowledge of nucleolar proteins is limited. However, a shared structural feature is the presence of modules thought to represent functional domains (3). Such modules include basic and acidic stretches, which possibly interact with ribosomal proteins (4), and the so-called RNA Recognition Motif (RRM) (5), which could bind either rRNAs at different stages of maturation or snoRNAs.
Ribosome biogenesis is tightly regulated in a close relationship with cell growth or differentiation. It has long been shown that changing cell culture conditions in a way that either improves the growth rate or induces a differentiation program results in enhanced versus decreased ribosome biosynthesis, respectively (6). Moreover, mutations that disrupt gene functions necessary for ribosome assembly often result in growth defects. In yeast or mammalian cell lines, requirement for cell growth has been demonstrated for several snoRNAs (7) and for all the components of RNase MRP, the best characterized snoRNP (8). In Drosophila, several classes of mutations cause growth alteration during development, such as the Minute class (9), widely believed to affect ribosomal protein genes (10), and mini and bobbed mutations, which alter rRNA production (11).
The Drosophila modulo gene (mod) has first been characterized as a dominant suppressor of Position Effect Variegation (PEV) (12). PEV occurs in experimental strains where genes placed close to constitutive heterochromatin are randomly turned on or off, leading to mosaic adult structures. The products of genes that modify PEV are believed to change the local chromatin structure, which leads, when they are mutated, to an increased (suppression of variegation) or a decreased (enhancement of variegation) expression of neighboring genes (13)(14)(15)(16). The dominant PEV suppression phenotype of mod, therefore, suggests that its product participates in the formation of multimeric protein complexes that package the DNA, promoting chromatin compaction and inactivation (12). We have recently reported that mod, apart from its role in chromatin structuration, has a second cellular function and is required in nucleolus activity and protein synthesis capacity (17). First, cell clones deficient for mod express phenotypic traits characteristic of Minute mutations. Second, the protein, while * This work was supported by the CNRS, grants from l'Association pour la Recherche sur le Cancer and la Ligue Contre le Cancer (to J. P.), and le Ministère de la Recherche et de la Technologie et la Ligue Nationale Contre le Cancer doctoral fellowships to (L. P. and P. L.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  actually associated to condensed chromatin and heterochromatin sites, is also abundantly found at the nucleolus. Consistent with this localization, Mod displays a modular organization which is often found in nonribosomal nucleolar proteins, including an acidic stretch, two basic regions located at both ends of the molecule, and four reiterated RRMs in the remaining core portion (18,19).
The goal of the investigation reported here was to gain insight into the nucleic acid interactions established by Mod at the nucleolus and chromatin and to determine the molecular mechanism that controls the distribution and balances the function of the protein in the two sub-nuclear compartments. The data show that the two basic domains independently contact DNA without sequence specificity, whereas RRMs provides a sequence-specific RNA-binding activity. We also report that nucleolar Mod is phosphorylated and associated to an RNA moiety, whereas the chromatin-bound Mod appears to be unmodified. A functional scheme is finally proposed in which modification by phosphorylation modulates Mod subnuclear distribution and balances its function at the nucleolus and chromatin.
SDS-PAGE, Bidimensional Gel Analysis, and Western Blotting-For SDS-PAGE, proteins were fractionated on 7.5% gels and electroblotted onto nitrocellulose. Bidimensional gel analysis was performed as described in (21). Nuclear proteins from embryos (0 -18 h) were treated (30 g) in isoelectric focusing buffer (Biolyt, pH 3.0 and pH 10, from Bio-Rad). After SDS-PAGE in the second dimension and electrotransfer, Mod was revealed with anti-Mod monoclonal antibody (mAb) LA9 and anti-mouse alkaline phosphatase-conjugated secondary antibody (Promega).
