Discovery of molecular and catalytic diversity among human diphosphoinositol-polyphosphate phosphohydrolases. An expanding Nudt family.

The turnover of the "high energy" diphosphoinositol polyphosphates by Ca(2+)- and cyclic nucleotide-modulated enzymes is considered a regulatory, molecular switching activity. Target processes may include intracellular trafficking. Following our earlier identification of a prototype human diphosphoinositol-polyphosphate phosphohydrolase (hDIPP1), we now describe new 21-kDa human isoforms, hDIPP2alpha and hDIPP2beta, distinguished from each other solely by hDIPP2beta possessing one additional amino acid (Gln(86)). Candidate DIPP2alpha and DIPP2beta homologues in rat and mouse were also identified. The rank order for catalytic activity is hDIPP1 > hDIPP2alpha > hDIPP2beta. Differential expression of hDIPP isoforms may provide flexibility in response times of the molecular switches. The 76% identity between hDIPP1 and the hDIPP2s includes conservation of an emerging signature sequence, namely, a Nudt (MutT) motif with a GX(2)GX(6)G carboxy extension. Northern and Western analyses indicate expression of hDIPP2s is broad but atypically controlled; these proteins are translated from multiple mRNAs that differ in the length of the 3'-untranslated region because of utilization of an array of alternative (canonical and noncanonical) polyadenylation signals. Thus, cells can recruit sophisticated molecular processes to regulate diphosphoinositol polyphosphate turnover.

The turnover of the "high energy" diphosphoinositol polyphosphates by Ca 2؉ -and cyclic nucleotide-modulated enzymes is considered a regulatory, molecular switching activity. Target processes may include intracellular trafficking. Following our earlier identification of a prototype human diphosphoinositol-polyphosphate phosphohydrolase (hDIPP1), we now describe new 21-kDa human isoforms, hDIPP2␣ and hDIPP2␤, distinguished from each other solely by hDIPP2␤ possessing one additional amino acid (Gln 86 ). Candidate DIPP2␣ and DIPP2␤ homologues in rat and mouse were also identified. The rank order for catalytic activity is hDIPP1 > hDIPP2␣ > hDIPP2␤. Differential expression of hDIPP isoforms may provide flexibility in response times of the molecular switches. The 76% identity between hDIPP1 and the hDIPP2s includes conservation of an emerging signature sequence, namely, a Nudt (MutT) motif with a GX 2 GX 6 G carboxy extension. Northern and Western analyses indicate expression of hDIPP2s is broad but atypically controlled; these proteins are translated from multiple mRNAs that differ in the length of the 3-untranslated region because of utilization of an array of alternative (canonical and noncanonical) polyadenylation signals. Thus, cells can recruit sophisticated molecular processes to regulate diphosphoinositol polyphosphate turnover.
The considerable energy burden that is required to sustain rapid, ongoing cellular turnover of the "high energy" diphosphoinositol polyphosphates (PP-InsP 4 , 1 PP-InsP 5 , and [PP] 2 - Fig. 1) has been rationalized as representing a molecular switching activity with important regulatory consequences (1,2). The physiological importance of this process is underscored by its being under the control of signaling cascades. For example, cAMP and cGMP modulate the interconversion of PP-InsP 5 and [PP] 2 -InsP 4 in a receptor-dependent but unusual manner that bypasses the participation of both protein kinases A and G (3). In addition, turnover of PP-InsP 4 and PP-InsP 5 is regulated by a specific mode of intracellular Ca 2ϩ mobilization (4).
Molecular switching by diphosphoinositol polyphosphates may contribute to regulating intracellular trafficking (1,2). For example, PP-InsP 5 represents the most potent known inhibitor of AP-3-dependent assembly of clathrin cages, a key step in the endocytic retrieval of discharged synaptosomal vesicles (5). Dephosphorylation of PP-InsP 5 to InsP 6 may switch AP-3 to a more active state (5). There are other proteins that participate in intracellular trafficking that also bind PP-InsP 5 very tightly, such as coatomer (6,7) and AP-2 (see Ref. 8). The high affinity with which PP-InsP 5 binds to myelin proteolipid protein may be important for its vesicular delivery to the myelin sheath (9). We (10) have also suggested, based on the work of others (11), that diphosphoinositol polyphosphates may participate in the regulation of mRNA export from the nucleus.
