ire-1-dependent transcriptional up-regulation of a lumenal uridine diphosphatase from Caenorhabditis elegans.

Lumenal ecto-nucleoside tri- and di-phosphohydrolases (ENTPDases) of the secretory pathway of eukaryotes hydrolyze nucleoside diphosphates resulting from glycosyltransferase-mediated reactions, yielding nucleoside monophosphates. The latter are weaker inhibitors of glycosyltransferases than the former and are also antiporters for the transport of nucleotide sugars from the cytosol to the endoplasmic reticulum (ER) and Golgi apparatus (GA) lumen. Here we describe the presence of two cation-dependent nucleotide phosphohydrolase activities in membranes of Caenorhabditis elegans: one, UDA-1, is a UDP/GDPase encoded by the gene uda-1, whereas the other is an apyrase encoded by the gene ntp-1. UDA-1 shares significant amino acid sequence similarity to yeast GA Gda1p and mammalian UDP/GDPases and has a lumenal active site in vesicles displaying an intermediate density between those of the ER and GA when expressed in S. cerevisiae. NTP-1 expressed in COS-7 cells appeared to localize to the GA. The transcript of uda-1 but not those of two other C. elegans ENTPDase mRNAs (ntp-1 and mig-23) was induced up to 3.5-fold by high temperature, tunicamycin, and ethanol. The same effectors triggered the unfolded protein response as shown by the induction of expression of green fluorescent protein under the control of the BiP chaperone promoter and the UDP-glucose:glycoprotein glucosyltransferase. Up-regulation of uda-1 did not occur in ire-1-deficient mutants, demonstrating the role of this ER stress sensor in this event. We hypothesize that up-regulation of uda-1 favors hydrolysis of the glucosyltransferase inhibitory product UDP to UMP, and that the latter product then exits the lumen of the ER or pre-GA compartment in a coupled exchange with the entry of UDP-glucose, thereby further relieving ER stress by favoring protein re-glycosylation.

The occurrence of the eukaryotic endomembrane system has brought forth a need for a mechanism to coordinate the metabolic flux of nucleotides, nucleotide sugars, and nucleotide sul-fate between the cytosol and the lumen of secretory organelles. E-type ATPases play an important role in this function. They have been conserved through evolution with their catalytic site in the ecto-position, facing the outer surface of the plasma membrane or its topological equivalent, the lumen of intracellular organelles such as the endoplasmic reticulum (ER), 1 Golgi apparatus (GA), and lysosomes/vacuoles. These enzymes, members of the ecto-nucleoside triphosphate diphosphohydrolase (ENTPDase) family, hydrolyze nucleoside tri-and/or diphosphates in the presence of cations and share, at the amino acid sequence level, five conserved motifs called apyrase-conserved regions (ACRs) (1).
Extracellular nucleotides such as ATP and ADP are intercellular signaling molecules in virtually every tissue where, modulated by ecto-apyrases, they participate in a broad range of biological processes such as the regulation of immune responses (2) and modulation of signaling by neuronal cells (3). Specifically, mammalian ENTPDase1/CD39 is responsible for the inhibition of ADP-induced platelet aggregation (4), whereas ENTPDase2 participates in the specification of neural cell migration during brain development by directly modulating ATPreceptor-mediated cell communication (5).
The physiological requirement for intracellular nucleotide diphosphatases in the lumen of the GA became apparent some years ago with the realization that nucleoside diphosphates (which are generated in this compartment as reaction products after glycosylation, sulfation, and phosphorylation of secretory and membrane proteins), proteoglycans, and lipids are inhibitors of glycosyltransferases. Therefore, nucleoside diphosphates, which do not cross the GA membrane, do not accumulate in the lumen of this organelle and are rapidly converted to nucleoside monophosphates by lumenal nucleotide diphosphatases. Nucleoside monophosphates, which inhibit glycosyltransferases to a much lesser extent, exit the lumen of the GA by means of a coupled exchange with entry from the cytosol of (additional) nucleotide sugars, nucleotide sulfate, or ATP, giving rise to a transport/antiport cycle present in every eukaryotic cell studied so far (6).
