Identification of the activator-binding residues in the second cysteine-rich regulatory domain of protein kinase Cθ (PKCθ)

Abstract

PKC (protein kinase C) θ is predominantly expressed in T-cells and is critically involved in immunity. Design of PKCθ-selective molecules to manage autoimmune disorders by targeting its activator-binding C1 domain requires the knowledge of its structure and the activator-binding residues. The C1 domain consists of twin C1 domains, C1A and C1B, of which C1B plays a critical role in the membrane translocation and activation of PKCθ. In the present study we determined the crystal structure of PKCθC1B to 1.63 Å (1 Å = 0.1 nm) resolution, which showed that Trp²⁵³ at the rim of the activator-binding pocket was orientated towards the membrane, whereas in PKCδC1B the homologous tryptophan residue was orientated away from the membrane. This particular orientation of Trp²⁵³ affects the size of the activator-binding pocket and the membrane affinity. To further probe the structural constraints on activator-binding, five residues lining the activator-binding site were mutated (Y239A, T243A, W253G, L255G and Q258G) and the binding affinities of the PKCθC1B mutants were measured. These mutants showed reduced binding affinities for phorbol ester [PDBu (phorbol 12,13-dibutyrate)] and diacylglycerol [DOG (sn-1,2-dioctanoylglycerol), SAG (sn-1-stearoyl 2-arachidonyl glycerol)]. All five full-length PKCθ mutants exhibited reduced phorbol-ester-induced membrane translocation compared with the wild-type. These results provide insights into the PKCθ activator-binding domain, which will aid in future design of PKCθ-selective molecules.

INTRODUCTION

PKC (protein kinase C) θ is a serine/threonine kinase [1]. In humans it is encoded by the PRKCQ gene and is highly expressed in T-cells and thymocytes [2]. PKCθ plays a crucial role in T-cell activation and is critically involved in immunity. PKCθ is linked to the TCR (T-cell receptor) signalling complex, which in turn activates transcription factors such as NF-κB (nuclear factor κB) and AP-1 (activator protein 1), mediating immune responses [3,4]. An overactive immune system is responsible for different immunity-related complications such as arthritis [5], asthma [6], multiple sclerosis [7,8], Type 1 diabetes [9] etc. PKCθ-knockout mice were found to be protected from asthma or arthritis in response to antigen challenge, indicating the importance of PKCθ in the pathway of downstream immune responses [5,7,10,11].

In the stepwise activation process of T-cells, macrophages first respond to APCs (antigen-presenting cells), which present the antigens derived from bacteria, viruses, parasites, prions etc. The antigenic proteins generated by the digestion of APC debris bind to the MHC present on the macrophage cell surface. T-cells are then activated by the stable interaction of TCR molecules with the MHC-bound antigenic proteins. Activated TCR molecules, upon clustering at the immunological synapse, trigger the activation of PLCγ 1 (phospholipase Cγ 1) at the T-cell membrane, leading to formation of DAG (diacylglycerol) and Ins(1,4,5)P₃ from PtdIns(4,5)P₂ [12]. DAG binds to PKCθ,which results in the translocation of active PKCθ to the immunological synapse. Activated PKCθ at the immunological synapse, in turn, mediates the activation of the downstream transcription factors associated with the immune response, such as NF-κB, NFAT (nuclear factor of activated T-cells) and AP-1, and production of interleukin-2 [3,4,13].

The PKC superfamily comprises 11 subtypes, which have been categorized as typical and atypical. Typical C1 domains bind the endogenous activator DAG or the ultra-potent phorbol esters, whereas atypical ones are insensitive to them. The typical PKCs are further divided into the conventional (α, βI, βII and γ) and novel (δ, ε, θ and η) classes, each having four primary domains. The N-terminal regulatory region has C1 and C2 domains, whereas the highly homologous C-terminal kinase region consists of C3 and C4 domains (Supplementary Figure S1 at http://www.biochemj.org/bj/451/bj4510033add.htm) [14]. DAG is sufficient to activate the novel PKCs, whereas the conventional PKCs additionally require Ca²⁺ for their activation. DAG is a lipid second messenger that selectively interacts with proteins containing a C1 domain and induces their translocation to discrete subcellular compartments such as the plasma membrane, Golgi etc. In both conventional and novel PKCs, the DAG/phorbol-ester-responsive C1 domain consists of a tandem repeat of highly conserved cysteine-rich zinc-finger subdomains known as C1A and C1B. These subdomains show significant differences in their binding affinities for phorbol esters and DAG [15]. Atypical PKCs (ζ and ί/λ) have a single non-DAG-binding C1 domain [14,16,17].

