The isolation and characterization of cDNA encoding the mouse bifunctional ATP sulfurylase-adenosine 5'-phosphosulfate kinase.

Biosynthesis of the activated sulfate donor, adenosine 3'-phosphate 5'-phosphosulfate, involves the sequential action of two enzyme activities: ATP sulfurylase, which catalyzes the formation of adenosine 5'-phosphosulfate (APS) from ATP and free sulfate, and APS kinase, which subsequently phosphorylates APS to produce adenosine 3'-phosphate 5'-phosphosulfate. Oligonucleotide primers were derived from a human infant brain-expressed sequence tag putatively encoding a portion of APS kinase. Using these primers, reverse transcriptase-polymerase chain reaction was performed on mRNA from neonatal normal mice resulting in amplification of a 127-bp DNA fragment. This fragment was subsequently used to screen a mouse brain lambda gt11 cDNA library, yielding a 2.2-kb clone. Primers were designed from the 5'-end of the 2.2-kb clone, and 5'-rapid amplification of cDNA ends was used to obtain the translation start site. Sequence from the overlapping clones was assembled into a 2475-bp composite sequence, which contains a single open reading frame that translates into a 624-deduced amino acid sequence. Northern blots of total RNA from neonatal mice yielded a single message species at approximately 3.3 kb. Southern blot of genomic DNA digested with several restriction enzymes suggested the gene is present as a single copy. Comparison against sequence data bases suggested the composite sequence was a fused sulfurylase-kinase product, since the deduced amino acid sequence showed extensive homology to known separate sequences of both ATP sulfurylase and APS kinase from several sources. The first 199 amino acids corresponded to APS kinase sequence, followed by 37 distinct amino acids, which did not match any known sequence, followed by 388 amino acids that are highly homologous to known ATP sulfurylase sequences. Finally, recombinant enzyme expressed in COS-1 cells exhibited both ATP sulfurylase and APS kinase activity.

Biosynthesis of the activated sulfate donor, adenosine 3-phosphate 5-phosphosulfate, involves the sequential action of two enzyme activities: ATP sulfurylase, which catalyzes the formation of adenosine 5-phosphosulfate (APS) from ATP and free sulfate, and APS kinase, which subsequently phosphorylates APS to produce adenosine 3-phosphate 5-phosphosulfate. Oligonucleotide primers were derived from a human infant brain-expressed sequence tag putatively encoding a portion of APS kinase. Using these primers, reverse transcriptase-polymerase chain reaction was performed on mRNA from neonatal normal mice resulting in amplification of a 127-bp DNA fragment. This fragment was subsequently used to screen a mouse brain gt11 cDNA library, yielding a 2.2-kb clone. Primers were designed from the 5-end of the 2.2-kb clone, and 5-rapid amplification of cDNA ends was used to obtain the translation start site. Sequence from the overlapping clones was assembled into a 2475-bp composite sequence, which contains a single open reading frame that translates into a 624-deduced amino acid sequence. Northern blots of total RNA from neonatal mice yielded a single message species at approximately 3.3 kb. Southern blot of genomic DNA digested with several restriction enzymes suggested the gene is present as a single copy. Comparison against sequence data bases suggested the composite sequence was a fused sulfurylase-kinase product, since the deduced amino acid sequence showed extensive homology to known separate sequences of both ATP sulfurylase and APS kinase from several sources. The first 199 amino acids corresponded to APS kinase sequence, followed by 37 distinct amino acids, which did not match any known sequence, followed by 388 amino acids that are highly homologous to known ATP sulfurylase sequences. Finally, recombinant enzyme expressed in COS-1 cells exhibited both ATP sulfurylase and APS kinase activity.
