Structural insights into molecular function of the metastasis-associated phosphatase PRL-3.

Phosphatases and kinases are the cellular signal transduction enzymes that control protein phosphorylation. PRL phosphatases constitute a novel class of small (20 kDa), prenylated phosphatases with oncogenic activity. In particular, PRL-3 is consistently overexpressed in liver metastasis in colorectal cancer cells and represents a new therapeutic target. Here, we present the solution structure of PRL-3, the first structure of a PRL phosphatase. The structure places PRL phosphatases in the class of dual specificity phosphatases with closest structural homology to the VHR phosphatase. The structure, coupled with kinetic studies of site-directed mutants, identifies functionally important residues and reveals unique features, differentiating PRLs from other phosphatases. These differences include an unusually hydrophobic active site without the catalytically important serine/threonine found in most other phosphatases. The position of the general acid loop indicates the presence of conformational change upon catalysis. The studies also identify a potential regulatory role of Cys(49) that forms an intramolecular disulfide bond with the catalytic Cys(104) even under mildly reducing conditions. Molecular modeling of the highly homologous PRL-1 and PRL-2 phosphatases revealed unique surface elements that are potentially important for specificity.

PRL (for phosphatase of regenerating liver) phosphatases constitute a novel class of small tyrosine phosphatases involved in the modulation of cell growth. Initial studies identified PRL-1 as an intermediate-early gene expressed in the early response of regenerating liver tissue to mitogens (1). Overexpression of this protein was shown to lead to cellular transformation (1)(2)(3). The biological role of PRL-1 is tissue-dependent. Its overexpression is associated with cell proliferation in the liver (1) but with differentiation of epithelial cells in the digestive system (4). The closely related phosphatases PRL-2 and PRL-3 are also involved in growth regulation, proliferation, and cell invasion (3,5,6). All three proteins are prenylated at their C terminus, which critically affects their cellular localization and function (6 -8). As shown for the human PRL-2, the role of PRLs is associated with the regulation of progression through mitosis, and their cellular localization is likely controlled by the cell cycle (8).
PRL phosphatases are widely distributed in eukaryotes. In humans, PRL-1 and PRL-2 are ubiquitously expressed in various tissues (6), whereas PRL-3 is normally expressed in cardiac and skeletal muscles (5). Comprising typically only 140 -180 amino acids, PRLs are among the smallest phosphatases. They consist of a single catalytic domain lacking any auxiliary docking/regulatory domains other than the prenylation site at the C terminus. PRLs contain the protein-tyrosine phosphatase (PTPase) 1 active consensus motif HCXXGXXR, referred to as the P-loop; however, their primary sequence shows only remote similarity to phosphatases in other regions. Tyrosine-specific phosphatases as well as dual specificity phosphatases (DSP), enzymes capable of dephosphorylating both phosphotyrosine and phosphothreonine/serine residues, share the same general catalytic mechanism. The key structural elements include the positively charged phosphate-binding region of the P-loop and a catalytic cysteine residue that possesses an unusually low pK a of ϳ5 such that its side chain exists as a thiolate at a physiological pH. During catalysis, this cysteine acts as a nucleophile to form a thiophosphoryl enzyme intermediate, and a conserved aspartic acid in a neighboring loop participates in both the formation and the hydrolysis of the phosphoenzyme intermediate. The conserved arginine in the P-loop is important for the stabilization of the transition state (for a review, see Ref. 9). In many PTPases, correct positioning of the flexible loop containing the catalytic aspartate is critical for activation and catalytic performance of these enzymes (10).
Recent interest in PRL phosphatases relates to their role in cell proliferation, including promotion of cell migration, invasion, and metastasis (2-4, 7, 11, 12). SAGE (serial analysis of gene expression) experiments showed that PRL-3 is massively overexpressed in colon tumors metastasizing to the liver but not in nonmetastatic tumors and in normal colorectal epithelium (13). Further support for the involvement of PRL-3 in metastasis was provided by the finding of gene amplification in a significant fraction of metastatic lesions from different pa-tients (13). Because of the massive levels of overexpression, PRL-3 constitutes a useful marker for metastasis and possibly a new therapeutic target. More studies at both the physiological and biochemical level are needed to better understand the function of PRL phosphatases and, in particular, to evaluate their involvement and role in metastasis.
