The pseudokinase MLKL (mixed lineage kinase domain-like) was identified recently as an essential checkpoint in the programmed necrosis or ‘necroptosis’ cell death pathway. In the present study, we report the crystal structure of the human MLKL pseudokinase domain at 1.7 Å (1 Å=0.1 nm) resolution and probe its nucleotide-binding mechanism by performing structure-based mutagenesis. By comparing the structures and nucleotide-binding determinants of human and mouse MLKL orthologues, the present study provides insights into the evolution of nucleotide-binding mechanisms among pseudokinases and their mechanistic divergence from conventional catalytically active protein kinases.

INTRODUCTION

The cell death signalling pathway, programmed necrosis or ‘necroptosis’, has emerged in the past decade as a pathway that functions in place of, or parallel to, apoptosis in scenarios when caspase activity has been compromised. The physiological role of necroptosis remains unknown, but it is clear that it is involved in the pathology of inflammatory diseases [14]. Constituents of the necroptosis pathway are still being identified and, in 2012, the pseudokinase MLKL (mixed lineage kinase domain-like) was first implicated in necroptotic cell death [5,6]. Subsequent studies using cells derived from mice genetically deleted for Mlkl have unequivocally established MLKL as an essential checkpoint protein in necroptotic signalling [7,8]. We recently described MLKL as a ‘molecular switch’ that is triggered by phosphorylation of the necroptotic protein kinase RIPK3 (receptor-interacting protein kinase 3) [7], a key effector that operates downstream of TNF (tumour necrosis factor) receptor 1 or TLR (Toll-like receptor) activation [79]. The molecular details of how MLKL is activated by RIPK3 phosphorylation or pseudoactive site mutation and how such events might instigate cell death remain unclear. Previous studies had implicated the mitochondrial proteins PGAM5 (phosphoglycerate mutase family member 5) and Drp1 (dynamin-related protein 1), as effectors activated by MLKL [10]; however, reduction in the levels of these proteins does not always block necroptosis, suggesting that additional or alternate as yet undiscovered effectors of necroptosis exist [7] (D. Moujalled, W.D. Cook, J.M. Murphy and D.L. Vaux, unpublished work).

MLKL has been shown to be composed of an N-terminal four-helix bundle domain tethered to the C-terminal pseudokinase domain by a two helix linker [7]. Pseudokinase domains are so-named because they resemble protein kinase domains topo-logically, but lack one of three essential catalytic residues: the VAIK lysine from the β3 strand, known to position the α- and β-phosphates of ATP during phosphoryl transfer; the HRD aspartic acid in the catalytic loop, known as the catalytic residue; and the DFG aspartic acid in the activation loop, which binds the divalent cations that are crucial to catalysis [1214]. The absence of any one of these motifs is sufficient to compromise phosphoryl transfer activity, but the consequences are typically more severe. For example, 18 of the 31 pseudokinases examined in a recent study did not detectably bind ATP, ADP, AMP, GTP or the ATP analogue AMP-PNP (adenosine 5′-[β,γ-imido]triphosphate; also known as p[NH]ppA) [15].

We recently reported the crystal structure of full-length mouse MLKL, which not only lacks two of the three essential catalytic residues, but also contains an atypical helix within the activation loop that contributes to unusual ‘pseudoactive’ site geometry. Despite the absence of key catalytic residues and the novel geometry, mouse MLKL robustly bound to ATP, ADP or the ATP analogue AMP-PNP, albeit only in the absence of divalent cations [7]. Intriguingly, human MLKL exhibited similar nucleotide-binding propensities, even though the sequences of mouse and human MLKL pseudokinase domains have diverged markedly in their pseudoactive site motifs, especially in the glycine-rich and catalytic loops [15]. These observations led us to investigate the hypothesis that, unlike conventional protein kinases, pseudokinases are not subjected to selective pressure to maintain catalytic activity, thereby permitting residues surrounding the nucleotide-binding cleft to evolve into non-canonical nucleotide-binding motifs.

In the present paper, we report the X-ray crystal structure of the human MLKL pseudokinase domain solved to a resolution of 1.7 Å (1 Å=0.1 nm) and, in conjunction with nucleotide-binding studies of structure-based mutants, we define the key nucleotide-binding determinants within the human MLKL pseudoactive site. In particular, we identified Lys331 within the HGK sequence of the human MLKL catalytic loop as a key nucleotide-binding residue, although this sequence differs markedly from the catalytic motif, HRD, found in conventional protein kinases and the HRN sequence present in mouse MLKL. We also show that similar to in mouse MLKL, the lysine residue in the VAIK motif (typically thought of as a catalytic rather than nucleotide-binding residue in conventional protein kinases) is a key nucleotide-binding residue in human MLKL. These data illustrate that not only have pseudokinases evolved non-canonical mechanisms of nucleotide binding relative to their catalytically active protein kinase counterparts, but also evolution between orthologues has imparted plasticity in nucleotide-binding mechanisms that could not be rationalized in the absence of their crystal structures.

EXPERIMENTAL

Recombinant protein expression and purification

A synthetic cDNA encoding human MLKL was purchased from DNA2.0 and used as a template for PCR amplification of the sequence encoding residues 190–471. Mutations were introduced into human MLKL using oligonucleotide-directed PCR mutagenesis, cloned into pFastBac HTb (Life Technologies) to generate constructs encoding an in-frame TEV (tobacco etch virus) protease cleavable N-terminal His6 tag. The sequences of all inserts were verified by Sanger sequencing (Micromon Facility), before bacmid and virus preparation according to established protocols [16,17]. Sf21 cells were infected with the virus and harvested after 48 h, as described previously [15,18], before resuspension in lysis buffer {0.5 M NaCl, 20 mM Tris (pH 8), 20% (v/v) glycerol, 10 mM imidazole (pH 8) and 0.5 mM TCEP [tris-(2-carboxyethyl)phosphine]}, supplemented with Complete™ EDTA-free protease inhibitor cocktail (Roche) and 1 mM PMSF, and lysed by sonication. The supernatant was clarified by centrifugation at 45000 g at 4°C for 30 min, before incubation with HisTag (Roche) Ni-NTA (Ni2+-nitrilotriacetate) resin at 4°C for >1 h with gentle rotation. The resin was collected by centrifugation, washed extensively with lysis buffer and wash buffer [0.5 M NaCl, 20 mM Tris (pH 8), 20% (v/v) glycerol, 40 mM imidazole (pH 8) and 0.5 mM TCEP], before the protein was eluted in three rounds of 5 ml of elution buffer [0.5 M NaCl, 20 mM Tris (pH 8), 20% (v/v) glycerol, 250 mM imidazole (pH 8) and 0.5 mM TCEP]. The proteins used for thermal shift assays were concentrated by centrifugal ultrafiltration and subjected to Superdex 200 10/300 gel filtration chromatography (GE Healthcare) in a buffer comprising 200 mM NaCl, 20 mM Tris, 10% (v/v) glycerol and 0.5 mM TCEP (pH 8.0). Wild-type MLKL was further purified for crystallization trials by cleaving the His6 tag by incubation with 200 μg of TEV protease per 5 mg of MLKL at 22°C for 2 h followed by dialysis at 4°C and further nickel-chromatography to eliminate uncut material and the TEV protease, according to previous protocols [19]. Detagged MLKL was concentrated by centrifugal ultrafiltration and applied to a Superdex 200 10/300 gel filtration column (GE Healthcare) with elution of 0.4 ml fractions in 200 mM NaCl, 20 mM Hepes and 5% (v/v) glycerol (pH 7.5). Fractions containing human MLKL were concentrated to 8–10 mg/ml by centrifugal ultrafiltration, divided into 0.1–0.3 ml aliquots and snap frozen in liquid N2 before storage at −80°C until required. The protein concentration was estimated on the basis of absorbance at 280 nm and a molar absorption coefficient calculated on the basis of the protein sequence.

