Cloning and Characterization of a Mammalian Lithium-sensitive Bisphosphate 3′-Nucleotidase Inhibited by Inositol 1,4-Bisphosphate*

Discovery of a structurally conserved metal-dependent lithium-inhibited phosphomonoesterase protein family has identified several potential cellular targets of lithium as used to treat manic depression. Here we describe identification of a novel family member using a “computer cloning” strategy. Human and murine cDNA clones encoded proteins sharing 92% identity and were highly expressed in kidney. Native and recombinant protein harbored intrinsic magnesium-dependent bisphosphate nucleotidase activity (BPntase), which removed the 3′-phosphate from 3′-5′ bisphosphate nucleosides and 3′-phosphoadenosine 5′-phosphosulfate with K m andV max values of 0.5 μm and 40 μmol/min/mg. Lithium uncompetitively inhibited activity with aK i of 157 μm. Interestingly, BPntase was competitively inhibited by inositol 1,4-bisphosphate with aK i of 15 μm. Expression of mammalian BPntase complemented defects in hal2/met22 mutant yeast. These data suggest that BPntase’s physiologic role in nucleotide metabolism may be regulated by inositol signaling pathways. The presence of high levels of BPntase in the kidney are provocative in light of the roles of bisphosphorylated nucleotides in regulating salt tolerance, sulfur assimilation, detoxification, and lithium toxicity. We propose that inhibition of human BPntase may account for lithium-induced nephrotoxicity, which may be overcome by supplementation of current therapeutic regimes with inhibitors of nucleotide biosynthesis, such as methionine.

Lithium is a major drug used to treat manic depression, yet its molecular mechanism of action has not been conclusively elucidated. Insight into lithium's pharmacological activity has come with the identification of a magnesium-dependent phosphomonoesterase family whose members are inhibited by lithium at subtherapeutic concentrations (1). Despite relatively low overall sequence similarity, in some cases undetectable, the family shows a conserved three-dimensional core structure. Functional and structural studies of three members, fructose 1,6-bisphosphatase (fbptase) 1 , inositol monophosphatase (im-pase), and inositol polyphosphate 1-phosphatase (1ptase), indicate that residues involved in metal binding and catalysis are conserved within a sequence motif D-X n -EE-X n -DP(i/l)D(s/g/ a)T-X n -WD-X 11 -GG (2)(3)(4)(5)(6)(7). Of interest, several studies indicate that lithium interacts at one or more of the metal binding sites (6,8,9). 2 Together, these data support the idea that the signature motif confers the common enzymatic characteristics of the family, including lithium sensitivity.
Direct evidence links physiological effects of lithium to members of this phosphomonoesterase family. Treatment of cells with lithium results in accumulation of inositol mono-and polyphosphates, suggesting that impase and 1ptase are inhibited in vivo. Dichtl et al. (17) recently showed that the mechanism of lithium toxicity in yeast involves inhibition of the cytosolic RNA processing enzyme Xrn1p due to the inhibition of Hal2p and subsequent accumulation of PAP. Concurrent PAPindependent lithium-mediated inhibition of RNase MRP leads to an accumulation of immature ribosomal RNA molecules (17). In addition, the data of Acharya et al. indicate that 1ptase is a target of lithium in Drosophila melanogaster (18). Defects in neuronal function observed in Drosophila 1ptase mutants were phenocopied precisely by the administration of lithium to wildtype flies (18).
The conservation of the sequence motif of this family is important for identifying novel family members from human genome sequencing efforts. To this end, we have used this motif to clone a novel mammalian metal-dependent lithium-inhibited bisphosphate 3Ј-nucleotidase (BPntase) from expressed sequence tag (EST) data bases. Characterization and tissue distribution of a mammalian BPntase suggest that it is involved in nucleotide metabolism, sodium homeostasis, and the physiological effects of lithium.

MATERIALS AND METHODS
Protein concentrations were determined by the method of Bradford (19), using bovine serum albumin (BSA) as a standard or by measuring A 280 (assuming an extinction coefficient of 1 cm/ml/mg). SDS-PAGE was run using the Laemmli method (20). Gels were silver-stained as described within the Silver Stain Plus system (Bio-Rad). DNA-manipulating enzymes were from Life Technologies, Inc. or Roche Molecular Biochemicals. All other materials were reagent grade, typically purchased from Sigma. Polyclonal antibodies were generated by injecting PAP-agarose-purified enzyme as described below into rabbits following a standard protocol (Alpha Diagnostic, San Antonio, TX). The Western blot procedure followed a standard protocol and was developed with a phosphorescent reagent activated by horseradish peroxidase conjugated to donkey-anti-rabbit IgG (Amersham Pharmacia Biotech).
Clone Identification and Sequencing-Mouse and human EST data bases were searched with BLASTP, gapped BLASTP, or TBLASTN (21) with appropriate query sequences. Candidate cDNAs were obtained from Research Genetics, Inc. (Huntsville, AL) and completely sequenced on both strands by dideoxynucleotide chain termination (22) with Sequenase version 2.0 (Amersham Pharmacia Biotech) and/or by fluorescent terminator sequencing (Duke University Medical Center Genome Facility, Durham, NC).
Plasmid Construction-Full-length mouse and human cDNA's 439033 and 645079 were obtained from Research Genetics, Inc. (Huntsville, AL) as inserts in pBluescriptSK(Ϫ). The mouse coding sequence was inserted into pGEX3X (Amersham Pharmacia Biotech) to generate an in-frame glutathione S-transferase (GST)-439033 fusion expression construct. The pBluSK439033 plasmid was digested with DraIII and BclI, and a fragment representing residues 5-308 was purified. Residues 1-5 were rebuilt with the complementary linker oligonucleotides 5Ј-AATTCGGGATCCCCATCATGGCTCCAGCCACACC and 3Ј-GCCC-TAGGGGTAGTACCGAAGGTCGGTGTGG. Triple body ligation of the 1-kb fragment, linker adapter, and pGEX3X/BamHI cut was performed to create pGEX439033. The non-fusion 439033 protein was made by generating recombinant 439033 baculovirus DNA. The baculovirus transfer vector pVL1393 (PharMingen, San Diego, CA) was digested with BamHI, and the BamHI/BclI fragment of pGEX439033 was inserted to create pVL439033. For yeast complementation studies, the BamHI-NotI fragment of pVL439033 was ligated into BamHI/NotIlinearized pRS426GAL (23) (obtained from Susan Wente, Washington University, St. Louis, MO) to create pRS439033, a yeast high copy plasmid with expression under control of a galactose-inducible promoter.
