Crystal Structures of the Novel Cytosolic 5′-Nucleotidase IIIB Explain Its Preference for m7GMP

5′-nucleotidases catalyze the hydrolytic dephosphorylation of nucleoside monophosphates. As catabolic enzymes they contribute significantly to the regulation of cellular nucleotide levels; misregulation of nucleotide metabolism and nucleotidase deficiencies are associated with a number of diseases. The seven human 5′-nucleotidases differ with respect to substrate specificity and cellular localization. Recently, the novel cytosolic 5′-nucleotidase III-like protein, or cN-IIIB, has been characterized in human and Drosophila. cN-IIIB exhibits a strong substrate preference for the modified nucleotide 7-methylguanosine monophosphate but the structural reason for this preference was unknown. Here, we present crystal structures of cN-IIIB from Drosophila melanogaster bound to the reaction products 7-methylguanosine or cytidine. The structural data reveal that the cytosine- and 7-methylguanine moieties of the products are stacked between two aromatic residues in a coplanar but off-centered position. 7-methylguanosine is specifically bound through π-π interactions and distinguished from unmodified guanosine by additional cation-π coulomb interactions between the aromatic side chains and the positively charged 7-methylguanine. Notably, the base is further stabilized by T-shaped edge-to-face stacking of an additional tryptophan packing perpendicularly against the purine ring and forming, together with the other aromates, an aromatic slot. The structural data in combination with site-directed mutagenesis experiments reveal the molecular basis for the broad substrate specificity of cN-IIIB but also explain the substrate preference for 7-methylguanosine monophosphate. Analyzing the substrate specificities of cN-IIIB and the main pyrimidine 5′-nucleotidase cN-IIIA by mutagenesis studies, we show that cN-IIIA dephosphorylates the purine m7GMP as well, hence redefining its substrate spectrum. Docking calculations with cN-IIIA and m7GMP as well as biochemical data reveal that Asn69 does not generally exclude the turnover of purine substrates thus correcting previous suggestions.

Within the HAD core domain, the canonical motifs I-III include the residues for magnesium ion coordination and catalysis [15][16][17].During catalysis of cytosolic 59-nucleotidases, the first aspartate of motif I (DXDX[T/V]) performs a nucleophilic attack on the phosphate moiety of the bound substrate involving a pentavalent phosphate transition state [18][19][20].Consequently, the phosphoryl group is transferred to the carboxyl group of this aspartate forming a covalent phospho-enzyme intermediate, which is subsequently resolved by hydrolysis.For that purpose, the second aspartate of this motif is thought to coordinate the attacking water molecule [13].During the reaction, a serine or threonine residue from motif II (WWW[S/T]), preceded by 3 hydrophobic W-residues, interacts with the phosphate group and stabilizes the pentavalent phosphate transition state.The conserved catalytic motif III (K(X) x D(X) 0-4 D) contains a lysine and two aspartate residues that stabilize the phosphorylated intermediate and coordinate the magnesium ion, respectively [5].Although the catalytic mechanism of 59-nucleotidases has been characterized in detail due to a number of available crystal structures and biochemical data, the structural basis for substrate recognition and discrimination remains elusive for some of them.
Recently, a novel member of the family of cytosolic 59nucleotidases, the cytosolic 59-nucleotidase III-like protein, has been characterized in humans and in Drosophila [21] and subsequently renamed cN-IIIB (UniProt IDs Q969T7 and Q9W197).Human cN-IIIB ( Hs cN-IIIB) exhibits a sequence identity and similarity of 59% and 80% to the human cytosolic 59-nucleotidase IIIA ( Hs cN-IIIA).Interestingly, although both enzymes share significant sequence similarity they appear to have different substrate specificities [21].Hs cN-IIIA is the main pyrimidine 59-nucleotidase in erythrocytes; its activity was thought to be restricted to (d)CMP, (d)UMP and dTMP [22].A lack of this enzyme in human erythrocytes results in hemolytic anemia, which probably is due to the accumulation of pyrimidine nucleotides and precipitation of unhydrolyzed ribosomal RNA [23][24][25].In contrast to human cN-IIIA, human cN-IIIB has a broad substrate specificity, efficiently converting CMP and UMP and, with significantly lower efficiency, GMP and AMP [21].Remarkably, cN-IIIB dephosphorylates the modified purine 7-methylguanosine monophosphate (m 7 GMP) with the lowest K m among all substrates tested.Due to the presence of the mRNA cap structure, 7methylguanosine nucleotides are released during degradation of all eukaryotic mRNAs: Either Dcp2 and related enzymes cleave off m 7 GDP, which is further converted to m 7 GMP, or m 7 GMP is released directly by the scavenger decapping enzyme DcpS [26][27][28].Additionally, 7-methylguanosine was shown to be present in several tRNA and rRNA species of archaea, bacteria and eukarya [29].Until recently, it was not known by which enzymatic activity m 7 GMP is further degraded after RNA decay.As a potential alternative to decay, accumulation of m 7 GMP in the cytosol and subsequent incorporation into nucleic acids seems possible but biologically highly undesirable [26,[30][31][32].
To gain insight into the broad substrate specificity of cN-IIIB and especially its unusual preference for m 7 GMP, we determined crystal structures of Drosophila melanogaster cN-IIIB ( Dm cN-IIIB) bound to different reaction products.The overall structure of Dm cN-IIIB is similar to cN-IIIA, however, critical residues conferring substrate specificity and determining the size of the substrate-binding pocket are different.The co-crystallized reaction product, 7-methylguanosine or cytidine, is bound in a deep cavity formed by the cap domain and the HAD core domain and is surrounded by three aromatic amino acids forming an aromatic slot.The structures explain why cN-IIIB prefers m 7 GMP over GMP as substrate but also efficiently converts pyrimidines.In combination with biochemical data from mutagenesis experiments of cN-IIIA and cN-IIIB and molecular docking calculations we define the structural basis for m 7 GMP dephosphorylation by 59nucleotidases.

