Developmental Regulation of dUTPase in Drosophila melanogaster*

dUTPase prevents uracil incorporation into DNA by strict regulation of the cellular dUTP:dTTP ratio. Lack of the enzyme initiates thymineless cell death, prompting studies on enzyme regulation. We investigated expression pattern and localization of Drosophila dUTPase. Similarly to human, two isoforms of the fly enzyme were identified at both mRNA and protein levels. During larval stages, a drastic decrease of dUTPase expression was demonstrated at the protein level. In contrast, dUTPase mRNAs display constitutive character throughout development. A putative nuclear localization signal was identified in one of the two isoforms. However, immunohistochemistry of ovaries and embryos did not show a clear correlation between the presence of this signal and subcellular localization of the protein, suggesting that the latter may be perturbed by additional factors. Results are in agreement with a multilevel regulation of dUTPase in the Drosophila proteome, possibly involving several interacting protein partners of the enzyme. Using independent approaches, the existence of such macromolecular partners was verified.

Faithful conservation and transmission of genetic information are crucial for living organisms. The enzyme dUTPase prevents uracil incorporation into DNA by hydrolysis of dUTP into dUMP and inorganic pyrophosphate and provides a unique and essential preventive DNA repair function. Lack of dUTPase leads to uracil-substituted DNA that perturbs base excision repair, resulting in DNA fragmentation and thymineless apoptosis of the cell. The physiological role of the enzyme argues for regulation of dUTPase presence, localization, and/or function, depending on cell status.
This implication was closely investigated in human cells. Two isoforms of the enzyme were identified: a mitochondrial isoform with constitutive expression, and a nuclear isoform with cell cycle-dependent expression (1). Both are transcribed from the same gene using alternative promoters and differ only in their N-terminal region responsible for adequate localization (2). dUTPase in mitochondria is necessary for integrity of mitochondrial DNA that replicates independently from the cell cycle. In mature lymphocytes, correlation between dUTPase presence and cell mitogenic status suggested enzyme up-regulation in stages associated with DNA synthesis, whereas in immature populations, enzyme levels are constitutive (3). The requirement of actively dividing cells for high dUTPase levels is also confirmed by studies in rat liver regeneration (4) and mitogen-activated T and B cells (5). Overexpression of dUTPase induces resistance to chemotherapeutic agents that target thymidylate biosynthesis (6). This suggests that in addition to inhibition of thymidylate synthase and dihydrofolate reductase, dUTPase targeting as a novel chemotherapeutic strategy may ensure synergistic effects.
Cancer cells are usually characterized by deficiency in most apoptotic pathways and are not easily induced into committing suicide. Programmed cell death pathways still operable after malignant transformations are therefore of high therapeutic value. Thymineless cell death was recently suggested to be independent from p53, a central factor in most apoptotic pathways (7,8). This finding underlines the importance and possible gains of targeting thymidylate metabolism in tumor cells.
In multicellular organisms, apoptosis plays a key role not only in prevention (and therapy) of malignant transformations but in other developmental processes as well. Apoptosis of T and B cells, triggered by receptor-ligand interactions, is essential in the development of the immune system to prevent autoimmunity as well as to facilitate selection of lymphocytes (9,10). Down-regulation of dUTPase in anti-IgM antibody-induced B-cell apoptosis was reported recently (11), indicating that the thymineless pathway might also be of importance in cell death during development. Putative accumulation of uracil-containing DNA under thymineless conditions in larval stages of Drosophila melanogaster may play a role in programmed cell death necessary for metamorphosis in the pupal stage (12). The role of dUTPase was considered to be central in this process; moreover, a developmentally induced heat-stable dUTPase inhibitor protein was also suggested to exist (13). Endogenous macromolecules specifically antagonizing with factors involved in apoptosis and/or DNA metabolism and repair are undoubtedly of great importance because these may provide a drug design lead molecule already tailored by evolution. Unfortunately, lack of knowledge about the molecular biology, physiology, and cellular status of fly dUTPase impeded a systematic assessment of its regulatory pathways.
To approach this problem, we have cloned the fly enzyme and performed kinetic and structural characterization of the recombinant protein (14). To describe the role of dUTPase in Dro-sophila that might also reveal important clues about eventual parallel mechanisms in Homo sapiens, we decided to proceed with physiological characterization of fly dUTPase.