These findings illustrate why it is of great interest to characterize the kinases and phosphatases that regulate the turnover of the diphosphoinositol polyphosphates. Recently, three InsP 6 kinases, two mammalian and one from yeast, have all been cloned (12,13). We (14) have previously cloned a human diphosphoinositol-polyphosphate phosphohydrolase (hDIPP) that dephosphorylates PP-InsP 5 and [PP] 2 -InsP 4 . The active site of this enzyme contains a Nudt (MutT) motif (14). This motif is also present at the active site of a family of enzymes that hydrolyze nucleoside phosphates (15), the metabolism of which may, therefore, sometimes compete with the diphosphoinositol polyphosphates. Such competition is apparent in hDIPP itself, which in addition to hydrolyzing PP-InsP 5 and [PP] 2 -InsP 4 , also attacks diadenosine polyphosphates such as Ap 6 A and Ap 5 A (16). The latter nucleoside phosphates are downstream effectors of a signaling cascade that regulates cardiac K ATP channels (17) and ryanodine receptors (18). Nudt motifs have also been identified in a transcriptional regulator (19) and in a protein regulating mRNA decay (20).
We have now discovered additional members of the DIPP subfamily of Nudt proteins. The properties of these new enzymes are described below. We also propose that the unique catalytic profile of each of the individual DIPP isoforms can provide flexibility in the response times for molecular switches.

EXPERIMENTAL PROCEDURES
Expression of Recombinant hDIPP1 and hDIPP2 Isoforms-hDIPP1 was obtained as described previously (14). PCR products encoding the appropriate open reading frame for hDIPP2␣ and hDIPP2␤, minus both * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EBI Data Bank with accession number(s) AF191649 -AF191655.
PCR Amplification of hDIPP2 mRNAs-Primary PCR reactions (25 l) contained 250 pg of Marathon-Ready human fetal kidney or adult heart cDNA (CLONTECH) and the 5Ј-primer GAAGTTCAAGC-CCAACC, which is homologous to nt 179 -195 of clone A (see Fig. 2). Antisense primers from increasingly distal portions of clone C (see Fig.  2) were used as the 3Ј-primer in each reaction. Generally, primary PCR reactions did not yield visible products by gel analysis. Therefore, 1-l aliquots of these reactions were used to prime secondary PCR reactions containing a pair of nested primers. The nested 5Ј-oligonucleotide used to amplify PCR␣ and PCR␤ (see Fig. 2) is homologous to nt 183-203 of clone A. The primary and nested 3Ј-primers in that reaction are complimentary to nt 4617-4634 and 4567-4587 of clone C, respectively. Individual products from secondary PCR reactions were gel purified, cloned into pCR-XL-Topo (Invitrogen), and analyzed by restriction enzyme digestion, PCR, and/or DNA sequencing. All sequencing was performed using AmpliTaq FS in a PCR cycling protocol with either dRhodamine or BigDye chemistry (Perkin-Elmer), and reactions were analyzed on an ABI 373 or 377 gel apparatus. The Wisconsin Package of programs (GCG Inc.) was used for all DNA sequence analysis.
3Ј-RACE-PCR reactions were performed as described above using human heart Marathon-Ready cDNA (CLONTECH) with primary and nested sense primers homologous to nt 3724 -3751 and 3750 -3777 of clone C, respectively. Primary and nested antisense primers were adaptor primers 1 and 2 (CLONTECH), respectively. "Touchdown" PCR cycling conditions were employed to improve specificity of amplification.
Individual PCR products were gel-purified and cloned as described above, and positive clones were identified by colony hybridization using a DIG-labeled probe spanning nt 3661-3829 of clone C. Positive clones were sequenced as described above.
Northern Blotting-Human multiple tissue Northern blots were purchased from CLONTECH. DIG reagents were from Roche Molecular Biochemicals. DIG-labeled DNA probes (see Fig. 2) were generated with a PCR labeling kit using specific primer pairs. Blots were hybridized overnight in DIG Easy Hyb at 50°C with ϳ50 ng of DIG-labeled DNA/ml of hybridization solution. Blots were washed up to 0.1 ϫ SSC at 50°C (Fig. 4A) or 0.5 ϫ SSC at 65°C (Fig. 4B) and analyzed with alkaline phosphatase-conjugated anti-DIG antibodies followed by chemiluminescent detection with CDP-STAR. Filters were stripped in boiling 1.0% SDS, rinsed, and rehybridized with a ␤-actin probe.