Results and Discussion
Crystallization and Structure Determination of Dm cN-IIIB Dm cN-IIIB was crystallized in the presence of the reaction products 7-methylguanosine (m 7 G) or cytidine, and a mixture of AlCl 3 and NaF.The crystallographic phase problem was solved by means of molecular replacement using the crystal structure of mouse cN-IIIA as starting model [13] (PDB ID 2G08).The structure model of Dm cN-IIIB bound to m 7 G was refined at a resolution of 1.65 A ˚to R and R free values of 15.9% and 19.4%, respectively.Diffraction data and refinement statistics are summarized in Table 1.
The final model of the Dm cN-IIIB m 7 G complex contains two molecules in the asymmetric unit with one continuous segment encompassing residues 13-311 of molecule A as well as residues 13-312 of molecule B, both superposing with a root mean square deviation (rmsd) of 0.86 A ˚for 288 common C a atoms (residues 13-103 and 115-311).After all amino acids had been placed, additional density was observed in the active center, which, based on its shape, could be assigned as an octahedrally coordinated magnesium ion, the co-crystallized product nucleoside m 7 G as well as a metal fluoride transition state analog.It was shown that for pH values higher than 7.0, enzyme-MgF 3 complexes rather than enzyme-AlF 3 complexes are formed [33,34].Since both aluminum and magnesium were present during crystallization, we interpreted the density as MgF 3 -.

Overall Structure of Dm cN-IIIB
The overall structure of Dm cN-IIIB consists of two domains: the globular core domain and a smaller cap domain.The globular core domain is a haloacid dehalogenase domain (HAD domain) and resembles a Rossmannoid-like a/b structure with a central seven-stranded b-sheet exhibiting the topology b1Qb9qb8qb7qb2qb3qb4q (Figure 1).Between two consecutive b-strands, one or more a-helices build right-handed crossover connections generating a b-sheet flanked by several ahelices on either side.The b-sheet is twisted in a way that the two outermost strands are arranged in an almost perpendicular orientation with respect to each other.The HAD core is distinguished from other Rossmannoid folds by two additional structural key features.The ''squiggle'' (amino acids Phe56-Thr61, Figure S1) is a single helical turn (in some structures resembling a p-helix) located immediately downstream of the first core b-strand (b2) and is followed by the ''flap'' -a short b-turn (amino acids Thr65-Gly68).Both structural motifs provide the structural flexibility required for nucleotidases to switch between open (substrate binding) and closed (catalytically active) conformations.The sequence of the HAD domain in Dm cN-IIIB includes two insertions, the caps, at two distinct sites.The first one lies between b2 and a-helix 9 and, due to its location, is termed C1-type cap [15].It comprises the a-helices 5-8 (amino acids 74-143) (Figure 1).The second insertion is named C2-type cap, compasses b5 and b6 (amino acids 195-216) and is positioned between b4 and a-helix 11.These two insertions protrude from the a/b Rossmannoid core of the structure and form the cap domain, which serves as a movable lid for the active site of the enzyme as suggested by different orientations of the domain in crystal structures of cN-IIIA [13,14] and cN-IIIB (reported here).Moreover, it represents the main substrate specificity determinant of cN-IIIB as it harbors critical residues for substrate binding (see below).