The HAL2 coding sequence was produced by polymerase chain reaction (PCR) amplification of genomic Saccharomyces cerevisiae DNA with primers 5Ј-GAGCTCCCGGGTGATCATATGGCATTGGAAAGAG and 5Ј-GGTACCTGATCAGCGGGCCGCAAGCTTAGGCGTTTCTTGA-CTG as described (24). Bacterial cysQ was cloned using oligonucleotide primers 5Ј-CATATGTTAGATCAAGTATGCCAGC and 5Ј-AAGCTTCT-GTTTCTGCCATCTGAATTTAG. PCR was performed in 100-l reactions using a GeneAmp PCR System 2400 temperature cycler (Perkin-Elmer). Reactions contained 100 ng of yeast genomic DNA or approximately 10 4 Escherichia coli DH5␣ cells, 100 ng of each primer, 0.2 l of Expand High-Fidelity Taq polymerase (Roche Molecular Biochemicals), and 1ϫ Expand PCR buffer. PCR proceeded for 30 cycles of 30 s of denaturation (94°C), 30 s of annealing (52°C), and 1 min of extension (72°C). The amplified products were gel-purified and cloned into pCR2.1 using the TA cloning kit (Invitrogen, San Diego, CA) to create pCRHAL2 and pCRcysQ. The BclI-NotI fragment of pCRHAL2 and the BamHI-NotI fragment of pCRcysQ were cloned into BamHI/NotI-linearized pRS426GAL to create pRSHAL2 and pRScysQ. The NdeI-BclI fragment of pCRHAL2 and the NheI-BamHI fragment of pCRcysQ were isolated and ligated into NdeI/BamHI-linearized pET11c or NheI/BamHI-linearized pET11c bacterial expression vector (Novagen, Madison, WI) to create pETHAL2 and pETcysQ. These plasmids were transformed into the E. coli strain BL21(DE3) for overexpression.
Recombinant virus was made by co-transfecting Spodoptera frugiperda (Sf9) cells (2 ϫ 10 6 ) with 2 g of sterile pVL439033 and 1 g of BaculoGold baculovirus DNA (PharMingen) as described in the Baculo-Gold user manual. Viable, recombinant viruses were harvested 4 days after infection and amplified to a 300-ml quaternary stock in Sf9 cells. Virus stocks had an approximate titer of 1 ϫ 10 8 and were stored at 4°C.
Northern Blot Analysis-Approximately 30 ng of the 430-bp StyI fragment of the human clone 251113 was radiolabeled with [␣-32 P]ATP to a specific activity of approximately 1 ϫ 10 9 cpm/g as described with the Random Hexamer primer labeling kit (Roche Molecular Biochemicals). A human multiple tissue Northern blot, prepared with 2 g of poly(A)-selected mRNA isolated from various tissues, was purchased from CLONTECH. The membrane was prehybridized and hybridized at 42°C in 10 ml of 50% formamide, 5ϫ SSPE (0.9 M NaCl, 50 mM phosphate, pH 7.4, 5 mM EDTA), 10ϫ Denhardt's solution, 0.1 mg/ml sheared salmon sperm DNA, 2% SDS. The membrane was washed twice with 2ϫ standard saline citrate (SSC), 0.05% SDS at room temperature and twice with 0.1ϫ SSC, 0.1% SDS at 50°C and then exposed to film. The membrane was stripped and reprobed with radiolabeled human actin DNA supplied with the blot.
Preparation of 5Ј [ 32 P]PAP Substrate-Nucleotide 3Ј-monophosphates (3Ј-AMP, -dAMP, -CMP, -GMP, -TMP, and -UMP) (Sigma) were 5Ј-labeled with 32 P isotope by incubating an excess of the unlabeled nucleotide with [␥-32 P]ATP (NEN Life Science Products) and T4 polynucleotide kinase (Roche Molecular Biochemicals) under conditions recommended by the manufacturer. The reaction was stopped by the addition of formate to a concentration of 0.01 M COOH, 0.5 M NH 4 COO. The mixture was applied to a 25-l Dowex (AG 1-X8 Resin, 200 -400mesh, formate form; Bio-Rad) column equilibrated in formate. Following extensive washing, the product was eluted with 0.03 M COOH, 1.05 M NH 4 COO. Ammonium formate was removed by repeated lyophilizing.
Sf9 cells were maintained in spinner flasks at 28°C in Grace's medium (Life Technologies, Inc.) at pH 6.1 supplemented with 3.3 g/liter yeastolate, 130 mg/liter lactalbumin hydrolysate, 50 g/ml gentamycin, and 10% heat-inactivated fetal bovine serum (FBS) until cell densities reached 1 ϫ 10 6 cells/ml (log phase). Viral infections were performed by incubating cells (1 ϫ 10 9 ) with approximately 15 ml of virus solution for 4 h at 28°C with periodic gentle agitation. The cells were then diluted to 1 liter with Grace's medium complete with 10% FBS. The infected cell suspension was incubated with stirring for 60 h. Cells were harvested by centrifugation at 6000 ϫ g for 15 min, followed by washing with phosphate-buffered saline. Cells were either frozen immediately at Ϫ80°C for later use or were lysed and purified at once.