The specificity and subcellular location of three different ENTPDases have been described until now in the mammalian secretory pathway. Two of these enzymes occur in the lumen of the ER and are UDP/GDPases; one of them is membranebound, and its activity depends solely upon Ca ϩ2 (7), whereas the other, ENTPD5, is soluble and is also active with Mg ϩ2 and Mn ϩ2 (8). The third mammalian ENTPDase is a GA lumenal apyrase (9).
Saccharomyces cerevisiae and Schizosaccharomyces pombe each have two GA enzymes that degrade nucleoside diphosphates: ScGda1p (10) and SpGda1p (11), which are solely nucleotide diphosphatases and have preferred activity toward GDP and UDP; ScYnd1p (12,13) and SpYnd1p (11) are apyrases and use nucleoside di-and tri-phosphates as substrates. Loss of ScGDA1 function results in a drastic reduction of GDP-mannose-dependent (as opposed to lipid-linked mannose) mannosylation of proteins and lipids (10). The ynd1gda1 double mutant is synthetically lethal in S. pombe but not in S. cerevisiae, although the latter cells grow poorly and show severe cell-wall defects (11,12).
Contrary to the GA, where numerous lumenal nucleotide sugar-dependent glycosyltransferases occur, the ER is the location of only two well described lumenal nucleotide sugar-dependent glycosyltransferases: the glucuronosyltransferases that occur only in the liver of higher animals and the ubiquitous UDP-glucose:glycoprotein glucosyltransferase (GT) (14,15) that, with the notable exception of S. cerevisiae, has been reported to occur in all cells studied so far. This enzyme plays a pivotal role in the quality control of glycoprotein folding, uses UDP-glucose as substrate, and is inhibited by the reaction product UDP. The fission yeast S. pombe has a well characterized glycoprotein-folding facilitation cycle mediated by GT, in which UDP is also generated as an inhibitory reaction product in the lumen of the ER.
Surprisingly, null mutants in the GA-localized SpGda1p show a 35-50% decrease in the GT-dependent glucosylation of N-linked oligosaccharides which occurs in the ER (11), whereas inactivation of SpYND1 had no effect upon this reaction. This unconventional mechanism for dealing with ER-generated, lumenal UDP by the S. pombe Golgi ENTPDase was hypothesized to be the result of antero transport of ER-derived vesicles containing UDP to the GA (11).
Several GA mutants in nucleotide sugar-transporter genes from mammals, yeast, Leishmania, Drosophila, and C. elegans have been identified and characterized. Most of these mutations are leaky, but all of them have serious physiological consequences (16).
C. elegans is an attractive model to study the relevance of intracellular ENTPDases in alleviating ER stress and regulating protein and lipid glycosylation. Its genome encodes at least three proteins with sequence similarity to the S. cerevisiae GA ENTDPase, Gda1p. A nucleotide diphosphatase functionally homologous to the yeast Ynd1p, encoded by the mig-23 gene, was recently described in C. elegans (17). Loss of MIG-23 function results in altered gonad morphogenesis, demonstrating the importance of this enzyme (17). Here we report the presence of cation-dependent nucleotide di-and tri-phosphatase activities in membranes of C. elegans and the characterization of UDA-1 and NTP-1; the former is related, in substrate specificity, to the yeast Gda1p, whereas the latter is an apyrase and is related to the yeast Ynd1p. Heterologous expression of the above two ENTPDases resulted in UDA-1 having its active site facing the lumen of vesicles of intermediate density between the ER and GA, whereas NTP-1 seems to colocalize with GA markers. Transcription of uda-1 but not ntp-1 and mig-23 was up-regulated by conditions causing ER stress and the accumulation of unfolded proteins such as tunicamycin, ethanol, and temperature. Up-regulation of uda-1 did not occur in ire-1 mutants, thus demonstrating the role of this pathway in this event. Our results show that metazoans, unlike yeast, can relieve ER stress not only by increasing the expression of GT as described previously, but also by simultaneously increasing transcription of a lumenal UDPase that converts UDP to UMP. We hypothesize that increase of lumenal UMP leads to the increase of UDP-glucose entry into the ER lumen under conditions of stress, thereby facilitating glucosylation and removal of improperly folded proteins in the ER lumen.