Catalytic Motifs and Active site of Dm cN-IIIB
Dm cN-IIIB contains all canonical sequence motifs typically present in cytosolic 59-nucleotidases, namely motifs I, II, III as well as motif S. A sequence alignment of the canonical catalytic motifs for all intracellular 59-nucleotidases completed by the new member cN-IIIB is shown in Figure 2. Motif I ( 55 DFDYTI 60 in Dm cN-IIIB) is located in the loop connecting b-strand 2 and a-helix 5 (Figure 1 and Figure S1) and contains two aspartate residues separated by a hydrophobic amino acid.In homologous enzymes, the first aspartate was shown to be the nucleophile responsible for the in-line attack on the a-phosphate of the bound nucleoside monophosphate.In the reaction, the cleaved-off phosphate is transferred to this aspartate, initially generating a pentavalent transition state and, later, a covalent phosphoenzyme intermediate.During subsequent hydrolysis of the phosphoenzyme intermediate, the side chain carboxyl of Asp57 is likely to be involved in donation of a proton to the bound phosphate, thus acting as the general acid-base residue in this reaction [13].In fact, this state is represented in the crystal structures with the transition state analog MgF 3 -coordinated by Asp55 and the attacking water occupying the second apical coordination site (Figure 3).The transferred proton may originate from the water molecule, which is further bound by the side chains of Asp57 (motif I) and Ser171 belonging to motif II in the crystal structure (see below).Alternatively, the proton may stem from the 59-hydroxyl of another bound nucleoside, which results in a phosphotransferase activity observed in some members of the cytosolic 59-nucleotidase family [22,35].Motif II is a four amino acid stretch with the sequence 168 LVFS 171 and is located in b-strand 3 of Dm cN-IIIB (Figure 1).Motif III is of variable length in the different nucleotidases and includes a highly conserved lysine residue as well as two invariant aspartates ( 219 K(X) 25 DSIGD 249 in Dm cN-IIIB).
In the crystal structure of Dm cN-IIIB bound to m 7 G, the magnesium ion is octahedrally coordinated by the side chain carboxyls of Asp55 and Asp245 (motif I and III) as well as by a water molecule and a fluorine atom occupying the four equatorial positions (Figure 3).The two apical coordination sites are occupied by one water molecule and the main chain carbonyl of Asp57, respectively.The magnesium ion of MgF 3 -makes an apical contact to the remaining oxygen of the Asp55 carboxyl group confirming the predicted role of this amino acid in the reaction (Figure 3).The three fluoride ions are in contact with the side chains of Ser171 of motif II, Lys219 of motif III and the magnesium ion, respectively.The position and function of Lys219 of Dm cN-IIIB is reminiscent of the basic residues termed arginine fingers that stabilize the negative charge on reaction intermediates in many other phosphohydrolases [15].Finally, the invariant residues Asp245 and Asp249 of motif III coordinate the magnesium ion (see above).

7-methylguanosine Binding by Dm cN-IIIB
It has recently been shown that the Michaelis constant (K m ) for the Dm cN-IIIB-mediated dephosphorylation of m 7 GMP (K m = 13 mM) is significantly lower than for AMP (32 mM), CMP (48 mM), UMP (91 mM) and GMP (102 mM).For the human ortholog Hs cN-IIIB, the preference for m 7 GMP in terms of the K m value is even more pronounced [21].At least to some extent, this also reflects a higher binding affinity for m 7 GMP compared to the unmodified purines and pyrimidines.However, the structural basis for this preference, unusual among 59-nucleotidases, and the broad substrate specificity in general has so far been unknown.
Overall, the C1-type cap containing the substrate specificity motif (motif S) together with the HAD core domain generates a substrate-binding pocket which neatly accommodates the bound 7-methylguanosine in a deep binding cleft (Figure 1).Both the base and the ribose moiety of m 7 G are in contact with diverse protein residues.In detail, the 7-methylguanine moiety of the nucleoside is stacked in a parallel displaced orientation between the aromatic side chains of Trp120 and Phe75, bridging typical stacking distances of 3.5 A ˚between the side chains and the base (Figure 4A).Besides a p-p stacking interaction resulting from the parallel orientation of the p-electron systems of the purine ring and the aromatic amino acid side chains, there is also a cation-p stacking coulomb interaction between the side chain p-electrons of Trp120 or Phe75 and the positive charge of m 7 G.It has been shown that the positive charge of m 7 G mainly localizes to nitrogen 7 [36], which interestingly is positioned above the center of the indole ring of the stacking Trp120.The cation-p stacking thus seems to be a major contribution and one of the structural reasons for the high affinity of Dm cN-IIIB for the monomethylated m 7 GMP and the discrimination against the unmethylated GMP.Notably, the binding pocket is completed by another aromatic side chain of Trp121, which packs by T-shaped edge-to-face stacking perpendicular against the methylated purine ring, bridging a distance of 3.8 A ˚between the methyl group of m 7 G and the plane of Trp121.This binding pattern results in the formation of an aromatic slot by Trp120, Phe75 and Trp121 neatly enclosing the base and shielding it from surrounding charged amino acids and water molecules (Figure 4A and 5A).This binding mode is reminiscent of 59-cap dinucleotide binding by the nuclear import adapter Snurportin 1 [37], where the hypermethylated 59-guanine of the snRNA is stacked between a tryptophan side chain and the second RNA base while it is shielded from the aqueous surrounding by a second tryptophan.Moreover, in Dm cN-IIIB the residues Ser124, Phe209, Trp120, Trp121 and Phe75 form a cavity that harbors the N7-methyl group of the modified base (Figure 5A).In addition, residues Phe78, Thr128, Leu199, Phe209 and Leu174 form a deep and elongated second cavity near O6 of the m 7 G, which would be able to accommodate diverse alkylated adenine nucleotides such as N6-isopentyladenosine, N6-glycinylcarbamoyladenosine or N6-N6-dimethyladenosine that have been found in different RNA species [29].Thus, cN-IIIB may not only convert m 7 GMP but also a number of additional modified or alkylated nucleotides derived from the degradation of various RNA species.
In addition to the base, the 29 and 39 hydroxyls of the ribose make polar contacts to the carboxyl group of Glu103 (Figure 4A).The ribose 59 hydroxyl makes contacts with the side chain of Asp57 as well as two water molecules, one of which would be in the position for the in-line attack on the phospho-aspartate 55 (see above).Moreover, O6 of the purine ring is in close contact with Ser124 of motif S as well as with two conserved water molecules.Furthermore, nitrogens N1, N2 and N3 of the base are in contact with several water molecules coordinated by protein residues.
In summary, m 7 GMP is converted with a ,10-fold lower K m compared to GMP for Dm cN-IIIB and with a 45-fold lower K m for the human ortholog.The different K m values for GMP and m 7 GMP underline the importance of the contribution of the cation-p interaction between the methylated base and the aromatic side chains.Moreover, the slightly higher affinity of human cN-IIIB for m 7 GMP compared to Dm cN-IIIB could be explained by the fact that one of the stacking residues in the Drosophila protein, Phe75, is replaced by a tyrosine (Tyr60) in the human counterpart (Figure S1).It has been shown for a viral m 7 G-cap binding methyltransferase (VP39) that the exchange of such a stacking tyrosine in the binding pocket by phenylalanine reduces the affinity for the m 7 G-cap by a factor of four [38].Whereas the human enzyme, compared to Dm cN-IIIB, has a 2-fold higher apparent affinity for m 7 G, it shows reduced affinities for the substrates CMP (2-fold higher K m ), GMP (4-fold higher K m ), AMP (14-fold higher K m ) and UMP (5-fold higher K m ).This may, at least in part, be a result of hydrogen bonding between the nucleobase