Crystallization of wild-type human MLKL pseudokinase domain

Human MLKL pseudokinase domain (150 nl at 5 mg/ml) was mixed with 150 nl of reservoir solution [25% (w/v) pentaerythritol propoxylate and 0.1 M Tris/HCl (pH 8.5)] in sitting drops (C3 Facility, CSIRO). Crystals initially grew after 3 days at 20°C and were harvested at 15 days by flash freezing in a cryobuffer composed of 30% (w/v) pentaerythritol propoxylate and 0.1 M Tris/HCl (pH 8.5).

Data collection and structure determination

X-ray diffraction data of human MLKL pseudokinase domain crystals were collected at the MX2 beamline at the Australian Synchrotron at 100 K. Data were processed with XDS [20] and the structure solved by molecular replacement with PHASER [21] using the pseudokinase domain from mouse MLKL (PDB code 4BTF [7]) as a search model. Subsequent rounds of building in COOT [22] and auto building and refinement in Phenix [23] yielded the final model. Structures were checked and validated with MolProbity [24] and using validation tools in COOT [22]. Summary statistics are shown in Table 1. All structure cartoons were drawn using PyMOL (http://www.pymol.org/).

Table 1
Data collection and refinement statistics for the human MLKL pseudokinase domain crystal structure

Statistics for the highest-resolution shell are shown in parentheses. Ramachandran and Clashscore statistics were obtained using MolProbity [24].

ParameterHuman MLKL (PDB code 4MWI)
Data collection  
 Wavelength (Å) 0.953700 
 Resolution range (Å) 50–1.70 (1.76–1.70) 
 Space group C2221 
 Unit cell (Å) 72.1, 74.7, 127.6 
 Total reflections 269586 (19272) 
 Unique reflections 38267 (3763) 
 Multiplicity 7.0 (5.1) 
 Completeness (%) 99.89 (99.50) 
 Mean II 21.87 (1.94) 
 Wilson B-factor 23.24 
Rmerge 0.0571 (0.884) 
 CC (1/2) 100 (65.9) 
Refinement  
Rwork 0.1646 (0.2627) 
Rfree 0.1944 (0.2883) 
 Number of atoms 2537 
  Macromolecules 2173 
  Ligands 31 
  Water 325 
 Protein residues 266 
 RMSD (bonds) (Å) 0.006 
 RMSD (angles) (°) 0.97 
 Ramachandran favoured (%) 99 
 Ramachandran outliers (%) 
 Clashscore 2.26 
 Average B-factor 26.80 
  Macromolecules 24.70 
  Ligands 65.60 
  Solvent 36.70 
ParameterHuman MLKL (PDB code 4MWI)
Data collection  
 Wavelength (Å) 0.953700 
 Resolution range (Å) 50–1.70 (1.76–1.70) 
 Space group C2221 
 Unit cell (Å) 72.1, 74.7, 127.6 
 Total reflections 269586 (19272) 
 Unique reflections 38267 (3763) 
 Multiplicity 7.0 (5.1) 
 Completeness (%) 99.89 (99.50) 
 Mean II 21.87 (1.94) 
 Wilson B-factor 23.24 
Rmerge 0.0571 (0.884) 
 CC (1/2) 100 (65.9) 
Refinement  
Rwork 0.1646 (0.2627) 
Rfree 0.1944 (0.2883) 
 Number of atoms 2537 
  Macromolecules 2173 
  Ligands 31 
  Water 325 
 Protein residues 266 
 RMSD (bonds) (Å) 0.006 
 RMSD (angles) (°) 0.97 
 Ramachandran favoured (%) 99 
 Ramachandran outliers (%) 
 Clashscore 2.26 
 Average B-factor 26.80 
  Macromolecules 24.70 
  Ligands 65.60 
  Solvent 36.70 

SAXS (small-angle X-ray scattering)

SAXS data collection was performed at the Australian Synchrotron SAXS/WAXS (wide-angle X-ray scattering) beamline [25] using an in-line gel filtration chromatography setup, essentially as described previously [17,26]. Summary statistics for data collection are presented in Table 2. Wild-type human MLKL pseudokinase domain (50 μl at 9.8 mg/ml) was injected on to an in-line Superdex 200 5/150 column (GE Healthcare) pre-equilibrated with 200 mM NaCl, 20 mM Hepes (pH 7.5) and 5% (v/v) glycerol, and eluted via a 1.5-mm glass capillary positioned in the X-ray beam. Scattering data were collected in 2 s exposures over the course of the elution and 2D intensity plots from the peak of the sizing exclusion chromatography run were radially averaged, normalized to sample transmission, with scattering profiles from earlier in the size exclusion chromatography elution averaged, and used to perform background subtraction of 1D profiles using the using the scatterBrain software (Stephen Mudie, Australian Synchrotron). Guinier analysis of each profile across the single elution peak showed a consistent Rg (radius of gyration), and a total of 12 scatter profiles were averaged using scatterBrain to generate the data presented in Supplementary Figures S1B–S1D (at http://www.biochemj.org/bj/457/bj4570369add.htm). Guinier analysis of data was performed using PRIMUS [27]; and indirect Fourier transform with GNOM [28] to obtain the distance distribution function P(r) and the maximum dimension of the scattering particle Dmax. Theoretical scattering curves were calculated from crystal structure atomic co-ordinates and compared with experimental scattering curves using CRYSOL [29].

Table 2
SAXS parameters and data analyses

q is the magnitude of the scattering vector which is related to the scattering angle (2θ) and the wavelength (λ) as follows: q=(4π/λ)sinθ. The theoretical envelope volume calculated from the crystal structure using CRYSOL. The experimental envelope volume estimated from scattering data using CRYSOL.