Characterization and Heterologous Expression of UDA-1 and NTP-1-One cDNA clone corresponding to the KO8H10.4 transcript (yk400d11) and another corresponding to the C33H5.14 open reading frame (ORF) (yk156b1) were kindly provided by Dr. Kohara (National Institute of Genetics, Mishima, Japan). The former ORF was sequenced and subcloned into the plasmid p426 (20) to obtain p426-uda-1. This plasmid was then used for heterologous expression into yeast gda1 mutant cells. The cDNA sequence of yk400d11 completely matched the predicted sequence of the ORF K08H10.4 which had been deposited in GenBank TM /EBI Data Bank with the accession number Z83113. The C33H5.14 ORF was sequenced and subcloned in the pCDNA3.1 (Invitrogen) to be transfected in COS-7 cells. The transfection was performed with LipofectAMINE (Invitrogen) according to the manufacturer's instructions. The sequence obtained from the clone yk156b1 completely matched the sequence found in the GenBank TM /EBI Data Bank with accession number U41007.
Northern Analyses-For Northern blotting, total RNA from mixedstage animals (21) was resolved on a 1% formaldehyde-containing gel, transferred onto a nylon membrane, and hybridized with 32 P-labeled cDNA probes using a random priming kit (Roche Applied Science). Bands were then quantitated by densitometry. The cDNA probes were obtained by amplification of full-length ORFs utilizing the expressed sequence tag clones yk400d11 (K08H10.4, uda-1), yk156b1 (C33H5. 14,ntp-1), and yk576a12 (mig-23). The UDP-glucose:glycoprotein glucosyltransferase probe was obtained by amplification of 500 bp at the 3Ј end of expressed sequence tag clone yk250b10.3. All clones were kindly provided by Dr. Kohara and converted to plasmids with a Rapid Excision kit (Stratagene).
Preparation of Yeast Vesicle Fractions-Vesicle fractions were prepared from gda1-null mutant cells transformed with the plasmid p426uda-1 or with the corresponding empty plasmid. Cells were grown and vesicles were obtained as described previously (22). Protein was determined using the BCA method (Bio-Rad).
Preparation of Membrane Fractions from COS-7 Cells-Crude membranes of transfected COS-7 cells were prepared by the method of Coppi and Guidotti (23). Protein concentration was determined using the BCA method (Bio-Rad).
Preparation of Membrane Fractions from C. elegans-Two ml of packed N2 mixed stage worms, grown in liquid culture, were suspended in 1 ml of PBS buffer supplemented with protease inhibitors mixture (Sigma) and disrupted four times (1 min each time) using a Mini-Beadbeater-8 cell disrupter (Biospec Products). One volume of this homogenate was diluted 20 times with PBS and centrifuged at 3000 rpm for 10 min. The resulting supernatant was centrifuged at 100,000 ϫ g for 40 min, and the pellet was suspended in buffer containing 10 mM triethanolamine acetic acid, pH 7.2, 0.8 M sorbitol, 1 mM EDTA, 1 g/ml leupeptin, 0.5 mM phenylmethylsulfonyl fluoride and stored in aliquots at Ϫ70°C. For measurement of nucleotide phosphatase activity, hydrolysis of nucleotides was measured in membrane fractions as described previously (22).
Sucrose Gradient Fractionation-Yeast gda1 mutant cells, carrying the p426-uda-1 plasmid, were grown overnight until exponential phase. Subsequent workup and sucrose gradient fraction was done as described (10). ␣-1, 2-mannosyltransferase activity was used as the GA marker (24), whereas NADPH cytochrome c reductase activity was used as the ER marker (25).
Western Analyses-Sucrose gradient fractions were loaded on SDS-PAGE and electroblotted onto a polyvinylidene difluoride membrane (Bio-Rad) in Towbin buffer at 100 V for 1 h. Antibody against Chs3p (kindly provided by Dr. H. Lucero, Boston University) was used at a 1:3000 dilution. Anti-Pep12 antibody (Molecular Probes) was used at a 1:1000 dilution. The chitinase detection was performed as described in Lopez-Avalos et al. (26); the chitinase antibody (kindly provided by Dr. C. Specht, Boston University) was used at a 1:3000 dilution. The FLAG antibody (Sigma), used to detect the NTP-1 in COS-7 cells, was diluted 1:1000. The secondary antibodies were anti-rabbit or anti-mouse IgG conjugated with peroxidase (Bio-Rad). The detection procedure was done with an ECL blot detection kit (Amersham Biosciences) according to the manufacturer's instructions.

