|Home | About | Journals | Submit | Contact Us | Français|
Interleukin-1 receptor–associated kinase-4 (IRAK-4) is an essential component of innate immunity in mice and humans. IRAK-4 is a bipartite protein composed of a death domain (DD) that mediates molecular recognition, and a catalytic kinase domain. Structure determination of the proteolytically stable, soluble IRAK-4 DD was hampered by poor diffraction quality. Addition of manganese (II) chloride to the crystallization solution produced significant improvements in diffraction, and the structure has been determined to 1.7-Å resolution. Examination of the IRAK-4 DD crystal structure reveals a single manganese ion coordinated to surface residues lysine-21 and aspartate-24. Coordination of the manganese ion resulted in a reduction in the surface entropy at this region of the molecule, by generating a contact-forming and conformationally homogenous surface patch. Prior studies have shown that surface entropy reduction by mutation of surface residues with large flexible side chains (i.e., Lys and Glu) to smaller side chains results in the production of diffraction-quality crystals. The intrinsic high surface entropy of Lys residues can also be decreased by reductive methylation. Our results suggest that screening of manganese ions as a crystallization additive may also facilitate ordered crystallization by reduction of surface entropy. Given the quick and inexpensive nature of screening, this technique is likely to be amenable to high-throughput methods such as those employed by Protein Structure Initiatives.
The immune system that mediates the cellular response is typically divided into two categories: (1) adaptive and (2) innate responses.1 The adaptive response is induced following recovery of an infection from a pathogen. In contrast, the innate response is activated immediately upon infection. Innate immunity utilizes germline-encoded, broadly specific receptors to recognize highly conserved molecular structures, called pathogen-associated molecular patterns, found in many different microorganisms.2 These molecular patterns can be specific to a particular type of pathogen, but are not found in the host. Lastly, the innate immune system does not retain a memory of prior exposure.3 In vertebrates, the innate immune response regulates the release of mediators of inflammation, such as cytokines (tumor necrosis factor-alpha, interleukin-1, etc.) and chemokines (interleukin-8).4
Activation of the mammalian innate immune response pathway is initiated by the direct recognition of pathogen-associated molecular patterns by the extracellular domain of a member of the Toll-like receptor family.5 Receptor activation results in the recruitment of a multi-protein signaling complex through interactions mediated by the cytoplasmic Toll/IL-1 receptor/resistance (TIR) domain of the Toll-like receptor and the TIR domain of the universal adaptor MyD88.6 In turn, the MyD88 adaptor recruits downstream kinases, including the interleukin receptor–associated kinase-4 (IRAK-4).7 This signaling cascade eventually culminates in the activation of NFκB and the production of proinflammatory cytokines. Given the role of the innate immune pathway in mediating inflammation,8 there has been extensive research on the components of the innate immune pathway. However, there have been few studies to investigate the structural details of these components. As a step towards structural characterization of key players within this pathway, we have determined the crystal structure of the death domain of IRAK-4.9
Initial crystallization screens identified promising conditions that were further refined to yield well-formed crystals. However, diffraction from these crystals was poor and was limited to a Bragg limit of 4 Å. Presumably, this weak diffraction was due to a high surface entropy introduced by either a dynamic tail, loop, or side chain. To reduce this surface entropy and subsequently produce crystals that had improved diffraction properties, we sought to identify a more stable protein construct by using limited proteolysis. Treatment of samples of the IRAK-4 death domain with various proteases demonstrated that the protein construct used for crystallization represented a stable core domain. Size-exclusion chromatography documented that the purified protein was monomeric and not prone to further aggregation. Screening of various additives yielded crystals grown in the presence of manganese chloride that showed a marked improvement in diffraction quality to a limiting resolution of 1.7 Å. The structure was solved by multiwavelength anomalous diffraction methods using a crystal transiently soaked in a high concentration of sodium bromide.10 Inspection of the final structure shows that a single manganese ion generates a contact-forming and conformationally homogenous surface patch consisting of a Lys and an Asp residue. To our knowledge, this is the first time manganese ions have been shown to reduce surface entropy of proteins. We propose that the use of manganese ions to limit surface entropy may be a generally feasible means of improving diffraction qualities of protein crystals.