In the present work, expression, localization, and developmental patterns of Drosophila dUTPase are described. Two isoforms of the enzyme were isolated by immunoprecipitation and identified by mass spectrometry. Discrepancy between protein and mRNA levels as well as cellular localization patterns suggested the possibility of regulation by additional factors.
Collection of Different Developmental Stages of Drosophila-Oregon-R wild-type flies were kept on cornmeal-yeast food, containing Nipagin as fungicide, at room temperature. Two-h early embryos, 12-h embryos, 36-h first larvae, 60-h second larvae, 96-h third larvae, and 8-day pupae, staged according to Ashburner (15), were collected, washed, and homogenized with a Potter-Elvehjem homogenizator and sonicated in phosphate-buffered saline (PBS). Ovaries were fractionated into immature and more mature populations by trypsinolysis and filtration. A heat-stable fraction of larval extract was prepared by 5-min incubation at 100°C in a boiling water bath of the crude extract, followed by centrifugation. Protein concentration measurements by Bradford assay indicated that ϳ5% of the total protein content stays in the solution phase after this heat treatment.
Subcellular Fractionation of Drosophila Embryos-Twelve-h embryos were collected, washed, and homogenized in 50 mM 1,4-piperazinediethanesulfonic acid buffer, pH 7.9, also containing 50 mM KCl, 5 mM EDTA, 2 mM MgCl 2 , 1 mM dithiothreitol, and 1 mM phenylmethylsulfonyl fluoride. The homogenate was subjected to differential centrifugation at 4°C, following the procedure described by Igaki et al. (16). The first pellet after centrifugation at 700 ϫ g for 10 min contained the nuclear fraction. Supernatant from this step was further centrifuged at 24,000 ϫ g for 10 min to sediment the mitochondrial fraction. The cytoplasmic fraction was obtained after a final centrifugation at 54,000 ϫ g for 1 h. Equal amounts of protein from each fraction were applied on SDS-PAGE and analyzed by Western blotting.
Preparation of Polyclonal Antibody for Drosophila dUTPase-Rabbits were immunized with recombinant full-length Drosophila dUTPase (14). Three immunizing shots were given, at time intervals of 2-3 weeks, first in complete then in incomplete Freund's adjuvant or in physiological saline (17)(18)(19). Serum was used at a dilution of 1:100,000 in Western blot on nitrocellulose membranes.
Western and Far-Western Blotting-Extracts were run using SDS-PAGE (20) and transferred to nitrocellulose membrane (Sigma). Blots were stained first with Ponceau dye and then developed with the antiserum, followed by staining with secondary antibody (alkaline phosphatase or horseradish peroxidase labeled anti-rabbit IgG, at 1:80,000 and 1:2,500 dilutions, respectively). For visualization, nitro blue tetrazolium/5-bromo-4-chloro-3-indolyl phosphate or the enhanced chemiluminescence kit of Amersham Biosciences was used. Monoclonal anti-␣-tubulin (Sigma) was used as loading control. For far-Western blotting, wet membranes were blocked with 5% bovine serum albumin in Tris buffered saline with 0.1% Tween 20 for 1 h and incubated with 30 g/ml of recombinant Drosophila dUTPase at 4°C overnight. After extensive washing with PBS, the procedure was continued according to Western blotting protocol.
Immunoprecipitation-dUTPase isoforms were immunoprecipitated from S2 cells with partially purified polyclonal antibody coupled to cyanogen bromide-activated Sepharose (Sigma). Antigen was eluted from the resin by boiling for 5 min in 0.1 M Tris buffer, pH 8.0, also containing 10% SDS (21).
Protein Identification by Mass Spectrometry-Proteins were in-gel digested by trypsin (Promega) after reduction and alkylation. The tryptic digests were analyzed on a Reflex III matrix-assisted laser desorption/ionization, time-of-flight mass spectrometer (MALDI-TOF MS, Bruker, Germany) unfractionated as well as after reversed-phase high-pressure liquid chromatography fractionation. 2,5-Dihydroxy benzoic acid was used as the matrix. All mass spectra were acquired in positive reflectron mode, with delayed extraction, using external calibration. A database search was performed on a National Center for Biotechnology Information database using ProteinProspector. 2 To obtain sequence information, post source decay (PSD) analysis of selected components was performed.
