We describe protocols for the structure determination of the IKK-binding domain of NEMO by X-ray crystallography. The methods include protein expression, purification and characterization as well as strategies for successful crystal optimization and structure determination of the protein in its unbound form.
NEMO is a scaffolding protein which plays an essential role in the NF-κB pathway by assembling the IKK-complex with the kinases IKKα and IKKβ. Upon activation, the IKK complex phosphorylates the IκB molecules leading to NF-κB nuclear translocation and activation of target genes. Inhibition of the NEMO/IKK interaction is an attractive therapeutic paradigm for the modulation of NF-κB pathway activity, making NEMO a target for inhibitors design and discovery. To facilitate the process of discovery and optimization of NEMO inhibitors, we engineered an improved construct of the IKK-binding domain of NEMO that would allow for structure determination of the protein in the apo form and while bound to small molecular weight inhibitors. Here, we present the strategy utilized for the design, expression and structural characterization of the IKK-binding domain of NEMO. The protein is expressed in E. coli cells, solubilized under denaturing conditions and purified through three chromatographic steps. We discuss the protocols for obtaining crystals for structure determination and describe data acquisition and analysis strategies. The protocols will find wide applicability to the structure determination of complexes of NEMO and small molecule inhibitors.
The NF-κB pathway is activated in response to a variety of stimuli, including cytokines, microbial products and stress, to regulate expression of target genes responsible for inflammatory and immune response, cell death or survival and proliferation1. Pathologies including inflammatory and autoimmune diseases and cancer2,3,4,5 have been correlated to hyperactivation of the pathway, which has made modulation of NF-κB activity a prime target for the development of new therapies6,7.
The canonical NF-κB pathway in particular is distinguished from the non-canonical pathway, responsible for lymphorganogenesis and B-cell activation, by the former's dependence on the scaffolding protein NEMO (NF-κB essential modulator8) for the assembly of the IKK-complex with the kinases IKKα and IKKβ. The IKK complex is responsible for the phosphorylation of IκBα (inhibitor of κB) that targets it for degradation, freeing the NF-κB dimers to translocate to the nucleus for gene transcription1 and is therefore an attractive target for the development of inhibitors to modulate NF-κB activity.
Our research focuses on the characterization of the protein-protein interaction between NEMO and IKKβ, targeting NEMO for the development of small molecules inhibitors of IKK complex formation. The minimal binding domain of NEMO, required to bind IKKβ, encompasses residues 44-111, and its structure has been determined in complex with a peptide corresponding to IKKβ sequence 701-7459. NEMO and IKKβ form a four-helix bundle where the NEMO dimer accommodates the two helices of IKKβ(701-745) in an elongated open groove with an extended interaction interface. IKKβ(734-742), also known as the NEMO-binding domain (NBD), defines the most important hot-spot for binding, where the two essential tryptophans (739,741) bury deeply within the NEMO pocket. The details of the complex structure can aid in the structure-based design and optimization of small molecule inhibitors targeting NEMO. At the same time, it is difficult that binding of a small molecule or peptide would recreate in NEMO the full conformational change (i.e., extensive opening of the NEMO coiled-coil dimer) caused by binding of the long IKKβ(701-745), as observed in the crystal, and the structure of unbound NEMO or NEMO bound to a small molecule inhibitor may represent a better target for structure-based drug design and inhibitor optimization.