Nuclear Fractionation and Immunoprecipitation Procedure-Isolation and fractionation of nuclei were carried out as described previously (23). Briefly, nuclei, purified under conditions that prevent loss of nuclear material, were successively extracted in 0.15 and 0.3 M NaCl, and chromatin fragments were solubilized by sonication in 0.45 M NaCl.
Western analysis of the recovered fractions was performed with a panel of monoclonal antibodies directed against chromatin-bound proteins like histone 2B, nucleosoluble components (S5 and P11 antigens), and Mod (17,23). The Mod protein was detected in 0.15 M and 0.30 -0.45 M NaCl fractions only. For Mod immunoprecipitation assays, sub-nuclear fractions were prepared from either 1 g of dechorionated embryos or 10 8 KC cells, diluted in lysis buffer (20 mM Tris-HCl, 0.15 M, 1 mM MgCl 2 , 1 mM CaCl 2 , 0.5% Nonidet P-40, pH 7.5) supplied with a mixture of protease inhibitors and preincubated for 1 h at 4°C with protein A-Sepharose beads (Amersham Pharmacia Biotech). Depleted samples (1 ml) were then incubated overnight at 4°C in the same buffer with 10 mg of protein A-Sepharose saturated with an excess mAb LA9 (1.6 mg of IgG purified from ascitic fluid for 100 mg of protein A-Sepharose beads). Beads were washed three times in 10 mM Tris-HCl, 0.5 M LiCl, 0.1% SDS, 2% Nonidet P-40, pH 7.4, then in 50 mM Tris-HCl, 50 mM NaCl, pH 8.0, and collected by centrifugation, and the pellet was prepared for SDS-PAGE.
KC Cells 32 P Labeling-KC cells were exponentially grown in D22 medium, centrifuged, and incubated for 1 h at 25°C in HN buffer (50 mM Hepes, 0.15 M NaCl, pH 7.6) at a density of 10 6 cells/ml. After centrifugation, cells were incubated again, 3 h at 25°C and 10 8 cells/ml, in HN buffer supplemented with 5 mM MgCl 2 , 2 mM L-glutamine, 1.8 mM glucose, 1 mCi/ml [ 32 P]orthophosphate, and, except when specified, with 10% fetal calf serum. Nuclei were purified and fractionated as described previously (23) but in the presence of an excess of sodium pyrophosphate (10 mM) and of phosphatase inhibitors (2 mM sodium vanadate, 0.1 M sodium fluorure, 20 mM EDTA, and 20 mM EGTA). Mod immunoprecipitation was also performed in the presence of sodium pyrophosphate and phosphatase inhibitors.
DNA Binding Assays-Purified GST fusion proteins (5 g) were added to dsDNA-cellulose beads (Amersham Pharmacia Biotech; 0.5 ml in NT2 buffer supplemented with 0.1 mg/ml bovine serum albumin) and incubated at 20°C for 2.5 h. After centrifugation and three washes in NT2 (0.5 ml each), bound proteins were stepwise eluted from the beads with 0.5 ml of NT2 containing increasing amounts of NaCl. Fractions, including the three washes, were precipitated with 5% trichloroacetic acid, washed with acetone, and analyzed by Western blotting. For embryonic Mod, nuclei were isolated and fractionated as described previously (24). DNA binding assays were performed using 85 and 250 g of nucleosoluble and chromatin-bound proteins, respectively.
SELEX Procedure-We used the synthetic RNA pool first described in Tsai et al. (24). SELEX procedure was essentially as described in Ghisolfi-Nieto et al. (25), with the following modifications. Purified GST fusion proteins (5 g) were incubated with 20 l of 50% glutathione-Sepharose beads (Amersham Pharmacia Biotech) for 30 min at 4°C in NT2 buffer (100 l final volume). After three washes in 0.5 ml of NT2 buffer, beads were suspended in 50 l of binding buffer (20 mM KCl, 150 mM NaCl, 50 mM Tris-HCl, 0.05% Nonidet P-40, 2.5% polyvinyl alcohol, 1 mM EGTA, 100 g/ml tRNA, 125 g/ml bovine serum albumin, 1 mM dithiothreitol, 0.8%, pH 7.4) and incubated with the RNA sample (0.5 g) for 5 min at 20°C. For the first two rounds, RNA samples were preincubated for 5 min at 20°C with Sepharose beads without protein prior to the selection step. After the third round, 0.5 M urea was added in the last wash. Polymerase chain reaction products were cloned and sequenced after seven rounds of selection/amplification.