Semiquantitative Reverse Transcription-Polymerase Chain Reaction (RT-PCR) and DNA Gel Electrophoresis-Total RNA was isolated from different developmental stages of Drosophila and S2 cells by TRIzol reagent (Invitrogen/Life Technologies, Inc., Grand Island, NY). Reverse transcription was performed with RNase H(Ϫ) (Promega) at 37°C for 1.5 h. For PCR amplification of dUTPase cDNAs, primers were designed to straddle the putative intronic region of the dUTPase gene (see Fig. 2). DNA gel electrophoresis was performed on 1.5% agarose gels using standard procedures.
mRNA in Situ Hybridization-cDNA was prepared by PCR on the dUTPase gene containing the pET-22b expression plasmid (14) using dUTPase-specific primers (forward, 5Ј-ATGCCATCAACCGATTTCGC-CGACATT; reverse, 5Ј-TTACGTAGCAACAGGAGCCGGAG-3Ј). The DIG Nucleic Acid Labeling and Detection Kit (Roche) was used for labeling the cDNA with digoxigenin and in situ detection. Ovaries were fixed, and hybridization was performed according to Ephrussi et al. (24). Samples were mounted in PBS buffer containing 90% (v/v) glycerol and analyzed with a Leica DMLS microscope.

RESULTS
Two Isoforms of dUTPase Protein in Drosophila-A monospecific antibody was required for dUTPase detection in Drosophila samples. Fig. 1A demonstrates that a highly specific polyclonal serum with no cross-reactivity to other cellular proteins was raised against recombinant full-length Drosophila dUTPase. In Drosophila S2 cell extract (Fig. 1A, lane 5), two distinct serum-reactive protein bands (with molecular masses of 23 and 21 kDa) suggested that two dUTPase isoforms might be present. Immunoprecipitated dUTPase proteins (Fig. 1B) were subjected to in-gel trypsinolysis and mass spectrometry ( Fig. 1, C and D). For the two putative isoforms, 64 and 59% of the peptide masses detected in the mass spectra showed clear matches to predicted dUTPase tryptic fragments, providing 72 and 66% coverage (Fig. 2, compare underlines and overlines) of the dUTPase sequence, respectively. Peptide mass fingerprints of the two digests are almost identical, excluding a unique peptide ((2-13) with m/z 1232.6) in the larger isoform (Fig. 1C, double asterisks).
Genomic data on Drosophila dUTPase predicted a unique C-terminal 28-residue extension (25). A peptide with m/z 1251.6 corresponding to the C-terminal tryptic segment of this region was detected in the mass spectra of both investigated protein samples (single asterisk in Fig. 1, C and D). PSD analysis (26) of this peptide from both 23-and 21-kDa samples confirmed the sequence as AAEPEGAAPAPVAT (residues in one-letter code, Fig. 1E). This indicates that the unique Ala-Pro-rich C-terminal segment is present in both putative isoforms. The potential significance of this result will be addressed below.
Having determined that C termini are identical in the two putative isoforms, N termini were investigated. PSD analysis of the peptide with m/z 1232.6, corresponding to the segment (2-13) of Drosophila dUTPase and present only in the mass spectrum of the 23-kDa isoform, provided independent sequence information (Fig. 1F). Interestingly, the determined sequence PSTDFADIPAAK argues for lack of the N-terminal methionine residue from the physiological dUTPase 23-kDa species. Similar posttranslational processing has been reported for the nuclear isoform of human dUTPase (1). The absence of this peptide from the tryptic digest of the 21-kDa isoform, even after reversed-phase high-pressure liquid chromatography fractionation, may indicate that the smaller species has a different N terminus. In this digest, the peptide with m/z 876.5, corresponding to residues 17-23 of the predicted dUTPase sequence, could be detected as closest to the N terminus. PSD spectrum of this peptide confirmed its sequence as IDTCVLR (Fig. 1G). These data indicate that N termini are in fact different in the two putative isoforms. This difference is in excellent agreement with the observed alteration in molecular masses of intact isoforms, as determined on SDS-PAGE gels (see Fig. 1B).
The lack of the indicated N-terminal segment in the shorter isoform does not influence the five conserved dUTPase sequence motifs, necessary for catalytic function. However, alteration of the N terminus results in breakdown of a putative nuclear localization signal (NLS) in the 21-kDa isoform ( Fig. 2 and "Discussion").