Full length NEMO and smaller truncation constructs encompassing the IKK-binding domain have proven intractable for structure determination in the unbound form via X-ray crystallography and nuclear magnetic resonance (NMR) methods10, which prompted us to design an improved version of the IKK-binding domain of NEMO. Indeed, NEMO (44-111) in the unbound form is only partially folded and undergoes conformational exchange and we therefore set to stabilize its dimeric structure, coiled-coil fold and stability, while preserving binding affinity for IKKβ. By appending three heptads of ideal dimeric coiled-coil sequences11 at the N-and C-termini of the protein, and a series of four point mutations, we generated NEMO-EEAA, a construct fully dimeric and folded in a coiled coil, which rescued IKK-binding affinity to the nanomolar range as observed for full length NEMO12. As an additional advantage, we hoped the coiled-coil adaptors (based on the GCN4 sequence) would facilitate crystallization and eventually aid in the X-ray structure determination via molecular replacement. Coiled-coil adaptors have been similarly utilized to both increase stability, improve solution behavior and facilitate crystallization for trimeric coiled coils and antibody fragments13,14. NEMO-EEAA is easily expressed and purified from Escherichia. coli cells with a cleavable Histidine tag, is soluble, folded in a stable dimeric coiled coil and is easily crystallized, with diffraction to 1.9 Å. The presence of the ordered coiled-coil regions of GCN4 could additionally aid in phasing the data from crystals of NEMO-EEAA by molecular replacement using the known structure of GCN415.
Given the results obtained with apo-NEMO-EEAA, we believe the protocols described here could also be applied to the crystallization of NEMO-EEAA in the presence of small peptides (like the NBD peptide) or small molecule inhibitors, with the goal of understanding the requirements for NEMO inhibition and structure-based optimization of initial lead inhibitors to high affinity. Given the plasticity and dynamic nature of many coiled-coil domains16, the use of the coiled-coil adaptors could find more general applicability in aiding structural determination.
1. Design of construct for crystallography
2. Large scale expression of His6 tagged NEMO-EEAA
3. Purification of His6 tagged NEMO-EEAA
4. His6 tag cleavage and purification
5. Sparse matrix screening
NOTE: The protocol performs crystallization trials using commercially available screens and setting up sitting drop experiments using a crystallization robot. Crystal images are collected automatically by an imager.
6. Seed stock generation
NOTE: We reproducibly obtain crystals for seed generation in 0.1 M Tris pH 8.0, 5% PGA-LM, 3.6% w/v PEG 20k. However, crystals will show high mosaicity and are unsuitable for data collection at this stage.
7. Fine screens
8. Generation of crystals for data collection
9. Determination of cryo-protectant
10. Crystal looping
11. Data collection
12. X-ray data processing
13. Structure solution
14. Structure refinement
Cloning, expression and purification of the IKK-binding domain of NEMO.
The protocol followed in this study to obtain the final sequence of NEMO-EEAA (Figure 1A), which produced diffraction quality crystals, involved the expression and characterization of all the intermediate constructs, including the addition of the coiled-coil adaptors at N- and or C-terminus, the mutations C76A, C95S and the mutations E56A, E57A. Figure 1B displays an SDS-PAGE gel of samples collected throughout the purification procedure as described in the Scheme in Figure 1C. The protein is overexpressed in E. coli and appears as a band approximately at the 14 kDa MW weight marker on the SDS-PAGE gel (lane 3, cells collected at harvest and lysed in Laemmli sample buffer supplemented by 8 M urea). The protein appears pure after the first IMAC column and displays a monomer and a dimer band at the level of the 14 and 28 kDa MW markers on the SDS-PAGE gel (lane 9). TEV cleavage is practically complete following the protocol and the protein elutes with the flow through during the second IMAC column almost entirely as a dimer, at the expected MW (band below 28 kDa). Size exclusion chromatography displays a single peak eluting between 60-65 mL and corresponding in our experience to the dimer (Figure 1D). The dimeric coiled coil always elutes earlier than expected on SEC due to the elongated shape of the coiled coil and the consequently large hydrodynamic radius10. NEMO-EEAA in fractions from SEC peak still appears as a monomer and a dimer on SDS-Page gel (lanes 14-15). Utilizing a stirred-cell concentrator is important to prevent sample possible aggregation and precipitation upon concentration.
Crystallization of NEMO-EEAA
Initial crystals were obtained from a commercial screen using PGA (see Table of Materials), utilizing 1.65 µg/mL of NEMO-EEAA in 2 mM Tris, 100 mM NaCl, 2 mM DTT, pH 8.0. Fine screening produced crystals in 0.1 M Tris pH 8.9, 5% PGA-LM, 3.6% PEG 20k (Figure 2A), which were utilized to produce a seed stock. Final crystals were obtained with seeding in 0.1 M Tris pH 8.0, 5.8% PGA-LM, 5.45% PEG 20k (Figure 2B).