Here, we determined the solution structure of PRL-3 and classify it as a member of the family of dual specificity phosphatases. The structure and site-directed mutagenesis experiments identify residues that are important for PRL-3 catalytic activity and reveal unique features that distinguish PRL-3 from other phosphatases.

PRL-3 Expression and Purification-
The human phosphatase PRL-3 sequence, comprising amino acids 1-169, was subcloned into the pET15b vector (Novagen Inc., Madison, WI) and overexpressed in Escherichia coli BL21(DE3) as a His-tagged fusion protein. The protein was purified as described previously (14). The resulting protein contains the N-terminal extension (Gly-Ser-His) from the cleavage site of the vector and does not include the C-terminal prenylation site (Cys-Cys-Val-Met). Isotopically enriched PRL-3 was prepared from cells grown on minimal M9 medium containing [ 15 N]ammonium chloride with or without [ 13 C 6 ]glucose (Cambridge Isotopes Laboratory, Andover, MA). The PRL-3 isoforms were expressed and purified as described above.
PRL-3 mutants were obtained by site-directed mutagenesis using PCR. The identities of the proteins were checked by DNA sequencing and further verified at the protein level by mass spectrometry (Sciex APII electrospray mass spectrometer; Thornhill, ON, Canada). All of the mutants were expressed and purified using the same protocol as for the wild-type PRL-3.
NMR Spectroscopy-NMR samples with a protein concentration of 3 mM were exchanged into 50 mM phosphate buffer, 100 mM NaCl, 10 -12 mM DTT, and 0.1 mM sodium azide, pH 6.80. NMR experiments were performed at 308 K. Backbone and side chain NMR signal assignments of PRL-3 were determined as described previously (14). NOE constraints for the structure determination were obtained from 15 N-edited and homonuclear NOESY obtained at Bruker DRX 500-MHz spectrometer and 13 C-edited NOESY acquired on a Varian Inova 800-MHz spectrometer. The mixing time was 110 ms for these experiments. 3
Structure Calculations-NOE restraints were obtained from 15 Nand 13 C-edited three-dimensional NOESY experiments and from twodimensional homonuclear NOESY spectra. The and torsion angles were derived from C ␣ , C ␤ , and H ␣ chemical shifts using TALOS (18) and compared with experimental values resulting from an HNHA experiment. The structures were calculated using the ARIA module (19) implemented in the program CNS (version 1.1) (20). The initial set of several hundred NOEs was assigned for the protein and used to calculate the first ensemble of structures. Manually assigned distance constraints were classified according to the peak intensities as strong (1.8 -3.0 Å), medium (1.8 -4.0 Å), and weak (1.8 -5.0 Å). Two-dimensional NOESY, three-dimensional 15 N NOESY, and three-dimensional 13 C NOESY spectra were used in the ARIA protocol to calibrate and assign NOE cross-peaks. The unambiguous distance restraints obtained after eight rounds of calculations were used to calculate the final set of structures. The quality of the obtained structures was assessed using PROCHECK (21). The statistics for the structure calculations are shown in Table I. The coordinates have been deposited with the Protein Data Bank (code 1R6H), and the chemical shift assignments have been deposited with the BioMagResBank (accession number 5455).