Materials-Diadenosine polyphosphates and diphosphoinositol polyphosphates were obtained as described previously (10,14). The EST cDNAs were obtained from the American Type Culture Collection, Manassas, Virginia. The clone C cDNA (see Fig. 2) was provided by Dr. T. Nagase, Kazusa DNA Research Institute, Chiba, Japan. PepTool v1.1 was purchased from BioTools Inc., Edmonton, Canada. tide sequence was used as a query to search the genome data bases for additional proteins that might also hydrolyze diphosphoinositol polyphosphates. A human EST cDNA, clone A ( Fig.  2), was identified and sequenced. This 1191-nt cDNA is 56% identical to hDIPP1 at the nucleotide level. Clone A contains a 543-nt open reading frame defined at the 5Ј-end by two tandem initiator codons in a context (TCTATGATGA) that fashions a relatively weak Kozak consensus. There is an inframe TGA stop codon 153 nt upsteam of the more 5Ј of the two candidate start codons (Fig. 2), indicating that the coding region of clone A is complete. The 5Ј-UTR is exceptionally G/C-rich (84%) and contains a perfect run of six CAG repeats, although this number is considered to be just below the threshold length that would make it a candidate for polymorphic expansion (21). The 5Ј-end of the nucleotide sequence of clone A terminates in a NotI site. Because the cDNA library from which clone A was isolated had been digested with NotI during construction, it is possible this cDNA may be truncated at the 5Ј-end. The 3Ј-UTR of clone A terminates with an 18-nt poly(A) tail, which begins 16 and 34 nt downstream of two ATTAAA polyadenylation signals. These motifs are slightly less efficient as polyadenylation signals compared with the canonical AATAAA (22).

Characterization
We also identified a 1334-nt EST cDNA, clone B, that substantially overlaps clone A (Fig. 2). The overlapping regions of these clones are identical, but clone B lacks most of the 5Ј-UTR that is present in clone A. However, clone B has an additional 292 nt of 3Ј-UTR, terminating in a poly(A) tail that begins 19 nt downstream of a rare, noncanonical polyadenylation signal (AGTAAA). The latter is about 30% as effective a polyadenylation signal compared with AATAAA, at least in vitro (22).
Clones A and B encode a putative 180-residue protein (20,306 Daltons) that we have named hDIPP2␣. This protein is 76% identical to hDIPP1 at the amino acid level (Fig. 3). Most of the differences between hDIPP2␣ and hDIPP1 are because of divergence in the 45-amino acid residues at the C terminus of hDIPP2␣. All but 3 of the 23 residues of the hDIPP1 Nudt motif are conserved in hDIPP2␣ (Fig. 3). Several catalytically essential residues in hDIPP1 that were recently identified in a mutagenic study (10) were also all conserved in hDIPP2␣, namely (i) the glycine triad at the N terminus of the Nudt catalytic motif, (ii) Glu 66 and Glu 70 within the Nudt motif, (iii) Gly residues within a GX 2 GX 6 G consensus (underlined in Fig.  3), and (iv) Phe 84 and His 91 , which are important primarily for [PP] 2 -InsP 4 hydrolysis. Finally, much of the predicted secondary structure of hDIPP2␣ and hDIPP1 is also conserved (Fig.  3).
Molecular Cloning and Characterization of Multiple hDIPP2 mRNAs-We hybridized a human multiple tissue Northern blot with a DNA probe (P1, see Fig. 2) that is specific for the coding region of hDIPP2␣ (and also for the hDIPP2␤ isoform described below). Signals of 1.5 and 1.8 kb were detected in heart, and, to a lesser extent, in skeletal muscle, kidney, placenta, and pancreas (Fig. 4A). Similar results (not shown) were obtained when we used a probe (P2, see Fig. 2) against the FIG. 2. Molecular clones for hDIPP2. The schematic describes several human cDNA clones for hDIPP2. All clones were fully sequenced in this study except clone C, which was characterized by others (23). The inset provides the color codes that highlight certain molecular features of the cDNAs. Clones A (GenBank TM accession number AF191649), B (AF191650), D (AF191654), E (AF191655), and F (AF191653) are all EST cDNAs (with original accession numbers of N31581, AA325102, AA085698, AA745892 and N58868, respectively). PCR products ␣ (AF191651) and ␤ (AF191652) were obtained from a human heart cDNA library as described under "Experimental Procedures." The 3Ј-RACE products were derived as described under "Experimental Procedures" and differ only in the length of their poly(A) tails. P1-4 describe the various probes that were used for Northern analysis. Nucleotide numbering begins with the 5Ј-end of clone A as a reference point.
common 3Ј-UTR of clones A and B.