Sulfate activation involves the transfer of a sulfate group to ATP by ATP sulfurylase (ATP sulfate adenylyltransferase, EC 2.7.7.4) to yield adenosine 5Ј-phosphosulfate (APS) 1 and pyro-phosphate. Subsequently, APS kinase (ATP adenosine-5Ј-phosphosulfate 3Ј-phosphotransferase, EC 2.7.1.25) transfers a phosphate group from ATP to APS to yield ADP and adenosine 3Ј-phosphate 5Ј-phosphosulfate (PAPS). Since the equilibrium for the ATP sulfurylase reaction is rather unfavorable (K eq ϭ 10 Ϫ8 ) in the physiologic direction, APS kinase plays an important role by continually removing APS, thus driving the overall sulfate activation pathway in the forward direction. Moreover, APS itself is structurally unstable and is subject to spontaneous degradation under physiologic conditions. PAPS is the sole source of sulfate for sulfate esters in mammals, and APS appears to be only an intermediate in the sulfate-activating pathway.
We have had a long-standing interest in the sulfate-activation pathway since characterizing the unique defect in the brachymorphic mouse, which affects both ATP sulfurylase and APS kinase (18,19). Attempts to purify these two enzyme activities from rat chondrosarcoma showed that they copurifed over 2000-fold, suggesting that the activities are inseparable (20). Recent studies have demonstrated that rat chondrosarcoma ATP sulfurylase and APS kinase reside on a single bifunctional enzyme (21), which uses a channeling mechanism to efficiently synthesize PAPS (22). Most recently, we demonstrated that the defect in the brachymorphic mouse results from a mutation that primarily alters the function of the novel coupling mechanism between the two active sites (23). Therefore, we have directed our efforts toward cloning and sequencing this unique enzyme to further understand the channeling function and eventually elucidate the brachymorphic defect. Here, we report the isolation of a 2475-bp cDNA from Mus musculus, which yields a 624-deduced amino acid sequence encoding both ATP sulfurylase and APS kinase activities.

EXPERIMENTAL PROCEDURES
Materials-Common reagents were high grade commercial products. Restriction endonucleases, with the exception of SacI, were purchased from New England Biochemicals. SacI, T4 DNA ligase, and Taq polymerase were obtained from Promega. The T7 DNA polymerase sequencing kit was purchased from U. S. Biochemical Corp. Prime-it II random hexamer labeling kit and pBluescript KSϩ were obtained from Stratagene. Radionucleotides were purchased from DuPont NEN. Oligonucleotides were synthesized on an Applied Biosystems 380B DNA synthesizer. An 18-day-old mouse brain gtll cDNA library was from ATCC. RNA or DNA was routinely obtained from neonatal (day 0 -2) C57B1 mouse brain.
DNA Isolation and Analysis-The techniques used for screening cDNA libraries, isolation of plasmid DNA, manipulation of DNA fragments, and transformation of E. coli were as described by Sambrook et al. (24). Mouse genomic DNA was isolated and analyzed by typical DNA blot analysis (24). Approximately 20 g of genomic DNA from neonatal mice was digested overnight with BamHI, EcoRI, HindIII, KpnI, PstI, SacI, and XhoI. DNA fragments were resolved on an agarose gel and blotted onto a nylon membrane via capillary action. The blot was prehybridized in 5 ϫ SSPE, 5 ϫ Denhardt's reagent, 50% formamide, and 200 mg/ml salmon sperm DNA at 42°C for 2 h and hybridized with the 32 P-labeled 127-bp or the 2.2-kb clone for 12-16 h at 42°C followed by three 1-h washes in 0.1% SDS and 0.5 ϫ SSC at 52°C. DNA sequencing was by the dideoxynucleotide chain termination method (24,25).
5Ј-Rapid Amplification of cDNA Ends-5Ј-rapid amplification of cDNA ends was performed according to the protocol provided by Life Technologies, Inc. Briefly, two antisense primers from the 5Ј-end region of clone SK9 -1 were synthesized. The primers were SK2 (5Ј-TCAG-GACTGAAGCCGAGGTTC-3Ј) and SK23 (5Ј-TGAGTCCTTGGCGGAT-GTTG-3Ј). First strand cDNA was synthesized from 1 g of neonatal mouse brain total RNA, using the SK2 primer, and RNA was removed with RNase H and cDNA purified by GlassMAX Spin cartridge. The anchor sequence was added to the 3Ј-end of the first cDNA strand using terminal deoxynucleotidyl transferase and dCTP, and PCR amplification was performed using a deoxyinosine-containing anchor primer and nested primer SK23. The 300-bp PCR product was subcloned and sequenced.