Cytidine Binding by Dm cN-IIIB
In contrast to other cytosolic 59-nucleotidases, cN-IIIB dephosphorylates both pyrimidines and purines (see above).To compare binding of purines and pyrimidines and explain this broad substrate specificity, we crystallized Dm cN-IIIB in complex with cytidine and solved the crystal structure by means of molecular replacement using the structure of the Dm cN-IIIB m 7 G complex as starting model.Remarkably, there was no density for MgF 3 -in the active site of the cytidine bound structure.The overall conformation of Dm cN-IIIB in complex with cytidine is highly similar to the m 7 G-bound structure with an rmsd of 0.32 A ˚for 293 common C a atoms of the molecules A. Differences between both structures are mainly restricted to the flap (amino acid Thr65-Gly68), which is not defined in the cytidine bound structure and thus most likely disordered.In contrast to the m 7 G-bound structure, several amino acids in the C1-type cap have poorly defined side chains, however all main chain atoms as well as the important residues for substrate binding are well defined in the electron density map.
Dm cN-IIIB binds cytidine in a similar fashion as m 7 G, using the same substrate-binding pocket, however the hydrogen-bonding pattern is different (Figure 4B).The pyrimidine ring is stacked in an almost coplanar fashion between Phe75 and Trp120, bridging plane distances of 3.5 and 3.4 A ˚, respectively.The side chain of Trp121 packs in an edge-to-face fashion with a distance of 3.5 A to the C5 atom of cytidine.When superimposed to m 7 G, the centroid of the pyrimidine ring is in close proximity to the methylated nitrogen 7 of the guanine base, thus the cytosine ring captures the position of the 5-membered imidazole ring of m 7 G.
Besides the described stacking interactions, the cytidine makes only one direct hydrogen bond to Ser124, which is in contact to the N4 of the base (Figure 4B).Consequently, the watermediated interactions of Asn79, Thr92, Trp120 and Ser171 of Dm cN-IIIB bound to m 7 G are missing in the cytidine-bound structure, partly explaining the lower affinity of the enzyme for pyrimidines.

Comparison of Open and Closed States of cN-IIIB and cN-IIIA
Although all cytosolic 59-nucleotidases share similar HAD core folds, overall the structures are different due to the structural variability of their cap domains, which confer substrate specificity (Figure S3).The modular architecture of nucleotidases, with the conserved HAD core domain as catalytic platform and the structurally variable cap domain, allows different substrate specificities while maintaining catalytic efficiency.The closest nucleotidase homolog to cN-IIIB is the cytosolic 59-nucleotidase IIIA (former cN-III, sequence alignment in Figure S2), which has been structurally and functionally characterized in detail [13,14].Hs cN-IIIA (originally called pyrimidine-specific 59-nucleotidase 1) catalyzes the hydrolysis of CMP and UMP [25].Several crystal structures of cN-IIIA in complex with the substrate UMP [14] or representing different states of the reaction mechanism [10,13] have been solved showing the molecular basis for substrate binding and catalysis.Superposition of the HAD core domains of cN-IIIA and cN-IIIB shows that the orientation of the cap domains with respect to the HAD core are slightly different (Figure 6).The crystal structures of the cN-IIIA UMP complex and the cN-IIIB m 7 G complex superpose with an rmsd of 1.19 A ˚for 218 common C a atoms.In the crystal structure of the cN-IIIA UMP complex, the enzyme completely encloses the substrate in a deep cavity between the HAD core and the cap domain [14].In contrast, in cN-IIIB bound to m 7 G, the tip of the cap domain (loop between ahelices 6 and 7) is shifted by 4 A ˚away from the HAD core representing a more open state (Figure 6 and Movie S1).This structural difference in the orientation of the cap domain relative to the HAD core may be due to the fact that cN-IIIA is bound to the reaction substrate including the phosphate (UMP), while cN-IIIB is bound to a reaction product missing the phosphate (7methylguanosine).Notably, there are several direct contacts of the UMP phosphate with Lys213, Ser164 and Asp51 of the HAD core domain as well as with the active site magnesium ion, which stabilize the closed conformation of cN-IIIA.Instead, the corresponding residues Ser171, Lys219 and Asp55 in cN-IIIB are in direct contact with the transition state analog MgF 3 -, which mimics the phosphate moiety already transferred to Asp55 (Figure 3).As a consequence, bond cleavage between the aphosphate, coordinated in part by the HAD core, and the nucleoside, which is tightly bound by the cap domain, allows both domains to move away from each other and to subsequently release the reaction products (Movie S1).