Hybridization with P1 or P2 also revealed an intense mRNA signal of 4.2 kb in heart, skeletal muscle, kidney and pancreas, with weaker expression in brain, placenta, lung and liver (Fig.  4A). To pursue the nature of this 4.2-kb mRNA, we noted that the 3Ј-terminal 270 bp of clone B overlaps the 5Ј-end of a 6.4-kb human brain cDNA (clone C, Fig. 2) (23). Therefore, we hybridized a second human multiple tissue Northern blot with a probe (P3; Fig. 2) homologous to nt 1494 -1792 of clone C, i.e. down-stream of the 3Ј termini of clones A and B. The P3 probe detected a 4.2-kb mRNA with the same pattern of expression as was seen with probes P1 and P2, but the 1.5-and 1.8-kb DIPP2 mRNAs were not detected (Fig. 4B). A more distal clone C probe (P4; Fig. 2) gave similar results (data not shown). These data suggest that the 4.2-kb message is a hDIPP2 mRNA comprised of clones A or B fused to ϳ3 kb of the 5Ј-end of clone C. We tested this prediction by PCR amplification experiments using several 5Ј-primers from the proximal end of clone B and various, progressively more distal, 3Ј-primers complimentary to segments of clone C (see "Experimental Procedures"). Human fetal kidney and adult heart cDNA were used as templates. All of the primer pairs that were tested generated single PCR fragments of the expected size based on our prediction of the structure of the 4.2-kb mRNA. Several PCR fragments were cloned and, when sequenced (for example, the 4.4-kb PCR␣ clone, Fig. 2), confirmed that the 4.2-kb mRNA is comprised of clones A and/or B fused to the 5Ј-end of clone C.
To map the 3Ј-end of the 4.2-kb hDIPP2 mRNA, we searched for functional polyadenylation signals at the appropriate location in clone C (see Fig. 2), by identifying homologous EST cDNA sequences containing poly(A) tails. These ESTs fell into two categories. The first group (which includes clone D in Fig.  2) has poly(A) tails attached to nt 3610, which is 20 nt downstream of a canonical AATAAA polyadenylation signal. The second set of EST cDNAs (which includes clone E in Fig. 2) has poly(A) tails attached to nt 4125. The candidate signal for this polyadenylation is the AATACA hexanucleotide (22,24), which begins 20 nt upstream of the poly(A) tail. Because this is a rare, noncanonical signal, we further studied its functionality by performing 3Ј-RACE reactions on human heart cDNA using a sense PCR primer homologous to nt 3750 -3777 of clone C (see "Experimental Procedures"). Seven independent 3Ј-RACE clones were identified and sequenced. All terminated in poly(A) tails (17-52 residues) at positions homologous to nt 4123-4125 of clone C (Fig. 2), strongly suggesting that the AATACA motif at nt 4105 is indeed a functional polyadenylation signal. Thus, we have obtained strong evidence that an array of alternate polyadenylation signals (Fig. 2) gives rise to multiple mRNAs for hDIPP2␣ with identical open reading frames that differ only in the length of their 3Ј-UTR.
Molecular Cloning of hDIPP2␤-Using the same human heart cDNA from which PCR␣ was identified, we amplified another 4.4-kb PCR product (PCR␤, Fig. 2). The latter was identical in sequence to PCR␣, except for the presence of an additional CAG codon in the hDIPP2␣ coding region (Fig. 2). The resultant putative protein, which we named hDIPP2␤ (Fig.  3), comprises 181 amino acids (calculated molecular mass, FIG. 3. Comparison of the predicted amino acid sequences of hDIPP1, hDIPP2␣, and hDIPP2␤. The figure shows an alignment of the predicted amino acid sequences of hDIPP1 and hDIPP2␣, generated by the GAP algorithm. The position of the additional Gln 86 residue in hDIPP2␤ is indicated. Identical amino acid residues are marked with a vertical line. The Nudt motif (GX 5 EX 7 REUXEEXGU, where U is usually either I, L, or V; (15)) is boxed, and the putative GX 2 GX 6 G signature for DIPP activity is underlined. The predicted secondary structure was assessed using the consensus method of Peptool v1.1; ␤-sheets are indicated in blue, and ␣-helices are indicated in yellow.