RNA Isolation and Analysis-Whole brains were extracted from freshly sacrificed neonatal mice and immediately frozen on dry ice, and total RNA was isolated as described previously (26,27). 30 g of total RNA was electrophoresed through an agarose-formaldehyde gel, which was subsequently stained with ethidium bromide, photographed, and then capillary blotted onto a Nytran filter. The filter was prehybridized at 42°C for 2 h in 5 ϫ SSPE, 5 ϫ Denhardt's reagent, 50% formamide, and 200 mg/ml salmon sperm DNA. Hybridization with the 32 P-labeled 127-bp and 2.2-kb clone at 42°C for 16 h was followed by three 1-h washes at 55°C. The filter was exposed for 5 days at Ϫ70°C with an intensifying screen.
Amplification of RNA Sequences (RT-PCR)-RT-PCR was performed as described previously (26) except as noted. To obtain the initial mouse cDNA, the PCR cycling conditions were 1 min at 94°C, 1 min at 72°C, and 1 min at 55°C. Primers were US (5Ј-AGAGAGAGGTACCGCTA-CACTCTGGATGGTG-3Ј) and DS (5Ј-AGAGAGAGAGCTCGCATCTG-CAAACAGTTTAGC-3Ј). The first 13 bases of each primer were designed to introduce KpnI (US) and SacI (DS) restriction sites and are underlined. Primers and conditions for RT-PCR of a nearly complete coding sequence are described in the next section.
Transient Expression Assays-RT-PCR was performed to amplify the sulfurylase-kinase coding sequence (starting from the second Met, residue 22) using 1 g total RNA from neonatal mouse brain. Primers used were SK35 (5Ј-GAATTCATATGCAGAGAGCAACCAACGTC-3Ј) and SK16 (5Ј-GGAAACGGCCTGATTTCAGG-3Ј). The amplified DNA fragment was subcloned with the TA cloning system (Invitrogen), and the full sequence was verified. The insert was excised with EcoRI, made blunt-ended with Klenow polymerase, and ligated into the eucaryotic cell expression vector pSVL (Pharmacia Biotech Inc.).
Transient transfections of COS-1 cells were performed by the calcium phosphate procedure as described (24). Briefly, the cells were transfected with 20 g of DNA and incubated at 37°C for 12 h in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. After 12 h, the growth medium was changed, and the cells were incubated at 37°C for an additional 48 h. The cells were then washed twice with phosphatebuffered saline and scraped with a rubber policeman into sucrosephosphate buffer (50 mM NaH 2 PO 4 -K 2 HPO 4 , pH 7.8, 0.25 M sucrose, 0.25 M KCl, 1 mM EDTA). The cells were sonicated briefly and centri-fuged at 10,000 rpm, and enzyme assays were performed on supernatant fractions.
Sequence Analysis-Sequence data base searches were performed using BLAST programs (30) on the data bases maintained by the National Center for Biotechnology Information at the National Library of Medicine. Nucleotide sequence from the mouse clones was assembled, and the deduced amino acid sequence was generated and analyzed using the computer programs SEQED, LINEUP, TRANSLATE, PEPTIDESORT, GAP, COMPARE, DOTPLOT, and PILEUP of the Wisconsin Package. 2

RESULTS
Complete Sulfurylase-Kinase cDNA Sequence-A portion of a putative APS kinase gene sequenced from a directionally cloned human infant brain cDNA library (31) was used as a source sequence for two primers US and DS, which included KpnI and SacI sites, respectively. These primers were used for reverse transcription of 2 g of mRNA from neonatal C57B1 mice, and the isolated 127-base pair fragment was subcloned into pBluescript KSϩ and sequenced. When compared to sequence data bases by BLAST searches, the 127-base pair sequence was seen to be 85% identical to the human cDNA sequence from which the RT-PCR primers were derived and to predict a peptide sequence having 58 -76% identity to various APS kinase peptide sequences. The insert was subsequently excised using KpnI and SacI, run on a 1% low melting agarose gel, purified, and radiolabeled with 32 P via the random hexamer labeling technique.