To exclude the possibility that the shorter isoform is generated by uncontrolled proteolysis in cell extracts, its existence was addressed at the mRNA level. The recent Drosophila Genome Release 3.1 reports an intronic region, conforming to the widely accepted GT-AG rule (27), in the single dUTPase gene denoted as CG4584. The resulting two expected mRNA transcripts are termed tr-A and tr-B in Fig. 2. For semiquantitative RT-PCR, forward and reverse primers (Fig. 2, dark gray boxes) were designed to straddle the putative intronic region of the dUTPase gene. RT-PCR with various Drosophila samples showed two bands at positions expected for the predicted 670and 578-bp lengths of the amplified region of tr-A and tr-B, respectively (Fig. 3E). These results confirmed the existence of two splice variants of dUTPase mRNA in Drosophila corresponding to two protein isoforms with different N termini.
Developmental Control of dUTPase Is Different at Protein and mRNA Levels-The two isoforms were followed at both protein and mRNA levels throughout development. Both protein isoforms showed a drastic decrease during larval stages (Fig. 3A), in agreement with the previous hypothesis (compare the Introduction of Deutsch (12)). In the pupal stage, the enzyme is again at a higher level. The ratio of the two protein isoforms is about 1:1 and does not change significantly during development (Fig. 3A). Adult flies express dUTPase mostly in their ovaries; male flies do not contain the enzyme at detectable levels (Fig. 3B). Mature as compared with immature follicles become significantly enriched in the short 21-kDa variant (Fig.  3C). This phenomenon parallels a change in subcellular localization of the enzyme (see below).
In addition to developmental regulation of dUTPase expression, enzyme levels were also suggested to change during the cell cycle in human cells (1). To investigate whether this control may also apply for fruit fly, we analyzed dUTPase levels in cycling and resting cells. Extracts of cells in the logarithmic growing stage showed high dUTPase levels, whereas the enzyme in overgrown cell populations was beyond detection limit (Fig. 3D).
In striking contrast to the strong developmental control of the dUTPase protein presence, mRNA levels show a constitutive character during fly development (Fig. 3E). Relative amounts of transcripts show some alteration in different stages (Fig. 3E). The considerable difference between changes of the mRNA and the protein levels during development suggests the existence of a posttranscriptional regulatory mechanism of the dUTPase isoforms in Drosophila. Subcellular Localization of dUTPase in Different Tissues of Drosophila-Between the two isoforms, the full-length dUTPase contains a putative NLS (PAAKKMKID; Fig. 2 and "Discussion"). However, the shorter isoform does not possess any potential subcellular localization signal. To investigate the presence of the two isoforms in different cellular organelles, subcellular fractionation experiments of Drosophila embryos were carried out. Fig. 3F demonstrates that: i) the nuclear fraction contains both dUTPase iso-forms in 1:1 ratio; ii) the mitochondrial fraction does not contain a detectable amount of the enzyme; and iii) the cytoplasmic dUTPase composition is comparable to that found in the fraction sedimenting with the nuclei.
To complement the above experiments performed in homogenized samples with cellular studies, immunohistochemistry was performed (Fig. 4). Follicles passing throughout maturation in the ovary were stained for dUTPase and DNA (Fig. 4A). In immature follicles, developing cells showed uniformly nuclear dUTPase staining, and the oocyte could not yet be distinguished (Fig. 4A, asterisk). During further development, nurse cells and the oocyte were visualized as separate compartments (Fig. 4A, arrow), and dUTPase stain was more diffuse and less intensive. In more mature follicles (Fig. 4A, cross), there was a clear distinction between large-nuclei nurse cells and the egg cell with its small nucleus (Fig. 4A, dashed arrow), which is positioned at one of the poles of the follicle. Within these nurse cells, no dUTPase stain was observed in the nuclei but the enzyme is present exclusively within the cytoplasm at a low intensity, comparable to dUTPase stain observed in the oocyte. Results indicate alteration in dUTPase localization during follicle development. The observed shift of dUTPase stain from nucleus to cytoplasm may be a prelude to protein transport from nurse cells to oocyte, a well-known mechanism for maternal origin of other proteins and mRNAs. In situ hybridization confirmed that nurse cells pump dUTPase mRNA into the oocyte (Fig. 5). Fig. 4B shows an early embryonic stage, collected from young flies that were allowed a maximum of 2 h for egg deposition. Morphology of the presented embryo corresponds clearly to the stage where just a limited number (about seven to eight) of nuclear mitotic cycles has yet passed (see Fig. 9.4 in Ref. 22). In this stage, the organism fully depends on maternal protein and mRNA and has not yet started zygotic gene expression (28). The maternal dUTPase pool is visible, mostly associated with the nuclei (Fig. 4B). In the late embryonic stages with zygotic expression where gastrulation is already present, dUTPase retains nuclear localization but its distribution becomes heterogenic within the whole embryo (not shown). During larval stages, dUTPase expression is confined to actively proliferating imaginal discs (Fig. 4C), corresponding to a decrease of the enzyme level observed on Western blot (Fig. 3A). Within the imaginal discs, the enzyme shows mainly nuclear localization (Fig. 4D).