Data collection and structure determination
NEMO-EEAA crystals suffer from mosaicity and anisotropy. Crystals formed in the P 1 21 1 space group, with data resolution varying in the a*, b* and c* axis (1.88 Å, 2.10 Å and 2.55 Å). Examples of diffraction profiles are in Figure 3. Data was acquired at the AMX (17-ID-1) beamline of NSLS II, with a beam size of 7 x 5 μm2. The small beam size was essential to focus on the desired portion of the crystal (Figure 2B) and ensure data quality.
The successful protocol for structure determination requires anisotropic truncation of the data using the STARANISO27 server (http://staraniso.globalphasing.org/cgi-bin/staraniso.cgi), followed by phasing by molecular replacement using MRage21 within PHENIX22 and the structure of dimeric GCN4 (PDB: 4DMD)15 as a search model. In the solution of this structure phasing was initially attempted by labeling the native methionine with SeMet. The anomalous signal was too weak, probably due to the solvent exposed nature of Met95 in the unbound form of NEMO (SeMet95 was successfully used for phasing of the NEMO / IKKβ structure9).
Initial data analysis was attempted with spherical truncation of the data to 2.3 Å. This data set could not be successfully phased by molecular replacement but a solution was obtained using MR-ROSETTA25 and the structure of NEMO(44-111) in complex with IKKβ(701-745) as a search model (PDB: 3BRV)9. This initial model could not be refined successfully. We utilized the Diffraction Anisotropy Server at UCLA26 to elliptically truncate the data and to remove anisotropy by anisotropic scaling. The newly processed dataset could be phased by molecular replacement. To further improve the data, we employed the STARANISO27 server for anisotropic truncation of the data and amplitudes were corrected with an anisotropic correction factor with Bayesian estimation28. This data, resulting in an increase in the number of unique reflections to 19,560, was used for structure refinement. Composite omit maps where calculated with PHENIX, excluding 10% of the atoms at a time, and compared with the final structure model, to confirm that the structure was not biased by the atomic model. The crystallographic model is complete except for the first and last residues in chain A of NEMO-EEAA, for which no electron density was observed.
NEMO-EEAA is a homo-dimeric, irregular, parallel coiled coil of ~175 Å in length. The regular coiled-coil region encompasses the ideal coiled-coil adaptor sequence at the N-terminus (residues 20-50) and the first two heptads of the NEMO proper sequence (residues 51-65). A regular coiled coil is also present at the C-terminus, starting at NEMO residue 97 and encompassing the C-terminal ideal coil-coil adaptor (Figure 4A). The central portion, encompassing NEMO residues 66-98 displays larger interhelical distances (interhelical spacing goes from an average value of 7.6 Å in the regular coiled-coil structure to a maximum of 11.5 Å in the irregular region) and discontinuous hydrophobic interface compared with an ideal coiled coil. This region of NEMO also represents the interface for binding IKKβ and undergoes a conformational change upon ligand binding. The IKKβ-bound structure displays a more open coiled-coil conformation to accommodate the ligand with larger interhelical spacing by 1.0 to 2.2 Å in this region (Figure 4B)12. In the apo structure residues Leu93, Met94, Lys96, Phe97 and Arg101 are shifted toward the center of the coiled coil to invade the ligand binding pocket (the Cα-Cα distance for Phe97 is tighter by 2.9 Å), with the overall effect of closing the large hydrophobic cleft which hosts the ligand in the IKKβ-bound structure.
Figure 1: Purification of NEMO-EEAA. (A) Sequence alignment of NEMO from Homo Sapiens, Mus Musculus and Bos Bovis and the engineered NEMO-EEAA. The coiled-coil adaptors sequence is underlined, and the mutations are highlighted in orange. (B) SDS-PAGE gel of fractions collected during expression and purification (as labeled and referenced in flow chart 1C). 10% Acrylamide, MES buffer. (C) A flow chart for the purification of NEMO-EEAA. (D) Size-exclusion profile indicating the dimer of NEMO-EEAA (blue). In green the molecular weight markers (kDa) are indicated. Please click here to view a larger version of this figure.