Purification of the GST fusion protein was performed using glutathione agarose as described in the manual provided by Amersham Pharmacia Biotech. Purification of the non-fusion recombinant mouse BPntase initially involved phenyl-Sepharose high performance liquid chromatography (HPLC). Cell pellets were resuspended in lysis buffer (25 ml/5 g wet cells) containing 50 mM HEPES, pH 7.5, 5 mM MgCl 2 , 3 mM NaCl, 1 mM EGTA, and 1 mM phenylmethylsulfonyl fluoride (PMSF). Cells were lysed in a French pressure cell, and cell debris was removed by centrifuging at 50,000 ϫ g for 20 min. The supernatant was filtered through a 0.22-m cellulose acetate filter and applied to Bio-Rad Q and S ion exchange cartridges connected in series and equilibrated in 50 mM HEPES, pH 7.5, 3 mM MgCl 2 (buffer A). The flowthrough and one column volume wash were collected and combined. Ammonium sulfate was added gradually to 30%, and the 50,000 ϫ g supernatant was applied to a 2-ml phenyl-Sepharose HPLC (Bio-Rad) column equilibrated with 30% ammonium sulfate in buffer A (buffer B). The column was washed with 10 column volumes of buffer B, and the protein was eluted with a 15-column volume reverse linear gradient from 30% to 0% ammonium sulfate saturation in buffer A. The protein eluted in 18 -22% ammonium sulfate saturation.
In order to improve the yield and reliability of the purification, an affinity purification step was developed based on the enzyme's activity on PAP. PAP-agarose resin (Sigma) was swelled in PAP-agarose buffer (50 mM HEPES, pH 7.5, 10 mM CaCl 2 , and 50 mM KCl) and poured into a 10-ml disposable polypropylene chromatography column (Bio-Rad). A 1-ml bed volume was sufficient to bind up to 6 mg of enzyme. Calcium, a potent inhibitor of homologous phosphomonoesterases, was added to satisfy the metal requirement for substrate binding while preventing the hydrolysis of immobilized PAP. Cell pellets were resuspended in PAP-agarose buffer containing 1 mM PMSF, and the cells were disrupted in a French pressure cell and centrifuged as above. The supernatant was applied to the column by gravity flow, and the column was washed with 15 column volumes of PAP-agarose buffer plus 0.5 M NaCl and then reequilibrated with PAP-agarose buffer containing no additional salt. The enzyme was eluted with PAP-agarose buffer containing 300 M 2Ј/3Ј,5Ј-PAP (Sigma). The protein was concentrated and the PAP removed by dialysis in Microcon centrifugal concentrators, 30,000 nominal molecular weight limit (Millipore; Bedford, MA). The protein was purified further from small molecules by size exclusion chromatography on Sephadex G50 (fine) QuickSpin columns (Roche Molecular Biochemicals). The protein was stored in 50 mM NaHEPES, pH 7.5, 100 mM KCl, 3 mM MgCl 2 , and 0.02% NaN 3 . Aliquots were frozen at Ϫ80°C and were stable for at least 3 months with no detectable loss of activity.
Native protein was also isolated with PAP-agarose affinity purification. Kidneys, lungs, liver, and heart from a freshly dissected mouse were homogenized separately with a Dounce homogenizer in 1 ml of ice-cold PAP-agarose buffer containing 1 mM PMSF. The homogenates were spun at 22,800 ϫ g in a 4°C microcentrifuge. The soluble fractions were applied to 50-l PAP-agarose columns, which were washed and eluted as above. The eluants were analyzed by SDS-PAGE and visualized with silver staining and Western blotting with the anti-BPntase antibody. Eluted enzyme from the kidney preparations was pooled, and BSA was added to a final concentration of 0.1 mg/ml. The protein was dialyzed approximately 50,000-fold against a reaction mixture containing 50 mM NaHEPES, pH 7.5, 100 mM KCl, 1 mM EGTA, and 3 mM MgCl 2 (HEKM) with Microcon centrifugal concentrators (Millipore) prior to performing kinetic analyses.
Determination of Substrate Specificity-Purified GST-439033 fusion protein was utilized to investigate reactions catalyzed by the protein.
Sequence homology with the consensus Li ϩ -sensitive phosphomonoesterase family active site motif led us to predict that the enzyme hydrolyzed small phosphorylated molecules, possibly including inositols and nucleotides, in a Mg 2ϩ -dependent manner. Trace molar amounts of [ 3 H] inositol standards (American Radiolabel Corp., St. Louis, MO) were incubated with protein in HEKM. Following incubation at 37°C, the reaction was stopped with 1 volume of 20 mM ammonium phosphate (AP) and was loaded onto a 4.6 ϫ 250-mm PartiSil 10 SAX-HPLC strong anion exchange column (Whatman, Clifton, NJ). Pre-reaction inositol standards and post-reaction potential inositol products typically were separated with linear gradients from 10 mM to 1.7 M AP over 50 min. Radioactivity of the eluant was measured continuously using a Beta-RAM in-line detector (INUS Systems, Tampa, FL).
Alternatively, GST-439033 was used to treat PAP and PAPS as potential substrates. Enzyme was incubated with both 10 M unlabeled PAPS (Sigma) and approximately 10 5 cpm 5Ј-[ 32 P]PAP. Following incubation at 37°C in HEKM, the reaction was stopped as above and loaded onto the PartiSil 10 SAX column. In this case, the gradient used was 10 mM to 1.02 M AP over 60 min. One-milliliter fractions were collected and monitored for absorbance at ϭ 260 nm and then for 32 P radioactivity by scintillation counting. The experimental chromatogram was compared with a standard chromatogram containing commercially obtained 5Ј-adenosine phosphosulfate (APS), 5Ј-AMP, PAP, and PAPS (Sigma).