Several PKC isoforms (α, β1, δ, ε, η, θ and ζ) are expressed in T-cells, but only PKCθ stably translocates to the immunological synapse. Other PKCs translocate to other regions of the plasma membrane. Selective inhibition of PKCθ, therefore, could be an attractive approach in the management of a hyperactive immune system, leading to suppression of related disorders [18].

Design of PKCθ inhibitors targeting the kinase region faces a major challenge in achieving selectivity because there is a high degree of homology in the kinase region among the more than 500 kinases in the human genome [8,19]. On the other hand, there are fewer C1 domains (less than 30 phorbol ester/DAG-responsive C1 domains). Furthermore, PKCθ is unique among the conventional and novel PKC isoforms in that PKCθC1B is the only C1B subdomain that binds to DAG with higher affinity than the C1A subdomain [20]. Likewise, PKCθC1B plays the predominant role in the membrane translocation and activation process of PKCθ [21].

In the present study, we report the crystal structure of the PKCθC1B subdomain at 1.63 Å (1 Å = 0.1 nm) resolution. To identify the activator-binding residues, we performed virtual molecular docking to the PKCθC1B domain of a small library of DAG and phorbol ester analogues using the PKCθC1B crystal structure. This docking predicted that five residues, Tyr²³⁹, Thr²⁴³, Trp²⁵³, Leu²⁵⁵ and Gln²⁵⁸, were particularly important for ligand binding. A sequence alignment of PKCθC1B with PKCδC1B showed that all of these five residues were present in the homologous activator-binding region of PKCδC1B [22]. We confirmed that mutation of PKCθC1B in these residues, specifically Y239A, T243A, W253G, L255G and Q258G, reduced affinity for both phorbol ester and DAG. Likewise, in the full-length PKCθ, these mutations caused significantly reduced membrane translocation by phorbol ester treatment.

EXPERIMENTAL

Chemicals, reagents and antibodies

Escherichia coli BL21-Gold (DE3) competent cells were from Stratagene. SOC (super optimal culture) medium, LB (Luria-Bertani) broth medium and TB (terrific broth) medium were from Invitrogen. Glycerol, IPTG (isopropyl β-d-thiogalactopyranoside), zinc sulphate, ampicillin, lysozyme, Triton X-100, polyethyleneimine, sodium chloride, γ-globulins (G5009), Tween 20 and methanol were from Sigma–Aldrich. Thrombin, glutathione–Sepharose 4B and Superdex 75 prepacked columns were from GE Healthcare Biosciences. PMA and PDBu (phorbol 12,13-dibutyrate) were from LC Laboratories. DOG (sn-l,2-dioctanoylglycerol), SAG (sn-1-stearoyl 2-arachidonyl glycerol), PS (L-α-phosphatidylserine) and PC (l-α-phosphatidylcholine) were from Avanti Polar Lipids. Ammonium sulfate, poly(ethylene glycol) 6000 and BSA were from EMD Chemicals. Sapintoxin D was from Alexis Biochemicals. [20-³H]PDBu was custom synthesized by PerkinElmer Life Sciences. DMEM (Dulbecco’s modified Eagle’s medium) high glucose, FBS (fetal bovine serum) and antibiotic/antimycotic concentrate used for the cell culture were from Gibco. Rabbit anti-GFP (green fluorescent protein) antibody and HRP (horseradish peroxidase)-conjugated anti(rabbit IgG) antibody used for the Western blot analysis were purchased from Cell Signaling, and SuperSignal West Femto Maximum Sensitivity Substrate used for chemiluminescence was obtained from Thermo Scientific.