The S2 cell line constitutes cells of diverse morphology. In addition to the major fraction of smaller sized, epithelial-like cells (29), some large cells of possible macrophage origin are also observable. dUTPase localization is also heterogeneous (Fig. 4E). In most cells, the enzyme is confined to the nucleus; however, cytoplasmic localization is also evident. Results from a quantitative analysis indicate that dUTPase localizes to nuclei in ϳ85% of S2 cells.
Immunohistochemical results indicate that both cytoplasmic and nuclear localization of dUTPase can be observed in a variety of Drosophila tissues. In actively dividing imaginal discs and embryos, localization is mainly confined to the nucleus (Fig. 4, B-D). In other larval tissues, the enzyme is either absent (Fig. 4C) or shows weak diffuse staining in the cytoplasm (not shown).
Existence of Interacting Macromolecular Partners of Drosophila dUTPase-Results on the developmentally constant 1:1 ratio of the two dUTPase isoforms (Fig. 3, A, B, and D), among which only one possesses a potential NLS (Fig. 2), are not straightforward to reconcile with the immunohistochemistry data obtained on immature follicles, early embryos, and larvae (Fig. 4, A-D), where the protein seems to be present mostly in the nucleus. Although the dUTPase protein pool of early embryos is derived from the cytoplasm of maternal nurse cells, the protein becomes quickly enriched in the embryonic nuclei. Beyond the putative NLS-driven process, this phenomenon argues for a different transport mechanism that possibly involves interacting protein partners of dUTPase.
In addition, interacting protein partner(s) were suggested to regulate enzyme activity (13). Experiments performed in our laboratory also indicated that several fractions of S2 cell extract have significant modulatory effect on the recombinant 23-kDa dUTPase isoform. 3 These observations initiated investigations by two experimental approaches toward endogenous macromolecules capable of binding to dUTPase. Fig. 6, A-D, shows results of surface plasmon resonance measurements. Recombinant dUTPase lacking the extra C-terminal 28-residue segment was immobilized on a sensor chip, and heat-stable larval extract (13) was injected upon this surface. In this experimental set-up, significant binding of macromolecules was observed (Fig. 6B). This binding is clearly dependent on total protein concentration in the extract, according to thermodynamic expectations on association-dissociation processes. An increase in the baseline signal after injection, also dependent on total protein concentration, indicates that a portion of the bound macromolecules was retained on the dUTPase-conjugated chip. Specificity of the binding was checked by injection of serum albumin (not shown) and dUTPase (Fig. 6A) over naive dUTPase-conjugated chip, not yet probed with Drosophila larval extract. No binding was observed in either of these control experiments, arguing against nonspecific adsorption. Interestingly, when such a control experiment was repeated by injecting dUTPase upon a dUTPase-conjugated chip probed previously with Drosophila larval extract, some binding was clearly observed (Fig. 6C). In view of partial retention of Drosophila proteins on the dUTPaseconjugated chip, these results may indicate that the retained macromolecules (or their network) are multivalent dUTPase partners.
To check whether the Drosophila-specific C-terminal 28-residue segment has a major role in binding macromolecular partners from cellular extract of the larvae, immobilization of both full-length and truncated recombinant Drosophila dUTPase was performed on another sensor chip. Injection of heat-stable larval extract upon these two enzyme surfaces in parallel experiments showed practically the same characteristics (Fig.  6D), arguing against an exclusive role of the C terminus in the interaction processes.
In the above experiments, heat-stable larval extract was used to prevent the possibility of nonspecific precipitation. Having determined that in the heat-stable extract some specific binding may occur, experiments with injections of total extract were also performed. Binding phenomena were observed in these injections as well, although the markedly different protein compositions of the total and heat-stable extracts precluded direct comparisons. Detailed analysis of binding phenomena using such a heterogeneous population of possible binding partners is not straightforward in surface plasmon resonance techniques. Further quantitative experiments with partially purified dUTPase partner macromolecules are planned in our laboratory.