Assays of Enzyme Kinetics-Phosphatase activity was analyzed in the following manner. Unless otherwise noted, reactions were carried out in HEKM buffer containing 0.4 mg/ml BSA. For inhibitor studies, the reaction mixtures contained appropriate concentrations of LiCl, unlabeled PAPS, or unlabeled Ins(1,4)P 2 . Enzyme was added to reaction mixtures containing appropriate concentrations of trace-labeled 3 H-inositols or 5Ј-32 P-bisphosphorylated nucleotides. Reactions were performed at 37°C for specific times and were stopped with 20 volumes of appropriate formate buffer (see below). The stopped reaction mixture was applied to a 200-l bed volume of Dowex equilibrated in formate buffer. Reaction products were eluted by washing with 20 column volumes of formate buffer whereas higher phosphorylated product re-mained bound to the column. Quantitation of hydrolysis was performed by liquid scintillation counting of the eluent. Formate buffer for Ins(1,4)P 2 assays contained 0.05 M NH 4  HAL2 Disruption and Complementation-Yeast were propagated in standard rich (YPD) medium, or in complete minimal (CM) medium lacking appropriate nutrients to maintain plasmids. The entire HAL2 open reading frame was disrupted from S. cerevisiae W303 and replaced with LEU2. LEU2 was amplified from pRS305 (23) by PCR with the sense primer 5Ј-GGCATTGGAAAGAGAATTATTGGTTGCAACTCAA-GCTGTACGAAACAGATTGTACTGAGAGTGC and the antisense primer 5Ј-GGCGTTTCTTGACTGAATGACATCGCATGATGTAGACA-CCACCAACCTTACGCATCTGTGCGG (where underlined bases correspond to the HAL2 coding sequence and the remainder is common to the pRS series of plasmids). The resulting hal2::LEU2 PCR product was transformed into diploid strain W303 (MATa/MAT␣), and stable transformants were selected on defined media plates lacking leucine as described (26). Gene disruption was confirmed by PCR using a sense primer complementary to the chromosomal region 5Ј of the HAL2 locus (5Ј-GACCAGCAATATCACCTGTTG) and an antisense primer corresponding to a sequence in LEU2 (5Ј-CCATGCCACGGTTCTGCT-C). Heterozygous diploids were sporulated and dissected as described (26). The hal2::LEU2 spores segregated 2:2 (data not shown) and were identified by growth on -Leu as well as PCR analysis as above.
Complementation analysis of the methionine auxotrophy of hal2::LEU2 was performed by transforming appropriate plasmids into haploid yeast containing the disruption using a standard lithium acetate protocol (24). Cells were grown initially in CM-ura containing the non-repressing sugar raffinose (2% w/v) and induced at mid-log phase by the addition of 2% galactose. Cells were counted with a hematocytometer and inoculated into CM-ura lacking methionine at 10 4 cells/ml. Cells were grown with shaking at 30°C, and growth was determined by measuring OD 600 .

Cloning and Tissue Distribution of a Putative Lithium-sensitive Phosphomonoesterase Family
Member-The wealth of sequence information obtained through genome and EST data bases has been invaluable in identifying and extending protein families. In order to "computer clone" novel members of the structurally conserved lithium-sensitive phosphomonoesterase protein family, basic local sequence alignment tool (BLAST) (21) comparisons were initiated. A query sequence of human 1ptase (accession number L08488) was used to search human and mouse EST data bases specifically looking for conservation of the motif "D-X n -EE-X n -DP(i,l)D(g,s)T-X n -WD-X 11 -GG," which contains the signature residues important for metal/lithium binding and catalysis. A partial sequence of the murine EST clone 439033, accession number AA008240, was found to have weak similarity to 1ptase and contained part of the above motif, "D-X n -EE-X n -DPlDgT." The entire 439033 cDNA clone was sequenced and found to be 1.4 kb and had a predicted open reading frame of 308 amino acids (data not shown; deposited into Gen-Bank accession number AF125043). Importantly, the coding sequence was found to also contain the "WD-X 11 -GG" motif, indicating that it is a bona fide member of this family. In addition, several truncated human clones were initially found, and subsequent searches have identified a full-length human cDNA, clone 645079, that has a length of 2.1 kb and a predicted coding sequence of 309 amino acids (deposited into Genbank accession number AF125042). Polyadenylation signals and A n tails were found indicating the 3Ј end of both clones had been identified. The reason for the different size 3Ј untranslated sequences between the mouse and human cDNA's, 300 versus 1000 nucleotides, is uncertain. Alignment of the human and mouse primary structures showed 92% identity and 96% similarity (data not shown).
The core structure of the lithium-sensitive family encompasses 160 residues and is composed of 5 ␣-helices, 11 ␤-strands, and at 2 two metal binding sites. A stereo image of the 1ptase ␣-carbon core is shown in Fig. 1B. To characterize the sequence/structure relationship, the secondary structure elements were mapped onto the alignment as shown by the gray arrows (strand) and coils (helix) in Fig. 1A. Remarkably, the MATCHBOX alignment identified 11 out of 16 core structure regions, providing further evidence that clone 645079 is structurally and functionally related to the lithium-sensitive protein family.
In order to determine the expression pattern and size of the 645079 transcript, a multitissue Northern blot analysis was performed. A radiolabeled 0.4-kb region of the human cDNA was used to probe 2 g of human messenger RNA from a variety of tissue sources, and the results are shown in Fig. 2. A single 2.5-kb message was visible at various levels in all tissues examined, indicating that the clone 645079 was approximately full-length. As a control for amount of mRNA loaded, the same blot was reprobed with ␤-actin probe, as shown in the lower panel. Relative expression was determined by comparing the  (27). Conserved regions among these sequences were identified by the program are displayed in lowercase letters and visually enhanced with gray boxes. Secondary structure elements, determined from the high resolution structure of bovine 1ptase (1), are mapped onto the alignment as arrows (strands) and coils (helices). B, the ␣-carbon trace of the core of 1ptase is displayed as a stereo image with conserved structural elements, 5 ␣-helices, and 11 ␤-strands, labeled. ratio of 645079 and actin radioactivity (as quantified by phosphorimage analysis). The lowest level of expression was observed in lung and was assigned a value of 1.0 arbitrary units to which the other tissues were normalized. The highest level of expression was observed in kidney, which has 90.5-fold increased level relative to lung. Additionally, Northern analysis of murine kidney mRNA using a radiolabeled fragment of clone 439033 demonstrated a single band of 1.4 kb consistent with size observed from sequencing of the clone (data not shown). Of interest, the relative amount of expression in brain was low.