RESULTS
Nucleotide Hydrolyzing Activities of C. elegans Membranes-We were interested in determining the occurrence of ENTPDases and their activities in C. elegans as an initial step toward the elucidation of their role in lumenal ER and/or GA posttranslational modifications of proteins. Crude membrane preparations of mixed stage, wild-type nematodes were assayed for nucleotide mono-, di-, and tri-phosphatase activities, substrate specificities, and ionic requirements. In the presence of Ca ϩ2 , C. elegans membranes showed high UDPase and GDPase activities, while CDP and ADP were only marginally hydrolyzed (Fig. 1). When a mixture of nucleoside diphosphates was assayed as substrates, GDPase and UDPase activities were not additive ( Fig. 1), suggesting that both activities may be catalyzed by a single polypeptide, as reported for yeast and mammalian GDP/UDPases (8,9,10,26,27). The nucleotide diphosphatase activity was strictly cation-dependent, as shown by the negligible hydrolysis in the presence of EDTA (Fig. 1). In contrast, a significant portion of the nucleoside triphosphate hydrolysis was cation-independent ( Fig. 1).
Members of the nucleotide phosphatase family share a high degree of amino acid sequence similarity, particularly at their ACRs. To identify nematode proteins responsible for the GDP/ UDPase activities, the C. elegans genome database was searched for homologs of S. cerevisiae Gda1p, the principal yeast GDP/UDPase. This search resulted in the identification of two C. elegans ORFs, C33H5.14 and K08H10.4, in addition to the recently reported mig-23 (17). The K08H10.4 ORF has a high similarity to S. cerevisiae and S. pombe GDA1 and to a gene product encoding a rat ER UDPase (14), ENTDP5 (Fig. 2). This C. elegans ORF will be subsequently referred to as uda-1 and its product, which contains the 5 ACRs that are the hallmark of the ENTPD family, as UDA-1 (Fig. 2). The Kyte-Doolittle hydrophobicity plot of this C. elegans ORF (28) suggests the presence of one N-terminal transmembrane domain between residues 1 and 21 (Fig. 3A). The other ORF, C33H5.14, whose gene product is NTP-1, also has five ACR domains but is more closely related to yeast and human apyrases, Ynd1p and ENTPD5 (9,11,12,13), than to UDA-1 (Fig. 3C). The Kyte-Doolittle hydrophobicity plot of this other C. elegans ORF (Fig.  3B) suggests the presence of one C-terminal transmembrane domain between residues 442 and 459.
Substrate Specificity of UDA-1-The substrate specificity and other biochemical properties of UDA-1 were studied after expression in S. cerevisiae gda1-null mutants because of their low background in nucleotide phosphatase activities. Nucleotide diphosphatase activities were significantly increased in membranes from gda1 mutant cells transformed with the vector containing uda-1 cDNA versus those transformed with the empty vector. Expression of uda-1 resulted in a 6-fold increase in UDPase activity and a 3-fold increase in GDPase activity (Fig. 4). A small increase in CDP hydrolysis was also observed, whereas no significant changes were detected when either nucleoside mono-or tri-phosphates were used as substrates (Fig.  4). The catalytic activity of UDA-1 toward UDP was highest when the pH of the reaction was 8.0 (data not shown), whereas for GDP the maximum was at pH 7.5 (data not shown). The K m for UDP at pH 8 in the presence of 10 mM Ca 2ϩ was 180 Ϯ 7 M (mean Ϯ S.E., n ϭ 3). UDPase and GDPase activities were optimally and similarly stimulated at 10 mM Ca 2ϩ , Mg 2ϩ , or Mn 2ϩ ; however, some cation selectivity was detected at 2 mM, where Mg 2ϩ Ͼ Ͼ Mn 2ϩ ϭ Ca 2ϩ . No enzymatic activity could be measured in the presence of EDTA (data not shown). The substrate specificity profile determined for the recombinant UDA-1 (Fig. 4) is almost identical to the pattern of nucleotide diphosphate hydrolysis produced by a total membrane fraction obtained from mixed-stage nematodes (Fig. 1), suggesting that UDA-1 may be the major membrane-bound enzyme responsible for hydrolyzing UDP and GDP in C. elegans as Gda1p is in S. cerevisiae (10).