This probe was then used to screen approximately 500,000 recombinants of an 18-day-old mouse brain gt11 cDNA library. Recombinants were replica-plated onto nitrocellulose filters and probed with the 127-bp fragment. A single 2.2-kb clone (SK 9-1) was identified and subsequently plaque-purified, subcloned and sequenced, and was found to lack a probable initiation codon. Therefore, two new primers were designed from the SK 9-1 5Ј-end, and 5Ј-rapid amplification of cDNA ends was used to obtain a 300-bp cDNA, which was subcloned and sequenced and found to contain the translation initiation codon. Multiple sequence determinations were made for all primers used, and uncertainties in the gel patterns were resolved by sequencing with a different polymerase, sequencing the complementary strand, or both.
Sequence from the overlapping clones was assembled into a 2475-bp composite sequence, which contains a single open reading frame that translated into a 624-deduced amino acid sequence having a calculated molecular mass of 70.7 kDa, the longest obtainable from our data (Fig. 1). This composite contains a region that correlates closely to an ATP-GTP binding motif (P-loop) (32, 33) from amino acid residue 59 to 65, flanked by cysteine residues at position 53 and 78 (and 83), the FISP sequence at residues 131-134 (3), and a PAPS-dependent enzyme motif from residue 175 to 186 (34,35). Toward the C-terminal end, a sequence commensurate with the recently described PP-motif found in ATP sulfurylases and PAPS reductases (36) is also present (residues 521-525).
Sequence Comparison among Species-When compared against the non-redundant combined set of SWISS-PROT, PDB, PIR, and GenPept data bases, the deduced amino acid sequence shows extensive homology to known separate sequences of both ATP sulfurylases and APS kinases from several sources. Pairwise GAP alignments of the kinase (residues 1-220) or sulfurylase (220 -624) portions of the mouse amino acid sequence to individual APS kinase or ATP sulfurylase sequences from other species revealed homologies to both enzyme types. The APS kinases (listed in descending order of their similarity to mouse) include the putative human APS kinase (partial sequence, 97% identity and similarity), S. cerevisiae MET14 (55% identity, 71% similarity), E. coli CysC (51%, 68%), A. thaliana APS kinase (47%, 67%), R. meliloti NodQ (44%, 64%), and Azospirillum brasilense NodQ (45%, 62%). Overall, the kinase peptide sequences appear well conserved from bacteria to the mouse. The ATP sulfurylases exhibited a wider range of relatedness to the mouse sequence, i.e. from strong similarities in the plant proteins from A. thaliana (57% identity, 75% similarity) and Solanum tuberosum (58%/ 74%), through the fungal/yeast enzymes P. chrysogenum Aps (26%, 51%) and S. cerevisiae MET3 (24%, 46%), to the low correspondence observed against the bacterial peptide sequences of E. coli CysD (20%, 46%), E. coli CysN (16%, 44%), R. meliloti NodP (16%, 45%), and A. brasilense NodP (19%, 44%). 3,4 The Wisconsin Package program COMPARE was used for pairwise comparisons between the putative sulfurylase-kinase sequence and each of the known individual sulfurylase and kinase sequences from several sources, both to localize the kinase and sulfurylase domains and to check for the presence of repeating elements or internal rearrangements. The results were displayed with the program DOTPLOT, and outcomes of pairwise comparisons with Arabidopsis and Saccharomyces sequences are shown in Fig. 2. Each plotted point represents a register of alignment and window location at which 15 of 30 residues in the window matched; identical or very similar colinear sequences result in a single diagonal line with a slope of 1. Repeats occurring in both sequences produce pairs of shorter diagonals paralleling the main register line, and a gap in one sequence of a pair is seen as a break or displacement of the main line. The APS-kinase comparisons (Fig. 2, A and C) reveal the greatest amount of colinear similarity across species, as represented by the dominant diagonal line. The ATP sulfurylase comparisons show more variability among species, with greater similarity to Arabidopsis than to Saccharomyces (Fig.  2, B and D).