Substrate Specificity Determination-Computer cloning represents a unique "reverse" biochemistry problem, i.e. determining the substrate specificity for a clone of unknown function. Sequence homology with proteins of the lithium-sensitive phosphomonoesterase family led us to predict that the enzyme would be metal-dependent and hydrolyze small phosphorylated molecules, possibly including phosphorylated inositols and nucleotides. To investigate this possibility, the coding sequence from the murine clone 439033 was inserted into a pGEX vector and expressed in bacteria generating a GST-439033 fusion protein. Recombinant GST-439033 was readily purified using glutathione affinity resin and incubated in Mg 2ϩ buffer containing radiolabeled Ins(1,4)P 2 , Ins(1,3,4)P 3 , and Ins(1,4,5)P 3 . Reactants from GST control, GST-439033, or pure 1ptase were separated by strong anion HPLC, and representative radiograms of eluted compounds are shown (Fig. 3A, lower, middle, and upper traces, respectively). GST and untreated (not shown) samples show three peaks, B, D, and E, corresponding to Ins(1,4,)P 2 , Ins(1,3,4)P 3 , and Ins(1,4,5)P 3 , respectively. Treatment with 1ptase provides additional standards Ins(4)P (peak A) and Ins(3,4)P 2 (peak C), consistent with removal of the 1-position phosphate. Treatment with GST-439033 shows qualitatively similar results in that peaks A and C have appeared at the expense of disappearance of peaks B and D. In both cases, Ins(1,4,5)P 3 is not hydrolyzed, as the size of peak E is unchanged. As GST-439033 does not metabolize the Ins(4)P product to inositol nor does it hydrolyze Ins(1)P (data not shown), we conclude it does not function as an impase. Additionally, Ins(3,4)P 2 is not converted, indicating that GST-439033 is not acting as a 3-or 4-phosphatase. Furthermore, other inositol polyphosphates that were tested included Ins(1,3,4,5)P 4 , Ins(1,3,4,5,6)P 5 , and InsP 6 , none of which was dephosphorylated. Together, these data demonstrate that GST-439033 functioned qualitatively as an inositol polyphosphate 1-phosphatase.
Given the sequence similarity to HAL2, cysQ, and to a lesser extent SAL1, we tested GST-439033 using PAP and PAPS substrates. GST-439033 or GST protein was incubated in the presence of Mg 2ϩ , 5Ј-[ 32 P]PAP, and unlabeled PAPS, the reactants were separated by HPLC, and fractions were collected and monitored by absorbance at 260 nm and scintillation counting (Fig. 3B). Upon reaction, the PAP and PAPS peaks disappear, whereas quantitative peaks at the elution times of 5Ј-AMP and 5Ј-APS appear. Thus, the enzyme has phosphatase activity on PAP and PAPS. Because the substrate is labeled at the 5Ј position and no free 32 P i appears, we conclude that the 3Ј-position phosphate is hydrolyzed.
To qualitatively address whether or not the nucleotide played a role in determining substrate selectivity, several additional bisphosphorylated nucleotides were tested as potential substrates. As described under "Materials and Methods," 3Јmonophosphates of 2Ј-dAMP, CMP, GMP, TMP, and UMP were 5Ј-labeled with polynucleotide kinase and [␥-32 P]ATP. The purity of labeled nucleotides was analyzed by PartiSphere SAX HPLC to ensure that no contaminating PAP was present. Although no standards are commercially available, purity of the nucleotides was inferred because, under conditions of HPLC, their elution times differed individually and from that of PAP. Only the purity of deoxy-PAP remained ambiguous due to its co-elution with the 2Ј-oxy form. The labeled nucleotides were incubated with purified GST-439033 or GST alone and subjected to Dowex anion exchange chromatography as described under "Materials and Methods" using PAP reaction conditions. In each case, product was detected upon reaction FIG. 2. Distribution of clone 645079 mRNA in human tissues. A human multitissue Northern blot was probed for the expression of the 645079 message. The blot was probed with a radiolabeled fragment of the human gene and exposed to film for 32 h at Ϫ80°C with two intensifying screens (top). The same filter was then stripped and reprobed with a radiolabeled actin cDNA fragment and exposed to film for 2 h at room temperature (bottom). Radioactivity for each band was quantified using a PhosphorImager, and expression levels of 645079 are provided relative to actin and normalized to levels observed in lung (assigned a value of 1). RNA molecular weight standards (kb) are displayed. with GST-439033, indicating that each of the nucleotides tested is a substrate for the enzyme (data not shown).
Purification of Non-fusion Protein-In order to determine detailed biochemical parameters of mammalian nucleotide/inositol phosphatase, purification of non-fusion protein was performed. The open reading frame of the mouse clone was inserted into a baculovirus cloning vector. Recombinant virus was amplified and used to infect Sf9 cells in medium containing 10% FBS. Typical infections resulted in production of approximately 10 mg of soluble enzyme/liter of cells, representing around 4% of the total soluble protein.
Initial investigations indicated that the protein flowed through both S-Sepharose cation exchange columns and Q-Sepharose anion exchange columns at pH 7.5. Therefore, a negative purification step involving both ion exchange columns in series was employed. Subsequently, ammonium sulfate was added to 30%, and soluble proteins were applied to an HPLC phenyl-Sepharose column. The protein eluted in 18 -22% ammonium sulfate in a reverse linear gradient from 30% to 0% ammonium sulfate. Following concentration and removal of ammonium sulfate, typical yields were on the order of 20% with purification of approximately 25-fold.