Enzymatic Assays-Enzymatic assays were carried out with 3-Omethylfluorescein phosphate (OMFP) as a substrate using absorbance at 450 nm for detection of the product (22). All of the assays were performed at 22°C in TBA buffer (40 mM Tris-HCl, 150 mM NaCl, pH 6.15), containing 6 mM DTT, 5% Me 2 SO. The substrate concentration ranged from 0 to 200 M. Enzyme concentrations for the wild-type protein and the D71A and C49A mutants were determined from the "burst" amplitude (B) as 2.1, 4.5, and 2.5 M, respectively. Concentration of the A111S mutant was calculated to be 0.5 M. Hydrolysis was initiated by the addition of the enzyme to the reaction mixture and monitored for 90 min using a SpectraMAX-220 (Molecular Devices, Sunnyvale, CA) plate reader.
The raw data were fitted to Equation 1, where V 0 is the initial rate, k is the reaction constant (burst constant), and V s is the steady-state rate for the linear portion of the reaction plot. The burst amplitude (B) can be obtained by intersecting the linear part of the reaction plot with the y axis (product To dissect k 2 ϩ k 3 sum of the rate constants, k 2 was derived after fitting B versus [S] to Equation 3. When k 2 Ͼ Ͼ k 3 , i.e. dissociation of the E . . . . P to E ϩ P is ratelimiting, k 3 ϭ k cat , and k cat /K m represents the catalytic efficiency of the phosphatase (22).
Molecular Modeling-Three-dimensional models of phosphatases PRL-1 and PRL-2 were constructed by homology modeling using the program Modeler 6v1 (23), using the experimentally determined structure of PRL-3 as a template. Sybyl 6.4 software (Tripos, St. Louis, MO) was used for further structure refinement and analysis. Structural refinement was performed by energy minimization using an AMBER 4.1 all-atom force field (24) with the Powell minimizer, a distance-dependent (4r) dielectric constant and an 8 Å nonbonded cut-off. The N and C termini were blocked with acetyl-and methylamino groups, respectively, and hydrogen atoms were added explicitly.

RESULTS
Solution Structure of PRL-3-We determined the structure of the human phosphatase PRL-3, the first of the PRL protein family. The previously reported resonance assignments (14) were used to assign NOEs from heteronuclear 15 N-and 13 Cedited three-dimensional NOESY and homonuclear two-dimensional NOESY experiments. The excellent quality spectra yielded a large number of distance constraints. TALOS was applied to generate dihedral angle constraints based on C ␣ , C ␤ , and H ␣ chemical shifts. These values were confirmed by experimentally determined coupling constants from an HNHA experiment. On average, 15.4 constraints/residue in the PRL-3 structured region (Ala 8 -Arg 153 ) were used to calculate the PRL-3 structure. The 20 lowest energy structures out of 60 calculated were chosen to represent the final ensemble. The structural statistics are shown in Table I. PRL-3 is comprised of a five-stranded ␤-sheet and six ␣-helices (Fig. 1). Strand ␤1 (Val 10 -Ser 13 ) is antiparallel with respect to the remaining four parallel strands ␤2 (Met 17 -Thr 22  This arrangement of the secondary structure elements and overall fold is typical for DSP, structurally classifying PRL-3 as a member of this class of phosphatases. A comparison using DALI server reveals the closest similarity to the dual specificity phosphatases VHR (Protein Data Bank code 1vhr), PTEN (the phosphatase and tensin homologue; 1d5r), and KAP (1fpz) with Z scores of 10.4, 10.1, and 9.9, respectively. The root mean square deviation value between PRL-3 and each phosphatase is ϳ2 Å for the 60 -80 ␣-carbons in the conserved elements of secondary structure. Structure-based sequence alignment with other dual specificity phosphatases and tyrosine phosphatases shows sequence identity below 20%, which is typical for this relatively divergent class of enzymes (Fig. 2).