A multiple alignment of the mouse sequence and representative APS kinases and ATP sulfurylases was done to examine the detail of the molecules' similarities (Fig. 3). Based on those alignments, we postulate that the first 199 amino acids (nucleotides 34 -630) correspond to APS kinase activity. This region 3 A. brasilense NodP and NodQ sequences were translations from the GenBank entry AZSNODPQ (accession no. M94886), R. meliloti NodP was from RHMNODPQA (M68858), and E. coli CysD and CysN were from ECOCYSDNC (M74586). The remaining amino acid sequences were obtained as described in the legend for Fig. 2. 4 While performing the sequence analyses for this manuscript, we became aware of a recent entry in GenBank, UUNPASY (accession no. L39001), an as yet unpublished combined ATP-sulfurylase-APS-kinase coding sequence from the worm Urechis caupo. Comparison via GAP reveals high overall relatedness (70% identity, 83% similarity) to the mouse amino acid sequence.
FIG. 1. cDNA and deduced amino acid sequences of mouse brain ATP sulfurylase APS kinase. The composite cDNA sequence of 2475 nucleotides is shown with the deduced amino acid sequence of 624 residues beginning at nucleotide 34. The first 1-199 residues corresponding to know APS kinase, and residues 237-624 corresponding to known APS sulfurylase sequences are shaded. The P-loop motif (residues 59 -65) and the PAPS-dependent enzyme motif (residues 175-186) are boxed, as are two highly conserved sequences (residues 411-431 and 499 -523) found in other ATP sulfurylases. Invariant Cys flanking the P-loop (residues 53, 78, and 83) and several invariant Lys, His, and Arg found in other ATP sulfurylases are circled.
is followed by a 37-amino acid stretch (nucleotides 631-742), which is not similar to either ATP sulfurylase or APS kinase. Subsequent amino acids from 237 to 624 (nucleotides 743-1905) are highly homologous to sequences with known ATP sulfurylase activity. As mentioned, the region that correlates closely to an ATP-GTP binding motif (P-loop) and the FISP sequence are identified only in the putative APS-kinase sequence and are also flanked by several highly conserved cysteine residues, as previously reported in Arabidopsis (35). There are also two invariant lysines and seven invariant arginines, often as part of homologous sequences, across the APS kinases that have been cloned and sequenced. As well, conserved regions containing invariant Arg and His residues analogous to those found in previously cloned ATP sulfurylases and proposed to be involved in MgATP and SO 4 Ϫ2 binding (14) are present in the APS sulfurylase portion of the new mouse sequence at residues 411-431 with invariant Arg-421, His-425, and His-428 and 499 -523 with invariant His-506, Arg-510, and Arg-522. These specific features, as well as overall homology to known ATP sulfurylase and APS kinase sequences from multiple species, suggest that this new deduced sequence represents a fused sulfurylase-kinase product.
Northern Blotting-The cDNA insert from the gtll clones was labeled with 32 P and used as a probe in Northern blotting (Fig. 4). Both the 127-bp fragment (data not shown) and the 2.2-kb clone recognize a single message species of approximately 3.3 kb in neonatal mouse brain (Fig. 4, lane 1). The size of the transcript indicates that the composite cDNA clone lacks approximately 0.8 kb of 5Ј-or 3Ј-untranslated region. The presence of intact RNA was confirmed by hybridization with a 32 P-labeled glyceraldehyde-3-phosphate dehydrogenase probe (Fig. 4, lane 2).
Southern Blotting-To assess the sulfurylase-kinase gene copy number, Southern blot analysis was performed. Genomic DNA digested with several restriction enzymes and probed with the 127-bp fragment reveals a simple pattern having only a single band in each of several restriction digests (Fig. 5). These results suggest there is a single functional gene for this sulfurylase-kinase and therefore provides no evidence for multiple genes or pseudo genes.