RNA Structure Probing and Footprint-RNAs were transcribed in vitro by T7 RNA polymerase, 5Ј-end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase, purified on a 10% polyacrylamide, 8 M urea gel, eluted overnight, and precipitated with ethanol. Labeled RNA (50,000 cpm) was renatured in reaction buffer at 37°C for 15 min and complex formation carried out at 20°C for 15 min in the presence of increasing amounts of Mod. Enzymatic cleavage reactions by RNase T1 (10 Ϫ2 unit), RNase V1 (0.2 unit), RNase T2 (0.2 unit), or nuclease NC (0.2 unit) were performed at 20°C for 5 min in 10 l of 50 mM Tris-HCl, 5 mM MgCl 2 , 150 mM NaCl, and 1 g of carrier tRNA, pH 7.5. Iron-EDTA reactions were done as in (26). Reactions were stopped by phenol/ chloroform extraction, and RNA fragments recovered by ethanol precipitation were analyzed on 20% polyacrylamide, 8 M urea gels. RNase T1 and alkaline ladders served for cleavage site assignments.
Polytene Chromosome Squashes and RNase Treatment-RNase A salivary gland treatment was as in Richter et al. (27). Briefly, salivary glands were preincubated for 2 min in phosphate-buffered saline, 0.1% Triton X-100, and then for 8 min in either phosphate-buffered saline alone (as a control) or phosphate-buffered saline supplemented with 0.5 mg/ml RNase A. Polytene chromosome squashes and staining were performed as described in Clark et al. (28). Mod was detected by the mAb LA9 at 20 g/ml and Maleless (Mle) by affinity purified rabbit anti-Mle antibodies (gift of M. Kuroda) at a dilution of 1/500. Anti-mouse-fluorescein isothiocyanate-conjugated and anti rabbit-tetramethylrhodamine isothiocyanate conjugated secondary antibodies (Jackson) were used at a dilution of 1/150. Slides were examined under an Axiophot microscope (Zeiss).

Mod DNA-binding Is Mediated by the N-and C-terminal
Domains-Mod has been previously shown to bind DNA both in vitro and in vivo (12,18). Cross-linking/immunoprecipitation experiments moreover demonstrated interaction with several repetitive elements, consistent with the protein association to heterochromatin and the dominant phenotype of PEV suppression (17).
In order to identify the domains in Mod involved in DNA binding, GST fusions of various truncated versions of the protein were tested on dsDNA-cellulose chromatography. The results showed that the deletion of the two positively charged Nand C-terminal domains is required to abolish the ability of Mod to bind DNA in this assay, whereas truncated mutant proteins of only one of these domains still exhibit some activity (not shown). We further develop SELECT experiments (29) to ask the question of sequence specificity in Mod DNA binding. No obvious consensual motif was revealed by sequence comparison of 40 clones obtained after five rounds of selection/amplification performed using the full-length protein. Thus, the two terminal domains in Mod likely interact with DNA without sequence specificity.