Expression vector construction

A bacterial expression vector of PKCθC1B with an N-terminal GST (glutathione transferase) tag was generated by subcloning the domain sequence (amino acids 232–281) from mouse PKCθ cDNA into the pGEX2TK vector (GE Healthcare Biosciences) between the restriction sites BamHI and EcoRI [23]. The GST-fused protein contained a thrombin cleavage site between GST and the C1B subdomain. A mammalian expression vector of mouse PKCθ with an EGFP (enhanced GFP) tag at the C-terminus was generated by subcloning the corresponding genes with the spacer sequence GGNSGG into the pEGFPNl vector (Addgene) between restriction sites BamHI and XhoI as described previously [24]. Dr Amnon Altman (La lolla Institute for Allergy and Immunology, La lolla, CA, U.S.A.) provided the original PKCθ full-length vector.

Generation of PKCθC1B and PKCθ mutants

The single point mutations (Y239A, T243A, W253G, L255G and Q258G) in the truncated PKCθC1B subdomain as well as in the full-length PKCθ were introduced by PCR using the QuikChange® Site-Directed Mutagenesis kit (Stratagene) [25–27]. Oligonucleotide primers were custom synthesized by Seqwright. The mutations were checked for their correct sequences using the sequencing facility at Seqwright.

Protein expression, purification and characterization

The PKCθC1B subdomain and its mutants fused with N-terminal GST were expressed in the E. coli strain BL21-Gold (DE3) (Stratagene). Expression and purification were performed following the methods described previously [28–30]. The purity of the proteins was checked by SDS/PAGE (15% gel). The exact molecular mass of the proteins was measured by MALDI–TOF-MS (matrix-assisted laser-desorption ionization–time-of-flight MS; Voyager DE-STR Biospectrometry Workstation; Applied Biosystems). Proteins were further characterized by their binding to SAPD (sapintoxin D; a fluorescent phorbol ester) using fluorescence spectroscopy (QuantaMaster™ 30; PTI) [31] in the absence of any added lipid.

Cell culture for Western blotting

A HEK (human embryonic kidney)-293 cell line expressing the ecdysone receptor (Invitrogen) was used to express the full-length PKCθ and its C1B subdomain mutants. The cells were cultured in plastic Petri dishes in DMEM supplemented with 10% FBS and 1% antibiotic/antimycotic solution at 37 °C in 5% CO₂ and 98% humidity. The confluent cells (90%) were transfected by adding 1 ml of unsupplemented DMEM containing 6 μg of endotoxin-free plasmid and 9 μl of Plus reagent (Invitrogen) and 9 μl of Lipofectamine™ reagent (Invitrogen) for 12 h at 37°C. The transfection medium was removed and the cells were washed with 1× PBS followed by the addition of antibiotic-free DMEM supplemented with FBS. Protein expression was induced for 48 h from the time of transfection before the cells were ready for membrane translocation studies [21,32]. The expression of the full-length PKCθ and its mutants was detected by Western blot analysis of the cell lysate 48 h post-transfection using rabbit anti-GFP primary antibody and HRP-conjugated anti-(rabbit IgG) secondary antibody.

Crystallization, data collection and refinement

The hanging-drop vapour-diffusion method was used to grow PKCθC1B crystals. The protein was concentrated to 25 mg/ml in gel filtration buffer (50 mM Tris, 150 mM NaCl, 2 mM dithiothreitol and 50 μM ZnSO₄, pH 7.2) and was mixed at a 1:1 ratio of reservoir solution consisting of 0.1 M lithium acetate and 20% poly(ethylene glycol) 3350 at 20°C. Protein crystals appeared in 7–10 days. Crystals were cryo-protected using 10% glycerol in mother liquor followed by flash-freezing in liquid nitrogen [33]. Data were collected at the home source RIGAKU FR-E⁺ (Rigaku) at Baylor College of Medicine (Houston, TX, U.S.A.). Data were processed using Mosflm and programs implemented in the CCP4 suite [34]. The space group of the crystal was confirmed by the program POINTLESS [35]. The initial phase problem was solved with the molecular replacement program PHASER [36] using the available structure (PDB code 1PTQ) followed by automated model building and solvent addition using ARP/wARP [37]. Further model refinement was carried out with Refmac (CCP4) [38] and COOT [39]. Final co-ordinates were deposited into the PDB under the accession code 4FKD.