Substrate Specificities of cN-IIIA and cN-IIIB
On the basis of structural data and mutagenesis experiments, it was suggested that the substrate-binding pocket of cN-IIIA is too small to accommodate purines and Asn69 would sterically clash with such substrates (Figure 5B), thus limiting the substrate spectrum of cN-IIIA to pyrimidine nucleotides [14].To characterize the different substrate specificities of cN-IIIA and cN-IIIB and to explain them by structural data we performed mutagenesis studies and activity assays.Our preliminary experiments with a SUMO-tagged version of Hs cN-IIIA were hampered by solubility problems, but suggested that m 7 GMP could be dephosphorylated, although with a much lower efficiency compared to CMP [21].With a modified purification procedure, we have now been able to obtain soluble untagged full-length Hs cN-IIIA, and steady-state kinetic parameters for the dephosphorylation of CMP and m 7 GMP could be determined (Table 2).While CMP was the preferred substrate, m 7 GMP was converted with a surprisingly high efficiency, with k cat /K m only 6-fold lower than the specificity constant for CMP.The K m for m 7 GMP was about 5-fold lower than for CMP and only about 2-fold higher than the K m of Hs cN-IIIB.Overall, the substrate specificities of Hs cN-IIIA and -B are more similar than previously thought, especially with respect to the efficient conversion of m 7 GMP by both enzymes.
The ability of Hs cN-IIIA to dephosphorylate m 7 GMP with high efficiency is inconsistent with the proposed role of Asn69 in preventing binding of purine nucleotides [14].This proposal was further tested by mutagenesis studies.For the enzyme variants, the following code was used: Amino acids Phe75-Gly76-Trp120-Trp121 of Dm cN-IIIB, forming the binding pocket for the nucleobase, are abbreviated FGWW (Figure 5).Asn69 in Hs cN- IIIA corresponds to Gly76 in Dm cN-IIIB, and the amino acids of Hs cN-IIIA corresponding to the FGWW motif were abbreviated HNWY (His68-Asn69-Trp113-Tyr114).When Gly76 of the Drosophila enzyme was replaced by an asparagine (variant FNWW), the specificity constants for CMP and m 7 GMP changed very little, but the apparent affinity for m 7 GMP was increased rather than decreased (Table 3).Introduction of the entire HNWY motif into Dm cN-IIIB had modest effects, partially converting the substrate preferences of the Drosophila enzyme to those of its human counterpart.When, in this context, Asn was replaced by Gly (variant HGWY), the K m for both CMP and m 7 GMP was decreased and k cat /K m was strongly increased only for m 7 GMP due to different effects on k cat (Table 3).In a comparison of FNWY and FGWY, the glycine variant lowered K m and increased k cat for both substrates, leading to a preference for m 7 GMP (Table 3).Thus, the effect of the amino acid occupying the position corresponding to Hs cN-IIIA Asn69 is context-dependent; a general function of this asparagine in this position to exclude purine substrates is not supported by the data.
To further test the hypothesis that Asn69 of cN-IIIA does not discriminate against purine substrates and explain the ability of cN-IIIA to convert m 7 GMP, molecular docking calculations were performed.The crystal structure of Mus musculus cN-IIIA (PDB ID 4FE3) was used to dock the potential substrate m 7 GMP as well as UMP into the ''side-chain flexible'' substrate-binding pocket, while UMP served as a control of the docking procedure since it is the co-crystallized substrate (see Materials and Methods for details).As expected, docking of UMP to Mm cN-IIIA led to a model that was virtually identical to the one revealed by the complex crystal structure.It had been suggested previously that Asn69 of cN-IIIA prevents binding of purine substrates for sterical reasons because it restricts the size of the substrate-binding pocket.However, when m 7 GMP was docked into the substrate-binding pocket of cN-IIIA, the side chain of Asn69 rotated by 180 degrees, adopting the conformation of another common rotamer (Figure 5C).Due to this movement, the size of the binding pocket was significantly increased, neatly accommodating the purine ring of m 7 GMP.Taken together, the results from activity assays and the docking analyses strongly suggest that Asn69 of cN-IIIA is not discriminating against purine nucleotides.
A second important residue in the substrate-binding pocket of cN-IIIB is Phe75, which is one of the aromates stacking against the nucleobase.Phe75 is replaced by a histidine (His68) in cN-IIIA decreasing the contribution of the p-p stacking interaction significantly although this position is occupied by a tyrosine or histidine in cN-IIIB orthologs of other species as well.Our set of mutants contains three pairs differing in the presence of phenylalanine or histidine at this position: FGWW versus HGWW; FGWY versus HGWY; and FNWY versus HNWY.In all three pairs, and with both CMP and m 7 GMP as substrates, the apparent affinities were consistently higher with the variants containing a phenylalanine at this position (Table 3).This fact underlines the importance of p-p stacking in substrate binding by 59-nucleotidases and may explain in part the generally lower affinities of cN-IIIA for its substrates.