Mod Is a Phosphoprotein-The Mod sequence contains a number of putative phosphorylation sites (18). Metabolic incorporation of 32 P in KC cells, followed by immunoprecipitation, Western blotting and autoradiography, reveals that Mod gives rise to a single band of 78 kDa (Fig. 1A, lane 2), indicating that the protein actually is phosphorylated in vivo. Consistently, treating Western blots with alkaline phosphatase progressively reduces and finally abolishes the 32 P labeling (Fig. 1A,  lanes 3 and 4). Next, by comparing the 32 P incorporation in starved and serum-stimulated cells, we noticed that serum addition raised the specific incorporation in Mod by a factor of 16, whereas the increase in crude cell extracts and purified nuclei was only 2.7-or 4.7-fold, respectively (Fig. 1B). As internal control to these experiments, we compared the Western signals from serial dilutions of cell extracts grown in the presence or absence of 10% fetal calf serum to show that the serum has no significant effect on Mod de novo synthesis (not shown). These data strongly suggest that serum enhances Mod phosphorylation.
To investigate for Mod phosphorylation in vivo, protein isoelectric variants present in embryos were analyzed by dimensional PAGE followed by a Western blot analysis. Nuclear extracts were prepared either in the presence of a mixture of phosphatase inhibitors or digested with phosphatases. Mod from phosphatase-treated extracts runs as a single spot with an apparent pI of 7.2, indicating that the multiple isoforms differing by negative charges (pI from 7.2 to 5.2) observed from non-digested nuclei, correspond to stepwise phosphorylation events on several amino acid residues (Fig. 1C).
Mod Differential Phosphorylation at Chromatin and Nucleolus-We previously reported a procedure that combines stepwise salt extraction with low and high speed centrifugations, to isolate distinct subnuclear fractions from purified nuclei, and a comparative Western analysis demonstrating that nucleosoluble proteins and components firmly bound to chromatin were recovered in distinct subfractions; the former by extraction at low salt concentration (0 -0.15 M and then 0.15-0.3 M NaCl) and the latter by sonication after raising NaCl concentration from 0. 30  whether this differential distribution could be related to different phosphorylation states of the protein, we analyzed 32 P incorporation in Mod from sub-nuclear fractions. The protein was immunoprecipitated from chromatin and nucleosoluble fractions of metabolically labeled KC cells and revealed by Western blotting and autoradiography. As shown in Fig. 2A, nucleosoluble Mod is highly labeled, whereas the chromatinbound protein does not have detectably incorporated radioactive phosphates. These data suggest that Mod is present in two different states in the nucleus: the nucleolar population is phosphorylated, whereas its chromatin-bound counterpart is not or is poorly phosphorylated.
Phosphorylation Down-regulates Mod DNA-binding Activity in Vitro-These data also suggested the hypothesis that phosphorylation could control Mod DNA-binding activity. To test the hypothesis, nucleolar and chromatin fractions prepared from embryos were assayed on dsDNA-cellulose chromatography. Data in Fig. 2B indicate that Mod extracted from chromatin strongly binds DNA and is eluted at NaCl concentrations of 0.5-1.0 M, whereas the nucleosoluble protein shows significantly weaker affinity and is eluted at NaCl concentrations of 0.2-0.3 M. Significantly, Mod from a nucleolar extract digested by phosphatase exhibits a substantially improved DNA-binding activity (elution at NaCl concentrations of 0.3-1.0 M). We therefore conclude that phosphorylation likely down-regulates Mod DNA-binding activity in vitro.
Identification of RNA Aptamers of Modulo-The presence of four RRMs in Mod led us to investigate the RNA binding properties of the protein. We first tested the affinity of GST-Mod or of a truncated version consisting only of the four RRMs (GST-RRM) to homopolymeric oligoribonucleotides. Both are able to bind poly(G) and poly(U), to a lesser extent poly(A), but not poly(C) (not shown). Second, we used GST-Mod and GST-RRM proteins in a selection-amplification procedure known as SELEX (30) to look for nucleotide specificity in RNA binding. PCR products were cloned and sequenced after seven rounds of selection/amplification of a 25-nucleotide-long random sequence (24). Out of 43 sequences selected from the GST-Mod, 24 obey the consensus (UUAC(N) x GU(A/G)G(U/A)(M) x ), where N and M are complementary nucleotides putatively able to form a stem, and "x" is comprised between 4 and 6 (Fig. 3). The other clones lack the consensus and do not efficiently bind Mod in gel shift controls (not shown). Their selection can originate from unspecific interaction with the basic terminal domains present in the full-length protein. This hypothesis is supported by the second SELEX experiment performed with GST-RRM. 10 clones from this selection were sequenced, and all were found to match with the consensus defined in Fig. 3. This experiment first confirms the consensual motif derived from the selection by GST-Mod and secondly indicates that RNA binding specificity is only provided by the RRM-containing domain.