Parameter Value
Data collection parameter  
 Instrument Australian Synchrotron SAXS/WAXS beamline [25
 Beam geometry wavelength (Å) 120 micron point source, 1.033 Å 
 q range (Å−10.0110–0.400 Å−1 
 Exposure time 2 s exposures 
 Protein concentration 9.8 mg/ml wild-type human MLKL pseudokinase domain injected on to in-line size exclusion chromatography 
 Temperature 15°C 
Structural parameters  
 I(0) (cm−1) [from P(r)] 0.01926±0.00011 
 Rg (Å) [from P(r)] 24.55±0.22 
 Dmax (Å) 85 
 I(0) (cm−1) (from Guinier) 0.01919±0.00014 
 Rg (Å) (from Guinier) 24.10±0.29 
 Theoretical envelope volume (Å−344530 
 Experimental envelope volume (Å−342037 
 Fit of monomer crystal structure to experimental data (from CRYSOL) χ=0.682 
Software employed  
 Primary data reduction scatterBrain (Australian Synchrotron) 
 Data processing PRIMUS [27] and GNOM [28
 Computation of model intensities CRYSOL [29
Parameter Value
Data collection parameter  
 Instrument Australian Synchrotron SAXS/WAXS beamline [25
 Beam geometry wavelength (Å) 120 micron point source, 1.033 Å 
 q range (Å−10.0110–0.400 Å−1 
 Exposure time 2 s exposures 
 Protein concentration 9.8 mg/ml wild-type human MLKL pseudokinase domain injected on to in-line size exclusion chromatography 
 Temperature 15°C 
Structural parameters  
 I(0) (cm−1) [from P(r)] 0.01926±0.00011 
 Rg (Å) [from P(r)] 24.55±0.22 
 Dmax (Å) 85 
 I(0) (cm−1) (from Guinier) 0.01919±0.00014 
 Rg (Å) (from Guinier) 24.10±0.29 
 Theoretical envelope volume (Å−344530 
 Experimental envelope volume (Å−342037 
 Fit of monomer crystal structure to experimental data (from CRYSOL) χ=0.682 
Software employed  
 Primary data reduction scatterBrain (Australian Synchrotron) 
 Data processing PRIMUS [27] and GNOM [28
 Computation of model intensities CRYSOL [29

Thermal shift assay for nucleotide binding

Thermal shift assays were performed as described previously [7,15] using a Corbett real-time PCR machine with proteins diluted in 150 mM NaCl, 20 mM Tris (pH 8.0) and 1 mM DTT to 2–5 μM in a total reaction volume of 25 μl. SYPRO® Orange (Molecular Probes) was used as a probe with fluorescence detected at 530 nm. Nucleotide concentrations of 0.2 mM and divalent cation concentrations of 1 mM were used in each experiment, except in titrations where the ATP concentration was varied from 0 to 0.8 mM. A positive ΔTm value indicates that ligand binds the protein and therefore protects the protein from denaturation. A minimum of two independent assays was performed for each protein and representative data are shown for each.

RESULTS AND DISCUSSION

Human MLKL pseudokinase domain shares the topology of a conventional protein kinase domain

We determined the structure of the human MLKL pseudokinase domain (residues 190–471) in its apo form to 1.7 Å resolution using X-ray crystallography (Table 1 and Figure 1A). Similar to the structure of mouse MLKL [7] (Figure 1B), the human MLKL pseudokinase domain topologically resembles conventional protein kinase domains, which are typified by a bilobal structure comprising a smaller N-lobe, composed of five antiparallel β-strands and one α-helix (termed helix αC), and the larger C-lobe, containing primarily α-helices. In contrast with the mouse MLKL pseudokinase domain [7] (Figure 1B), the human MLKL pseudokinase domain was crystallized in a closed conformation: a disposition typically associated with the catalytically active conformation of a protein kinase [30]. As shown in Figure 1(A), the side chains of Ile265, Met254, Phe350 and His329 (represented as blue sticks from top to bottom) form a hydrophobic regulatory ‘R’-spine synonymous with the structures of catalytically active protein kinases in their closed active conformations [31]. In contrast, despite counterparts of the residues that compose the R-spine in the human MLKL structure being conserved in the mouse sequence, the mouse MLKL pseudokinase domain structure (Figure 1B) does not contain an intact R-spine due to the presence of an atypical helix within its activation loop that buttresses against, and displaces, the helix αC and associated R-spine residues.

The crystal structure of the human MLKL pseudokinase domain resembles that of a catalytically active protein kinase

Figure 1
The crystal structure of the human MLKL pseudokinase domain resembles that of a catalytically active protein kinase

(A) The human MLKL pseudokinase domain crystal structure at 1.7 Å resolution. The side chains of the hydrophobic R-spine residues, Ile265, Met254, Phe350 and His329 (represented as blue sticks from top to bottom). (B) The mouse MLKL pseudokinase domain structure (PDB code 4BTF [7]) shown in the same orientation as human MLKL in (A). (C) Superimposed human (green) and mouse (blue) MLKL pseudokinase domain structures. (D) Magnification of superimposed human and mouse MLKL structures to illustrate the differing activation loop positions. All structural cartoons were prepared using PyMoL. Hs, Homo sapiens; Mm, Mus musculus.

Figure 1
The crystal structure of the human MLKL pseudokinase domain resembles that of a catalytically active protein kinase

(A) The human MLKL pseudokinase domain crystal structure at 1.7 Å resolution. The side chains of the hydrophobic R-spine residues, Ile265, Met254, Phe350 and His329 (represented as blue sticks from top to bottom). (B) The mouse MLKL pseudokinase domain structure (PDB code 4BTF [7]) shown in the same orientation as human MLKL in (A). (C) Superimposed human (green) and mouse (blue) MLKL pseudokinase domain structures. (D) Magnification of superimposed human and mouse MLKL structures to illustrate the differing activation loop positions. All structural cartoons were prepared using PyMoL. Hs, Homo sapiens; Mm, Mus musculus.

The pseudokinase domain structures of human (green) and mouse (blue) MLKL are superimposed in Figure 1(C), with a magnified view of their overlaid pseudoactive sites shown in Figure 1(D). Whereas in the mouse MLKL structure, the region after Gln343 adopts the atypical activation loop helix, the equivalent region of the human MLKL structure (after Gln356) was largely unstructured with no electron density evident for residues 357–368. Intriguingly, the positions of the phenylalanine side chains (Phe350 in human and Phe337 in mouse) of the GFE sequence at the start of the human and mouse activation loops closely overlap (Figure 1D) and only diverge thereafter. As a result, it is conceivable that the structures solved for human (Figures 1A and 2A) and mouse (Figures 1B and 2C) MLKL pseudokinase domains may represent two different conformational states for MLKL, in keeping with the notion that this protein is a highly dynamic molecular switch. Further support for the idea that human MLKL may transition between an active conformation, as reported in the present paper and by others recently [32], and an inactive conformation that contains an α-helix at the start of its activation segment, such as that observed in mouse MLKL structures [7,32], is suggested by previous studies of protein kinases, such as LOK (lymphocyte-oriented kinase), for which structures of active and inactive conformers have been solved [33]. However, contrary to LOK, which undergoes dimerization as a result of activation segment exchange, recombinant human MLKL pseudokinase domain showed no evidence of dimer formation in size exclusion chromatography (Supplementary Figure S1A) and SAXS experiments (Table 2 and Supplementary Figures S1B–S1D). SAXS data were consistent with the existence of human MLKL pseudokinase domain as a monomer in solution, with excellent agreement between the experimental SAXS profile and the theoretical scatter pattern calculated from the monomeric human MLKL structure using CRYSOL [29] (χ=0.682; Table 2 and Supplementary Figure S1B). Nevertheless, how RIPK3-mediated phosphorylation of Thr357 and Ser358, two residues within the unstructured portion of the human MLKL activation loop, leads to propagation of the necroptotic signal and whether MLKL oligomerization underlies this phenomenon is yet to be determined and remains the subject of ongoing investigation.