To determine whether the catalytic domain of UDA-1 was oriented toward the lumen of an organelle, we measured the UDPase activity in intact and permeabilized vesicles from gda1 yeast cells expressing the enzyme. Only vesicles in which membrane permeability had been disrupted with either detergents or the ionophore alamethicin showed 2-to 6-fold higher activity compared with intact ones (Fig. 5). This result suggests that the active site of UDA-1 is lumenal. Vesicles prepared from gda1 yeast cells, carrying an empty vector, showed significantly lower activity when permeabilized with TX-100 than with alamethicin or digitonin, which is consistent with the previous observation that Ynd1p, responsible for the background activity in gda1 mutant cells, is inhibited by TX-100 (13).
We then studied, by sucrose gradient fractionation, the subcellular localization of UDA-1 expressed in S. cerevisiae gda1 mutants. Migration of UDA-1 was detected by its UDPase activity and localized to a compartment having a density in between that of the yeast ER and GA enzymatic markers, NADPH cytochrome c reductase and ␣-1,2-mannosyltransferase, respectively (data not shown). The UDPase peak did not coincide with markers of the endosomal compartment such as Pep12p, which is implicated in vesicular protein transport, or the catalytic subunit of chitin synthase complex III, Chs3p (data not shown).
The glycosylation defects present in gda1 yeast mutants are the result of a severe reduction in GDP hydrolysis in the Golgi apparatus. Any phenotypic correction detected upon expression of UDA-1 in this yeast strain would indicate that at least some of the C. elegans protein reached the GA. We analyzed the Oglycosylation profile of chitinase secreted by gda1 cells because its under-glycosylation results in a faster migration on SDS-PAGE than chitinase secreted by wild-type yeast. No difference in the migration of chitinase secreted by gda1 cells transformed with the plasmid-containing uda-1 or the vector alone were seen, indicating no correction of the O-glycosylation defects and presumably no GA localization (Fig. 6). The global cell wall alteration present in gda1 mutants (as revealed by sensitivity to ␤1,3-glucanase-induced lysis; Ref. 26) also could not be suppressed by the expression of uda-1 (data not shown). Presumably, the nematode enzyme did not reach the GA of yeast.
Because a heterologous expression system was used, the specific organelle where the UDPase resides in C. elegans can only be inferred. Nevertheless, two additional lines of evidence suggest an intracellular location for UDA-1: (i) the enzymatic activity is latent as shown above (Fig. 5), and (ii) the substrate specificity is very similar to that of intracellular ENTPDases, Substrate Specificity of NTP-1-To determine enzymatic activity of NTP-1, we first attempted heterologous expression in yeast; this resulted in no significant increase in enzymatic activity of the transformants. Transient transfection into COS-7 cells was then performed. Western blots against the epitope tag localized at the C terminus of the ORF showed expression of a protein of ϳ63 kDa (data not shown). The substrate specificity of NTP-1 was then measured in crude FIG. 2. Sequence comparison of C. elegans K08H10.14 (UDA-1) and C33H5.14 (NTP-1) with related UDP/GDPases. Sequence alignment was performed using the BOXSHADE program. Underlined residues represent the five ACR (apyrase conserved region) domains. The GenBank TM accession numbers of the sequences are given in parentheses: Gda1p, Saccharomyces cerevisiae (U18799); MIG-23, Caenorhabditis elegans (U39652); ENTPDase6, Rattus norvegicus (AJ238636); K08H10.4, Caenorhabditis elegans (Z83113); ENTPDase5, Homo sapiens (AF016032); C33H5.14, Caenorhabditis elegans (U41007). membranes of COS-7 cells transfected transiently with the heterologous cDNA and compared with the activity of cells transfected with the empty vector. Expression of ntp-1 resulted in a significant increase in calcium-dependent hydrolysis of nucleoside triphosphates and a marginal increase in the hydrolysis of nucleoside diphosphates (Fig. 7). Immunostaining of the expressed protein showed a perinuclear distribution with partial overlap, with fluorescent staining with an anti-GOS28, a Golgi apparatus protein, suggesting perhaps a GA localization for NTP-1 (data not shown).