Band shift assays were performed to confirm that selected sequences actually interact with Mod fusion proteins and to get an estimate of their affinities (K d values are given in Fig. 3). M8 and M12 were found to correspond to the aptamers of highest affinity, with K d values of 25 nM for GST-Mod and of 10 nM for GST-RRM. The other selected RNAs present weaker affinities, with K d values ranging from 100 nM to 1000 nM. Fig. 4 reports only the data obtained with M12 aptamer and the various controls done. The aptamer is actually shifted by GST-Mod and GST-RRM, whereas a ribonucleotide lacking the consensus (non-selected sequence, NS) does not (Fig. 4A). Fig. 4B provides other controls showing that anti-Mod mAb LA9 inhibits the interaction (lane 2), GST alone is ineffective (lane 3), and competition with unlabeled M12 progressively abolishes the shift whereas competition with ribonucleotide NS has no effect (lanes 4 -8). Taken together these data indicate that Mod is a sequence-specific RNA binding protein and support the idea that the bipartite motif defined here constitutes a high affinity site for the protein.
Probing the RNA Aptamer Conformation and Mod Footprint-The conformation of M12 was investigated using several enzymatic probes (31) such as RNase V1 (specific for paired nucleotides; Fig. 5B, lane 9), RNase T1 (specific for unpaired guanines; Fig. 5B, lane 1), RNase T2 (preferential cut after unpaired adenines; Fig. 5B, lane 5), and nuclease from Neurospora crassa (specific for unpaired nucleotides; Fig. 6A). The results are reported on the secondary structure model derived from the enzymatic probing (Fig. 5C). The data support the existence of a stable hairpin structure presenting an external loop (nucleotides G27-U31) and two internal loops (nucleotides A13-C21/A38 -A41 and A47-A55). The external loop is well defined by the presence of an RNase T1 cut at G30 and of several RNase T2 or nuclease NC cleavages at U29 and A28 (see Fig. 5, A and B). Interestingly, G27 is not accessible to RNase T1, indicating that this peculiar guanine is either stacked within the loop or base paired with U31. Many hairpin loops closed by non-canonical base pairs (G-A or G-U) have been described (32). In most of the selected aptamers, position U31 is predominantly U or A (see Fig. 3). Other single-stranded specific RNase cleavages are mainly located in the regions A15-U19, U48 -A52, and G64 -A67 (Fig. 5, A and B). The existence of helices I and II is supported by the presence of several RNase V1 cleavages at positions 9 -11, 22-24, 36, 44 -  1 and 2), has not detectably incorporated 32 P (lanes 4 and 5). B, DNA affinity chromatography of chromatin-bound and nucleolar Mod from embryos. Protein samples were incubated with dsDNA cellulose, eluted by increasing NaCl concentration, and resulting samples were analyzed by Western blotting. 45, and 59 -60 (Fig. 5B). These data indicate that the two conserved sequences (UUAC and GU(A/G)G(U/A)) are located likely in single-stranded regions.
