Crystal structure of the catalytic domain of Haspin, an atypical kinase implicated in chromatin organization

Results and Discussion

Haspin Has High Affinity for Its Substrates.

We bacterially expressed and purified several fragments of the catalytic domain of Haspin. A proteolytically resistant fragment, encompassing residues 452–798 (Haspin^452–798), was chosen for an in vitro kinase assay with histone H3 as substrate. This segment also undergoes significant autophosphorylation (see SI Results and Fig. S1 A and B). Kinetic analysis of the phosphorylation reactions employing a NADH-coupled photometric assay revealed that the K_M of Haspin^452–798 for ATP is 180 μM (Fig. 1B). Previously, it has been reported that full-length Haspin has a K_M for histone H3 of 98 nM (15). Although limitations in sensitivity of our assay prevented us from measuring a K_M for histone H3, we were able to examine the binding of Haspin^452–798 to an immobilized substrate, the N-terminal segment of histone H3 (H3^1–50-GST), with surface plasmon resonance (SPR). The Haspin kinase domain bound to the immobilized histone H3 with an observed K_d of 0.7 μM (Fig. 1C).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Haspin^452–798 kinase domain catalytic parameters. (A) A scheme of human Haspin is shown below a disorder probably plot calculated with PrDOS (24). The red dotted line indicates 50% probability. (B Upper) NADH oxidation was monitored as a function of time, converted to nanomoles of ATP consumed and plotted against time (minutes) at different ATP concentrations. (Lower) Velocities were plotted against ATP concentration. (C Upper) BIAcore sensorgrams for the binding of Haspin^452–798 to H3^1–50-GST immobilized on a glutathione-coated chip. Recombinant Haspin^452–798 was applied to H3^1–50 at various concentrations. (Lower) Analysis of equilibrium dissociation constants, with response units at equilibrium (RU eq) plotted as a function of Haspin concentration. RU eq is derived from the steady-state plateau of the association phases of the sensorgrams. Equilibrium dissociation constants were determined with BIAevaluation 4.0.1 (Biacore).

Overall Description of the Haspin Catalytic Domain.

We determined the structure of Haspin^452–798 by the single anomalous dispersion (SAD) method from crystals of a selenomethionine-containing derivative (Table 1). The structure reveals the classical bilobate ePK fold, with N- and C-lobes, and a catalytic cleft between them containing residues involved in catalysis (Fig. 2). Despite overall low sequence identity, most critical ePK residues implicated in Mg²⁺ and nucleotide binding, and additional residues required for catalysis, are conserved in Haspin. On the other hand, the structure reveals at least four main Haspin-specific features. First, the N-lobe is entirely buried under an additional layer created by an N-terminal extension and two conserved insertions (described below). Second, the overall organization of the activation segment and of its surroundings is unconventional and contributes to creating an unusual substrate-binding site. Third, there is a conserved 20-residue insertion between the β7 and β8 loop. Fourth, a deletion in the large lobe removes the αG helix, which in other ePKs occupies a prominent position on the front face of the domain.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Overall structure of Haspin^452–798. (A) Multiple sequence alignment of Haspin-like kinases. Secondary structure elements and degree of conservation are indicated. Three insertions described in the text are marked by a red bar. (B) Cartoon representation of two rotated views of Haspin^452–798 showing N-terminal extension (orange), β3-αC loop (green), αC (blue), αC′ (dark red), activation segment (red), and β7′-β7′ hairpin (yellow). (C) The additional layer on the N-lobe is visualized after a ≈45° rotation from the view in B. Side chains of residues making critical interactions as described in the text are displayed. (D) p21 activated kinase (Pak1) (pdb:1YHV) was colored as in B to emphasize the presence of an N-terminal extension analogous to that observed in Haspin (17). Note that the additional insertions are absent in Pak1.

Reorganization of the N-Lobe.

The core of the N-lobe of Haspin consists of a twisted 5-stranded anti-parallel β-sheet and the αC helix (Fig. 2B). A specific feature of Haspin is that the core of the N-lobe is entirely buried under an additional layer built from three distinct structural elements. The first element is an N-terminal extension (residues 452–487) that precedes the glycine-rich loop (β1–β2). The first part of the extension folds as a α helix (αA, residues 452–467). It surmounts the αC helix and contributes to its stabilization (see below). Before connecting to the β1 strand, the extension meanders irregularly on the surface at the far end of the N-lobe, forming two additional short helical segments (αA′ and αA″). Conserved residues in this region, including Gln-469, Phe-475, and Leu-477, contribute to its stabilization with hydrogen bonds (H-bonds) and hydrophobic interactions (Fig. 2C).