Figure 2: Morphology and size of NEMO-EEAA crystals. (A) A crystal of NEMO-EEAA from the initial sparse matrix screen, unseeded. This type of crystal was used for seeding for subsequent crystallization attempts (0.1 M Tris pH 8.0, 5% PGA-LM, 3.6% PEG 20k; with a protein solution in 2 mM Tris, 100 mM NaCl 2 mM DTT, pH 8.0). (B) Crystal used for data collection (0.1 M Tris pH 8.0, 5.8% PGA-LM, 5.45% PEG 20k, with a protein solution in 2 mM Tris, 100 mM NaCl 2 mM DTT, pH 8.0). The approximate area used for data collection is circled in red. The scale bar for 100 µm length is defined in the figure. Please click here to view a larger version of this figure.
Figure 3: Crystallographic data obtained from NEMO crystals. (A) X-ray diffraction profile of early NEMO crystal: elongated streaks indicate mosaicity in the crystal. (B) X-ray diffraction profile of an optimized NEMO-EEAA crystal. The resolution ring is drawn at a spatial resolution of 2.5 Å. Please click here to view a larger version of this figure.
Figure 4: Structure of unbound NEMO. (A) The NEMO-EEAA dimer is shown as a ribbon, light blue = coiled-coil adaptors, blue = NEMO(51-112). (B) Superposition of apo NEMO-EEAA structure (blue, PDB: 6MI3) and NEMO(44-111) from the IKKβ-bound structure (grey, PDB: 3BRV, IKKβ is not displayed), shown as ribbons; the structures are aligned on chain A, region 44-111 only. This figure has been modified from previous publication12. Please click here to view a larger version of this figure.
Crystallization attempts of NEMO in the unbound form were unsuccessful, including attempts using the full-length protein and several truncation constructs encompassing the IKK-binding domain. Our biophysical characterization of the IKK-binding domain of NEMO (residues 44-111) by circular dichroism, NMR spectroscopy and fluorescence anisotropy indicated that the construct, albeit able to bind IKKβ, existed in a state of conformational exchange, not suitable for crystallization9,10. Our approach involved engineering a construct of the IKK-binding domain of NEMO that adopted a more stable, folded and native-like conformation, eliminating the flexibility that prevented crystallization. Ideal coiled-coil domain adaptors fused to the N- and C-termini of the NEMO(44-111) sequence offer the advantage of stabilizing the dimeric coiled-coil fold of the native NEMO and facilitating crystallization14. Behavior of a series of protein constructs was monitored at each step through SDS-PAGE, size exclusion chromatography, circular dichroism, fluorescence anisotropy and NMR spectroscopy, as described earlier10,12. Despite the progressive improvement in all desired parameters no construct yielded diffraction quality crystals until the mutations E56A, E57A were introduced. The mutations were the highest impact mutations suggested by the Surface Entropy Reduction Server SERp29 that did not involve residues implicated in hot-spots of binding for IKKβ, as in the NEMO / IKKβ complex structure. Although other constructs stabilizing the ordered structure of the IKK-binding domain of NEMO through cysteine disulfide linkage and binding IKKβ with high affinity have been described30, the protocol here described represents the first successful approach to structure determination of the IKK-binding domain of NEMO in the unbound form.
Protein production and purification followed standard procedures in our laboratory. Upon cell lysis a considerable portion of the protein is present in the insoluble fraction, even when growing the cells after induction at 18 °C, therefore the protein was solubilized in 8 M urea after cell lysis and prior to purification, and refolded while bound to the IMAC column. Extensive washing of the column bound protein both under denaturing conditions and after refolding are critical for a pure product after the first purification step. The pure NEMO-EEAA has a higher tendency to precipitate than the previous constructs but is still sufficiently soluble under crystallization conditions.