Closer inspection of the pseudoactive site of human MLKL (Figure 2A) and a comparison with the ATP-bound active site of the prototypic kinase PKA (protein kinase A) (Figure 2B), reveals a number of important features. First, ATP is readily accommodated in the pseudoactive site of human MLKL, with the only potential hindrance arising from the packing of the side chain from the glycine-rich loop residue, Arg210, in the putative adenosine-binding pocket in the structure. Secondly, Lys230 of the VAIK motif forms an ion pair with the helix αC Glu250 (Figure 2A): a bond typical of a protein kinase in a catalytically active conformation as illustrated by the analogous positions of their counterparts in PKA, Lys72 and Glu91 (Figure 2B). Thirdly, the D184FG motif of PKA is replaced by GFE351 in human MLKL. In the human MLKL structure (Figures 2A and 2B), Glu351 is inappropriately positioned to contribute to conventional divalent cation binding: a finding consistent with our observation that, unlike most conventional protein kinases, human MLKL does not require divalent cations for maximal nucleotide engagement. Finally, the position of the human MLKL ‘catalytic’ loop differs markedly from that of PKA. The human MLKL sequence lacks the catalytic aspartic acid usually present within the HRD motif of conventional protein kinases (YRD166 in PKA) and instead contains HGK331. Notably, the position of human MLKL Lys331 does not overlap with Asp166 from PKA, but, as discussed below, the Lys331 side chain is directed towards the putative nucleotide-binding cleft, suggesting a potential functional role.

Divergent evolution within the pseudoactive sites of human and mouse MLKL

Figure 2
Divergent evolution within the pseudoactive sites of human and mouse MLKL

(A) The human MLKL pseudoactive site with side chains of key residues shown as sticks. (B) The pseudoactive site of human MLKL (green), in the same orientation as in (A), superimposed with ATP–(Mn2+)2 (magenta sticks) and side chains from the active site of PKA (yellow sticks; PDB code 1ATP [44]). (C) The mouse MLKL pseudoactive site with side chains of key pseudoactive site residues shown as sticks. Lys219, Asn318, Ser323 and Glu338 correspond to the human MLKL residues Lys230, Lys331, Asn336 and Glu351 shown in (A) and (B). (D) An alignment of catalytic motifs from PKA with the corresponding sequences in a subset of MLKL orthologues. G-loop, VAIK, HRDXXXXN and DFG motifs are boxed. Residues known to mediate catalysis within conventional protein kinases are shaded yellow. A cartoon depicting the secondary structure of a prototypic protein kinase and Hanks subdomain nomenclature [45] are shown above the alignment.

Figure 2
Divergent evolution within the pseudoactive sites of human and mouse MLKL

(A) The human MLKL pseudoactive site with side chains of key residues shown as sticks. (B) The pseudoactive site of human MLKL (green), in the same orientation as in (A), superimposed with ATP–(Mn2+)2 (magenta sticks) and side chains from the active site of PKA (yellow sticks; PDB code 1ATP [44]). (C) The mouse MLKL pseudoactive site with side chains of key pseudoactive site residues shown as sticks. Lys219, Asn318, Ser323 and Glu338 correspond to the human MLKL residues Lys230, Lys331, Asn336 and Glu351 shown in (A) and (B). (D) An alignment of catalytic motifs from PKA with the corresponding sequences in a subset of MLKL orthologues. G-loop, VAIK, HRDXXXXN and DFG motifs are boxed. Residues known to mediate catalysis within conventional protein kinases are shaded yellow. A cartoon depicting the secondary structure of a prototypic protein kinase and Hanks subdomain nomenclature [45] are shown above the alignment.

Of particular note, the human MLKL pseudoactive site (Figures 2A and 2B) is markedly different from that of mouse MLKL (Figure 2C). The most prominent difference is that the VAIK motif Lys219 in mouse MLKL is positioned by Gln343 of the unusual activation loop helix, rather than the conventional helix αC glutamic acid. We showed previously that the Lys219–Gln343 interaction regulated MLKL activation, because mutation of either of these sites resulted in an autoactivated MLKL that killed cells in the absence of RIPK3. This suggested that perturbation of the pseudoactive site by mutation of either Lys219 or Gln343 or by RIPK3-mediated phosphorylation of the adjacent activation loop residues, Ser345/Ser347/Thr349 is the molecular trigger for necroptosis [7]. In the human MLKL structure, in contrast, the activation loop does not contain the atypical helix observed in mouse MLKL, even though both the lysine residue in the VAIK motif and the glutamine residue in the activation loop are highly conserved between MLKL orthologues, suggesting that this interaction is of intrinsic importance to MLKL's activation mechanism.

The other major difference between the pseudoactive sites in human and mouse MLKL structures is that Asn318 of the HRN ‘catalytic’ loop sequence and Glu338 of the GFE activation loop motif in the mouse protein are not positioned to participate in nucleotide binding (Figure 2C), whereas the side chains of their human counterparts, Lys331 of the HGK sequence and Glu351 of the GFE sequence, are both directed towards the putative nucleotide-binding cleft. This difference led us to examine the contributions of Lys331 and Glu351 to human MLKL's nucleotide-binding capacity.

Examination of nucleotide-binding determinants within the human MLKL pseudoactive site

We expressed and purified recombinant wild-type, K230M, K331N and E351K mutant human MLKL pseudokinase domains from insect cells and qualitatively examined their binding to ATP, ADP, AMP, AMP-PNP and GTP, in the presence and absence of Mg2+ or Mn2+, using a thermal stability shift assay (Figure 3A–3D). We used thermal shifts to detect ligand binding by these pseudokinase domains because it is a highly sensitive, robust and proven technique [7,15,3437]. Wild-type human MLKL pseudokinase domain bound to ATP, ADP, GTP and AMP-PNP, but, exactly like murine MLKL [7], nucleotide binding was severely diminished in the presence of Mg2+ or Mn2+ (Figure 3A), as shown in a recent study [15]. The K230M human MLKL mutant exhibited diminished thermal shifts for all ligands (all ΔTm<3°C) consistent with a lower affinity for ATP, ADP and GTP than wild-type human MLKL. However, unlike wild-type human MLKL, the K230M mutant did not measurably bind AMP-PNP, since its thermal denaturation curve was virtually indistinguishable from the buffer control condition (Figure 3B).