Despite the lack of quantitative analysis in the above surface plasmon resonance experiments, sensograms argued strongly in favor of the presence of macromolecules capable of physical interaction with dUTPase in Drosophila larval extract. These experiments, however, cannot decide whether the observed binding phenomenon is attributable to one single component or to a mixture of dUTPase-binding macromolecules. Far-Western blotting was therefore used additionally with the aim of separating the potential interacting components. This approach is limited to identification of those protein-protein interactions that persist under the relatively harsh experimental conditions. Comparative Western and far-Western blots (Fig. 6E) of extracts from Drosophila embryos and first-stage larvae show several distinct protein bands that appear only on the far-Western blots. A different pattern of far-Western blots of embryos and first-stage larvae is evident.
The two independent approaches consonantly indicate the presence of dUTPase partners in Drosophila that may depend on development. Co-immunoprecipitation and affinity experiments are in progress in our laboratory with the aim of identifying the interacting proteins and characterizing their potential regulatory functions.

DISCUSSION
Two Isoforms of dUTPase-Authentic localization signals for nuclear and mitochondrial isoforms of human dUTPase, as well as human uracil-DNA glycosylase, responsible for excising uracil from DNA, have been described (1, 30 -32). Mammalian base excision repair enzymes are present in both nuclei and mitochondria (33)(34)(35)(36)(37) for independent repair of organellar DNA. Drosophila dUTPase partially conforms to this situation; it is shown here to possess two physiological isoforms with different N termini. The 23-kDa isoform contains a putative NLS (PAAKKMKID), which is a close homolog of both the experimentally confirmed human c-myc NLS (PAAKRVKLD) (38,39) and human RanBP3 NLS (PPVKRERTS) (conserved residues shown in bold) (40). The homologous SPSKRARPA signal is present in the human dUTPase N terminus, wherein mutation of basic residues completely abolishes nuclear localization (2). The similarities in these NLS sequences together with the experimental evidence available for three different proteins, including the human ortholog, strongly suggest that the PAAKKMKID segment is indeed the authentic NLS of Drosophila dUTPase.
The 21-kDa isoform is an N-terminally truncated version lacking this NLS segment; however, it does not contain any other known transport signal. Cellular fractionation experiments (Fig.  3F) indicated that isolated mitochondria from Drosophila embryos do not contain detectable amounts of dUTPase. In this respect, it is important to point out that mitochondrial base excision repair enzymes were mostly investigated in mammalian systems, and our knowledge on the generation and metabolism of uracil-DNA in Drosophila is not yet complete. The lack of the uracil-DNA glycosylase homolog, uracil-DNA glycosylase, in in- sects (41) does, in fact, indicate some possible differences as compared with mammalian organisms.
Nuclear Localization Does Not Show a Complete Dependence on the Putative NLS-Immunohistochemical staining of actively dividing S2 cells demonstrated that most dUTPase colocalizes with nuclear DNA. However, Western blot analysis of these cells showed a 1:1 ratio of the two isoforms; only one of them possessed the putative NLS, indicating possible additional transporter factors.
In agreement with the above results, the capability of dUTPase translocation both into and out of the nuclei is further supported by immunohistochemical and Western blot experiments on developing ovaries and embryos (compare Figs. 3 and 4, A and B). dUTPase translocation from the cytoplasm of nurse cells into the nuclei of the early embryo cannot be rationalized purely on the basis of the NLS signal because the very same dUTPase population is shown to change its intracellular localization. On both the protein and mRNA level, a surplus of the shorter isoform lacking the putative NLS is evident from Western blot of mature follicles and mRNA analysis of ovaries (Fig. 3, C and E). This would suggest mostly diffuse cytoplasmic localization for embryonic dUTPase, in contrast to the experimental results (compare Fig. 4B). Translocation of both dUTPase isoforms either into or out off the nucleus alludes to the existence of an additional transport mechanism, probably involving specific dUTPase-binding proteins.