The cDNA of the Mus musculus IRAK-4 gene, mouse 30017484 pSPORT1, was purchased from the American Type Culture Collection (Manassas, VA). The polymerase chain reaction (PCR) using Pfx polymerase (Invitrogen, Carlsbad, CA) amplified the region corresponding to the death domain of IRAK-4, amino acids 1–113. PCR was carried out using the following primers (Integrated DNA Technologies, Coralville, IA): forward, 5′-CAG-CAGCCGCATATGAACAAGCCGTTGACACC-3′; reverse, 5′-CCGCTCGAGTTACTAAGGCAGGCTTT-3′, restriction enzyme recognition sites underlined. The amplified PCR product was purified using the Qiaquick PCR purification kit (Qiagen, Valencia, CA) and the fragment was digested with NdeI and XhoI (New England Biolabs, Ipswich, MA) overnight at 37°C. The digested PCR product was ligated into a modified pGEX6p1 vector (Amersham Biosciences, Piscataway, NJ) using T4 DNA ligase (Invitrogen). The ligation mixture was then used to transform chemically competent DH5α E. coli cells. Samples positive for insertion were then sent for dideoxy sequencing to confirm that the proper gene was cloned into the vector and that the insert was in frame. The resultant fusion plasmid produces the IRAK-4 death domain as a carboxy-terminal fusion with glutathione S-transferase (GST-IRAK4 DD).
The GST-IRAK4 DD vector was used to transform BL21 (DE3) E. coli chemically competent cells. An expression check was carried out by growing transformants 4 h in 2 mL of Luria broth (LB) containing 200 μg/mL ampicillin at 37°C. A small sample was removed and the remaining culture was induced using 1 mM (final concentration) isopropyl-β-d-thiogalactoside (IPTG). After 3 h at 30°C, 500 μL of both the induced and uninduced samples were pelleted by centrifugation, and the media were removed. The pellets were resuspended in 100 μL of H2O by vortexing. Equal volumes of the resuspension and 2X sodium dodecyl sulfate (SDS) sample loading buffer were mixed, typically 15-μL volumes, heated at 95°C for 5 min, and loaded onto a 10% SDS–polyacrylamide gel electrophoresis (PAGE) gel for analysis.
For large-scale expression, a small-scale starter culture containing ampicillin (200 mg/L) was inoculated with single colonies of BL21 (DE3) bearing the GST-IRAK4 vector and incubated with vigorous shaking at 37°C. Approximately 5 h later, the starter cultures were used to inoculate 6 L of LB bearing the same concentration of ampicillin. This culture was then induced with a final concentration of 0.5 mM IPTG when the optical density at λ = 600 nm reached 0.6–0.8. Protein expression continued for 16–18 h at 18°C. The cells were harvested by centrifugation and resuspended using 20 mM HEPES (pH 7.5), 100 mM KCl, 10% glycerol, DNAase, and a protease inhibitor cocktail purchased from Sigma (St. Louis, MO). After resuspension, the cells were flash frozen in liquid nitrogen and stored at −80°C until needed.
Frozen cells were thawed and disrupted by multiple passes through an Avestion C5 French pressure cell. Following lysis, cellular debris was removed by centrifugation (15,000 rpm, 1 h). The supernatant was then applied to a glutathione sepharose (Amersham Pharmacia) column that had been equilibrated in 1X phosphate-buffered saline (PBS) (pH 7.4), 500 mM KCl, at a flow rate of 0.5 mL/min. The column was then washed at a flow rate of 2 mL/min for 150 min with the same equilibration buffer. The sample was then eluted using 20 mM reduced glutathione, 200mM Tris (pH 8.3), and 500 mM KCl. Fractions (10 mL) were then analyzed by SDS-PAGE. Fractions containing GST-IRAK-4 DD were then dialyzed in 20 mM MES (pH 6.5), 100 mM KCl (buffer A), using a 6000–8000 molecular-weight cutoff (MWCO) dialysis bag. Rhinoviral 3C protease (~100 μg) was added to the dialysis bag and the sample was then allowed to dialyze overnight at 4°C. The sample was then analyzed for protease cleavage by SDS-PAGE. Upon complete cleavage, the dialysate was loaded onto a HiTrap SP FF 5-mL column (Amersham Biosciences), previously equilibrated with buffer A, at a flow rate of 2 mL/min. After loading and washing with buffer A (~30 mL), a linear gradient to 1 M KCl over 24 column volumes was used to elute the untagged protein. The fractions were analyzed by SDS-PAGE to determine the fractions containing untagged IRAK-4 DD. Fractions containing IRAK-4 DD were pooled and stored in 10% (v/v) glycerol, aliquoted, flash frozen in liquid nitrogen, and stored at −80°C.