A critical step in the structure determination was the use of seeding. In an approach similar to microseed matrix screening31, seeds from crystal obtained during sparse matrix screening were used in an additional screening of conditions, followed by a fine screening round, in order to determine final crystallization conditions. Most of the crystals grown in the final conditions display high mosaicity and are not suitable for data collection. Crystals were screened over several visits to the beamline and best data quality was achieved by collecting datasets at the edge of the crystal, where the newest growth had occurred. As the region which showed lattice uniformity in cross-polarized images was small, it was instrumental to have access to the AMX beamline at NSLS II, due to the small beam size.
We successfully determined the structure of NEMO-EEAA by molecular replacement using as a search model the structure of the dimeric coiled-coil of GCN415, which mapped to the ideal coiled-coil adaptors fused at the N- and C-termini of the NEMO sequence. The molecular replacement was successful as the GCN4 adaptors maintain their native structure when fused to NEMO. At the same time, we could verify that NEMO maintains its native like structure by comparing the portions not involved in IKK-binding with the corresponding structure region in the NEMO / IKKβ complex structure. As an alternative, it could be possible to use the structure of NEMO in the NEMO / IKKβ complex (PDB: 3BRV)9 as a molecular replacement search model, focusing on monomer B, which corresponds to the chain which undergoes the smallest conformational change upon ligand binding. This strategy showed some initial success using the MR-ROSETTA module of PHENIX.
Finally, it was essential for a successful structure determination to utilize all available data in the dataset, by resorting to an anisotropic cut-off of merged intensity data, as provided by the STARANISO27 server or the Diffraction Anisotropy Server at UCLA26.
The essential role played by the NEMO/IKK complex in the NF-kB pathway makes it a desirable target for modulation of the pathway for therapeutic intervention. Protein-protein interactions, especially when involving a large binding interface, are challenging to inhibit with small peptides or small molecules, and structure-based inhibitor design can offer a significant advantage. The constructs of the IKK-binding domain of NEMO that we developed overcome the limits of the flexible NEMO(44-111) to facilitate crystallization and structure determination. As the construct easily crystallizes in the unbound form, we envision that the same protocol can be successfully applied in the crystallization of the complex of NEMO with small molecule inhibitors, to determine a structure that would provide details on binding modes and allow further ligand improvements.
An analysis of crystal packing in the crystallization conditions achieved in this work indicates that ligand co-crystallization may be preferred to ligand soaking into apo-protein crystals. While crystal packing provides some space around chain B of the dimer (6 to 10 Å to the nearest chain) and on one face of the ligand binding site (approximately 13 Å in the hot-spot region between Phe82-Phe87), the symmetric binding pocket is completely occluded by nearby chains, preventing ligand binding.
Coiled coils are present in a significant proportion in the proteome and are essential in their function as molecular spacers, scaffolds and in communicating conformational changes16. Despite their abundance and relevant roles, there are relatively few structures of full-length coiled-coil proteins. The use of coiled-coil adaptors may aid in the stabilization of structural domains of coiled-coil proteins while preserving their native structure and aid in the structural determination and elucidation of their function.
The authors have nothing to disclose.
We thank Prof. D. Madden, for many helpful discussions throughout this project. We thank Prof. D. Bolon for the gift of the plasmid containing the optimized GCN4 coiled coil. We thank Dr. B. Guo for NEMO plasmids. We thank Christina R. Arnoldy, Tamar Basiashvili and Amy E. Kennedy for demonstrating the procedure. We thank the BioMT Crystallography Core Facility and the departments of Chemistry and of Biochemistry & Cell Biology at Dartmouth for the use of the crystallography equipment and the BioMT personnel for their support. This research used the AMX beamline of the National Synchrotron Light Source II, a U.S. Department of Energy (DOE) Office of Science User Facility operated for the DOE Office of Science by Brookhaven National Laboratory under Contract No. DE-SC0012704. We thank the staff at NSLS II for their support. This work was funded by NIH grants R03AR066130, R01GM133844, R35GM128663 and P20GM113132, and a Munck-Pfefferkorn Novel and Interactive grant.