uda-1 Is Transcriptionally Up-regulated in an ire-1-Dependent Manner-Accumulation of unfolded/misfolded proteins within the ER lumen is deleterious for cell survival. A signaling pathway between the ER and the nucleus, the unfolded protein response (UPR), has evolved in all eukaryotes to mitigate such stress. A mechanism has also been described for retaining not yet properly folded N-glycan-bearing proteins and for increasing their folding efficiency in the ER (14). Monoglucosylated oligosaccharides, which originated as trimming intermediates, mediate the interactions of folding glycoproteins with the ER resident lectin/chaperones, calnexin and calreticulin. Deglucosylation by glucosidase II releases glycoproteins from their chaperone anchors. GT, which uses UDP-glucose as a substrate, is an ER sensor, together with lectin chaperones, of glycoprotein conformations, giving rise to monoglucosylated N-glycans only if the oligosaccharides are linked to incom-pletely folded glycopeptides. Conditions that promote accumulation of unfolded proteins in the ER have been reported to significantly induce synthesis of GT mRNAs in S. pombe and mammals (14).
It has been hypothesized that a rat Ca ϩ2 UDPase, ENTDP6, which remains soluble in the lumen of the ER and is devoid of transmembrane domains and KDEL-related ER localization sequences (8), promotes reglucosylation reactions involved in glycoprotein folding and quality control by hydrolyzing UDP, which is an inhibitory product of the UDP-glucose:glycoprotein glucosyltransferase (8,29). This ER UDPase may also facilitate reglucosylation by generating the antiporter, UMP, which is required for entry into the ER lumen of the substrate, UDPglucose. In mammals, this nucleotide sugar enters the lumen of the ER by means of a specific transporter in a coupled, equimolar exchange with UMP (30). No in vivo supporting evidence for this hypothesis has yet been provided.
The amino acid sequence and predicted topology for UDA-1 is very similar to that experimentally demonstrated for the above Ca ϩ2 UDPase. If indeed UDA-1 participates in glycoprotein reglucosylation reactions, then stress conditions that may promote the accumulation of misfolded proteins, such as high temperature, addition of tunicamycin or ethanol, should result in an increased mRNA encoding for this enzyme. To investigate this point, RNA from mixed stage, wild-type nematodes, before and after treatment of the animals with the above effectors, was extracted and analyzed by Northern blotting. A 6-h shift from 16 to 25°C resulted in a 3-to 4-fold induction of the expression of-uda-1 mRNA (Fig. 8). A similar effect was observed upon the addition of either tunicamycin or ethanol (Fig.  8). No modification in the expression of an unrelated mRNA, actin, was observed (Fig. 8).
The C. elegans genome predicts two proteins that are homologous to all canonical GTs, both in term of size and amino acid sequence conservation. The probe utilized here was derived from the 3Ј region, which contained the terminal 300 amino acids where the identity is highest and would, therefore, recognize both GT homologs, which differ by 300 bases in length. Similar, up-regulation of the mRNAs of uda-1 and GT was observed upon exposure of nematodes to tunicamycin or ethanol (Fig. 8); the latter treatment, as well as high temperature, were previously reported to induce expression of the mRNA of calreticulin (31), an abundant calcium-binding lectin chaperone of the ER and pre-GA compartment (34). Importantly, the transcript levels of the mig-23 apyrase and that of the apyrase encoded by ORF C33H5.14, ntp-1, did not change under any of the above experimental conditions (Fig. 8).
Induction of the UPR under the above experimental conditions was confirmed by detection of increased expression of hsp4, a C. elegans homolog of the mammalian ER chaperone BiP. Incubation in liquid medium containing tunicamycin or ethanol of a C. elegans transgenic strain in which the hsp4 promoter drives the expression of green fluorescent protein resulted in significant induction of green fluorescent protein expression compared with controls ( Fig. 9). The localization of hsp4::gfp expression was consistent with previous reports (32). The above results demonstrate that uda-1 transcription is up-regulated by the same effectors leading to induction of the UPR.