A Sephadex 75 (Amersham Biosciences) gel filtration column was equilibrated with 120 mL of buffer B. Purified IRAK-4 DD was thawed in water at 25°C, and 500 μL of the solution was applied to the column. Buffer B was used at a flow rate of 1 mL/min to elute (2-mL fractions) IRAK-4 DD from the column. A 12% SDS-PAGE gel was used to analyze peak fractions.
A 4-mL aliquot of IRAK-4 DD was thawed in water at room temperature and subsequently placed on ice. An Amicon Ultra centrifugal filter device (Millipore, Cork, Ireland) with a 5000-Da MWCO was equilibrated with 20 mM HEPES (pH 7.5), 100 mM KCl (buffer B) at 4°C in a Sorvall H6000A swinging-bucket rotor at 3500 rpm. The thawed sample of IRAK-4 DD was exchanged into buffer B by ultra-filtration, and concentrated to a final volume of ~200 μL. The concentrated sample was placed in a 1.5-mL microcentrifuge tube and spun (16,100 RCF, 4°C, 10 min) to remove a white precipitate that formed during the final concentration step. The pelleted precipitate was discarded and the concentration of protein in the supernatant was determined by the Bradford assay (Biorad, Richmond, CA). The protein was adjusted to a final concentration of 10 mg/mL and used for crystallization screens. An initial screen using 400 different conditions from commercially available screens (Hampton Research, Aliso Viejo, CA, and Emerald Biosystems, Bainbridge Island, WA) was used to find conditions amenable to crystallization for IRAK-4 DD. All crystallization experiments were performed using the hanging-drop method of vapor diffusion. A 150-μL reservoir volume was used in 48-well VDX plates (Hampton Research). Equal volumes of the reservoir (1.2 μL) and protein (1.2 μL) solutions were mixed and suspended over a reservoir volume of 150 μL at 25°C and stored at 8°C. Well-formed crystals were observed using a reservoir solution composed of 100 mM HEPES (pH 7.5) and 25% PEG 3350. Using a home source, the diffraction from these initial crystals was poor. Various additives were screened in order to improve the diffraction limits of the crystals. The addition of MnCl2·4H2O (Hampton Research), to a final drop concentration of 10 mM, resulted in the formation of crystals that diffracted Cu-Ka X-rays to better than 2.5 Å. For cryoprotection, crystals were soaked for 30 min in solutions containing 100 mM HEPES (pH 7.5) and 25% PEG 3350 with 10%, 20%, and 30% increments of glycerol. Phases were determined by the multiwavelength anomalous diffraction method, utilizing crystals soaked in 30% glycerol, 100 mM HEPES (pH 7.5), 25% PEG 3350, 1 M NaBr, for 30 sec.
Crystals were harvested within a week of forming, mounted on 0.1- to 0.2-mm nylon loops (20-μm thickness, Hampton Research) and plunged into liquid nitrogen for storage. X-ray diffraction data from IRAK-4 DD native crystals were collected using a MAR CCD detector (120-mm distance) at 12.667 keV under standard cryogenic conditions at synchotron beamline 14-ID at Advanced Photon Source (APS), Argonne National laboratory (Argonne, IL). Diffraction data on the bromide-soaked crystals of IRAK-4 DD were collected on a MAR CCD detector (130-mm distance) at 13.480, 13.484, and 13.584 keV at synchotron beamline 32-ID at APS. The native crystal was rotated through a total of 200° with 1.0° oscillation per frame (3-sec exposures) and the bromide-soaked crystal was rotated through a total of 360° with 1.0° oscillation per frame (2-sec exposures). Initial phase-angle determination utilized the dispersive and anomalous signals from the multiwavelength anomalous diffraction data sets. Phase improvement by solvent flattening and phase extension to 1.7 Å produced a readily interpretable map. Details of the structure, including refinement statistics and PDB deposition, are reported elsewhere.9
As part of our ongoing efforts towards the biochemical and biophysical characterization of key mediators in the innate immune pathway, we have determined the high-resolution crystal structure of the death domain of IRAK-4.9 Production of the protein as a glutathione S-transferase fusion yielded soluble samples when overexpressed in E. coli (Figure 1A1A).). The use of the GST fusion also facilitated purification by affinity chromatography. Although the removal of the fusion tag slightly decreased the solubility of the IRAK-4 DD, the purified protein was not prone to further aggregation, as determined by size-exclusion chromatography (Figure 1B1B).