Two additional elements are inserted between β3 and αC (insertion 1, residues 515–528) and between β4 and β5 (insertion 2, residues 567–598). Insertion 1 forms an irregular but rigid structure that replaces the short αB helix often observed in other ePK family members such as PKA (Fig. 2C). Insertion 2 is also rigid. Its first part (residues 571–584) folds as an α helix (named αC′) and is locked in place by interactions with the ceiling of the core of the N-lobe mediated by the side chains of Tyr-569 and Pro-570 (Fig. 2C). The side chain of Trp-577 on the αC′ helix buries a tripartite interaction involving the underlying side chains of Lys-489, Glu-492, and Glu-497 in the Gly-rich loop (Fig. 2C). A charged interaction analogous to the Lys-489–Glu-497 interaction is important for catalysis in several ePK family members, probably due to stabilization of the Gly-rich loop (16). Being surmounted by insertion 2, the Gly-rich loop is further stabilized and is therefore highly ordered (average temperature factor of 48.5 Ų against an average temperature factor of 49.2 Ų for the entire model). The second part of insertion 2 meanders on the surface of the N-lobe, contributing with the side chains of Phe-593 and Phe-594 to a hydrophobic core that also includes Val-521 from insertion 1 and several aliphatic and aromatic side chains from the top surface of the β-sheet of the N-lobe (Fig. 2C). Together, the N-terminal extension and insertions 1 and 2 form a continuous system and they reinforce each other through additional interactions, most notably between the side chains of Gln-526 and Gln-598.

Haspin Adopts the Active Conformation.

The structure of Haspin^452–798 likely depicts the catalytically active conformation. The side chains of Lys-511 and Glu-535 (in the β3-strand and αC helix, respectively), two conserved residues whose side-chains are almost invariably in close contact in active kinases but not in inactive kinases, are only 2.7 Å away (Fig. 3A), indicating that the αC helix (on which Glu-535 is located), which is usually a very mobile element of kinases, has adopted a position compatible with catalysis.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Role of αA helix in Haspin activation. (A) A representation of the hydrophobic environment around αC (blue) reveals stabilization by the activation segment (red) and by the αA helix in the N-terminal extension (orange). (B) Different Haspin variants were separated by SDS/PAGE and stained with Coomassie. (C) Indicated Haspin variants were tested to determine the ability to phosphorylate histone H3 in vitro in the presence of radiolabeled ATP. The reactions were then separated by SDS/PAGE and visualized by autoradiography.

Two distinct sets of interactions, the first with residues in the αA helix (N-terminal extension) and the second with a hydrophobic cluster from the activation segment, contribute to the stabilization of the active conformation of the αC helix (Fig. 3A). To dissect the role of the αA helix, we tested the relative kinase activity of N-terminal variants of the Haspin kinase domain. In thermal denaturation experiments, all variants showed similar protein stability (Fig. S2). Haspin^406–798 and Haspin^452–798, both containing the αA helix, efficiently phosphorylated histone H3. Conversely, constructs lacking the αA helix (Haspin^468–798) were significantly less active (Fig. 3 B and C). A similar decrease in activity was observed when Val-463, a residue at the hydrophobic interface between the αA and αC helices, was replaced with Asn (Haspin^452–798 V463N) (Fig. 3). In contrast, abolishing the salt bridge between Arg-695 and Glu-466 (Haspin^452–798 E466A), which connects the activation segment to the αA helix, did not affect significantly the ability of Haspin to phosphorylate histone H3, indicating that this feature is not essential for kinase activity (Fig. 3). In an accompanying paper, Knapp and colleagues report the independent structure determination of a construct of human Haspin (465–798) lacking the αA helix (PDB ID 2VUW). Haspin^452–798 and Haspin^465–798 are structurally very similar (Fig. S3), suggesting that the differences in activity, reported in Fig. 3 B and C, are associated with modest structural changes (see SI Discussion).

Other kinases contain an αA helix preceding the small lobe. In p21-activated kinase 1 (Pak1, PDB ID code 1yhv) (17), whose structure is rather closely related to Haspin's (overall rmsd for 198 C_α atoms is 3 Å, Fig. 2D), the N-terminal αA helix packs against the surface formed by the αC helix and the β4–β5 strands. An equivalent helix (αL₁₆, residues 338–353) is also present in the structure of the extracellular signal-regulated kinase 2 kinase domain (18). In other kinases, such as the CDKs, the αC helix is pushed into the active conformation by equivalent interactions with cyclins and cyclin-like proteins (19).

Conformation of the Activation Segment.

