Structure-Based Analysis of Catalysis and Substrate Definition in the HIT Protein Family

See allHide authors and affiliations

Science  10 Oct 1997:
Vol. 278, Issue 5336, pp. 286-290
DOI: 10.1126/science.278.5336.286


The histidine triad (HIT) protein family is among the most ubiquitous and highly conserved in nature, but a biological activity has not yet been identified for any member of the HIT family. Fragile histidine triad protein (FHIT) and protein kinase C interacting protein (PKCI) were used in a structure-based approach to elucidate characteristics of in vivo ligands and reactions. Crystallographic structures of apo, substrate analog, pentacovalent transition-state analog, and product states of both enzymes reveal a catalytic mechanism and define substrate characteristics required for catalysis, thus unifying the HIT family as nucleotidyl hydrolases, transferases, or both. The approach described here may be useful in identifying structure-function relations between protein families identified through genomics.

Human FHIT and PKCI proteins are members of the ubiquitous protein family HIT, which denotes a conserved histidine triad (His-x-His-x-His) sequence motif (1). HIT protein sequences have been identified independently and through genomics in Prokaryae, Archae, and Eukaryae [(1, 2) and references therein]. Members can be aligned to reveal at least two subfamilies. One subfamily, of which human PKCI is a member, is found throughout evolution and is characterized by a highly conserved COOH-terminal sequence and greater than 94% amino acid sequence identity overall between known mammalian homologs. We previously described the identification, cloning, and structure determination of human PKCI in its unliganded form (3). Human FHIT, whose gene resides in a fragile locus on human chromosome 3 (4), is a member of a divergent eukaryotic HIT subfamily that differs significantly from the more conserved subfamily (∼20% identity) by deviations at NH2- and COOH-termini. Disruption of the FHIT gene is associated with human cancers (4), but definitive evidence supporting its role as a tumor suppressor has yet to be elucidated. We recently reported the three-dimensional structure of FHIT and its complex with a nucleoside ligand (2). A core domain of ∼100 amino acids is similar among all HIT family members; PKCI and FHIT share 20.7% identity over a 110–amino acid overlap.

Characterized HIT proteins exist as homodimers. The HIT protomer structure can be described as a general α + β type and further subclassed as an α + β meander fold (2, 3). The HIT protomer contains a core of five antiparallel β strands and a central α helix. Two protomers are brought together in the homodimer through interactions between central helices and the joining of β strands into a 10-stranded antiparallel sheet. Histidine residues are prominent features in the two equivalent active sites of the HIT homodimer.

No enzymatic activity has been reported for PKCI or related subfamily members, although the rabbit PKCI HIT homolog has been shown to bind purine monophosphate nucleotides (5). However, FHIT and a related Schizosaccharomyces pombe HIT protein have been shown to cleave asymmetrically a broad range of dinucleotide polyphosphates in the presence of a divalent metal ion. These enzymes prefer an adenine base in the substrate with cleavage always resulting in the release of a monophosphate nucleotide (6). We now report that PKCI follows saturable kinetics in cleavage reactions with adenosine diphosphate (ADP) but not with Ap3A or adenosine triphosphate (ATP), clearly showing a preference for an adenine base in the substrate (7). FHIT also cleaves ATP, Ap3A, and Ap4A with saturable kinetics, but not ADP (Table 1) (8). A metal binding site has not been identified in structures of either FHIT or PKCI (2, 3, 5), so we suspected that catalysis should occur in the absence of a divalent metal. Indeed, FHIT hydrolysis of Ap3A proceeds, albeit at a slower rate, even in the presence of 1.0 mM EDTA (Table 1). If a divalent metal were required for hydrolysis, chelation would reduce the catalytic rate constant k cat to an undetectable level, presumably zero. Nucleotide polyphosphates exist in the cell primarily as complexes with divalent cations. The metal ion effects observed here may result in part from the impact of divalent cations on nucleotide conformation.

Table 1

Kinetic parameters and substrate preference for FHIT and PKCI (7). K m, Michaelis constant; –, no product detected.

View this table:

Because a metal is not required and because of the conserved histidyl residues in the active site, we surmised that catalysis proceeds through a covalent nucleotidyl phosphohistidyl intermediate, as has been observed for many other enzymes (9). If this is true, a radiolabeled nucleotide [adenosine monophosphate (AMP)] should be found transiently and covalently attached to the protein. To confirm this hypothesis, we trapped a covalent FHIT-nucleotidyl phosphohistidyl intermediate in a hydrolase reaction using α-32P– and 2,8-3H–labeled ATP (10) (Fig.1).

Figure 1

Analysis of the covalent FHIT-nucleotide monophosphate intermediate by SDS-PAGE and autoradiography. Lane 1, cold unlabeled Coomassie blue–stained FHIT protein; lane 2,32P radiolabel coincident with the FHIT protein; lane 3,3H radiolabel coincident with the FHIT protein; lane 4, 10-kD ladder molecular weight markers. Details of the trapping reaction are summarized in the text (10).