20% w/v γ-PGA (Na+ form, LM) | Molecular Dimensions | MD2-100-150 | For fine screen crystallization of NEMO-EEAA |
3.5 kDa MWCO Dialysis Membrane | Spectra/Por | 132724 | For dialysis removal of imidazole |
Amicon Stirred Cell | Millipose Sigma | UFSC 05001 | For protein concentration |
Ammonium Chloride | Millipore Sigma | G8270 | For minimal media labeling |
Benzonase Nuclease | Millipore Sigma | 9025-65-4 | For digestion of nucleic acid |
BL21-CodonPlus (DE3)-RIL Competent Cells | Agilent Technologies | Model: 230245 | TEV expression |
CryoPro | Hampton Research | HR2-073 | Cryo-protectants kit |
D-Glucose (Dextrose) | Millipore Sigma | A9434 | For minimal media labeling |
Difco Terrific Broth | ThermoFisher | DF043817 | For culture growth |
Dithiothreitol > 99% | Goldbio | DTT25 | For reduction of disulfides |
E. coli: Rosetta 2 (DE3) | Novagen | 71400-3 | Expression of unlabeled NEMO-EEAA |
FORMULATOR | Formulatrix | Liquid handler/ screen builder | |
HCl – 1.0 M Solution | Hampton Research | HR2-581 | For fine screen crystallization of NEMO-EEAA |
HiLoad 16/600 Superdex 75 pg | GE Healthcare | 28989333 | For size exclusion purification |
HisTrap HP 5 mL column | GE Healthcare | 17524802 | For purification of His-tagged NEMO-EEAA |
HT 96 MIDAS | Molecular Dimensions | MD1-59 | For sparse matrix screening of NEMO-EEAA |
HT 96 Morpheous | Molecular Dimensions | MD1-46 | For sparse matrix screening of NEMO-EEAA |
Imidazole | ThermoFisher | 288-32-4 | For elution from His-trap column |
Isopropyl-beta-D-thiogalactoside | Goldbio | I2481C5 | For induction of cultures |
MRC2 crystallization plate | Hampton Research | HR3-083 | Crystallization plate |
NT8 – Drop Setter | Formulatrix | Crystallization | |
pET-16b | Millipore Sigma | 69662 | For cloning of NEMO-EEAA |
pET-45b | Millipore Sigma | 71327 | For cloning of NEMO-EEAA |
Phenylmethylsulfonyl fluoride | ThermoFisher | 36978 | For inhibition of proteases |
Polycarbonate Bottle for use in Ultracentrifuge Rotor Type 45 Ti | Beckmann Coulter | 339160 | Ultracentrifuge bottle |
Polyethylene Glycol 20,000 | Hampton Research | HR2-609 | For fine screen crystallization of NEMO-EEAA |
pRK793 (TEV) | Addgene | Plasmid 8827 | For TEV production |
QuikChange XL II | Agilent Technologies | 200522 | Site directed mutagenesis |
Required Cap Assembly: | Beckmann Coulter | 355623 | Ultracenttrifuge bottle cap |
ROCK IMAGER | Formulatrix | Crystallization Imager | |
Seed Bead Kit | Hampton Research | HR2-320 | Seed generation |
Sodium Chloride ≥ 99% | Millipore Sigma | S9888 | For buffering of purification solutions |
TCEP (Tris (2-Carboxyethyl) phosphine Hydrochloride) | Goldbio | TCEP1 | Reducing agent |
The Berkeley Screen | Rigaku | MD15-Berekely | For sparse matrix screening of NEMO-EEAA |
The PGA Screen | Molecular Dimensions | MD1-50 | For fine screen crystallization of NEMO-EEAA |
Tris – 1.0 M Solution | Hampton Research | HR2-589 | For fine screen crystallization of NEMO-EEAA |
Ultrapure Tris Buffer (powder format) | Thermofisher | 15504020 | For buffering of purification solutions |
Urea | ThermoFisher | 29700 | For denaturation of NEMO-EEAA |