Heterologous Expression of Sulfurylase-Kinase-A DNA construct in the pSVL vector (see "Experimental Procedures") that contains a version of sulfurylase-kinase sequence was expressed in COS-1 cells. This construct lacks the first 21 amino acids of the predicted protein, initiating the coding sequence at the second methionine. Cell lysates expressing this construct were found to contain both ATP sulfurylase and ATP kinase activities (Fig. 6), when assayed for individual sulfurylase (28), kinase (29), or the overall reaction (22). Extracts from COS-1 cells that were not transfected, transfected with expressing vector only, or transfected with vector containing the construct in the opposite orientation exhibited base-line activity (Fig. 6). In contrast, cells transfected with the construct in the correct   4. Northern blot analysis of ATP sulfurylase APS kinase RNA. 30 g of total RNA from neonatal mouse brain was loaded on a 1% agarose gel. The blot was hybridized with a 32 P-labeled 2.2-kb probe to yield a single message species at approximately 3.3 kb (lane 1). As a control, the same blot was stripped and hybridized with a mouse glyceraldehyde-3-phosphate dehydrogenase cDNA probe (lane 2). The location of RNA size markers (in kb) is shown on the left. Autoradiograph was exposed for 5 days; prolonged exposure did not reveal any additional message species. reading frame exhibited a 3-6-fold increase in sulfurylase, kinase, and overall specific activities. These data indicate that the recombinant enzyme has the same specificity as the native bifunctional enzyme (22). Furthermore, the crude cell extracts exhibited specific activities comparable to the native enzyme (20,28,29); thus, it would appear that the single recombinant protein was properly folded and processed in the COS-1 expression system. DISCUSSION ATP sulfurylase and APS kinase are essential enzymes in sulfate activation. In lower organisms, these enzymes appear to be relatively small, contained on separate proteins, and supply PAPS mainly for cysteine biosynthesis. Three gene products are required for sulfate activation in E. coli (11): CysD and CysN encode ATP sulfurylase and CysC encodes APS kinase. ATP sulfurylase is found as a heterodimer of 27-and 62-kDa subunits (4), in which the 27-kDa peptide is the catalytic subunit (CysD), while CysN forms the 62-kDa GTP binding/hydrolyzing subunit (11). Leyh and Suo (11) have studied the gene organization of the sulfate activation locus in E. coli and noted that the translational termination and initiation sequences of the CysD-CysN and CysN-CysC gene pairs overlap. The gene order is therefore ATP sulfurylase followed by APS kinase. Translational coupling in which the termination and initiation sequences of the CysD-CysN and CysN-CysC gene pairs overlap has been implicated in maintaining stoichiometry as well as enhancing translation efficiency (37,38).
Interestingly, Foster et al. (14) have demonstrated via cloning and sequencing that the C-terminal domain of P. chrysogenum ATP sulfurylase is similar to the nucleotide sequence of APS kinases from several organisms. Although this portion of the ATP sulfurylase does not confer APS kinase activity on the enzyme, Foster et al. (14) postulate that a large region of the ATP sulfurylase may have evolved from APS kinase and that, therefore, this region may contain the allosteric PAPS binding site. Second, Schwedock et al. (17) have recently shown that the nodulation gene nodP encodes an ATP sulfurylase, while the product of the nodQ gene has APS kinase activity in addition to its role in ATP sulfurylase in R. meliloti. The findings of fused domains for ATP sulfurylase and APS kinase in these two systems are both evolutionarily and mechanistically relevant to our findings of a mammalian bifunctional sulfurylase-kinase enzyme (21), which also uses a channeling mechanism to efficiently synthesize PAPS (22).