To elucidate the structural basis for HIT protein catalytic activity and to define characteristics of in vivo HIT ligands, we prepared a series of complexes representing key steps in the catalytic reaction. The PKCI and ADP catalytic cycle was characterized with the use of the nucleotide adenosine α/β-methylene diphosphate (AMP-CP) as an analog of the substrate complex. A transition-state analog was prepared by reacting the enzyme with adenosine and tungstate to produce the pentacovalent enzyme complex. Finally, the nucleotide product complex was produced by reacting the enzyme with AMP. Crystal structures of both FHIT and PKCI were determined for each of these complexes and were analyzed in comparison with the corresponding apo structures (11) (Table2).

Table 2

Crystallographic data and refinement statistics (11).

View this table:

The apo unliganded structures (Figs. 2A and 3A) show FHIT His96 and PKCI His112 stabilized by a conserved structural motif. The carbonyl oxygen atom of FHIT His94 or PKCI His110 is hydrogen-bonded in a close interaction with the Nδ atom of FHIT His96 or PKCI His112, suggesting the proper activation of these residues as nucleophiles for an in-line attack on the α phosphate group as it enters the binding pocket.

Figure 2

Stereo views of PKCI in (A) apo, (B) AMP-CP substrate analog, (C) adenosine-tungstate transition-state analog, and (D) AMP product-bound forms. The region shown was selected to highlight interactions between the ligand and protein residues surrounding one of the two equivalent ligand-binding sites in the HIT homodimer. A subset of residues is shown superimposed on the Cα backbone cardinal spline of each respective structure. Hydrogen bonds are denoted by dotted lines. The tryptophan shown is in the COOH-terminal tail of the other protomer. Figure generated with Setor (15).

Figure 3

Schematic diagram of FHIT in (A) apo, (B) AMP-CP substrate analog, (C) adenosine-tungstate transition-state analog, and (D) adenosine-sulfate product complexed forms. As in Fig. 2, hydrogen-bonding interactions are depicted by dotted lines. A subset of residues is shown superimposed on the Cα backbone cardinal spline of each respective structure.

The substrate analog structures (Figs. 2B and 3B) of PKCI and FHIT in the presence of AMP-CP at 2.0 Å and 2.3 Å, respectively, show PKCI His112, FHIT His96, and ligand in an optimal orientation for attack at the α phosphate position. The distances to the phosphorus center from the Nɛ of PKCI His112 and from FHIT His96 are only 3.15 Å and 3.43 Å, respectively, whereas the distances to the proximal α phosphate oxygen atoms are nearly equal—an ideal situation for an in-line attack on the α phosphate group. The β phosphate group in the FHIT AMP-CP structure was disordered.

The transition-state analog structures (Figs. 2C and3C) of PKCI and FHIT at 1.8 Å and 2.6 Å, respectively, show an adenosine base and tungstate ions bound in an orientation similar to that of the substrate analog. The transition-state analog structures are the remarkable products of reacting the enzymes with adenosine and sodium tungstate, which resulted in the formation of three new covalent bonds: one between the Nɛ of PKCI His112 or FHIT His96 and the tungstate ion in the α phosphate position (bond length less than ∼2.5 Å), one between the two tungstate ions, and one between the tungstate ion in the α phosphate position and the 5′-OH of the adenosine base. Two oxygen atoms are eliminated in the process. The tungstate ion in the α phosphate position is pentacovalent and has trigonal bipyramidal geometry (Fig.4), thus mimicking the transition state for the α phosphate inversion during the first step of catalysis (12). As in the FHIT AMP-CP complex, the tungstate ion in the β phosphate position was disordered in the corresponding FHIT structure.

Figure 4

Stereo view of the 1.8 Å omit map (blue at 1.0σ) and Bijvöet difference map (red at 3.5σ) superimposed on the final PKCI substrate analog adenosine-tungstate structure. Note the trigonal bipyramidal geometry of the tungstate ion in the α phosphate position.

The product-bound structures (Figs. 2D and 3D) of PKCI and FHIT show characteristics of the final catalytic step. Although there are no gross differences in binding of the adenosine base among the three complexes of PKCI, there are differences in the binding of the α and β phosphate groups. In the 1.5 Å PKCI-AMP structure, the distance between Nɛ of His112 and the phosphorus atom in the α phosphate position is >3.5 Å. The orientation and binding of the oxygen atoms in the α phosphate position indicate interactions alternative to those observed in the previous two structures. Three structurally conserved water molecules (1W to 3W), located between the β strands behind the active-site histidines, are conserved in all FHIT and PKCI structures. An additional water within hydrogen-bonding distance to a phosphate oxygen, 4W, is found only in product (AMP) complexes of PKCI and FHIT.