FIG. 4. Multiple human tissue Northern analysis of hDIPP2
mRNA. Northern analysis with DIG-labeled DNA probes was performed as described under "Experimental Procedures." A, hybridization with the hDIPP2 coding region probe P1 (see Fig. 2). B, hybridization with the 3Ј-UTR probe P3 (see Fig. 2). The lower image in each panel represents the results of subsequent hybridization of each blot with a ␤-actin control probe. 20,434 Daltons). The additional CAG codon is predicted to code for Gln 86 in hDIPP2␤ (Fig. 3). An identical open reading frame for hDIPP2␤ was identified in a PCR product amplified from human fetal kidney cDNA (data not shown). Furthermore, a 915-bp human EST that encodes hDIPP2␤ (clone F; Fig. 2) was identified and sequenced. Putative DIPP2␣ and DIPP2␤ open reading frames also occur in a number of independently isolated mouse and rat EST cDNA clones (Table I). Therefore, there is considerable evidence for the existence of molecular clones for both DIPP2␣ and DIPP2␤ in three mammalian species. Purified, recombinant hDIPP2␣ (Fig. 5, lane 2) hydrolyzed the ␤-phosphate from PP-[ 3 H]InsP 5 , yielding [ 3 H]InsP 6 as the product (Fig. 6A). The values of the kinetic parameters for this reaction, most notably the exceptionally high affinity for PP-InsP 5 (K m ϭ 4.2 nM), were very similar to those for hDIPP1 (Table II). However, purified, recombinant hDIPP2␤ (Fig. 5,  lane 3) was 5-fold less active against PP-InsP 5 compared with both hDIPP2␣ and hDIPP1 (Table II). Fig. 6B shows that hDIPP2␣ is also active against [PP] 2 -InsP 4 (which is formed by further phosphorylation of PP-InsP 5 , see Fig. 1). Note that in these reactions the PP-InsP 5 reaction product did not accumulate (Fig. 6), because hDIPP2␣ rapidly dephosphorylated this material to InsP 6 (see above). We had only limiting amounts of [PP] 2 -InsP 4 , so we could not determine the K m and V max for this reaction. Instead, we determined the value of the first order rate constants for [PP] 2 -InsP 4 hydrolysis. In these assays hDIPP2␤ was again found to be considerably (3-fold) less active than was hDIPP2␣ (Table III).

Dephosphorylation of PP-InsP 5 and [PP] 2 -InsP 4 by hDIPPs-
Clearly, the additional Gln 86 residue in hDIPP2␤ has a substantial negative impact upon its catalytic activity toward PP-InsP 5 and [PP] 2 -InsP 4 . The analysis of the effects of Gln 86 extends earlier studies showing the importance for catalysis of regions of the DIPP proteins that lie outside the Nudt consensus sequence (10). Moreover, the hydrolysis of diphosphoinositol polyphosphates by the hDIPP2 proteins extends the exception, introduced by hDIPP1, to the paradigm that Nudt motifs impart hydrolytic specificity solely toward nucleoside diphosphates (15). Assays were quenched, neutralized, and analyzed by HPLC as described under "Experimental Procedures." Differences in elution times between the runs shown in the upper and lower panels reflects the use of different HPLC columns. QE vector controls expressed no significant hydrolytic activity (data not shown), as we previously described (14). Vertical arrows mark the elution positions of standards, which were determined in parallel HPLC runs.  substituted by a phosphate group, because it is synthesized from inositol 1,3,4,5,6-pentakisphosphate (Fig. 1). Moreover, the interconversion of inositol 1,3,4,5,6-pentakisphosphate with PP-InsP 4 takes place in a metabolic compartment discrete from that in which InsP 6 , PP-InsP 5 , and [PP] 2 -InsP 4 are interconverted (25). We now show for the first time that PP-InsP 4 is actively dephosphorylated by hDIPP1 and by both hDIPP2 proteins (Table III). Again, the additional Glu 86 in hDIPP2␤ reduced the catalytic activity 2.5-fold compared with hDIPP2␣ (Table III). We used HPLC to determine that hDIPP2␣ (Fig. 7), as well as hDIPP2␤ and hDIPP1 (data not shown), cleaved the ␤-phosphate from the diphosphate group of PP-InsP 4 , yielding inositol 1,3,4,5,6-pentakisphosphate, as is the case when PP-InsP 4 is dephosphorylated in vivo (25). Thus, even though PP-InsP 4 is a structurally unique inositol phosphate that exists in a separate metabolic compartment, it can compete with PP-InsP 5 and [PP] 2 -InsP 4 for hydrolysis by hDIPPs in vivo.