The Mod footprint was next studied using RNases T1, T2, V1, and nuclease NC as enzymatic probes. We also used iron-EDTA which generates, in the presence of hydrogen peroxide, reactive hydroxyl radicals cleaving ribose moieties irrespective of secondary structure (33). Experiments are shown in Fig. 5, A  and B and summarized on the predicted secondary structure of M12 depicted in Fig. 5C. Mod induces strong protection against single-stranded specific RNases at positions that correspond to the two conserved sequences, in the hairpin loop (at U28, A29, and G30) and in one internal loop (at U18, U19 and A20). Reduced RNase V1 cleavages are also observed on both sides of helix II and at positions 9 -11, whereas significant enhance-ments occur at positions 44 -45. Furthermore, iron-EDTA footprint experiments revealed protections at riboses 29 to 35 (Fig.  5C). It has to be noted that no information was obtained for the nucleotides A15-A20 and A59 -A63 from iron-EDTA experiments, owing to the presence of several nonspecific cleavages. Mod does not induce reactivity changes in the bottom of helix I, indicating that the two non-random oligonucleotides used for the selection do not directly participate in the RNA binding site.
The folding program of Zuker (34) was used to generate suboptimally folded structures for the other selected aptamers. Interestingly, M8 RNA, which binds Mod with the same affinity as M12, is predicted to fold in a very similar structure. The hairpin loop GUGGA and the single-stranded sequence UUAC in M8 are separated by a stem formed by five base pairs (see  4, 5, and 6), whereas NS RNA added to the same concentrations has no effect (lanes 7 and 8). Fig. 3). For the other examined RNAs which present weaker binding affinities, helix I appears to be shorter (4 base pairs in M19) or is prolonged by several base pairings involving the conserved sequence UUAC (M4, M19, M29, and M37). M9 RNA is of particular interest because this aptamer lacks the conserved motif UUAC but efficiently binds Mod (Fig. 3). RNase probing and footprint experiments showed that M9 can adopt two conformations and that Mod binding to the GUGGU conserved motif stabilizes a stem-loop structure that resembles that of M12 (not shown).
These data indicate that the essential determinants for RNA recognition by Mod are located within the hairpin loop having the conserved sequence GU(A/G)G(U/A). The presence of the second conserved UUAC sequence located in a single-stranded region and at an appropriate distance from the hairpin loop increases the efficiency of binding. Our data also suggest that Mod makes specific contacts with nucleotides in the hairpin loop but also with the ribose-phosphate backbone.
Mod Belongs to a Ribonucleoprotein Complex in Vivo-We reasoned that if Mod is involved in a ribonucleoprotein (RNP) complex, the electrophoretic pattern in native conditions should be changed after digestion by RNase. The question was addressed using soluble nuclear extracts from Schneider cells. Western blot analysis of RNase-treated and non-treated extracts run on native PAGE (Fig. 6) actually shows that RNase treatment strongly modifies the migration of Mod, consistent with an association to RNA molecule(s). Moreover only one migrating species is revealed, indicating that most of the nucleosoluble Mod protein is involved in a single complex. Noteworthy, preincubation with an excess of M12 aptamer shifts Mod toward a faster migrating form (Fig. 6, lane 3). Thus, M12 can displace RNA molecule(s) associated to Mod in vivo. This effect is clearly sequence-specific because competition with a large excess of ribonucleotide NS does not change the Mod migration pattern (Fig. 6, lane 2). These results suggest that M12 RNA aptamer binds to the same or overlapping site as the in vivo cognate RNA(s). We have previously reported that Mod only from nucleolus and not the chromatin-bound protein is recovered in soluble nuclear extracts (17). Thus, the evidence provided here, that Mod belongs to a RNP complex in vivo, regards the nucleolus-associated form only.