Direct evidence that UPR signaling is involved in uda-1 up-regulation by the above effectors was obtained by measuring its transcription level in nematodes unable to signal through IRE-1. This is an ER transmembrane glycoprotein containing both kinase and RNase activities in its cytoplasmic domain. ER stress leads to its autophosphorylation and the subsequent activation of its RNase activity and, as a consequence, the splicing of XBP1 mRNA. In this way, the activation of the transcription factor allows the expression of the UPR genes (32). We found that a null mutation causing loss of ire-1 function abolished up-regulation of uda-1 transcription by the above listed experimental conditions, thus directly demonstrating a role for ire-1 in the up-regulation of uda-1 transcription. (Fig. 10).

DISCUSSION
In this study, we have shown the occurrence of cation-dependent nucleotide di-and tri-phosphatase activities in membranes of C. elegans. We cloned and expressed the genes for two novel ENTPDases, uda-1 and ntp-1. Although UDA-1 causes hydrolysis of UDP and GDP and, to a much lesser extent, CDP, it does not hydrolyze other nucleoside diphosphates, triphosphates, or monophosphates. This finding is in contrast to the behavior of NTP-1, which hydrolyzes both nucleoside di-and tri-phosphates. The substrate specificity of UDA-1 is very similar to the first purified and cloned ecto-nucleoside diphosphatase, Gda1p of S. cerevisiae, which, we have shown previously in in vivo and in vitro experiments (6), plays a pivotal role in the nucleotide sugar transport/antiport cycle. Recently a different ENTPDase of C. elegans, MIG-23, was described (17). Although this protein has activity toward UDP and GDP, it also hydrolyzes ADP, which UDA-1 does not; whether or not MIG-23 hydrolyzes nucleoside tri-or mono-phosphates is not known. Gene expression data 2 indicate that the transcript for uda-1 (K08H10.4), being high in the embryos, can be detected in all developmental stages of the nematode; it decreases during the first two larval stages and then rises to maximum in larva 4 stage and adult hermaphrodites.
We and others have previously demonstrated (6,12,13) that UDP/GDPases play crucial roles in protein and lipid glycosylation reactions that occur in the lumen of the GA. The sugar donors for these reactions, nucleotide sugars, must first be transported by a specific transporter from the cytosol, their site of synthesis, into the lumen of the GA, where sugars are transferred by specific transferases to proteins and lipids (6,33,37). These transporters have recently been described in C. elegans, where SQV-7 is involved in vulva invagination and early embryonic development (38,39). Several GA lumenal glycosyltransferases, all of which yield nucleoside diphosphates as products (40 -45), have also been identified in C. elegans. These can only exit the GA lumen after hydrolysis to nucleoside monophosphates. MIG-23 is such a lumenal nucleotide diphosphatase; it affects gonad morphogenesis through glycosylation of the MIG-17 ADAM protease (17). It is also likely that other ENTPDases from C. elegans will have similar functional roles 2 Available on the World Wide Web at www.wormbase.org. Northern blot analysis of total RNA extracted from wild-type worms grown under different conditions. A, temperature stress was induced by keeping cultures at 25°C for 6 h. B, ethanol at 7% or 5 g/ml tunicamycin was added at 16°C for 6 h. Specific probes are described under "Experimental Procedures." C, densitometric ratios of intensities of Northern blots.
FIG. 9. Induction of UPR. SJ4005 worms, containing an integrated copy of hsp-4::gfp, a green fluorescent protein transcriptional reporter driven by hsp-4 promoter, were grown at 16°C. A, untreated. B, tunicamycin addition. C, ethanol addition. Treatments were performed as described for Fig. 8. Fluorescent images were obtained using the same exposure for all treatments.
FIG . 10. uda-1 up-regulation is IRE-1-dependent. Northern blot analysis of total RNA extracted from ire-1 RE666 worms, having a deletion in the ire-1 gene, were grown under different conditions. Tunicamycin or ethanol was added for 6 h at 16°C as described under "Experimental Procedures." in generating nucleoside monophosphates necessary for the nucleotide sugar transport/antiport cycle and glycosylation of proteins and lipids.