Are Diadenosine Polyphosphates Substrates for hDIPP2?-We recently discovered that diadenosine polyphosphates, particularly Ap 6 A and Ap 5 A, compete with PP-InsP 5 and [PP] 2 -InsP 4 for the active site of hDIPP1 (16). Ap 6 A and Ap 5 A are the downstream effectors of a signaling cascade that regulates cardiac K ATP channels (17) and ryanodine receptors (18). It is, therefore, significant that there was no detectable hydrolysis of Ap 5 A by hDIPP2␣ (Table IV). Additionally, hDIPP2␣ hydrolyzed Ap 6 A 10-fold less actively than did hDIPP1 (Table IV). Thus, compared with hDIPP1, hDIPP2␣ is more specialized for the dephosphorylation of diphosphoinositol polyphosphates, particularly PP-InsP 5 .
hDIPP2␤ is less active than hDIPP2␣ toward diphosphoinositol polyphosphates (see above) yet, remarkably, hDIPP2␤ was more active than hDIPP2␣ against Ap 5 A (Table IV). This result indicates that the lower catalytic efficiency of hDIPP2␤ toward diphosphoinositol polyphosphates is not because of any general deterioration of the protein's catalytic activity. Additionally, our data (Table IV) show that the specificity of the hDIPP2 isoforms toward diadenosine polyphosphates is altered by the additional Gln residue in hDIPP2␤.
Detection of hDIPP2 Protein in Human Cells-Initial attempts to produce a DIPP2-specific antibody raised against a human C-terminal peptide were unsuccessful. Therefore, we used antipeptide antibodies to the C terminus of mouse DIPP2␣ (see "Experimental Procedures"). These antibodies cross-reacted with recombinant hDIPP2␣ and hDIPP2␤ (Fig. 8, lanes G and H, respectively), but did not detect recombinant hDIPP1 (Fig. 8, lane I). The DIPP2-specific antibodies were then used for Western analysis of protein extracts prepared from K562 cells (a human chronic myelogenous leukemic cell line) and SW480 cells (a human colonic adenocarcinoma cell line). These cells were chosen because RNA dot blotting indicated that they express hDIPP2 mRNA (data not shown). A single protein band from both the K562 and SW480 extracts cross-reacted with the anti-DIPP2 antibodies (Fig. 8, lanes A  and B, respectively). The size of these bands is slightly smaller than that of the recombinant hDIPP2 isoforms (Fig. 8, lanes G  and H) as expected, because the recombinant enzymes each have a poly(His) tag. No signal was detected when the cell extracts were probed with preimmune serum (Fig. 8, lanes C  and D). Furthermore, the hDIPP2 signal was eliminated when the antipeptide antibodies were blocked with the peptide itself (Fig. 8, lanes E and F). These data demonstrate that DIPP2␣ and/or DIPP2␤ are expressed in human cells. DISCUSSION Prior to this study, only one mammalian isoform of DIPP was known (i.e. DIPP1), and only two diphosphorylated inositol phosphates were recognized to be its substrates (PP-InsP 5 and [PP] 2 -InsP 4 ) (14). We have now shown that PP-InsP 4 , a metabolically unrelated polyphosphate (25), is also an important DIPP substrate. We have also described two new and catalytically distinct human DIPP isoforms (hDIPP2␣ and hDIPP2␤).   Rather complex molecular processes are likely to regulate the expression of the hDIPP2 proteins, as they are translated from multiple mRNAs, apparently generated through the use of an array of alternate (canonical and noncanonical) polyadenylation signals. Why is diphosphoinositol polyphosphate metabolism this eclectic? To help understand these complexities, consider first the likely significance of there being multiple mRNAs for hDIPP2 enzymes. The 4.2-kb hDIPP2␣ and/or hDIPP2␤ mRNA is the most abundant and generally expressed message that we have detected (Fig. 4). Even larger mRNAs for both the ␣ and ␤ isoforms are predicted to occur, both by the existence of clone C and by our amplification of the 4.4-kb PCR␣ and PCR␤ products (Fig. 2). Indeed, a faint 4.7-kb mRNA band was observed during Northern blotting (Fig. 4, A and B). Smaller 1.5 and 1.8 kb mRNAs were also detected, most prominently in the heart (Fig. 4). We have shown that these differently sized mRNAs contain the same ␣ and/or ␤ open reading frame but vary considerably in the length of the 3Ј-UTR because of the use of alternate polyadenylation signals. As such, this represents a relatively uncommon method of transcriptional regulation, as most eukaryotic transcription units possess a single polyadenylation signal (26). Even more unusual is the conscription of an array of noncanonical polyadenylation signals such as AATACA, ATTAAA, and AGTAAA, some of which, in vitro at least, are considerably less efficient than the canonical AATAAA (22). However, it has been pointed out previously (24) that mRNAs polyadenylated near noncanonical signals can be much more abundant than the in vitro data alone predict, indicating that other factors must also regulate polyadenylation efficiency.