Examination of nucleotide binding by wild-type and mutant human MLKL pseudokinase domains using a thermal stability shift assay

Figure 3
Examination of nucleotide binding by wild-type and mutant human MLKL pseudokinase domains using a thermal stability shift assay

Thermal denaturation curves (upper panels) and charts of ΔTm (lower panels) for wild-type (A), K230M (B), K331N (C) and E351K (D) human MLKL in the presence or absence of ATP, ADP, AMP, GTP and AMP-PNP in the presence or absence of Mg2+ or Mn2+. The curves and charts are colour coded according to the assay condition, with the key shown on the right. Each plot and chart is representative of data obtained in a minimum of two independent experiments performed on separate days using different aliquots of the same protein preparations. Data shown for wild-type MLKL (A) are consistent with previously reported data for this protein [15], and were generated independently from these previously published experiments. (EH) The affinity of each human MLKL variant was quantified by performing ATP titrations in the thermal stability shift assay. The plots and Kd values shown represent means±S.E.M. for two independent experiments. (I) A comparison of the thermal denaturation curves of wild-type, K230M, K331N and E351K human MLKL pseudokinase domains in the absence of ligands. K230M is less thermostable than wild-type, whereas K331N and E351K exhibit elevated thermal stability relative to wild-type human MLKL. Data are plotted as means±S.E.M. for a minimum of two independent experiments.

Figure 3
Examination of nucleotide binding by wild-type and mutant human MLKL pseudokinase domains using a thermal stability shift assay

Thermal denaturation curves (upper panels) and charts of ΔTm (lower panels) for wild-type (A), K230M (B), K331N (C) and E351K (D) human MLKL in the presence or absence of ATP, ADP, AMP, GTP and AMP-PNP in the presence or absence of Mg2+ or Mn2+. The curves and charts are colour coded according to the assay condition, with the key shown on the right. Each plot and chart is representative of data obtained in a minimum of two independent experiments performed on separate days using different aliquots of the same protein preparations. Data shown for wild-type MLKL (A) are consistent with previously reported data for this protein [15], and were generated independently from these previously published experiments. (EH) The affinity of each human MLKL variant was quantified by performing ATP titrations in the thermal stability shift assay. The plots and Kd values shown represent means±S.E.M. for two independent experiments. (I) A comparison of the thermal denaturation curves of wild-type, K230M, K331N and E351K human MLKL pseudokinase domains in the absence of ligands. K230M is less thermostable than wild-type, whereas K331N and E351K exhibit elevated thermal stability relative to wild-type human MLKL. Data are plotted as means±S.E.M. for a minimum of two independent experiments.

The side chain of Lys331 is directed towards the putative nucleotide-binding cleft in the human MLKL structure (Figures 2A and 2B), and this residue is the counterpart of the catalytic aspartic acid of the HRD motif in conventional protein kinases (Figure 2D). We hypothesized that the positive charge of the Lys331 side chain might pair with the negatively charged phosphates of nucleotides and thus play a novel role in nucleotide binding. To test this possibility, we substituted Lys331 for an asparagine residue, the corresponding residue in mouse MLKL (Figure 2D). Using the thermal shift assay, we observed that K331N mutant human MLKL was still able to bind ATP, ADP, GTP and AMP-PNP. Strikingly, the K331N mutant was intolerant of the presence of Mg2+ or Mn2+ and no thermal shifts were detected for any of the nucleotides in the presence of these cations.

The thermal stability of the K331N mutant in the absence of any nucleotide was significantly enhanced compared with the wild-type (ΔTm of ~5°C) (Figure 3C), possibly due to a shift to a more ‘closed’ conformation. This led to smaller shifts in the melting point following nucleotide addition and made it difficult to draw any conclusions as to the comparative affinity of nucleotide binding by this mutant. Therefore, in order to quantify ATP binding, we performed ATP titrations of wild-type human MLKL, K331N and the other mutants. As described below, we observed a >3-fold reduction in ATP affinity for the K331N mutant relative to wild-type human MLKL (Figure 3F).

E351 is located within the GFE sequence, a highly conserved motif throughout MLKL orthologues (Figure 2D), but distinct from the DFG motif typically found in a protein kinase's activation loop. The effect of the E351K mutation was of particular interest to us because it had been identified in a human lung carcinoma specimen [38]. We speculated that the mutation of Glu351 to a positively charged lysine residue might enhance nucleotide binding because, on the basis of our structure of human MLKL, this side chain would be directed towards the negatively charged phosphate moieties of nucleotides. Indeed, we observed thermal shifts of a similar magnitude to those for ATP, ADP, GTP and AMP-PNP binding to wild-type human MLKL, but also an unanticipated enhancement of binding to AMP.

An increased Tm value following the addition of nucleotide corresponds to an increase in protein stability, which we have used to identify nucleotide binding. Although the magnitude of the shift will be proportional to the amount of bound nucleotide, any attempt to quantify this can be confused by factors such as differences in the thermal stability of apo or bound mutant proteins compared with the wild-type (summarized in Figure 3I). To overcome this, nucleotide titrations were performed and Tm values against nucleotide concentrations were plotted (Figures 3E–3H). These titration curves can be used to determine Kd in a manner that does not depend on the magnitude of the thermal shift and therefore are independent of any changes in the intrinsic thermal stability of each mutant. Using this approach, we determined that wild-type human MLKL bound ATP with a Kd value of 27 μM (Figure 3E), which is comparable with that determined by isothermal titration calorimetry in a previous study [15]. In contrast, the Kd values for ATP binding by K230M and K331N were >300 μM and >90 μM respectively (Figures 3F and 3G), confirming the importance of these lysine residues in nucleotide engagement. Intriguingly, E351K human MLKL bound to ATP with an enhanced affinity (Kd of 8 μM) relative to wild-type human MLKL (Figure 3H). The functional consequences of this mutation to necroptotic signalling and whether this mutation might be oncogenic remain the subjects of ongoing investigations.

Nucleotide-binding residues of human MLKL have divergently evolved from those of mouse MLKL and conventional protein kinases

The ability to predict nucleotide binding among pseudokinase domains remains a formidable challenge owing to the structural diversity evident within their pseudoactive sites. This diversity is exemplified by comparing the pseudoactive sites of human and mouse MLKL orthologues (Figure 2), where the only uniting feature is the proximity of the VAIK motif lysine residue to mediate nucleotide engagement. The role of the lysine residue in the VAIK motif in nucleotide binding is a matter of some controversy because typically, in catalytically active ‘conventional’ protein kinases, this lysine residue is widely thought to position the α- and β-phosphates of ATP for phosphoryl transfer rather than mediate ATP binding [39]. However, as we learn more about the binding determinants that underlie nucleotide engagement by pseudokinases, it has become evident that the rules of nucleotide engagement established from studies of bona fide protein kinases cannot be applied prescriptively to pseudokinases. This is demonstrated in the present study, where the lysine residue in the VAIK motif of both human (Figure 3) and mouse [7] MLKL is required to bind nucleotide and not, as in conventional kinases, for positioning the nucleotide for catalytic transfer. Such an observation is not without precedent: a previous study of the pseudokinase, HSER (heat-stable enterotoxin receptor)/GUCY2C (guanylate cyclase 2C), also implicated the VAIK motif lysine residue as an essential determinant of ATP binding [40]. Although we cannot exclude the possibility that Lys230 serves a structural role in human MLKL by virtue of the hydrogen bond to Glu250 in helix αC, the observation that K230M mutant MLKL shows a modest decrease in thermal stability relative to wild-type human MLKL (Figure 3I) suggests that this is unlikely to account for the loss of nucleotide binding observed (Figure 3B). Further examination of K230M human MLKL in titrations (Figure 3F) showed definitively that this mutation compromises ATP binding, since these experiments enable ATP-binding affinity to be measured independently of thermal shift magnitudes and/or stability of the apoprotein. In the case of mouse MLKL, the analogous mutation K219M did not affect thermal stability, thereby indicating that structural perturbation does not underlie the observed loss of nucleotide binding in the mouse orthologue [7].