When C. elegans was exposed to elevated temperature, tunicamycin, or ethanol, conditions known to cause stress, transcriptional up-regulation of uda-1 but not of ntp-1 and mig-23 occurred. This specific transcriptional up-regulation is mediated by the ire-1 pathway, as mutants unable to signal by the ER stress sensor IRE-1 failed to elicit such up-regulation. We postulate that the principal role of transcriptional up-regulation of uda-1 is to increase the availability of UDP-glucose in the lumen of the ER or pre-GA (see below) under conditions of ER stress, thereby increasing the glucosylation of improperly folded proteins. Up-regulation of GT transcription under ER stress has been reported previously in S. pombe (46) and human cells (47) and seems to occur together with that of UDA-1 (Fig. 8). Synergy between GT and UDPase activities is postulated to bring fast relief of stress to the ER lumen: the increased rate of glucosylation of misfolded glycoproteins, as more GT becomes available, would then promote their recognition together with lectin chaperones which, in turn, provide them with additional opportunities to acquire a mature conformation and exit the ER. A direct consequence of increased GT activity could be the accumulation of its inhibitory product, UDP, in the lumen of the ER. Increase in UDA-1 would generate UMP, which can exit the ER or pre-GA as antiporter, thus promoting entry of additional UDP-glucose, the substrate of GT, from the cytosol.
Several lines of evidence support our hypothesis, although important questions discussed further below remain to be answered. UDA-1 is a nucleotide diphosphatase with its active site facing the lumen of an organelle having a density in between that of the classical ER and GA markers when expressed in S. cerevisiae. Although one must be cautious in making physiological inferences of protein localization in heterologous systems, the results are consistent with recent studies showing that GT, glucosidase II, as well as CNX and CRT, are not only localized in the ER per se, but also in a pre-GA compartment of rat hepatocytes, pancreas, as well as salivary glands of Drosophila (34). Previous studies from our and other laboratories have shown that vesicles from such compartments migrate on sucrose gradients at densities in between the ER and GA markers (35,36). Recently it was also shown by immunohistochemistry that the ER and pre-GA transitional elements from Chinese hamster ovary cells contain, in addition to the soluble apyrase ENTPDase5, a membrane-associated Ca ϩ2 -dependent UDPase. An open reading frame with high sequence similarity to the latter protein has been identified in C. elegans (F08C6.6) but not in yeast (7); it will be interesting to determine the subcellular location and membrane association of this enzyme and its possible induction by the UPR. As mentioned above, among the important caveats of our hypothesis that have not yet been answered is a direct demonstration that UDA-1 and/or GT and the supply of lumenal UDP-glucose are rate-limiting during protein glucosylation in the ER, and that the up-regulation of transcription of these proteins, as also perhaps the UDP-glucose transporter, will result in a decrease of misfolded proteins in the ER lumen under conditions of stress.
The ATPase activity of eukaryotic intracellular ENTPDases, which is absent in UDA-1 but not in NTP-1 and most likely also MIG-23, may play an important role in maintaining lumenal ATP concentration within different compartments such as the ER for protein import and processing, the GA for the phosphorylation of secreted and membrane proteins (6) and for acidification of lysosomes/vacuoles. Nucleoside tri-and diphosphates cannot cross the lysosomal membrane and must be converted to monophosphates to return to the cytosol and balance the nucleotide metabolic flux. A human apyrase, LALP70, is likely to participate in this function (48).
The only apyrase of S. cerevisiae and S. pombe, Ynd1p, occurs in the yeast GA and participates in the glycosylation process (11)(12)(13). The ATPase activity of ScYnd1p is normally down-regulated by the binding of its cytoplasmic C terminus to Vma13p, a subunit of the vacuolar H ϩ -ATPase (49). It has been reported recently (50) that after ATP uptake into the GA lumen, vesicular transport leads to the extracellular release of ATP, and that both processes are regulated by the vacuolar ATPase, thus implicating the yeast GA apyrase not only in organellar acidification, but also in ATP transport/utilization along the secretory pathway. Because C. elegans has two apyrases, NTP-1 and most likely MIG-23, which are highly similar to GA apyrases from yeast and mammals, it is tempting to speculate that these may also be important in similar events. C. elegans is an attractive model to study the role of these ENTPDases in multicellular eukaryotes.