In order to increase yield, an improved affinity purification method based on previously reported PAP-agarose chromatography (28) was developed. Calcium exhibits micromolar inhibition of phosphomonoesterase activity of other members of the lithium-sensitive family. Cell lysates therefore were applied to PAP-agarose equilibrated in 10 mM CaCl 2 as described under "Materials and Methods" in order to prevent hydrolysis of the immobilized substrate. Typically, a 1-ml resin bed volume was sufficient to bind approximately 6 mg of enzyme. The resin was washed with buffer containing up to 500 mM NaCl to elute nonspecifically bound proteins, and the enzyme was eluted in approximately 300 M free PAP. The single-step purification resulted in a high yield of homogeneous protein (Fig. 3). This protein was concentrated and dialyzed to remove calcium and free PAP prior to enzymatic assays. Aliquots were stored frozen at Ϫ80°C for up to 3 months with no apparent loss of activity.
In order to confirm the in vivo expression, we isolated it from freshly harvested mouse tissues. Following the success of using a PAP-agarose affinity matrix to obtain purified baculovirusexpressed 439033 protein in a single step, we employed this technique to isolate the native enzyme. The soluble fraction from centrifuged mouse kidney homogenate was applied directly to a PAP-agarose column equilibrated in 10 mM CaCl 2 . As shown by the silver stain in Fig. 4B, a single band comigrating with recombinant BPntase appears in the column elution fraction. In addition to similar PAP-agarose binding and SDS-PAGE migration, polyclonal antibodies to recombinant 439033 protein react with the 37-kDa band (Fig. 4C), giving further evidence that the native enzyme has been isolated from fresh mouse tissues. Several mouse tissues, including kidney, lung, heart, and liver, were analyzed (Fig. 4C). Western blot data corroborated Northern blot analysis (Fig. 2), indicating highest levels of expression in the kidney and little expression in lung.
Kinetics of Phosphomonoesterase Activity-Previously reported kinetic parameters for bisphosphate nucleotidase activity were determined using non-radioactive PAP and/or PAPS substrate. Due to relative insensitivity of this assay, Michaelis constants for all nucleotides were approximated. As a means to accurately determine the affinity constant and catalytic efficiency of mammalian protein, a radiolabeled nucleotide assay was developed. [ 32 P]PAP labeled in the 5Ј position was generated simply by incubating 3Ј-AMP with [␥-32 P]ATP and T4 polynucleotide kinase. In this way, nucleotide concentrations in the nanomolar range were easily assayed. Similarly, radioassays for inositol phosphatase activity were used as described previously (15).
Phosphatase activity on PAP, Ins(1,4)P 2 , and Ins(1,3,4)P 3 is linearly dependent on amount of enzyme and time within a range of 10 -50% hydrolysis (data not shown). Therefore, dose and time were adjusted within experiments to keep fraction of substrate hydrolyzed around 30%. Plots of reaction velocity versus substrate concentration for recombinant 439033 protein are displayed in Fig. 5 (A-C). The Michaelis-Menten constants were determined from Lineweaver-Burk analyses (Fig. 5, insets) and are displayed in Table I, part A. PAP is hydrolyzed to 5Ј AMP with a K m of 520 nM and a V max of 42 mol/min/mg of enzyme. Ins(1,4)P 2 is hydrolyzed to Ins(4)P with a K m of 113 M and a V max of 11 mol/min/mg. Ins(1,3,4)P 3 is hydrolyzed to Ins(3,4) P 3 with a K m of 100 M and a V max of 8 mol/min/mg. According to these parameters, PAP is clearly the favored substrate under these conditions, and the protein is thus named bisphosphate 3Ј-nucleotidase, or BPntase. The k cat /K m for the hydrolysis of PAP, 5.0 ϫ 10 7 s Ϫ1 M Ϫ1 , is approximately 1000-fold higher than the k cat /K m for the hydrolysis of Ins(1,4)P 2 and Ins(1,3,4)P 3 , which are 6.0 ϫ 10 4 s Ϫ1 M Ϫ1 and 8.1 ϫ 10 4 s Ϫ1 M Ϫ1 , respectively. Lithium-free PAPS was used to inhibit PAP phosphatase activity in order to determine an approximate association constant for the sulfated nucleotide. As expected, inhibition was competitive (data not shown), with a K i of 700 nM (Table I, part B).
Purified native protein was assayed for hydrolysis of PAP and Ins(1,4)P 2 . Native enzyme hydrolyzes PAP (K m ϭ 475 nM, V max ϭ 60 mol/min/mg) and Ins(1,4)P 2 (K m ϭ 130 M, V max ϭ 8 mol/min/mg) in a Li ϩ -sensitive manner and with kinetics similar to those determined for the recombinant protein (data not shown).
PAP hydrolysis is dependent on the presence of Mg 2ϩ . Optimum activity was obtained within a Mg 2ϩ concentration range of 2.5-3.5 mM. Similar to other enzymes in the lithium-sensitive family, activity is inhibited at high concentrations of Mg 2ϩ (Fig. 6A) (2). A Hill plot of data within a range of Mg 2ϩ concentrations of 0.01-0.2 mM gives a Hill coefficient of 1.60, indicating positive cooperativity but an undetermined number of metal binding sites (Fig. 6B). The pH optimum for PAP hydrolysis is approximately 7.5 (Fig. 6C). Similar results were obtained for the hydrolysis of Ins(1,4)P 2 and Ins(1,3,4)P 3 (data not shown).