This low sequence similarity leads to significant local structural differences. Interestingly, particularly large differences are observed in the loops surrounding the active site, responsible for both catalytic activity and substrate specificity. Fig.  3A shows an enlarged view of this region in PRL-3. The catalytic residues Cys 104 and Arg 110 are located at opposite ends of the catalytic P-loop. The hydrophobicity of conserved amino acids between Cys 104 and Arg 110 distinguishes PRL-3 from other dual specificity phosphatases. The amino acids in this region have been shown to contribute to substrate specificity for numerous phosphatases (25,26). The other distinct feature of the PRL-3 catalytic site is the lack of protruding loops (Fig.  3B), which often participate in substrate binding (10,27). The N terminus, preceding strand ␤1, which is usually structured in protein-tyrosine phosphatases and contributes to substrate binding, is unstructured and mobile in PRL-3. The region following strand ␤2, which recognizes binding partners for KAP phosphatase (27), is much shorter and flat. The loop following strand ␤3 is also much shorter and flatter than in the PTP1B, VHR, PYST, and KAP phosphatases (10,27,28). The resulting catalytic pocket is extremely shallow and suggests a broad range of specificity for PRL-3.
Dephosphorylation is mediated by an acidic residue in the loop adjacent to the catalytic P-loop that serves as a general acid by contributing a proton to the leaving phenolate or serine/ threonine group (29). Sequence alignments suggest that Asp 72 in the ␤4-␣3 loop could serve as a general acid for PRL-3 (Fig.  2). This loop also includes four prolines and two glycines responsible for its unusual conformation and high mobility (data not shown). This loop does not seem to interact with other parts of the protein and may undergo conformational changes upon substrate binding. Such substrate-induced conformational adjustment was observed for PTP1B and PYST (10,30). It is possible that PRL-3 requires substrate activation, bringing the general acid residue closer to the catalytic cysteine to become fully active because poor activity was observed with synthetic substrates (see below).
Mutagenesis and Kinetic Analysis-The conserved catalytic residues Cys 104 and Arg 110 define the signature motif of PTPases and suggest a similar mechanism of dephosphorylation in PRL phosphatases. Because the C104A mutation was already shown to abolish the catalytic activity of PRL-3 (5), our mutagenesis experiments focused on the other conserved residues of the active site such as Ala 111 , Asp 71 , Asp 72 , and Cys 49 .
PRL-3 showed extremely low activity with phosphorylated peptides and the commonly used synthetic substrate, p-nitrophenyl phosphate (data not shown). For this reason, a more reactive substrate, OMFP, was chosen (22). Kinetic analysis of the wild-type protein showed a slow, two-step reaction with the formation of the phosphoenzyme intermediate and very slow, kinetically limiting release of phosphate group for regeneration of the free enzyme (Fig. 4). This differs from the fast burst phase of catalysis observed with other phosphatases (22,31).
The rate-limiting step for PRL-3 activity is the dephosphorylation (regeneration) of the enzyme. The value of k cat /K m obtained for the wild-type PRL-3 was 7.5 s Ϫ1 M Ϫ1 , which is 3 orders of magnitude lower than that for a typical DSP such as FIG. 1. Structure of PRL-3. A, stereo view of the backbone superposition of the 20 lowest energy structures for residues Ala 8 -Gln 156 . The unstructured N and C termini are not shown. B, ribbon representation of the average PRL-3 structure generated with MOLSCRIPT (41) and Raster3D (42). The secondary structure elements and N and C termini are labeled.
CDC25 and VHR (22). Surprisingly, both constants K S and K m (Table II) are very similar to those determined for CDC25 (23), implying that in the case of PRL-3 not substrate binding but stability of the phosphoenzyme (EP) is responsible for the slow kinetics.
A111S Mutant-The most striking feature in the catalytic site of PRL-3 is the presence of an alanine next to the catalytic arginine in the P-loop. Nearly all tyrosine and dual specificity phosphatases have a serine or threonine in this position, and the hydroxyl group of this conserved residue is important for the breakdown of the phosphoenzyme intermediate and for stabilizing the thiolate group of the catalytic cysteine (31,32). The alanine to serine mutation significantly improves both the burst rate and the dephosphorylation efficiency ( Fig. 4 and Table II). Under the experimental conditions used in this paper, it was possible to monitor only the steady-state portion of the kinetic curve for the A111S mutant. However, the size of the burst phase can be measured from the non-zero y intercept in the kinetic curve of Fig. 4. Increasing the enzyme concentration from 0.5 to 2.5 M linearly increased the size of the burst. The amplitude also showed substrate concentration dependence at concentrations below saturation (ϳ100 M). These results suggest that the A111S mutant follows two-phase kinetics but with a more rapid burst phase similar to that of CDC25 and VHR. Despite the improvement in catalytic performance of the A111S mutant, the k cat remained 2 orders of the magnitude lower than the same constant for VHR phosphatase (22). It appears that the presence of Ala 111 in the catalytic site of PRL-3 only partially explains the very low catalytic efficiency.