RNase Treatment Does Not Prevent Mod Association to Chromatin-On polytene chromosomes, Mod is associated to condensed chromatin sites, including a majority of bands on chromosome arms, and to the nucleolus (17). In order to test whether Mod at the chromatin might also interact with a RNA molecule, we analyzed the Mod pattern on polytene chromosomes from salivary glands previously digested with RNase. As a control, we simultaneously followed the distribution of the Mle protein, which is known to be released from the male X chromosome by RNase A (27). Male salivary glands digested or not by RNase A were squashed, and polytene chromosome preparations were immunostained for both Mod and Mle. Comparison of panels A to C and B to D in Fig. 7, indicates RNase digestion actually induces efficient release of Mle from the X chromosome and of Mod from the nucleolus, but does not affect the Mod chromosomal pattern. DISCUSSION Previous studies indicated that Mod likely fulfils two distinct nuclear functions, because the dominant suppression of PEV (12) and recessive Minute-like (17) phenotypes are suggestive for roles in chromatin compaction and regulation of nucleolus activity, respectively. The present findings provide molecular evidence supporting first that the protein is involved in distinct networks at nucleolus and chromatin and second that phosphorylation down-regulates DNA-binding and therefore controls distribution and function of Mod in the two sub-nuclear compartments.
The structure/function analysis performed using Mod trun-cated isoforms has clearly identified the domains involved in nucleic-acid binding. One major conclusion is that Mod is a sequence/motif-specific RNA binding protein. Two lines of evidence indicate that this property is provided by the RRMcontaining domain. First, the same consensus was derived from SELEX experiments performed with GST fusions of either the whole protein or the RRM moiety only. Second, the two protein variants show close K d values for M12 aptamer association. Structural probing and footprinting experiments predict that Mod stabilizes a hairpin-like structure and presumably establishes direct contacts within two single-stranded GU(A/ G)G(U/A) and UUAC sequences corresponding to the hairpin loop and part of a bulged loop and with the ribose-phosphate backbone as well. This RNA motif appears quite different from that of the few targets of RRM-containing proteins identified so far, which are usually constituted of hairpin loops of 8 to 10 residues (Ref. 25, and references therein). More generally, a search in data bases failed to reveal any significant resemblance between the Mod cognate motif and a previously identified RNA sequence. A second conclusion is that the two basic terminal domains of Mod are able to bind DNA in a nonsequence-specific manner. The decreased affinity of protein variants truncated of either the N or C terminus indicates that each domain can act independently. It is, however, conceivable that the two domains function synergistically in the native protein to improve DNA-binding activity. Mod therefore presents unique in vitro nucleic acid binding properties as it is able to directly contact DNA via the two tips and bind to a specific RNA motif via the central RRM domain. This leads up to consider three points regarding the situation in vivo. One can first wonder about the nature of the nucleic acids Mod is interacting with at the nucleolus and chromatin. Clear evidence that Mod is released from the nucleolus as a RNP complex is provided by the RNase-induced modification of nucleosoluble Mod migration in native gel electrophoresis. Also consistent with a direct interaction with a nucleolar RNA, M12 aptamer specifically displaces the protein from the RNP complex. This latter result in addition strongly suggests that the Mod RNA target at nucleolus likely presents a structure similar to M12. However, we did not find any motif related to the aptamer in the repertoire of Drosophila rRNA or snoRNA sequences available in data bases. Eukaryotic cells contain an extraordinarily complex population of snoRNAs (8,35), but only a few have been cloned in Drosophila. It is therefore tempting to assume that the Mod RNA target at the nucleolus corresponds to snoRNA sequences not identified so far. As the protein was previously shown not to bind rDNA (17), we conclude that in this compartment Mod associates to specific RNA molecule(s) but does not contact DNA. Interactions of Mod with nucleic acids at chromatin appear to be quite different. On one side, it is well established that the protein directly contacts genomic DNA (12,17). On the other side, digestion by RNase does not affect the Mod pattern on polytene chromosomes, while it induces efficient release of Mle from the X chromosome and of Mod from the nucleolus. The simplest explanation is that chromatin-associated Mod does not interact with RNA. Alternatively, in the process of chromatin compaction, Mod may recruit RNA together with additional protein factors to form highly condensed structures in which the RNA moiety is protected from digestion by RNase. Whether or not Mod requires a RNA partner in the process of chromatin compaction obviously needs further investigation.