Protein Expression and Purification for Crystallization
The cDNA of Drosophila melanogaster cytosolic 59-nucleotidase IIIB ( Dm cN-IIIB) (CG3362 on www.flybase.org)encoding the fulllength protein (amino acids 1-319) was cloned into the pET-SUMOadapt vector as described [21].The SUMO-Dm cN-IIIB fusion construct was expressed in Escherichia coli Rosetta 2(DE3) (Merck) at 24uC in kanamycin-and chloramphenicol-containing 2YT-medium.Expression was induced at an OD 600 of 1.0 by addition of 0.6 mM IPTG.The cells were harvested after 4 hours of induction (5,0006g, 20 min, 4uC) and resuspended in lysis buffer (300 mM NaCl, 50 mM Tris/HCl pH 7.5, 10 mM imidazole, 5 mM MgCl 2 and 2 mM b-mercaptoethanol).Cells were disrupted using a microfluidizer 110S (Microfluidics) and the lysate, which was clarified by centrifugation (30,0006g, 30 min, 4uC), was loaded onto a HisTrap column (GE Healthcare) equilibrated with lysis buffer.The column was washed with 2 column volumes (CV) of lysis buffer and eluted with a linear gradient of lysis buffer containing 500 mM imidazole.Afterwards, imidazole was removed from the solution by passing the eluate over a 50 ml desalting column (GE Healthcare) equilibrated in desalting buffer (300 mM NaCl, 50 mM Tris/HCl pH 7.5, 5 mM MgCl 2 and 2 mM b-mercaptoethanol).For cleavage, the SU-MO-Dm cN-IIIB fusion-protein was incubated with SUMO protease (Invitrogen) overnight at 4uC and in a 1: 100 molar ratio of protease : fusion protein.In order to remove the His 6 -SUMO tag as well as the protease, the solution was passed through a second HisTrap column.The flow through of this column, mainly containing Dm cN-IIIB, was finally purified using a Superdex S75 (26/60) gel filtration column (GE Healthcare) in a buffer containing 150 mM NaCl, 20 mM Tris/HCl pH 7.5, 5 mM MgCl 2 and 2 mM b-mercaptoethanol.The pure protein was concentrated to 15 mg/ml using Millipore concentrators with a molecular weight cut-off of 10,000 Da (Merck) and aliquots were frozen in liquid nitrogen and stored at 280uC.

Crystallization and Structure Determination
Full-length Dm cN-IIIB (amino acids 1-319) was crystallized by vapor diffusion in 24-well sitting drop crystallization plates at 20uC.Plate-shaped crystals were obtained in a condition containing 26% (w/v) PEG 4,000, 0.1 M HEPES/NaOH pH 7.5 and 0.2 M NaCl at a protein concentration of 12 mg/  ml and when the protein was previously mixed with 5 mM AlCl 3 , 30 mM NaF and with a 15-fold molar excess of m 7 G or cytidine, respectively.The crystals belong to space group P2 1 and were flash cooled in liquid nitrogen after soaking in reservoir solution containing additionally 8% (w/v) PEG 4,000 as cryo protectant.Diffraction images from crystals containing m 7 G or cytidine were collected at the beamline BL14.1 of the Berliner Elektronenspeicherring-Gesellschaft fu ¨r Synchrotronstrahlung (BESSY II, Berlin) [39].The datasets were integrated, scaled and merged using XDS and XSCALE [40], respectively.The structure of the Dm cN-IIIB m 7 G complex was solved by means of molecular replacement using MOLREP [41] as implemented in CCP4 [42] and the crystal structure of mouse pyrimidine 59-nucleotidase type 1 (P5N-1 = cN-IIIA, PDB ID 2G08) as starting model [13].Full-length mouse cN-IIIA and Dm cN-IIIB possess a sequence identity and similarity of 33% and 54%, respectively.Atomic coordinates of magnesium trifluoride as well as solvent molecules and the co-crystallized product 7-methylguanosine were added to the model, which was refined by iterative cycles of PHENIX [43] and manual building in Coot [44] to R and R free values of 15.91% and 19.42%.The crystal structure of the Dm cN-IIIB cytidine complex was solved by molecular replacement (MOLREP) using the previously solved Dm cN-IIIB m 7 G complex structure as starting model but with the nucleoside and MgF 3 -omitted.Interestingly, there was no density for MgF 3 -in this crystal structure, thus it was not modeled.Atomic coordinates of the cytidine were added to the model and the structure was refined (PHENIX) to R and R free values of 22.24% and 25.47% (see Table 1).

Site-directed Mutagenesis
The expression plasmid coding for Dm cN-IIIB [21] was mutated by site-directed mutagenesis (Stratagene QuikChange protocol) with specific oligonucleotides listed in Table S1.Mutations were verified by sequencing and by MALDI-TOF mass-spectrometric analysis of the purified proteins.