We have now isolated overlapping cDNA fragments whose composite sequence is postulated to encode a fused sulfurylasekinase product based on the following criteria. The encoded protein derives from a single reading frame and exhibits a molecular weight commensurate with that previously obtained for the native bifunctional enzyme (21); when compared to protein sequence data bases, the deduced amino acid sequence shows extensive homology to separate sequences of both ATP sulfurylases and APS kinases from several sources; the expressed recombinant protein catalyzed the synthesis of both APS (ATP sulfurylase activity) and PAPS (APS kinase activity) as well as overall activity (i.e. synthesizing PAPS from ATP and SO 4 Ϫ2 ). Furthermore, the mouse sulfurylase-kinase sequence contains highly conserved residues, which are found in all ATP sulfurylases that have been sequenced, and are postulated to be involved in MgATP and SO 4 Ϫ2 binding (14). There are also invariant Lys and Arg residues, often a part of a homologous sequence, across all six APS kinases that have been sequenced, including now the mouse sulfurylase-kinase. In addition, the kinase domain exhibits a region that correlates closely to a ATP-GTP binding motif (P-loop), as might be expected for a phosphate binding protein (33). The sequence (GXXGXGK(TT)) is identical to the pattern in thymidine kinase (39) and to the octapeptide signature (GESGAGKT) in myosin heavy chain (40). A PAPS-dependent enzyme motif (KAXAGXXXXFTG) (34,35) is also present in the N-terminal APS-kinase portion. A sequence resembling a portion of the recently proposed ATP pyrophosphatase PP motif (36), found in several ATP sulfurylases as well as PAPS reductase, was also found in the mouse brain sulfurylase-kinase sequence.
Although all these data strongly suggest that the fused mammalian sulfurylase-kinase is related to similar activities previously isolated and cloned from lower organisms, the gene organization found in E. coli and P. chrysogenum differs significantly from the organization of the product we have isolated. Our sequence shows strong homology in the N-terminal domain to known APS kinases and strong homology in the C-terminal domain to known ATP sulfurylases. This is the opposite orientation from the fungal ATP sulfurylase, which has 75% of APS kinase at its C-terminal sequence. In addition to resulting in interspecies structural differences, this reverse orientation (as well as the fact that most of the sulfurylases and kinases are clearly separate gene products) may contribute to significant mechanistic differences. Rat chondrosarcoma sulfurylase-kinase releases PP i followed by APS, with concomitant binding of the APS to the APS kinase activity, while P. chrysogenum sulfurylase releases APS first followed by PP i . The former specific order of product release and substrate addition may result in a more efficient pathway via substrate channeling of the APS intermediate, as we have demonstrated occurs in the rat chondrosarcoma enzyme system (22). In contrast, APS bound to the ATP sulfurylase does not serve as a substrate for the APS kinase of P. chrysogenum (39), suggesting a different mechanism pertains for the fungal ATP sulfurylase fused with a partial APS kinase sequence.
The combined ATP sulfurylase-APS kinase described in this paper may represent an evolutionary trend toward a more efficient sulfate activation pathway in higher organisms. Although procaryotes can achieve coordinate expression of consecutive reactions through linkage of different polypeptides in a single operon, eucaryotic coordinate expression is more often realized by linking multiple functions in a single polypeptide. Bazan et al. (41) have proposed that certain bifunctional enzymes such as 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase represent the gene fusion of catalytic units. The newly cloned mouse brain bifunctional sulfurylase-kinase may have followed such a combinatorial process to maintain stoichiometry of the two activities and to achieve the maximal functional efficiency necessary for optimal PAPS production.
The cloning of sulfurylase-kinase will allow its role in the PAPS activation pathway to be studied further in higher organisms. Homology of the single open reading frame-generated sequence to known ATP sulfurylases and ATP kinases and demonstrable activity of the recombinant protein verify our earlier work that the sulfurylase-kinase in mammals is represented by a single bifunctional enzyme (21). Further expression analysis of mutated forms will allow clarification of mechanistic and evolutionary structure-function relationships for this essential enzyme, as well as provide a library of phenotypic mutants for future pathophysiological studies and eventually a basis for elucidation of the molecular defect in the brachymorphic mouse.