Although we show in the present study that human MLKL has evolved the catalytic loop residue Lys331 as an additional nucleotide-binding residue, no such additional determinants are evident from inspection of the mouse MLKL structure (Figure 2C). In mouse MLKL, we noted previously that mutation of the VAIK motif Lys→Met led to a loss of detectable nucleotide binding [7]. In contrast, the comparable mutation in human MLKL, K230M, led to a deficit, but not complete abrogation, of nucleotide binding (Figures 3B and 3F); a finding consistent with the co-evolution of Lys331 within the catalytic loop as an additional nucleotide interactor. Although HGK331 has evolved in place of the conventional kinase HRD motif in primate MLKL sequences and augments nucleotide binding, this sequence is highly divergent in orthologous sequences (Figure 2D), consistent with MLKL serving only non-catalytic functions in cell signalling. Notably, in many species, the catalytic loop contains an additional lysine residue in place of the asparagine residue in the HRDXXXXN motif (where X is any amino acid) (Figure 2D), and on the basis of the human and mouse MLKL pseudoactive site structures, these lysine residues would be positioned to potentially interact with nucleotides and enhance binding.

Taken together with previous work, our findings provide support for the idea that pseudokinases have greater freedom to evolve non-canonical nucleotide-binding mechanisms than protein kinases, which by contrast are subject to strict selective pressures to maintain their active site geometry for efficient phosphoryl transfer. This notion is supported by our observations that, unlike conventional protein kinases, the human and mouse MLKL and HSER/GUCYC pseudokinases have evolved a dependence on their VAIK lysine residue for nucleotide binding. In addition, there is clear evidence to support the idea that lysine and arginine substitutions have evolved within pseudokinase pseudoactive sites to mediate engagement of nucleotides via ionic interactions with their negatively charged phosphates. Within human MLKL this is evident for Lys331 which has evolved in place of the HRD motif catalytic aspartic acid, whereas the DFG motifs of EphB6 and STRADα have evolved into RLG and GLR respectively, with the acquisition of these arginine residues now known to serve important nucleotide-binding roles, presumably by supplanting the conventional aspartic acid residue and its associated Mg2+ ligand to directly interact with phosphates from the nucleotide [15,41]. Further support for this theory has arisen from our characterization of the human MLKL activation loop mutation identified in a lung carcinoma specimen GFE351→GFK, where acquisition of the lysine residue leads to an enhanced affinity for ATP. Collectively, these studies illustrate the importance of structural studies to provide insights into the diverse nucleotide-binding mechanisms among pseudokinase domains and pseudoactive site heterogeneity, even between orthologues.

Owing to their crucial roles in regulating a number of inflammatory diseases [1,3,42,43] pseudokinases have emerged as potential therapeutic targets. Our structural studies of human and mouse MLKL clearly highlight their divergent nucleotide-binding site geometries relative to the catalytic clefts of bona fide protein kinases, an attractive feature that could be exploited for the development of inhibitors with greater specificity than traditional kinase inhibitors. We anticipate that future structural studies of pseudokinases will similarly provide important insights into the unique landscape of their nucleotide-binding clefts and lead to new opportunities for drug discovery.

Abbreviations

     
  • LOK

    lymphocyte-oriented kinase

  •  
  • MLKL

    mixed lineage kinase domain-like

  •  
  • Ni-NTA

    Ni2+-nitrilotriacetate

  •  
  • AMP-PNP

    adenosine 5′-[β,γ-imido]triphosphate (also known as p[NH]ppA)

  •  
  • PKA

    protein kinase A

  •  
  • RIPK3

    receptor-interacting protein kinase 3

  •  
  • SAXS

    small-angle X-ray scattering

  •  
  • TCEP

    tris-(2-carboxyethyl)phosphine

  •  
  • TEV

    tobacco etch virus

  •  
  • WAXS

    wide-angle X-ray scattering

AUTHOR CONTRIBUTION

James Murphy, Isabelle Lucet, Guillaume Lessene, Warren Alexander, Jeffrey Babon, John Silke and Peter Czabotar designed the study and analysed data. James Murphy, Isabelle Lucet, Joanne Hildebrand, Maria Tanzer, Samuel Young, Pooja Sharma, Jeffrey Babon and Peter Czabotar performed the experiments. James Murphy wrote the paper with input from all other authors.

We thank MX and SAXS/WAXS beamline staff at the Australian Synchrotron where diffraction and scattering data were collected, and the Monash University Protein Production Facility for access to the Corbett real-time PCR cycler for thermal shift assays. Crystallization experiments were performed at the Bio21 C3 Collaborative Crystallization Centre.

FUNDING

This work was supported by the NHMRC (National Health and Medical Research Council) [grant numbers 637342, 1016647 and 1046984] and fellowships to J.M.H., W.S.A. and J.S.; Australian Research Council fellowships [FT100100100, FT110100169 and FT0992105] to J.M.M., J.J.B. and P.E.C.; a Victorian International Research Scholarship to M.C.T.; with additional support from the Victorian State Government Operational Infrastructure Support and NHMRC IRIISS (Independent Medical Research Institutes Infrastructure Support Scheme) [grant 361646]. J.S. is a member of the Scientific Advisory Board of TetraLogic Pharmaceuticals.

References

References
1
Kaczmarek
 
A.
Vandenabeele
 
P.
Krysko
 
D. V.
 
Necroptosis: the release of damage-associated molecular patterns and its physiological relevance
Immunity
2013
, vol. 
38
 (pg. 
209
-
223
)
2
Green
 
D. R.
 
Pseudokiller, qu’est-ce que c’est?
Immunity
2013
, vol. 
39
 (pg. 
421
-
422
)
3
Murphy
 
J. M.
Silke
 
S.
 
Ars moriendi: the art of dying well – new insights into the molecular pathways of necroptotic cell death
EMBO Rep.
2013
 
in the press
4
Khan
 
N.
Lawlor
 
K. E.
Murphy
 
J. M.
Vince
 
J. E.
 
More to life than death: molecular determinants of necroptotic and non-necroptotic RIP3 kinase signaling
Curr. Opin. Immunol.
2014
, vol. 
26
 (pg. 
76
-
89
)
5
Sun
 
L.
Wang
 
H.
Wang
 
Z.
He
 
S.
Chen
 
S.
Liao
 
D.
Wang
 
L.
Yan
 
J.
Liu
 
W.
Lei
 
X.
Wang
 
X.
 