Initial screening attempts identified well-formed crystals using polyethylene glycol 3350 as a precipitant. However, the crystals diffracted only to 4 Å using X-rays from a Cu-Kα home source (Figure 2A2A).). Speculating that diffraction from these crystals was limited due to the presence of disordered regions within the polypeptide, the purified sample was subjected to limited proteolysis utilizing a battery of specific and non-specific proteases to identify unstructured regions. However, the purified sample was resistant to proteolysis under various conditions, suggesting that the IRAK-4 DD construct we had generated constituted a well-folded domain without any disordered regions.
Various additives were screened in order to identify compounds that would further stabilize the crystal lattice and improve the diffraction qualities of these crystals. Addition of MnCl2·4H2O to a final drop concentration of 10 mM resulted in a marked improvement in the diffraction limits from these crystals (Figure 2B2B)) with an identical crystal morphology, as observed previously. The structure was solved to 2.0-Å resolution by the multi-wavelength anomalous diffraction method, utilizing crystals transiently soaked in high concentrations of sodium bromide.10 The structure was further refined to 1.7 Å utilizing amplitudes from a native crystal lacking halides.9
Inspection of the 1.7-Å native structure revealed the presence of a 12 sigma peak in difference SIGMAA11 maps calculated with model phases (Figure 2C2C).). Based on the presence of a high concentration of manganese in the crystallization solution and the proximity of the protein ligands, a manganese ion has been assigned to this feature in the model (Figure 2C2C).). This manganese ion is engaged by protein residues Lys-21 (distance from Nζ = 2.08 Å) and Asp-24 (distance from Oδ2 = 2.7 Å). These distances are typical of those observed for interactions with manganese in other protein structures, phosphoenolpyruvate carboxykinase (PCK) (2.3 Å average12,13), and arginase-I (2.2 Å average14,15). Lys (pKa = 10.53) has been known to coordinate with manganese (II) as seen in the PCK structure.12,13 Although a neighboring residue in the IRAK-4 DD, Arg-20 (pKa = 12.48), does not coordinate with the manganese, it serves to lower the pKa of the Lys side chain and keep it in the deprotonated state. In the deprotonated state, Lys-21 is free to coordinate to the manganese cation. Lys-21 makes a true inner-sphere coordination interaction with manganese, as the NζMn+2 separation is 2.08 Å; however, Asp-24 makes an outer-sphere coordination with manganese.
We note that metal ions are not critical for the crystallization of IRAK-4 DD, as observed in cases where such additives have been shown to be essential to crystallization.16–20 Rather, the role of this ion is to engage the side chains of two flexible residues that are in close proximity. Ordinarily, without manganese present, these side chains would be in numerous conformations and would not support the formation of a stable crystal contact. The reduction in conformational heterogeneity of this surface patch by the manganese coordination, generates a favorable crystal contact (Figure 2D2D).). Reduction in conformational heterogeneity is a reduction in local surface entropy. The authors recognize that in order to lend further support to this hypothesis, structural analyses of point mutations of Lys-21 and Asp-24 to Ala residues would be helpful. However, these mutations could produce significant changes in the chemical properties of IRAK-4 DD, including changes in the stability, solubility, and aggregation state of the protein. The molecular details of the final structure have been published elsewhere.9
Our results with improvements in the diffraction qualities of IRAK-4 DD underscore the fact that not all well-folded, soluble, monomeric proteins yield high-quality crystals. Results from several Protein Structure Initiatives studies place the success rate for obtaining diffraction-quality crystals from well-folded, monodisperse proteins in the range of 40–50%.21 Successful crystallization of most macromolecules requires a homogeneous sample. This requires homogeneity in both the composition and the conformation of the sample. Hence, flexible regions of macromolecules may interfere with crystal-lattice formation because of conformational heterogeneity.22 In fact, if there is prior knowledge of the domain fold and a known region of high flexibility (i.e., a flexible loop, as in the SRβ-GTP:SRX structure23), one can use circular permutation as a tool to reduce surface entropy and lead to crystallization.22
In many instances, the presence of large, flexible regions in proteins can be routinely identified using limited proteolysis combined with high-resolution mass spectrometry.24 The identified domain boundaries are used to design a new construct for the protein of interest encoding the appropriate truncations, which represent a stable core. In addition to these large regions of heterogeneity, flexible side chains, such as Lys, Arg, and Glu, can introduce conformational heterogeneity and small regions of surface flexibility. These small regions of flexibility can be problematic if they are located in a crystal contact region. In these cases, an alternative approach towards reducing conformational heterogeneity has been the use of surface-entropy reduction mutagenesis.25,26 This approach entails the replacement of clusters of flexible side chains (i.e., Lys, Glu, and Arg) with short aliphatic residues (typically Ala) in order to induce local reduction in conformational heterogeneity.27,28 This technique has been shown to yield improvements in diffraction, as well as facilitate de novo crystallization.26,29 However, this methodology is time-consuming, as many such mutants may need to be screened in order to identify those that prove amenable to crystallization.