Developmental Regulation of Enzyme Level-The drastic decrease in expression of both dUTPase isoforms in larvae is paralleled with confinement of residual dUTPase to the imaginal discs (compare Fig. 3A and 4C). This stage-specific control is in line with earlier results (42), based on activity measurements. Lack of dUTPase in larvae was suggested to result in uracil-substituted DNA and induced thymineless cell death, contributing to apoptosis required during metamorphosis (12). The absence of dUTPase in larvae is expected to induce stable replacement of thymine by uracil in DNA, because the fruit fly genome codes only for mismatch-specific uracil-DNA glycosylases (41) and lacks the uracil-DNA glycosylase homolog, uracil-DNA glycosylase. However, the third larval stage was suggested to be associated with the expression of an effective uracil-DNA endonuclease (43) that may induce stage-specific DNA degradation. In agreement, we observed that imaginal discs, not sentenced to death in metamorphosis, retained dUTPase, whereas other larval tissues, subjected to apoptosis in the prepupal stage, did not. A critical assessment of the role of thymineless cell death in developmental apoptosis in Drosophila requires further studies involving characterization of the uracil content in larval DNA and phenotype analysis of transgenic mutant strains with perturbed dUTPase content. Such experiments are in progress in our laboratory.
This strict regulation of the dUTPase protein isoforms is in striking contrast with the constitutive mRNA levels (see Fig.  3). This contradiction underlines the importance of such parallel detection techniques in developmental investigations (44). Several mechanisms might be invoked to account for this discrepancy. First, mRNA lifetime may be unusually extended by complexation to control elements. Second, the productive translational polysome complex may be subjected to multilevel modulation, and third, posttranslational modifications may significantly shorten protein lifetime. According to available databases, the 3Ј-or 5Ј-untranslated dUTPase mRNA regions do not contain any known translational control elements. However, within the metabolic pathway of thymidylate biosynthe- sis to which dUTPase also belongs, thymidylate synthase regulates not only its own mRNA (by feedback) but also mRNA of dihydrofolate reductase (45)(46)(47). Lack of the methionine residue at the N terminus of the 23-kDa isoform may shorten protein lifetime (see N-end rule (48)).
Cell Cycle-dependent Regulation-In both Homo sapiens (30) and Candida albicans (49), dUTPase transcription is under cell cycle control. The present data suggest similar proliferationrelated expression of fly dUTPase (Fig. 3D). In addition to a Zeste site (14,50), the potential promoter region up to Ϫ250 bp of both transcripts contains two tandem Drosophila-specific DNA replication-related elements (TATCGATA) (51), present also in many cell cycle-controlled genes (e.g. proliferating cell nuclear antigen and DNA polymerase ␣) (52). This promoter region suggests same regulation for both isoforms. In contrast, only the nuclear isoform of human dUTPase is subjected to cell cycle-dependent regulation, whereas the mitochondrial isoform is constitutive, corresponding to the character of DNA synthesis in these two organelles.
A complete depletion of dUTPase in nonreplicating nuclei would seriously compromise constitutive (e.g. transcriptioncoupled) repair processes (53). Small constitutive amounts of dUTPase, escaping experimental detection and presumably coupled to DNA repair complexes (36,54), might resolve this problem. This underlines the importance of further investigations on dUTPase-interacting proteins, both in Drosophila and human cell lines.
Presence and Putative Role of the Extra C-terminal Region-The 28-residue Drosophila-specific C-terminal segment, present in both isoforms (see Fig. 1, C-E) is highly flexible and has no significant effect on the activity of the purified enzyme (14). However, several fractions of S2 cell extract modulated the activity of recombinant Drosophila dUTPase strictly, depending on the presence of the unique C terminus. 3 This unique region may therefore mediate regulation of dUTPase activity by other protein factors. dUTPases from various organisms contain species-specific, flexible N-or C-terminal extensions that provide interaction surface for cellular macromolecules (17)(18)(19). According to surface plasmon resonance experiments, the conserved dUTPase domain also participates in proteinprotein interactions (Fig. 6D). Far-Western data suggested a stage-specific pattern of dUTPase interacting proteins (Fig.  6E). Effector proteins of dUTPase are suggested to exist in Drosophila and in Bacillus subtilis (13,55).
In conclusion, we identified the Drosophila dUTPase isoforms both at mRNA and protein levels and characterized their expression, localization, and developmental patterns. In agreement with the observed multiple regulation of the enzyme, independent experiments supported the existence of endogenous macromolecular dUTPase partners. Further investigation and identification of dUTPase interacting proteins in Drosophila as well as in human cell lines is expected to provide novel insights into the mechanism of thymineless death.