Mixed lineage kinase domain-like protein mediates necrosis signaling downstream of RIP3 kinase
Cell
2012
, vol. 
148
 (pg. 
213
-
227
)
6
Zhao
 
J.
Jitkaew
 
S.
Cai
 
Z.
Choksi
 
S.
Li
 
Q.
Luo
 
J.
Liu
 
Z. G.
 
Mixed lineage kinase domain-like is a key receptor interacting protein 3 downstream component of TNF-induced necrosis
Proc. Natl. Acad. Sci. U.S.A.
2012
, vol. 
109
 (pg. 
5322
-
5327
)
7
Murphy
 
J. M.
Czabotar
 
P. E.
Hildebrand
 
J. M.
Lucet
 
I. S.
Zhang
 
J. G.
Alvarez-Diaz
 
S.
Lewis
 
R.
Lalaoui
 
N.
Metcalf
 
D.
Webb
 
A. I.
, et al 
The pseudokinase MLKL mediates necroptosis via a molecular switch mechanism
Immunity
2013
, vol. 
39
 (pg. 
443
-
453
)
8
Wu
 
J.
Huang
 
Z.
Ren
 
J.
Zhang
 
Z.
He
 
P.
Li
 
Y.
Ma
 
J.
Chen
 
W.
Zhang
 
Y.
Zhou
 
X.
, et al 
Mlkl knockout mice demonstrate the indispensable role of Mlkl in necroptosis
Cell Res.
2013
, vol. 
23
 (pg. 
994
-
1006
)
9
Kaiser
 
W. J.
Sridharan
 
H.
Huang
 
C.
Mandal
 
P.
Upton
 
J. W.
Gough
 
P. J.
Sehon
 
C. A.
Marquis
 
R. W.
Bertin
 
J.
Mocarski
 
E. S.
 
Toll-like receptor 3-mediated necrosis via TRIF, RIP3, and MLKL
J. Biol. Chem.
2013
, vol. 
288
 (pg. 
31268
-
31279
)
10
Wang
 
Z.
Jiang
 
H.
Chen
 
S.
Du
 
F.
Wang
 
X.
 
The mitochondrial phosphatase PGAM5 functions at the convergence point of multiple necrotic death pathways
Cell
2012
, vol. 
148
 (pg. 
228
-
243
)
11
Reference deleted
12
Boudeau
 
J.
Miranda-Saavedra
 
D.
Barton
 
G. J.
Alessi
 
D. R.
 
Emerging roles of pseudokinases
Trends Cell Biol.
2006
, vol. 
16
 (pg. 
443
-
452
)
13
Eyers
 
P. A.
Murphy
 
J. M.
 
Dawn of the dead: protein pseudokinases signal new adventures in cell biology
Biochem. Soc. Trans.
2013
, vol. 
41
 (pg. 
969
-
974
)
14
Manning
 
G.
Whyte
 
D. B.
Martinez
 
R.
Hunter
 
T.
Sudarsanam
 
S.
 
The protein kinase complement of the human genome
Science
2002
, vol. 
298
 (pg. 
1912
-
1934
)
15
Murphy
 
J. M.
Zhang
 
Q.
Young
 
S. N.
Reese
 
M. L.
Bailey
 
F. P.
Eyers
 
P. A.
Ungureanu
 
D.
Hammaren
 
H.
Silvennoinen
 
O.
Varghese
 
L. N.
, et al 
A robust methodology to subclassify pseudokinases based on their nucleotide-binding properties
Biochem. J.
2013
, vol. 
457
 (pg. 
323
-
334
)
16
Babon
 
J. J.
Kershaw
 
N. J.
Murphy
 
J. M.
Varghese
 
L. N.
Laktyushin
 
A.
Young
 
S. N.
Lucet
 
I. S.
Norton
 
R. S.
Nicola
 
N. A.
 
Suppression of cytokine signaling by SOCS3: characterization of the mode of inhibition and the basis of its specificity
Immunity
2012
, vol. 
36
 (pg. 
239
-
250
)
17
Kershaw
 
N. J.
Murphy
 
J. M.
Lucet
 
I. S.
Nicola
 
N. A.
Babon
 
J. J.
 
Regulation of Janus kinases by SOCS proteins
Biochem. Soc. Trans.
2013
, vol. 
41
 (pg. 
1042
-
1047
)
18
Ungureanu
 
D.
Wu
 
J.
Pekkala
 
T.
Niranjan
 
Y.
Young
 
C.
Jensen
 
O. N.
Xu
 
C. F.
Neubert
 
T. A.
Skoda
 
R. C.
Hubbard
 
S. R.
Silvennoinen
 
O.
 
The pseudokinase domain of JAK2 is a dual-specificity protein kinase that negatively regulates cytokine signaling
Nat. Struct. Mol. Biol.
2011
, vol. 
18
 (pg. 
971
-
976
)
19
Murphy
 
J. M.
Metcalf
 
D.
Young
 
I. G.
Hilton
 
D. J.
 
A convenient method for preparation of an engineered mouse interleukin-3 analog with high solubility and wild-type bioactivity
Growth Factors
2010
, vol. 
28
 (pg. 
104
-
110
)
20
Kabsch
 
W.
 
Integration, scaling, space-group assignment and post-refinement
Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010
, vol. 
66
 (pg. 
133
-
144
)
21
McCoy
 
A. J.
Grosse-Kunstleve
 
R. W.
Adams
 
P. D.
Winn
 
M. D.
Storoni
 
L. C.
Read
 
R. J.
 
Phaser crystallographic software
J. Appl. Crystallogr.
2007
, vol. 
40
 (pg. 
658
-
674
)
22
Emsley
 
P.
Lohkamp
 
B.
Scott
 
W. G.
Cowtan
 
K.
 
Features and development of Coot
Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010
, vol. 
66
 (pg. 
486
-
501
)
23
Adams
 
P. D.
Afonine
 
P. V.
Bunkoczi
 
G.
Chen
 
V. B.
Davis
 
I. W.
Echols
 
N.
Headd
 
J. J.
Hung
 
L. W.
Kapral
 
G. J.
Grosse-Kunstleve
 
R. W.
, et al 
PHENIX: a comprehensive Python-based system for macromolecular structure solution
Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010
, vol. 
66
 (pg. 
213
-
221
)
24
Chen
 
V. B.
Arendall, 3rd
 
W. B.
Headd
 
J. J.
Keedy
 
D. A.
Immormino
 
R. M.
Kapral
 
G. J.
Murray
 
L. W.
Richardson
 
J. S.
Richardson
 
D. C.
 
MolProbity: all-atom structure validation for macromolecular crystallography
Acta Crystallogr., Sect. D: Biol. Crystallogr.
2010
, vol. 
66
 (pg. 
12
-
21
)
25
Kirby
 
N. M.
Mudie
 
S. T.
Hawley
 
A. M.
Cookson
 
D. J.
Mertens
 
H. D. T.
Cowieson
 
N.
Samardzic-Boban
 
V.
 