The second point regards the molecular mechanisms that control Mod distribution between the two subnuclear compartments. We propose that posttranslational modification by phosphorylation down-regulates the DNA-binding activity and capacity of the protein to link chromatin and, therefore, modulates the equilibrium between chromatin versus nucleolus association. This model is supported by several lines of evidence: i) the chromatin-bound protein does not detectably incorporate 32 P in cell culture assays; ii) the nucleolar fraction is highly modified and cannot bind DNA in vitro; and iii) the DNA-binding activity is restored after digestion by phosphatase. In contrast, phosphorylation is unlikely to affect the capability of Mod to bind RNA because competition experiments have shown that the M12 aptamer, selected by a recombinant protein produced in Escherichia coli (i.e. unmodified), binds phosphorylated nucleolar Mod as well.
The third point to consider is how the molecular data reported here correlate to, and possibly improve our understanding of, functional data obtained from genetics. The dominant suppression of PEV phenotype unambiguously indicates that the encoded protein participates in the local assembly of high order chromatin structure and transcriptional silencing of neighboring genes. This function is thus clearly related to chromatin rather than nucleolus and to the DNA-binding activity of Mod. The lack of sequence specificity in DNA recognition that is exhibited by the two basic terminal domains in vitro contrasts with the protein distribution at specific heterochromatic sites on mitotic chromosomes (17). This suggests that Mod might associate to unknown factors that could provide specificity and direct the complex toward particular chromatin sites. An interesting possibility is that the distal domains in Mod interact with sequences lying relatively far apart from each other on the DNA fiber, which could favor and stabilize the formation of condensed chromatin structures.
Total loss of mod function results in the expression of several recessive phenotypes. Some have clearly been related to defect in ribosome biogenesis and protein synthesis capacity, such as the "Minute-like" phenotype of mutant cell clones induced in a wild type background (17) and defect in cell growth and proliferation of mutant imaginal tissues. 2 Other phenotypes, mela-notic tumor formation (12) and lymph gland hyperplasia, 2 which are highly reminiscent to phenotypes caused by mutations in the ribosomal protein S6 gene (36), are also consistent with a role for Mod in the regulation of ribosome assembly and cell growth. Interestingly, Mod phosphorylation is enhanced when cell growth is stimulated in cultures supplemented with serum. As our bidimensional gel analysis revealed a multiplicity of Mod phosphoisoforms in vivo, it is therefore tempting to assume that modification by phosphorylation at critical sites in Mod is required not only to prevent DNA-binding and chromatin compaction, but also to enhance nucleolus activity. Regarding the molecular function of Mod at the nucleolus, an involvement in rDNA transcription appears unlikely as the protein does not bind rDNA and is released from the organelle by low salt extraction of purified nuclei (17). Instead, it presumably participates in a subsequent step of the ribosome biosynthesis. Like most nucleolar proteins identified so far, Mod indeed presents a modular structure that combines four RRMs with acidic and basic domains and suggests interaction with nucleolar components, including ribosomal proteins, rRNA and/or snoRNAs, and possibly a chaperone function facilitating ribosome assembly (4).
Various aspects in the molecular function of Mod remain to be resolved, such as the nature of RNA target in vivo and its possible involvement in chromatin compaction, the identity of kinases/phosphatases that modify the protein or the nature of transcriptional chromatin targets. However, the results presented here, together with previously reported molecular and genetic data, led us to propose the following functional model for Mod. In response to physiological stimuli, mimicked by serum in cell culture assays, the protein becomes phosphorylated and is released from chromatin, inducing structural relaxation and PEV suppression, moves to the nucleolus where it associates RNA, and forms a RNP required to improve ribosome biogenesis and protein synthesis capacity. As far as we know, among ribosomal or nucleolar proteins thought to possess a second cellular activity apart from ribosome assembly and function (1,37), Mod is the first example in which the bifunctional character is proven by genetics and in which a regulatory molecular mechanism of how the two functions are coordinated in vivo is proposed.