D71A and D72A Mutants-One of the two aspartic acids Asp 71 and Asp 72 is expected to act as a general acid in the catalysis (Fig. 2). The D72A mutation lowered the k cat value to a level where it could not be measured with our experimental setup, confirming its catalytic role. On the other hand, the D71A mutant had a k cat /K m value similar to that of wild-type PRL-3, indicating that Asp 71 plays no role in catalysis.
C49A Mutant-The PRL-3 structure reveals the presence of Cys 49 in close vicinity to the catalytic Cys 104 . The disulfide bond formation between these cysteines provides another plau- sible explanation for the observed low catalytic activity. This is particularly important because a similar intramolecular disulfide bond in KAP phosphatase was extremely stable, even in the presence of 0.2 M DTT (27). Mutation of the noncatalytic cysteine residue in KAP prevented the formation of this disulfide bond and increased the enzymatic activity 40-fold. In PRL-3, the kinetics of the C49A mutant was identical to that of the wild-type protein, indicating that Cys 49 is not responsible for the low enzymatic activity of PRL-3.
Oxidation of the Catalytic Cysteine-Under less strongly re-ducing conditions than used for the kinetic studies, disulfide bond formation between Cys 49 and Cys 104 could be detected by NMR and gel electrophoresis. Fig. 5 shows a fragment of a 1 H-15 N heteronuclear single-quantum correlation spectra of 15 N-enriched PRL-3 under mildly (0.5 mM DTT; Fig. 5A) and strongly (15 mM DTT; Fig. 5B) reducing conditions. Two species are present in the first spectrum (Fig. 5A), corresponding to a mixture of the reduced and oxidized forms of PRL-3. The oxidized form disappears as DTT is added, and only the reduced form is detected at a high concentration of DTT (Fig. 5B). Gel The residues for which mutagenesis was done are colored in blue. B, stereo view of backbone superposition of PRL-3 (in red) and VHR phosphatase (Protein Data Bank code 1VHR; blue), which is based on 74 C ␣ atoms from secondary structure elements. Significant differences in the vicinity of the catalytic site reflect a different substrate specificity for these enzymes. The critical loops for substrate recognition are labeled. N-term , N terminus; C-term, C terminus. electrophoresis can also be used to monitor the oxidation of phosphatases because the reduced and oxidized forms have different electrophoretic mobilities (33,34). Characterization of the wild-type PRL-3 by gel electrophoresis showed two bands corresponding to the oxidized and reduced forms (data not shown). The C49A mutant produced a single band, confirming that Cys 49 is involved in disulfide bond formation. These results indicate that PRL-3 is capable of forming an intramolecular disulfide between Cys 49 and the catalytic Cys 104 in a similar fashion to PTEN and CDC25 (33,34).
Isoforms of PRL-3-Two alternative splicing variants of PRL-3 have the Glu 11 -Ile 34 fragment 2 or the Ala 111 -Gln 135 fragment (gi:13111875) deleted. Analysis of the PRL-3 structure indicates that these deletions would eliminate folded regions from strands ␤1-␤3 in the first isoform and the two helices ␣4 and ␣5 in the second. Interestingly, the ends of the deletions map to neighboring loops, which raised the intriguing possibility that the isoforms are structurally reduced versions of the dual specificity phosphatase fold. To investigate this, isoforms 2 and 3 of PRL-3 were expressed and analyzed by NMR and enzymatic assays. Both isoforms were unstructured and inactive, demonstrating that the deletions compromised the structural integrity of PRL-3, causing unfolding and a loss of activity. Although the in vivo expression of the shorter PRL-3 isoforms is yet to be confirmed, our results demonstrate that only the full-length PRL-3 is biologically active.