Purification of Proteins for Enzyme Assays
The expression clone containing the cDNA of Hs cN-IIIA (transcript variant 1; NM_001002010.2) in a pET-SUMOadapt expression vector [46] has been described [21].Escherichia coli Rosetta 2(DE3) cells (Novagen) were transformed with the plasmid and cultured in 800 ml TB medium [47] including kanamycin (30 mg/ml) at 37uC to an OD 600 of 0.9.Expression was induced by mixing the culture with 800 ml ice-cold TB medium including kanamycin (30 mg/ml) and IPTG (1 mM), and incubation was continued at 16uC for 5 hours.After harvesting, cells were resuspended in 25 ml lysis buffer (20 mM potassium phosphate buffer pH 7.6, 20 mM imidazole, 500 mM KCl, 2 mM MgCl 2 , 5% (w/v) glycerol) and incubated for 1 h at 8uC with 10 mg of lysozyme, 5 mg of DNase I and 30 mg PMSF.The cells were disrupted twice with a French Press and centrifuged at 30,0006g for 75 minutes.The supernatant was loaded onto a 2 ml Ni 2+ -NTA-agarose column (Qiagen) equilibrated in lysis buffer.After washing with lysis buffer, including a step with lysis buffer containing 2 M KCl, bound protein was eluted stepwise with lysis buffer containing increasing concentrations of imidazole (50-500 mM).Fractions containing His 6 -SUMO-cN-IIIA were identified by SDS-PAGE analysis and pooled.For SUMO-cleavage, 1.5 mg recombinant Ulp1 protease (expression plasmid obtained from Christopher D. Lima [48]), 5 mM DTT and 1 mM EDTA were added, and the mixture was dialysed for 4 h at 0uC against lysis buffer containing 10% (w/v) sucrose instead of glycerol.Cleaved protein was again applied to a Ni 2+ -NTA-agarose column, and the flow-through fraction was loaded onto a Heparin column (GE Healthcare) equilibrated in buffer A (20 mM potassium phosphate buffer pH 7.6, 75 mM KCl, 10% (w/v) sucrose).Bound proteins were eluted with 7 CV of a linear gradient from buffer A to buffer B (as buffer A, but containing 2 M KCl).Fractions containing Hs cN-IIIA were identified by SDS-PAGE analysis and pooled for dialysis against 20 mM HEPES Dm cN-IIIB variants were purified as described [21].

Biochemical Analysis and Colorimetric Activity Assay of cN-IIIB and cN-IIIA
Colorimetric nucleotidase assays were performed as reported [21].Substrates were from Jena Bioscience (m 7 GMP) and Sigma Aldrich (CMP).Substrate concentrations were varied between 1 mM and 2 mM.Human enzymes were assayed at 37uC, Drosophila enzymes at 25uC.In the case of Hs cN-IIIA and Hs cN-IIIB, enzymes were preincubated with 5 mM DTT for 5 minutes at room temperature.The reaction was started by enzyme addition and stopped by mixing an aliquot (2.5-100 ml per time point) with malachit green oxalate/ammonium molybdate reagent.A calibration curve with phosphate concentrations from 1-30 mM was used to estimate orthophosphate released in the enzymatic reaction.Enzymatic steady state parameters were determined as described [21].The measurements were repeated twice, with kinetic constants varying by up to 622%.R-values for the hyperbolic Michaelis-Menten fit ranged from 0.9703-0.9962with 8-16 points per fit.

Molecular Docking Calculations
Protein and ligand preparation was done using AutoDockTools 1.5.6 [49] and molecular docking studies were carried out with AutoDock Vina [50].The substrates m 7 GMP and UMP were docked to the active site of the crystal structure of Mus musculus cN-IIIA (PDB ID 4FE3).Water molecules and the substrate UMP were omitted and polar hydrogens as well as Gastiger charges were added prior to docking calculations for both docked molecules and the receptor.The charge of the active site magnesium ion was set to +1.2.Calculations were done using a box size of 18622622 A encompassing the ligand binding site as well as all residues that were kept flexible during docking (Thr66, His68, Asn69, Lys89, Trp113, Tyr114, Ser117).Exhaustiveness and energy range of the calculations were set to 1000 and 4, respectively, and docking calculations were repeated twice in order to verify reproducibility of results.

Accession Numbers
Coordinates and structure factors of the crystal structures of Dm cN-IIIB bound to 7-methylguanosine or cytidine have been deposited within the protein data bank (PDB) with the PDB IDs 4NV0 and 4NWI, respectively.

Figure 1 .
Figure 1.Overall structure of Drosophila melanogaster cN-IIIB.Dm cN-IIIB consists of the conserved central HAD core domain (grey b-strands and white a-helices) and a cap domain.The cap domain packs against the top of the HAD core and is composed of two regions, a C1-type cap (red) and a C2-type cap (blue).The ''squiggle'' and the ''flap'', both structural key features of 59-nucleotidases, are marked with an asterisk and a black star, respectively.Secondary structure elements as well as termini and the bound magnesium ion (green sphere) are labeled.The co-crystallized reaction product 7-methylguanosine (m 7 G) is bound between the cap domain and the HAD core and is represented as stick model (carbon in yellow, nitrogen in blue, oxygen in red).The lower panel shows a schematic representation of Dm cN-IIIB with the canonical 59-nucleotidase motifs I-III and S as well as the positions of the cap insertions highlighted as boxes and labeled.doi:10.1371/journal.pone.0090915.g001