Molecular docking

Three-dimensional chemical structures of DAG and phorbol ester analogues were downloaded from PubChem and were subjected to pre-dock energy minimization using the Powell method in Sybyl 8.0 (Tripos International) with an iteration of 100. These analogues were incorporated together into a single library ready to be docked into the crystal structure of PKCθC1B. A residue-based molecular docking was done in the SurflexDock module of Sybyl 8.0 using a threshold of 0.5, a bloat of 2.0 and a radius of 3 Å for the protomol generation. Protomol is the geometric constraint developed computationally where the ligands were docked. Residues Tyr²³⁹, Lys²⁴⁰, Ser²⁴¹, Pro²⁴², Thr²⁴³, Phe²⁴⁴, Leu²⁵¹, Leu²⁵², Trp²⁵³, Gly²⁵⁴, Leu²⁵⁵ and Glu²⁵⁸ were used to generate the protomol. These residues were selected by comparison with the phorbol ester binding site in PKCδC1B. Ring flexibility, soft grid treatment, pre-dock and post-dock energy minimization were also applied for the docking procedure. A high total score value represents better fitting. The surface area and the volume of the activator-binding pocket of C1B domains were measured using the program CASTp [40].

[³H]PDBu binding assay

Binding of PDBu, DOG and SAG to PKCθC1B and its mutants was measured using the poly(ethylene glycol) precipitation assay method as described previously [41]. K_d values were determined by fitting the data into a two-component model of saturable receptor binding plus linear non-specific binding [41]. Binding affinities of DOG and SAG to PKCθC1B and its mutants were determined by competition of [³H]PDBu binding using the above conditions, but with [³H]PDBu at 2–5-fold the concentration of the K_d for each mutant and with a series of concentrations of DOG or SAG (1–100000 nM). The DOG or SAG was dissolved in chloroform and added to the phospholipid. The lipid was then dried down, buffer was added, and the mixture was sonicated and finally added to the assay. Sonication was performed with a Model XL2020 sonicator (Misonix) at room temperature with four pulses of 5 s using a microtip sonicator probe at 5% power. For analysis of the dependence on the phospholipid composition, total phospholipid was maintained at 100 μg/ml. The final concentration of PS was varied from 0 to 100% with the remainder as PC [41].

Membrane translocation of PKCθ

HEK-293 cells expressing full-length PKCθ and its C1B mutants were treated with PMA or DOG for 2 h as described previously [42]. Harvested cells were then washed with ice-cold PBS and were resuspended in lysis buffer (20 mM Tris, pH 7.4) containing protease inhibitor cocktail. Cells were lysed by multiple cycles of extrusion from and re-uptake by a 1 ml of insulin syringe with a 16 gauge needle. The cell lysate was then subjected to ultracentrifugation (76000 rev./min; Beckman TLA 120.2 rotor) for 1 h at 4 °C. The supernatant was designated as the cytosolic fraction. The pellet was then resuspended again in lysis buffer with protease inhibitor and 1% Triton X-100 and kept on ice for 1 h. The mixture was sonicated three times with a 5 s pulse using a Branson digital sonifier model 250. The mixture was then further subjected to ultracentrifugation (76000 rev./min; Beckman TLA 120.2 rotor) for 1 h at 4°C. This supernatant was designated as the membrane fraction [43]. Western blot analysis of the cell fractions was performed as described previously [44]. The proteins were transferred to a nitrocellulose (0.45 μm) membrane and were detected using an anti-EGFP primary antibody (1:5000 dilution) followed by treatment with an HRP-linked secondary antibody (1:5000 dilution) for visualization. An enhanced chemiluminescence detection system from Thermo Scientific was used to visualize the labelled protein bands. The relative intensities of the protein bands were calculated using an AlphaImager from Alpha Innotech [42].

Confocal analysis of PKCθ wild-type and different mutants

Confocal analysis of the different GFP–PKCθ constructs transfected into the LNCaP prostate cancer cell line and quantification of the images was performed as described in [45].

Statistical analysis

The significance of the data obtained (for both PKCθ wild-type and its C1B mutants) from the activator-induced membrane translocation study was assessed using an unpaired two-tail t test. P < 0.05 was considered significant.