Reductive methylation has been utilized as an alternate approach to surface entropy reduction.30 Reductive methylation is thought to affect surface entropy reduction by increasing the hydrophobicity of Lys side chains, resulting in a decrease in thermal motion.30 This technique has been successful in yielding crystals of maltooligosyl trehalose synthase,31 pokeweed antiviral protein,32 Pvs25,33 and BmBKTtx1.34 Moreover, this technique is shown to be sufficiently gentle to facilitate the crystallization of multi-protein complexes such as the YopN-SycN-YscB ternary complex and its selenomethionine derivatives.35
Here we report the use of divalent metal ions, in particular manganese, as a successful alternate to the above approaches in reducing surface entropy. In the case of the IRAK-4 DD, the diffraction limits from initial crystals were presumably compromised by the presence of a cluster of flexible side chains in a loop region including Arg-20-Lys-21-X-X-Asp-24. Utilization of manganese salts as an additive resulted in a marked improvement in the diffraction limits from these crystals. Examination of the final refined structure reveals that a single manganese ion (as characterized by a 12σ peak in difference Fourier maps) engages Lys-21 and Asp-24 (Figure 2C2C).). This results in a fixed conformation for the side chains of both residues, which would presumably be otherwise flexible, resulting in a local reduction in conformational heterogeneity of the sample. This conformationally homogenous surface patch can then form a stable crystal contact (Figure 2D2D).). The use of manganese ions (and divalent metals in general) may be a generally applicable alternative or a suitable complement to methods such as surface entropy–reduction mutagenesis or reductive methylation towards the production of diffraction-quality crystals.
We note that manganese is not present in any of the commercially available crystallization screens from Hampton Research or Emerald Biostructures, but is available only as an additive optimization reagent (Hampton Research Additive Screen 1, Condition #7). However, magnesium is present in numerous screens. Mn+2, as a softer Lewis acid, has a much higher affinity for softer Lewis bases, such as amino groups, than Mg+2.13 In the case of the IRAK-4 DD, Mg+2 would have been able to coordinate to Asp-24 by virtue of the stronger Lewis base offered by the oxygen, but it is unlikely that Lys-21 would be able to coordinate to Mg+2. The greater affinity that Mn+2 has for nitrogen-containing coordination environments lends itself to reducing the conformational heterogeneity of Lys residues where the utility of Mg+2 might be limited. Thus manganese might offer a way to easily reduce surface entropy and increase the diffraction quality of crystals or favor crystallization. Given the rapid and inexpensive nature of screening additives, the use of manganese ions may also be a valuable asset in the context of high-throughput structural genomics initiatives.
We would like to thank the staff of beamlines 14-ID (K. Brister) and 32-ID (J. Brunzelle) at APS for their support. In addition, we are also grateful to D.L. Cudiamat for technical assistance and J.J. Yu and T.A. Graham for critical comments on this manuscript. This work was supported by start-up funds and the Institute for Genomic Biology, University of Illinois, Urbana-Champaign (to S.K.N.). M.V.L. was supported by an individual National Institutes of Health National Research Service Award 1 F30 NS 048779-01 from the National Institute of Neurological Disorders and Stroke.
Present affiliations: University of Chicago, Chicago, Illinois (S.M.K. and M.M.G.)