Inhibition of Phosphatase Activity-Based on its limited sequence similarities to other Mg 2ϩ -dependent phosphomonoesterases in the 1ptase family, we predicted that BPntase would also be uncompetitively inhibited by Li ϩ . The velocity of PAP phosphatase activity versus lithium concentration at various substrate concentrations was determined and displayed as a Dixon plot in Fig. 7A. Parallel lines in the Dixon plot indicate that lithium inhibition is uncompetitive with respect to PAP. The K i for this inhibition is 157 Ϯ 17 M LiCl (Table I, part B). Similarly, Li ϩ inhibition is uncompetitive with respect to Ins(1,4)P 2 , and the calculated K i is 540 M (data not shown).
As BPntase markedly favors nucleotides, we sought to determine if the inositol substrates compete for identical sites. The data are summarized in Table II presence of Ins(1,4)P 2 displays competitive inhibition with a K i of 15 M (Fig. 7B; Table I, part B). This provides a basis that Ins(1,4)P 2 may play a regulatory role via inhibition of phosphatase activity as opposed to hydrolysis in the action of BPntase in vivo. Functional Complementation of Yeast HAL2-The similar substrate selectivity of BPntase, SAL1, and Hal2p prompted experiments to determine if BPntase could functionally complement hal2 yeast mutants. The HAL2 gene therefore was disrupted in diploid W303 S. cerevisiae by single gene replacement with the LEU2 marker gene. The diploid yeast were sporulated, tetrads were dissected, and spores were found to segregate 4:0 on rich media consistent with a nonessential function. Previous studies show that the hal2 disruption confers methionine auxotrophy (10); therefore, spores with the disruption are expected to be methionine auxotrophic and leucine prototrophic. Replica plating tetrads on Leu Ϫ and Met Ϫ plates showed a 2:2 segregation of both leucine and methionine auxotrophy (data not shown).
Complementation of the hal2 disruption was examined by overexpressing mBPntase, Hal2p, cysQ protein, and human 1ptase. The genes were expressed from a high copy plasmid (pRS426) containing a galactose-inducible promoter. Following growth in media containing methionine and 2% galactose, cultures were inoculated at 10 4 cells/ml into synthetic media lacking methionine and containing 2% galactose. Fig. 8 displays the functional complementation of the hal2 knockout. Exogenous Hal2p, mBPntase, and cysQ complement at similar levels. As expected, the vector control displays methionine auxotrophy. In addition, 1ptase, which does not hydrolyze nucleotides, does not complement the defect. This is consistent with the idea that hydrolysis of bisphosphorylated nucleotides and not inositol polyphosphates is required for growth in the absence of methionine.
Enzyme overexpression was monitored by changes in specific activities. At 1 M PAP, the cultures had the following specific activities in nanomoles of PAP hydrolyzed/min/mg of total protein: HAL2, 23 Table I, part A.

DISCUSSION
The Lithium-sensitive Phosphomonoesterase Family-Lithium's role in biology is of intense interest, yet the mechanisms by which it exerts its effects in human disease and development remain to be firmly established. The identification of a family of structurally conserved lithium-sensitive phosphomonoesterases has provided insight into this mechanism in several ways. First, biochemical and biological characterization of these proteins demonstrated that potent inhibition of enzymatic activity occurs at submillimolar concentrations, indicating that they are relevant targets. In vivo studies have validated this point. Perhaps most compelling are genetic studies of Drosophila 1ptase mutants demonstrating that deletion of the ipp phenocopies treatment of normal flies with lithium (18). Additionally, lithium induces changes in the levels of substrates and products of several family members in vivo. Second, the initial connection of lithium to inositol signaling pathways was made through impase leading to an "inositol depletion" hypothesis (30). Discovery of other lithium-sensitive signaling and metabolic proteins, such as 1ptase, fbptase, Hal2p, SAL1, and now BPntase, has greatly expanded our view of pathways involved. The notion that manic depression is a polygenic disorder is consistent with the involvement of multiple signaling pathways. Third, the conservation of the structures of this family has provided a key demonstration that they have evolved from a common ancestor. The ability to identify members such as fbptase required knowledge of the three-dimensional structure, as the overall sequence similarity is undetectable by current local alignment strategies. Importantly, the comparison of the structures has resulted in the identification of the signature motif of this family: D-X n -EE-X n -DP(i/l)D(s/g/a)T-X n -WD-X 11 -GG.
The hypothesis that proteins having this motif will also function as metal-dependent lithium inhibited phosphomonoesterases has been validated by this study. This is of significant importance especially in the upcoming "post-genomic" era of modern biology. Repeating the strategy defined in this study on emerging or completed EST and genome data bases will facilitate rapid identification of novel human lithium targets. The demonstration that BPntase was uncompetitively inhibited by Li ϩ with the lowest inhibition constant described to date is significant. Additionally, structural information was obtained from multiple sequence alignments with 1ptase and mapping regions of conservation to the core elements. Of note, although the signature motif is able to definitively identify a protein's membership in the family and lithium sensitivity, it appears unable to predict a protein's substrate selectivity. This is evident from characterization of the mammalian BPntase that appeared more related to 1ptase than to nucleotidases, yet the substrate preference favored bisphosphate nucleotides. The Role of Nucleotidase Versus Inositol Phosphatase Activities of BPntase-The ability of BPntase to hydrolyze both nucleotide and inositol polyphosphates raises the question which activity is important biologically? Previous reports characterizing nucleotidase biochemical parameters were limited in interpretation due to the relative insensitivity of measuring phosphate release from unlabeled substrate and/or spectroscopic measurements. This was overcome by developing a simple radiolabeling procedure for making 5Ј-labeled PAP, and our data show that the Michaelis affinity constant for nucleotides was in the nanomolar range, an order of magnitude below previously estimated values (13,28,29,31). Previous biochemical studies of Hal2p nucleotidase showed that it did not possess 1-phosphatase activity (12). In contrast, our data clearly demonstrate that Hal2p does have 1-phosphatase activity, with similar selectivity and catalytic efficiency to mammalian and plant counterparts (13). We demonstrate that the catalytic efficiency of BPntase, Hal2p, and CysQ toward nucleotide substrates is highly favored. Furthermore, 1ptase, which does not possess nucleotidase activity, is unable to complement hal2 mutant yeast providing further biological support favoring the nucleotidase function. Thus, we conclude that the biologically relevant activity is the 3Ј-nucleotidase.