Molecular Models of PRL-1 and PRL-2-To address the differences in biological roles of PRLs, molecular modeling was used to generate models for the human PRL-1 and PRL-2 phosphatases based on their high sequence similarity to PRL-3 (83 and 78% amino acid identity). All three phosphatases display striking conservation of amino acids around the P-loop (Fig. 6, A and B). The only variable amino acid in the vicinity of the active site that can potentially contribute to substrate specificity is Ile 141 in PRL-3, which is phenylalanine in both PRL-1 and PRL-2. Nearly all amino acids, which are different in PRL-1, -2, and -3, cluster on the face opposite the active site. Fig. 6C shows the conserved residues in PRL phosphatases from a wide range of eukaryotes. This analysis reveals that the most highly conserved regions are in the immediate vicinity of the catalytic site along with a small number conserved, surfaceexposed amino acids in other regions. DISCUSSION PRLs are a distinct class of small phosphatases within the family of DSP. They consist of a catalytic domain and a unique C-terminal prenylation site that is necessary for their cellular localization (7,8). The family of DSPs shares a low sequence similarity, but PRL phosphatases are closely related, displaying over 40% sequence identity between lower eukaryotes and human and above 70% among mammals. Nearly all of the conserved amino acids have either a clear structural role or are part of the active site (Fig. 6). The first group includes residues from the hydrophobic core and numerous prolines and glycines in loops between elements of the secondary structure. The conserved catalytic regions, which participate in substrate binding and confer specificity, include the P-loop, the Asp 72containing general acid loop, and the ␣5-␣6 loop. The particularly strong sequence conservation around the catalytic site suggests a similar substrate specificity in PRLs. Two unique features of the P-loop should affect the catalytic mechanism and specificity of PRL phosphatases. First, the highly conserved amino acids Val 105 -Ala-Gly-Leu-Gly 109 define the hydrophobic character of the P-loop and may indicate a preference for more hydrophobic substrates than those of other phosphatases. Second, the C-terminal sequence of the P-loop is very unusual, Ala 111 -Pro 112 -Val 113 , and the proline residue likely provides a unique conformational restrain in this critical position.
PRL-3 does not have the N-terminal substrate recognition region, which is responsible for the phosphotyrosine specificity in PTP1B. Because of the unique flat conformation of the loops 2 D. Banville, unpublished results. a The analysis of the "burst" phase showed that k 2 ϩ k 3 ϭ k 2 , indicating that k 2 Ͼ Ͼ k 3 . b The k cat /K m for wild-type PRL-3 is 7.5 s Ϫ1 M Ϫ1 . c ND, not determined. surrounding the catalytic P-loop, the catalytic cleft of PRL-3 is the shallowest of all known phosphatases. The structure of the complex between KAP and phospho-CDK2 showed that, unlike tyrosine phosphatases, the substrate specificity of DSPs may rely on interactions distant from the active site of the catalytic domain (27). These interactions may also involve various loops, surrounding the catalytic P-loop. The loop between the two C-terminal helices ␣5 and ␣6 of PRL-3 is positioned similarly to the corresponding loop in the KAP structure and likewise may play a role in substrate recognition. On the contrary, the unique KAP antiparallel ␤-hairpin with the key Lys 54 at its tip that participates in substrate binding has no structural equivalent in PRL-3. Although sequence alignments suggest that CDC14 is the closest homologue, structural comparison shows that PRLs are more similar to VHR, PTEN, and KAP.