RESULTS

Expression, purification and characterization of PKCθC1B and its mutants

The C1B domains of PKCθ and its mutants were expressed as fusion proteins with an N-terminal GST tag in E. coli. SDS/PAGE of the whole-cell lysates confirmed high levels of expression of all of the GST-fused C1B proteins. After initial affinity purification through glutathione–Sepharose 4B columns and subsequent cleavage of the GST tag, the proteins were further purified by ammonium sulfate precipitation and gel-filtration chromatography. Purified PKCθC1B and the mutants were obtained in good yields. For PKCθC1B, Y239A, T243A, W253G, L255G and Q258G, the yields were 9.6, 3.6, 2.4, 15.4, 4.3 and 3.9 mg/l of TB medium respectively. Purified proteins were thoroughly characterized by SDS/PAGE, MALDI–TOF-MS and phorbol ester binding. All of the proteins showed a single symmetrical peak upon elution from the Superdex-75 column with retention times corresponding to their Stokes radii, indicating the purity as well as the proper folding of the proteins (illustrated for wild-type PKCθC1B in Supplementary Figure S2A at http://www.biochemj.org/bj/45l/bj4510033add.htm; results not shown for the mutants). Using SDS/PAGE, all the proteins showed a single sharp band near 7 kDa, indicating the purity of the proteins (Supplementary Figure S2B). The molecular masses of the proteins were further confirmed by MAFDI–TOF-MS (illustrated for wild-type PKCθC1B in Supplementary Figure S2C; results not shown for the mutants). For PKCθC1B, Y239A, T243A, W253G, L255G and Q258G, the molecular masses were found to be 7333, 7245, 7305, 7210, 7281 and 7272 Da respectively with an accuracy of ±0.04%.

Expression and characterization of PKCθ and its C1B subdomain mutants in HEK-293 cells

Western blot analysis of the whole-cell lysate of HEK-293 cells transiently transfected with plasmids encoding PKCθ and its various mutants showed a band near 100 kDa, confirming the expression of the full-length PKCθ and its C1B subdomain mutants (results not shown).

Crystal structure of PKCθC1B

The diamond-shaped crystals of PKCθC1B with an average dimension of 0.2 mm×0.1 mm×0.05 mm were obtained as described in the Experimental section. The crystals diffracted to a high resolution and data were collected and processed at 1.63 Å (Table 1).

Table 1. PKCθC1B crystal data processing and refinement statistics.

Values in brackets refer to the highest resolution shell. rms, root mean square.

Crystal parameters	Values
Space group	P61
Unit cell axis (Å)	a = b =65.85, c = 27.12
Resolution (Å)	28.51–1.63 (1.72–1.63)
Total number of reflections	45483
Total number of unique reflections	7559
Redundancy	6.0
Completeness (%)	88.39 (98.00)
l/σ	18.6(9.7)
Rmerge	0.044 (0.115)
Rwork/Rfree (%)	17.14/20.42
Wavelength	1.54178 Å (copper anode)
Distance (mm)	100
Number of water molecules	130
Wilson B-factor	15.2
B-factor	19.93
Number of metal atoms	2 Zn2+
Bond length (Å, rms)	0.01
Bond angle (°, rms)	1.03