An unexpected finding of our studies was that inositol 1,4bisphosphate potently inhibits nucleotidase activity, having an inhibitory constant in the physiological range. This provides a novel connection that, rather than functioning to hydrolyze inositol polyphosphates, BPntase is potently regulated by these potential messengers. Kidney tissue has been shown to possess ample levels of 1ptase, which has a catalytic activity 100 times higher than the 1-phosphatase activity of BPntase (16). Of interest, a previous report of Quintero et al. (13) suggested that the 1-phosphatase activity of SAL1 was important for mediating the sodium/lithium efflux response. This interpretation was based on the observation that both the inhibition of inositolspecific phospholipase C using the inhibitor 48/80, and the overexpression of SALI promote lithium efflux in cells. The authors concluded from these experiments that phospholipasemediated inositol signaling regulates cation efflux through ENA pumps and that SALI participates in this process by promoting turnover of messenger molecules. Another interpretation, consistent with our model, is that blocking phospholipase C activity reduces levels of inositol 1,4-bisphosphate, relieving inhibition of the nucleotidase activity, thereby mimicking the effects of overexpression. Further studies are needed to clarify which model is correct.
The ability of this family of nucleotidases to utilize either bisphosphate nucleosides or PAPS has further confounded determination of the physiologically relevant substrate. Muta- FIG. 8. Complementation of ⌬hal2 methionine auxotrophy. The native HAL2 gene in W303 was disrupted by replacement with a LEU2 cassette. Complementation analysis on haploid cells was performed with indicated gene products on a 2 plasmid (pRS426), selected by uracil prototrophy, and under control of a GAL promoter. Transformed yeast were inoculated at 10 4 cells/ml in synthetic media lacking uracil and methionine and containing 2% galactose. tions in either HAL2/MET22 or cysQ result in defective sulfur assimilation (10,14), providing evidence the PAPS nucleotidase activity is biologically relevant. Neuwald et al. suggested that PAPS or a derivative may be cytotoxic when allowed to accumulate (14). As support for this idea, they cite the fact that poor growth due to mutations in the PAPS-utilization pathway can be rescued by inhibiting the formation of PAPS with additional mutations in cysC, a 5Ј-APS 3Ј-kinase (14). Furthermore, Peng and Verma have shown that supplementation of media with methionine but not sulfite supports growth of hal2 mutants, indicating that the PAPS nucleotidase activity is most relevant (31).
Alternatively, due to their similar structures, PAP and PAPS may play cooperative roles in the sulfur assimilation pathway. For example, Ozeran et al. showed that a PAPS translocase transports PAPS across mitochondria membranes via an antiport mechanism with PAP as the returning ligand (32,33). BPntase may function to deplete the pool of PAP, thereby facilitating the transport of PAPS across membranes and into the vicinity of the sulfotransferase machinery. The PAPS translocase itself was found to be competitively inhibited by PAP with respect to PAPS (32,33). Therefore, accumulation of PAP via a decrease in BPntase activity could have drastic effects on sulfur assimilation pathways. In addition to a possible effect on sulfur assimilation, recent evidence points to a role of PAP nucleotidase activity in regulating RNA processing. Dichtl et al. (17) reported deletion of hal2 results in defects in Xrn1p-mediated RNA processing due to direct inhibition by PAP. This enzyme is not essential, but the redundant function is accomplished by RNase MRP, an enzyme that may itself be inhibited directly or indirectly by lithium. Therefore, Dichtl et al. propose that lithium toxicity, at least in yeast, is mediated by inhibition of RNase MRP and by concurrent inhibition of the cytosolic enzyme Xrn1p via inhibition of HAL2 and subsequent PAP accumulation (17). Under growth conditions containing high Na ϩ or Li ϩ concentrations, overexpression of the PAPmetabolizing enzymes HAL2 and SAL1 would rescue growth by an increase in enzyme activity, thus reducing accumulated PAP pools. Methionine supplementation would also rescue growth by down-regulating the production of PAP from PAPS (34). As it appears that BPntase is a true functional homologue of HAL2 and SAL1, our results augment the findings of Dichtl et al. BPntase is active on bisphosphorylated nucleotides other than PAP such as PGP and PCP. Thus, other bisphosphorylated nucleosides that may accumulate in the absence of nucleotide phosphatase activity may also inhibit Xrn1p.
The Role of BPntase in Kidney Function-It is particularly interesting that the levels of nucleotidase are highest in kidney, consistent with the notion increases in nucleotidase activity are associated with resistance to salt. Both SAL1 and HAL2 were identified on screens for proteins that, when overexpressed, conferred tolerance to cations such as sodium and lithium. Furthermore, the up-regulation of SAL1 expression in response to increased tonicity (13) suggests a physiological role for nucleotidase activity in osmoregulation.
In addition, one of the major side effects of lithium treatment is nephrogenic diabetes insipidus, characterized by polyuria and polydipsia. The root of this side effect may lie in the inappropriate inhibition of BPntase phosphatase activity by Li ϩ . Disruption of sodium extrusion machinery on either the lumenal or basal side of kidney cells could upset sodium balances. Such a disruption in sodium homeostasis could be manifested as the movement of large amounts of fluid across kidney cells, resulting in the voiding of copious amounts of liquid. Of particular clinical significance is the ability of methionine supplementation to suppress lithium toxicity in yeast, through a mechanism that down-regulates production of PAP and ultimately effects RNA processing, sulfur assimilation, and cation efflux pathways. We suggest that a similar strategy in which lithium is co-administered with methionine in patients treated for bipolar disorder may prove clinically useful in abrogating the harmful effects of nucleotidase inhibition.