Figure 2 .
Figure 2. Conserved catalytic sequence motifs of the intracellular human 59-nucleotidases and cN-IIIB.Sequence conservation of catalytic motifs of 59-nucleotidases completed by the new member cN-IIIB.The respective motifs of Hs cN-IIIB and Dm cN-IIIB have been deduced from the crystal structure and sequence analyses.Shading indicates the different degrees of conservation from absolutely conserved (black) to partly conserved (grey).Motif I, II and III contain conserved catalytic residues and motif S contains residues involved in substrate recognition.Amino acids are numbered and consensus sequences are given above the alignment.doi:10.1371/journal.pone.0090915.g002

Figure 3 .
Figure 3. Organization of the catalytic center of Dm cN-IIIB.The magnesium ion (green sphere) is octahedrally coordinated by several residues of Dm cN-IIIB shown as stick model (carbon in yellow, nitrogen in blue, oxygen in red) as well as several water molecules (red).One equatorial coordination position of Mg 2+ is occupied by a fluorine atom of the co-crystallized metal fluoride MgF 3 -(magnesium in green; fluorine in light blue).MgF 3 -is coordinated by Asp55, Asp57, Lys219, Ser171 as well as water molecules.Note that the coordination of MgF 3 -represents the pentavalent phosphate transition state in the nucleotidase reaction cycle.Contacts are represented by dashed lines and the model is defined by a representative 2 mF o -DF c electron density map (blue mesh) contoured at a sigma level of 2.0.doi:10.1371/journal.pone.0090915.g003

Figure 4 .
Figure 4. Detail view on the substrate-binding pocket of Dm cN-IIIB.(A) The co-crystallized reaction product 7-methylguanosine is shown in ball-and-stick mode (carbon in grey, nitrogen in blue, oxygen in red) and is defined by an mF o -DF c omit-electron density map (blue mesh) contoured at a sigma level of 3.0.m 7 G is stacked in a coplanar but off-centered fashion by Phe75 and Trp120 (colors as in Figure 3).Trp121 packs by T-shaped edge-to-face stacking against the bound product and binding is additionally supported by numerous direct or water-mediated hydrogen bonds (dashed lines with distances) to various protein residues of Dm cN-IIIB.(B) Binding of the reaction product cytidine by Dm cN-IIIB.In contrast to m 7 G, the nucleobase moiety of cytidine makes only a hydrogen bond to Ser124.doi:10.1371/journal.pone.0090915.g004

Figure 5 .
Figure 5. Structural comparison of Dm cN-IIIB and Mm cN-IIIA.Detail view on the substrate-binding pockets of Dm cN-IIIB bound to m 7 guanosine in red (A) and Mm cN-IIIA bound to UMP in dark blue (B).The left panel shows the binding pocket with labeled residues and substrates represented as sticks.Note that Phe75 of Dm cN-IIIB is replaced by a histidine in Hs cN-IIIA while the Trp121 of Dm cN-IIIB is replaced by a tyrosine in Hs cN-IIIA.In the right panel the substrate-binding pockets are shown as surfaces, which were calculated using a probe radius of 1.65 A ˚and for clarity reasons were clipped at the level of the nucleobases.Note that the substrate-binding pocket for m 7 G (cN-IIIB, red surface) is significantly larger than the UMP binding pocket of cN-IIIA (blue surface), which was thought to be the structural reason for its purine substrate exclusion.(C) Orientation of substrate binding residues and the corresponding pocket after docking m 7 GMP to Mm cN-IIIA (light blue).Note that Asn69 (marked by an asterisk) rotates by 180 degrees in comparison to (B) and adopts another common rotamer thus enlarging the binding pocket and neatly accommodating the m 7 GMP.doi:10.1371/journal.pone.0090915.g005

Figure 6 .
Figure 6.Overall structural comparison of Dm cN-IIIB bound to m 7 G with Mm cN-IIIA bound to UMP.Superposition of Dm cN-IIIB (HAD core in white; cap domain in red) and cN-IIIA (HAD core in grey; cap domain in blue) showing the overall structural similarity of both enzymes.Note that the cap domain of cN-IIIB is shifted by 4 A ˚when the structures are superposed via their the HAD core.Thus, the substrate-bound cN-IIIA adopts a closed conformation while the product-bound cN-IIIB represents an open conformation.A movie of the trajectory from the open-to-closed state is shown in Movie S1. doi:10.1371/journal.pone.0090915.g006

Figure
Figure S1 Multiple sequence alignment of cytosolic 59nucleotidase IIIB from Drosophila melanogaster, Xenopus laevis, Gallus gallus, Mus musculus, Rattus norvegicus and Homo sapiens.Identical residues are shown on red background, while similar ones are represented in red on white background.Secondary structure elements from the crystal structure of Dm cN-IIIB are shown above the alignment.Note that important residues from motifs I-III (marked with a black triangle) as well as both tryptophans necessary for substrate binding in

Table 1 .
X-ray diffraction data and structure refinement statistics of Drosophila melanogaster cN-IIIB bound to different reaction products.Values in parentheses indicate the specific values in the particular highest resolution shell.

Table 3 .
Kinetic properties of the cytosolic 59-nucleotidases IIIA, IIIB and variants of Dm cN-IIIB.