This comprehensive structural and biochemical study has identified essential elements of the HIT protein catalytic cycle, leading us to propose a general mechanism for the HIT protein family (Fig. 5). One step of the catalytic cycle is missing from our structural study, that of the tetracovalent nucleotidyl intermediate (Fig. 5C). Thus, groups involved in activation of a second nucleophile for release of the product from the covalent histidyl complex are not readily apparent. However, the trapped covalent nucleotidyl phosphohistidyl intermediate (Fig. 1) serves as evidence of this catalytic step.

Figure 5

Catalytic mechanism proposed for FHIT, PKCI, and (by analogy) all HIT family members. The α and β phosphate positions are depicted fully in the ligand. The positions of residues covalently attached to the phosphates are indicated by N, R, and R′, where N is the nucleoside, R is unknown for PKCI and known for several substrates for FHIT (Table 1), and R′ is a hydrogen in the hydrolysis reactions described for PKCI and FHIT (however, as observed for GalT, R′ may be another group in transferase reactions). (A) Substrate-bound form when the histidine acts as a nucleophile in an in-line attack on the α phosphate (AMP-CP structures). (B) Trigonal bipyramidal pentacovalent transition state of the inversion of the α phosphate position (adenosine-tungstate transition-state analog structures). (C) Covalently associated nucleotidyl phosphoprotein reaction intermediate (Fig. 1). (D) The product-bound form of the protein (AMP and adenosine-sulfate product-bound structures).

FHIT and PKCI bind ligand in a deep cleft in each protomer, the floor of which is formed from the five-stranded β sheet. Two structurally conserved loops located between the β strands of the floor are involved in direct contacts with the ligand. One loop is involved in side chain and backbone contacts with the phosphates of the ligand, and the other loop is involved in side chain contacts with the ribose and base of the ligand. The third side of the cleft is involved in contacts with the adenosine base and is formed by NH2-terminal strands in the FHIT structure and by an NH2-terminal helix in PKCI. The highly conserved COOH-terminal residues of PKCI mediate several PKCI-ligand interactions and form the fourth side of the cleft, a feature apparently absent in the FHIT structures. A loop between the last strand and helix of FHIT, although mostly disordered in the crystal structures, could be making PKCI-like contacts with the ligand during catalysis; such contacts are suggested by a comparison of the FHIT structures (2).

The highest degree of amino acid conservation between PKCI and FHIT is observed in the ligand-binding pocket and notably includes the conserved histidyl residues, two of which make direct contacts with the ligand α phosphate group. Although the nucleotide ribose and base occupy similar positions in PKCI and FHIT, few interactions observed in these complexes are mediated by conserved amino acid side chains. Analysis of the liganded states of these enzymes indicates that hydrogen-bonding distances between the protein and oxygen atoms in the α phosphate position are closer (by an average of ∼0.3 Å, Luzzati coordinate error of 0.23) in the transition-state analog complexes (Figs. 2 and 3). The presumably stronger interactions could stabilize the inversion of the α phosphate when it passes through the pentacovalent transition state during catalysis.

FHIT histidyl residues, which face into a more open solvent-accessible cleft, are found to flip in response to the liganded state of the enzyme, possibly affecting the rate of hydrolysis and the determination of the unique substrate specificity for these two human HIT proteins. Despite this difference, catalysis at the α phosphate position appears to proceed in a similar manner in the two cases.

A recent analysis of the Protein Data Bank (PDB) (13) revealed that the PKCI dimer was structurally similar to a protomer of galactose-1-phosphate uridylyltransferase (14), also known as GalT. Subsequent reports also described features of this similarity (2, 5). The active site of GalT is also similar, but not identical, to that in HIT proteins. GalT catalyzes a phosphotransferase reaction, mediated through a covalent nucleotidyl phosphohistidyl intermediate, which involves the exchange of a uridine 5′-monophosphate moiety between the hexose-1-phosphates of glucose and galactose. This implies a functional as well as structural similarity to the HIT proteins.

Our results suggest that distant members within the HIT superfamily share a common catalytic mechanism; however, substrate specificity is likely dictated by the extent of divergence in the COOH-terminal residues, which vary greatly between HIT subfamilies. The catalytic activity observed in vitro is consistent with the ability of these enzymes to hydrolyze their substrates, whereas the covalent enzyme intermediates reported here suggest an alternative nucleotidyl phosphotransferase activity. Our studies support a mechanism by which the α phosphate of respective substrates undergoes an inversion through a trigonal bipyramidal pentacovalent transition state, thus resulting in a transient, covalently attached nucleotidyl protein intermediate before hydrolysis or transfer of the phosphoramidate bond. Although the in vivo substrates for both FHIT and PKCI remain undiscovered, the highly conserved and ubiquitous nature of the PKCI HIT subfamily throughout Archae, Prokaryae, and Eukaryae suggests a fundamental role for these closely related enzymes. The comparative structural biology approach undertaken here has identified an adenosine diphospho component of the ligand and identified differences between FHIT and PKCI that likely contribute to their unique activities. Similar structure-based approaches will become increasingly useful in the identification of protein function and mechanism as other protein families are revealed through the application of genomics.


Stay Connected to Science

Navigate This Article