The significance of there being multiple mRNAs is typically interpreted in terms of their differing in either stability or translatability, to modulate gene expression in a tissue-or developmental stage-specific manner (26). Indeed, the 3Ј-UTRs of the longer hDIPP2 cDNAs (clone C, PCR␣ and PCR␤) contain Alu repeat sequences (Fig. 2), which are implicated in modulating message stability (27). The 5Ј-UTR of hDIPP2 mRNAs may also play a role in regulating protein expression in view of its exceptionally high (84%) G/C content (28). It seems more than a coincidence that this is a feature that the hDIPP2 mRNAs share with the 5Ј-UTR of the hDIPP1 mRNA (14).
We should also consider the potential value to the cell of its being able to express various isoforms of hDIPP with distinct catalytic properties. The rank order for the general reactivity of these enzymes is hDIPP1 Ͼ hDIPP2␣ Ͼ hDIPP2␤ (Tables II  and III). We propose that the differential expression of these catalytically distinct isoforms provides flexibility in the response times for the molecular switching activity, which is believed to underlie the ongoing turnover of the diphosphoinositol polyphosphates (see the Introduction). For example, hDIPP2 mRNA is most highly expressed in the heart (Fig. 4), whereas hDIPP1 mRNA predominates in the brain (14). Moreover, although hDIPP2␣ is generally less active than hDIPP1, the former retains relatively high activity toward PP-InsP 5 (Table II). Thus, hDIPP2␣ may be more specialized for the hydrolysis of PP-InsP 5 .
The ␣ and ␤ isoforms of hDIPP2 are distinguished by only one amino acid residue, the presence of Gln 86 in hDIPP2␤, but the latter is up to 80% less active than hDIPP2␣ against PP-InsP 5 (Table II). A nucleotide polymorphism in the hDIPP2 gene may account for the coexistence of hDIPP2␣ and hDIPP2␤. Alternately, the two isoforms might be products of different genes. A third possibility is that there are tandem CAG codons at an appropriate intron-exon junction that obey the GU/AG rule, thereby providing two alternate splice sites.
There is a precedent for the transcription of a pair of mRNAs that are identical except for the presence or absence of an in-frame CAG codon (29). However, in that case the resulting Gln residue in the protein had no observable functional impact. This is clearly not the case for hDIPP2␤. Thus, it is important to understand the process that gives rise to the ␣ and ␤ forms of hDIPP2. Whatever the mechanism, it is probably conserved across species, because both DIPP2␣ and DIPP2␤ isoforms are also present in mouse and rat (Table I).
We may also consider that the diphosphoinositol polyphosphates have certain analogies with the inositol lipids. Each of these lipids have different cellular functions that are dictated by subtle differences in the distribution of phosphate groups around their inositol ring (30). A diverse group of enzymes exercises careful control over the levels of the individual inositol lipids. In this context, it can be appreciated why the regulation of the expression of several catalytically distinct DIPP isoforms will be an important physiological process, providing cells with quite subtle abilities to tailor specific rates of turnover of individual diphosphoinositol polyphosphates in vivo.