The second block of interactions keeping the Haspin αC in place implicates residues in the activation segment. The tip of the activation segment (residues 692–700) forms a β-hairpin (β9–β9′) (Figs. 2–4). Leu-693 and Val-700 in the β9–β9′ hairpin contact a hydrophobic cluster of the αC helix containing Pro-534, Ile-537, and Ile-538 (Fig. 3A). Overall, the Haspin activation segment is shorter than in classical ePK folds, such as PKAs, and its sequence and conformation are uncommon (Fig. 4A). For instance, the segment starts with an unusual 687-Asp-Tyr-Thr-689 (DYT) motif, rather than with the usual Asp-Phe-Gly (DFG) motif (Figs. 2A and 4A). However, the side-chain conformation of the Asp and Tyr residues of Haspin are similar to the ones observed in typical DFG motifs in active kinases. Additionally, the Ala-Pro-Glu (APE) motif that normally closes the segment is absent (Fig. 4A).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Conformation of Haspin's activation segment. (A) Multiple sequence alignment of activation segments in the indicated kinases. The secondary structure of the loop in Haspin452–798 is Above the alignment, whereas the usual succession of structural elements in prototype kinases such as PKA is shown Below the alignment. (B) The activation segment and αEF helix in Haspin (Left) and PKA (Right). The αE and αF helices are shown as a reference as they superimpose very well in the two structures. The activation segment of Haspin is extruded into solvent due to several aromatic residues contributed by the segment itself and by the αEF helix. (C) Arg-648 in the HRD motif of Haspin (Left) coordinates the main chain of the αEF–αF loop, Right, Below the activation segment. In the phospho-CDK2/cyclin A complex (Right), the equivalent arginine is implicated in activation loop stabilization by binding to the activation loop phosphate (25).

After the β9–β9′ hairpin, the chain is extruded into solvent due to the presence of the bulky αEF helix (residues 716–729), which runs almost parallel to the αF helix (Fig. 4B). The chain then turns again and connects to αF. This region, which roughly corresponds to the P+1 loop and the aEF helix of other kinases (see Fig. 4A), is completely reorganized. Several conserved aromatic and hydrophobic residues, including Val-704, Leu-710, Phe-711, Phe-719, Tyr-722, and Trp-733 contribute to a hydrophobic core that, together with additional hydrophilic interactions with the core of the kinase, stabilizes the unusual conformation of this region. Remarkably, the side chain of Arg-648 in the His-Arg-Asp (HRD) motif, which in those kinases containing the HRD motif is usually implicated in the stabilization of the phosphorylated activation segment (see Fig. 4C for an example), in Haspin is H-bonded to the main chain carbonyl groups of Trp-733 and Glu-735, thus contributing to shape the αEF–αF loop (Fig. 4C).

Additional Structural Features.

A characteristic element of Haspin is the presence of a 20-residue insertion extending the β7–β8 loop (insertion 3). Insertion 3 contains two additional β-strands, β7′ and β7″, engaging in a β-hairpin. The hairpin extends toward the back of the catalytic domain, where it makes contacts with the αC–β4 loop, mainly through main-chain interactions. This might explain why the sequence of insertion 3 is poorly conserved within the Haspin family (Fig. 2 A and B). In typical kinases such as PKA (2), the αG helix lies at the front of the C-lobe, beside the activation loop (Fig. 4B), an excellent position to influence substrate recognition and specificity in protein kinases (20, 21). In Haspin, the αG helix is absent, and the chain progresses directly from the end of the αF helix to the beginning of the αH helix (Fig. 2A).

Haspin Binds and Phosphorylates Nucleosomes in Vitro.