A low-background-intensity focusing small-angle X-ray scattering undulator beamline
J. Appl. Crystallogr.
2013
, vol. 
46
 (pg. 
1670
-
1680
)
26
Griffin
 
M. D.
Billakanti
 
J. M.
Wason
 
A.
Keller
 
S.
Mertens
 
H. D.
Atkinson
 
S. C.
Dobson
 
R. C.
Perugini
 
M. A.
Gerrard
 
J. A.
Pearce
 
F. G.
 
Characterisation of the first enzymes committed to lysine biosynthesis in Arabidopsis thaliana
PLoS ONE
2012
, vol. 
7
 pg. 
e40318
 
27
Konarev
 
P.
Volkov
 
V. V.
Sokolova
 
A. V.
Koch
 
M. H. J.
Svergun
 
D. I.
 
PRIMUS: a Windows PC-based system for small-angle scattering data analysis
J. Appl. Crystallogr.
2003
, vol. 
36
 (pg. 
1277
-
1282
)
28
Svergun
 
D.
 
Determination of the regularization parameter in indirect-transform methods using perceptual criteria
J. Appl. Crystallogr.
1992
, vol. 
25
 (pg. 
495
-
503
)
29
Svergun
 
D.
Barberato
 
C
Koch
 
M. H. J.
 
CRYSOL: a program to evaluate X-ray solution scattering of biological macromolecules from atomic coordinates
J. Appl. Crystallogr.
1995
, vol. 
28
 (pg. 
768
-
773
)
30
Taylor
 
S. S.
Kornev
 
A. P.
 
Protein kinases: evolution of dynamic regulatory proteins
Trends Biochem. Sci.
2011
, vol. 
36
 (pg. 
65
-
77
)
31
Kornev
 
A. P.
Haste
 
N. M.
Taylor
 
S. S.
Eyck
 
L. F.
 
Surface comparison of active and inactive protein kinases identifies a conserved activation mechanism
Proc. Natl. Acad. Sci. U.S.A.
2006
, vol. 
103
 (pg. 
17783
-
17788
)
32
Xie
 
T.
Peng
 
W.
Yan
 
C.
Wu
 
J.
Gong
 
X.
Shi
 
Y.
 
Structural insights into RIP3-mediated necroptotic signaling
Cell Rep.
2013
, vol. 
5
 (pg. 
70
-
78
)
33
Pike
 
A. C.
Rellos
 
P.
Niesen
 
F. H.
Turnbull
 
A.
Oliver
 
A. W.
Parker
 
S. A.
Turk
 
B. E.
Pearl
 
L. H.
Knapp
 
S.
 
Activation segment dimerization: a mechanism for kinase autophosphorylation of non-consensus sites
EMBO J.
2008
, vol. 
27
 (pg. 
704
-
714
)
34
Eswaran
 
J.
Patnaik
 
D.
Filippakopoulos
 
P.
Wang
 
F.
Stein
 
R. L.
Murray
 
J. W.
Higgins
 
J. M.
Knapp
 
S.
 
Structure and functional characterization of the atypical human kinase haspin
Proc. Natl. Acad. Sci. U.S.A.
2009
, vol. 
106
 (pg. 
20198
-
20203
)
35
Scheeff
 
E. D.
Eswaran
 
J.
Bunkoczi
 
G.
Knapp
 
S.
Manning
 
G.
 
Structure of the pseudokinase VRK3 reveals a degraded catalytic site, a highly conserved kinase fold, and a putative regulatory binding site
Structure
2009
, vol. 
17
 (pg. 
128
-
138
)
36
Lucet
 
I. S.
Babon
 
J. J.
Murphy
 
J. M.
 
Techniques to examine nucleotide binding by pseudokinases
Biochem. Soc. Trans.
2013
, vol. 
41
 (pg. 
974
-
980
)
37
Christie
 
M.
Boland
 
A.
Huntzinger
 
E.
Weichenrieder
 
O.
Izaurralde
 
E.
 
Structure of the PAN3 pseudokinase reveals the basis for interactions with the PAN2 deadenylase and the GW182 proteins
Mol. Cell
2013
, vol. 
51
 (pg. 
360
-
373
)
38
Forbes
 
S. A.
Bhamra
 
G.
Bamford
 
S.
Dawson
 
E.
Kok
 
C.
Clements
 
J.
Menzies
 
A.
Teague
 
J. W.
Futreal
 
P. A.
Stratton
 
M. R.
 
The Catalogue of Somatic Mutations in Cancer (COSMIC)
Curr. Protoc. Hum. Genet.
2008
, vol. 
57
 (pg. 
10.11.1
-
10.11.26
)
39
Carrera
 
A. C.
Alexandrov
 
K.
Roberts
 
T. M.
 
The conserved lysine of the catalytic domain of protein kinases is actively involved in the phosphotransfer reaction and not required for anchoring ATP
Proc. Natl. Acad. Sci. U.S.A.
1993
, vol. 
90
 (pg. 
442
-
446
)
40
Jaleel
 
M.
Saha
 
S.
Shenoy
 
A. R.
Visweswariah
 
S. S.
 
The kinase homology domain of receptor guanylyl cyclase C: ATP binding and identification of an adenine nucleotide sensitive site
Biochemistry
2006
, vol. 
45
 (pg. 
1888
-
1898
)
41
Zeqiraj
 
E.
Filippi
 
B. M.
Goldie
 
S.
Navratilova
 
I.
Boudeau
 
J.
Deak
 
M.
Alessi
 
D. R.
van Aalten
 
D. M.
 
ATP and MO25α regulate the conformational state of the STRADα pseudokinase and activation of the LKB1 tumour suppressor
PLoS Biol.
2009
, vol. 
7
 pg. 
e1000126
 
42
Kawagoe
 
T.
Sato
 
S.
Matsushita
 
K.
Kato
 
H.
Matsui
 
K.
Kumagai
 
Y.
Saitoh
 
T.
Kawai
 
T.
Takeuchi
 
O.
Akira
 
S.
 
Sequential control of Toll-like receptor-dependent responses by IRAK1 and IRAK2
Nat. Immunol.
2008
, vol. 
9
 (pg. 
684
-
691
)
43
Zhou
 
H.
Yu
 
M.
Fukuda
 
K.
Im
 
J.
Yao
 
P.
Cui
 
W.
Bulek
 
K.
Zepp
 
J.
Wan
 
Y.
Kim
 
T. W.
, et al 
IRAK-M mediates Toll-like receptor/IL-1R-induced NFκB activation and cytokine production
EMBO J.
2013
, vol. 
32
 (pg. 
583
-
596
)
44
Zheng
 
J.
Trafny
 
E. A.
Knighton
 
D. R.
Xuong
 
N. H.
Taylor
 
S. S.
Ten Eyck
 
L. F.
Sowadski
 
J. M.
 
2.2 Å refined crystal structure of the catalytic subunit of cAMP-dependent protein kinase complexed with MnATP and a peptide inhibitor
Acta Crystallogr., Sect. D: Biol. Crystallogr.
1993
, vol. 
49
 (pg. 
362
-
365
)
45
Hanks
 
S. K.
Quinn
 
A. M.
Hunter
 
T.
 
The protein kinase family: conserved features and deduced phylogeny of the catalytic domains
Science
1988
, vol. 
241
 (pg. 
42
-
52
)

Author notes

The structural co-ordinate reported for human MLKL pseudokinase domain will appear in the PDB under code 4MWI.

Supplementary data