The most unusual conserved feature of PRLs is the alanine residue following the catalytic arginine. Only two other known phosphatases, CDC25 and the recently identified SKRP1 (35), do not contain a serine or threonine in this position. In CDC25, other serines in the P-loop apparently provide the required hydroxyl functionality because reintroduction of the conserved serine results in a decrease rather than an increase in the catalytic activity (36). In contrast, the active site of PRL-3 does not contain any serines or threonines. Our data clearly show that this is responsible for part of the very low catalytic activity of PRL-3 and suggest that the missing hydroxyl group is provided by the physiological substrate.
The distant positioning of the mobile loop containing the general acid Asp 72 is another cause of the poor activity toward synthetic substrates. For MKP-3, a phosphatase with a similarly inactive conformation of the general acid loop, a 10 6 increase in the k cat /K m value was observed when the peptide substrate was replaced by a protein substrate (30). PRL-3 likely undergoes a similar substrate-induced conformational rearrangement to bring Asp 72 closer to Cys 104 and enhance catalysis.
The conservation of Cys 49 in all PRL phosphatases suggests a functional role. There is a growing interest in the oxidation of the catalytic cysteine in phosphatases and its role in the regulation of signaling pathways in response to oxidative stress FIG. 6. Sequence similarity analysis of PRL phosphatases shows highest conservation of residues in the active site. The amino acids in the molecular models of human phosphatases PRL-1 (A) and PRL-2 (B) that are different than those in PRL-3 are colored in red. The catalytic residues in the P-loop and amino acids different from PRL-3 are labeled. Two orientations of each phosphatase are shown. C, C ␣ trace of PRL-3 is color-coded according to the phylogenetic conservation among 55 PRL phosphatases. An increase in sequence identity corresponds to the change in color from white to blue. The catalytic P-loop and general acid loop are labeled. The figures were generated with GRASP (43). (37). Recent data indicate there are two possible mechanisms for the oxidation of the catalytic cysteine side chain that involve its conversion to a sulfenic acid or formation of an intramolecular disulfide bond (38,39). A plausible role for this disulfide is to protect the catalytic cysteine from irreversible oxidation during oxidative stress. Our results show a potential regulatory role of the conserved Cys 49 in controlling the efficiency of dephosphorylation under physiological conditions.
Among the other conserved residues in the active site, Phe 70 likely contributes to the substrate binding/recognition. Two absolutely conserved amino acids, Asn 142 and Gln 145 , possibly correspond to invariant glutamines in PTPases (i.e. Gln 262 and Gln 266 in PTP1B). Gln 262 in PTP1B is important for phosphoenzyme formation and hydrolysis. In PRL-3, Tyr 53 is positioned near the active site and could participate in substrate binding, perhaps in a phosphorylation-dependent fashion as observed for Tyr 138 in the VHR phosphatase (40). The conserved Arg 47 is sandwiched between Asp 54 and Asp 67 and may help define the general acid loop orientation through interactions with Asp 67 at the beginning of this loop.
On the opposite side of the PRL-3 active site, a conserved cluster of charged amino acids is present. These include Lys 89 , Glu 121 , Lys 125 , Glu 127 , and a basic stretch of 11 amino acids at the C terminus of the protein in a conserved motif (R/K)X(R/ K)X(R/K)X(R/K)X(R/K)X(R/K), where X is an uncharged residue. This basic fragment, located next to the C-terminal prenylation site, likely participates in membrane binding via interactions with phospholipids.

CONCLUSIONS
The PRL-3 structure demonstrates that PRL phosphatases are structurally similar to dual specificity phosphatases. Our structural and mutagenesis studies explain why PRL-3 shows very low enzymatic activity. Absence of a serine/threonine residue in the active site is one reason, because the A111S substitution significantly increases activity. Second, the general acid loop requires a conformational change to bring Asp 72 closer to the catalytic Cys 104 . These impediments are likely overcome in the presence of physiological substrates to reveal catalytic activity similar to that observed with other phosphatases. Finally, the redox-dependent disulfide bond between Cys 49 and the catalytic Cys 104 may potentially play a regulatory role and deserves further investigation.