In vitro, Haspin^452–798 phosphorylates histone H3. It also phosphorylates a single band, which we identify as histone H3, within purified chicken erythrocyte nucleosomes (Fig. S4). Because Haspin binds a peptide substrate with high affinity (Fig. 1), we asked whether Haspin physically interacts with nucleosomes. When loaded onto a Superdex-200 size-exclusion chromatography (SEC) column, Haspin^452–798 and nucleosomes eluted as expected for globular structures of 35 kDa and 280 kDa, respectively (Fig. S4). When Haspin^452–798 and nucleosomes were combined stoichiometrically and then analyzed by SEC, Haspin^452–798 coeluted with nucleosomes in a stoichiometric or near-stoichiometric complex, suggesting high affinity binding (Fig. 5A). When Haspin^452–798 and nucleosomes were combined with Mg²⁺ and ATP, we only observed a partial release of Haspin from nucleosomes, possibly suggesting that Haspin retains the ability to interact with a phosphorylated substrate, albeit at lower affinity (Fig. 5B). To corroborate this idea, we prephosphorylated H3^1–50-GST (P-H3^1–50-GST) (Fig. 5C). Haspin^452–798 (and a catalytically impaired version of the same kinase) bound to H3^1–30-GST, H3^1–50-GST, and P-H3^1–50-GST (Fig. 5D). Using SPR, we determined that Haspin^452–798 binds to P-H3^1–50-GST with a K_d of 7 μM, a 10-fold decrease in binding affinity relative to the nonphosphorylated substrate (Figs. 5E and 6E).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Haspin binds nucleosomes in vitro. (A) Haspin kinase and nucleosomes were mixed, separated by SEC, and analyzed by SDS/PAGE. A stoichiometric amount of Haspin coelutes with nucleosomes, indicative of complex formation. (B) Haspin kinase and nucleosomes were mixed with Mg²⁺ and ATP, and later separated by SEC and analyzed by SDS/PAGE. (C) H3^1–50-GST was phosphorylated in vitro in a cold kinase assay with Haspin^452–798 (see Methods). Incorporation of phosphate on Thr-3 is demonstrated by an anti-P-Thr-3 monoclonal antibody. The preparation was subjected to a further step of phosphorylation in the presence of γ-[³²]ATP. No radioactivity was incorporated in the prephosphorylated sample, indicating that it has been phosphorylated to completion in the first round with cold ATP. (D) GST, H3^1–30-GST, H3^1–50-GST, or P-H3^1–50-GST on GSH beads were incubated with the Haspin^452–798 or Haspin^452–798-D687A (KD). After a washing step, SDS/PAGE and Coomassie staining were used to visualize bound species. (E) Binding of Haspin^452–798 to immobilized H3^1–50-GST. Response units at equilibrium (RUeq) were measured on a Biacore device and plotted as a function of protein-kinase concentration. K_d values were calculated by measuring association (k_a) and dissociation (k_a) rate constants from Biacore sensorgrams. K_d values were calculated as K_d = k_d/k_a (see Fig. 6E).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 6.

The substrate-binding pocket. (A) Electrostatic potential mapped onto the molecular surface of the Haspin kinase domain. Red: negative; blue: positive. (B) Sequence conservation was mapped onto the molecular surface. (C) Analysis of equilibrium dissociation constants for Haspin kinase wild type (WT) and indicated mutants. Response units at equilibrium (RUeq) are plotted as a function of protein-kinase concentration. (D) Surface representation of the residues tested in this study for substrate binding properties. (E) K_d, k_a, and k_d values for several Biacore binding experiments involving Haspin^452–798 and its mutants as mobile ligands, and H3^1–50-GST or P-H3^1–50-GST as an immobilized receptor.

Identification of Residues Implicated in Substrate Recognition.

Besides autocatalysis (Fig. S1), the only known substrate of Haspin is Thr-3 of histone H3. Thr-3 lies within the conserved sequence 1-Ala-Arg-Thr-Lys-Gln-5 (10). Arg-2 and Lys-4 are both required for efficient H3 binding and phosphorylation by Haspin (10). As expected for a positively charged substrate, there is a negative charge distribution around the catalytic cleft (Fig. 6A), which is lined by residues that are highly conserved in the Haspin family (Fig. 6B). Because we have been unable to cocrystallize Haspin with a peptide substrate, we applied mutational analysis to identify Haspin residues involved in H3 substrate recognition. Wild type (WT), catalytically inactive (D687A), and different point mutants of Haspin (Fig. S5A) were tested for their ability to phosphorylate histone H3 in vitro (Fig. S5B). Because Haspin has a low K_M (15) and a low K_d for histone H3 (Fig. 1), we used histone H3 at a concentration of 100 nM in our reactions to maximize the likelihood of observing an effect of the mutations on kinase activity. In these qualitative experiments, substitution of Glu-492, His-651, Asp-707, Asp-709, Gly-713, and Asp-716 with other residues markedly affected histone H3 phosphorylation (Fig. S5B). To gain a quantitative understanding of the effects of these mutations, we investigated their effect on the enzymatic parameters of Haspin. We determined catalytic efficiency (k_cat) and K_M for ATP in a kinase assay in which the H3 substrate was fixed at a concentration of 5 μM. Three mutants, D611L, H651A, and D716L, also had reduced k_cat, suggesting either a role in catalysis or reduced stability of the enzyme. Besides a reduced k_cat, D716L also had an increased K_M for ATP (Fig. S5C). At least four mutants, E492A, D707L, D709N, and G713F, displayed normal k_cat and K_M for ATP (Fig. S5) but reduced affinity for H3 (Fig. 6 C–E), suggesting that they are important for substrate recognition. Consistently, all four mutants were severely impaired in H3 phosphorylation when assayed at nonsaturating concentrations of substrate (Fig. S5B). The mutated residues line the catalytic pocket of Haspin (Fig. 6D). His-651, identified above for having a reduced catalytic efficiency, also displayed a severely reduced affinity for the substrate (Fig. 6 C–E). Thus, mutations in this residue appear to affect both substrate recognition and catalytic efficiency. This result is consistent with the position of His-651, whose side chain flanks the side chain of the catalytic Asp-649 (Fig. 6D).