The overall structure (Figure 1) consists of two β-sheets and a short α-helix at the C-terminal end, identical with the crystal structure of PKCδC1B (PDB code 1PTQ) [22]. The two β-sheets formed a V-shaped activator-binding groove. Two Zn²⁺ ions were detected in the structure, each of which co-ordinates with three cysteine residues and a histidine residue. One Zn²⁺ co-ordinates with Cys²⁶¹, Cys²⁶⁴, Cys²⁸⁰ and His²³¹, whereas the other co-ordinates with Cys²⁴⁷, Cys²⁴⁴, Cys²⁷² and His²⁶⁹. A total number of 129 water molecules were present in the structure. PKCθC1B has a high sequence homology (80%) and structural similarity with the crystal structure of PKCδC1B (PDB code 1PTQ) with an rmsd (root mean square deviation) of 0.66. However, the activator-binding pocket opening of PKCθC1B is slightly narrower (7.56 Å) than that of PKCδC1B (8.08 Å) owing to the upward orientation of the Trp²⁵³ at the rim of the activator-binding pocket opening. Superimposition of the crystal structure of PKCδC1B (PDB code 1PTR), complexed with phorbol 13-O-acetate (blue), with that of the PKCθC1B (pink) showed that all the homologous residues were superimposable except for the Trp²⁵³ residue (Figure 2). In PKCθC1B the Trp²⁵³ was orientated towards the membrane-binding region of the C1B subdomain, whereas the corresponding homologous Trp²⁵² in PKCδC1B was orientated away from the membrane-binding region of the C1B subdomain (Figure 8A). Binding of the activator to the CI domain caps the hydrophilic activator-binding cavity and makes the membrane-binding surface more hydrophobic, aiding in the membrane anchoring. The upward orientation of Trp²⁵³ in PKCθC1B was not due to the crystal packing effect and this has been further confirmed by the recently described NMR structure of PKCθC1B (PDB code 2ENZ) in which the Trp²⁵³ was also orientated upwards. While comparing our PKCθC1B crystal structure (PDB code 4FKD) with the NMR structure, it was found that the activator-binding pocket opening and the depth of pocket in the NMR structure (8.72 Å and 20.75 Å respectively) are slightly bigger than in the crystal structure (7.56 Å and 19.49 Å respectively).

Figure 1. Structure of PKCθC1B.

(A) Electron density map (2F₀ − F₀) of the PKCθC1B subdomain showing upward orientation of Trp²⁵³ towards the membrane. Expanded view of the electron density fit for Trp²⁵³ is shown in the top left-hand corner. (B) Ribbon structure of PKCθC1B superimposed on the contour surface. The blue area on the contour surface represents the polar residues, whereas the white to orange-red represents the most hydrophobic residues. Extra residues added to the N- and C-terminus form a β-sheet structure.

Figure 2. Structure and sequence comparison between PKCδC1B and PKCθC1B for identifying activator-binding residues.

Overlay of the crystal structure of PKCδC1B (blue) bound with phorbol-13-O-acetate and PKCθC1B (pink) shows that they are superimposable. Phorbol-13-O-acetate forms five hydrogen bonds in the activator-binding groove of PKCδC1B; two bonds with Gly²⁵³, two bonds with Thr²⁴², and one bond with Leu²⁵¹ (denoted with a dotted red line). PKCθC1B consists of the same homologous activator-binding residues, indicating similar binding could be possible with phorbol-13-O-acetate upon co-crystallization. Residues underlined in the PKCδC1B sequence constitute the activator-binding regions and the residues in red are the direct activator-binding residues reported in the co-crystallized structure (PDB code 1PTR). Residues marked with an arrow in PKCθC1B are the residues that were mutated. Only the activator-binding portions of the C1B domains are shown in the Figure.

Figure 8. Orientation of the tryptophan residue in PKCδC1B, PKCθC1B and Munc 13.1 C1.

(A) Orientation of Trp²⁵² of PKCδC1B (blue), Trp²⁵³ of PKCθC1B (green) and Trp⁵⁸⁸ of Munc13.1C1 (red) at consensus position 22 in the overlaid structure of PKCδC1B (PDB code 1PTQ), PKCθC1B (PDB code 4FKD) and Munc 13-1 C1-domain (PDB code 1Y8F) respectively. Trp²⁵² is orientated away from the membrane, whereas Trp²⁵³ is orientated towards the membrane. Trp⁵⁸⁸ of the Munc 13-1 C1 domain is projected inside the activator-binding pocket. (B) Possible π stacking interaction between Trp²⁵² and His²⁶⁹ and cation–π interaction between Trp²⁵² and Lys²⁷¹ in PKCδC1B, and (C) possible cation–π interaction between His²⁷⁰ and Arg²⁷² in PKCθC1B.

The crystal structure also revealed that the nine extra residues (GSRRASVGS) at the N-terminus and the six extra residues (EFIVTD) in the C-terminus, which were added during subcloning of the corresponding gene in the pGEX2TK vector, formed two parallel β-sheets. These extra residues presumably stabilized the overall PKCθC1B structure for crystal formation. A similar observation was made recently with the PKCδC1B crystals [46].

Although an activator-bound structure of PKCθC1B would be able to identify the protein residues directly involved in activator binding, attempts with several hundred precipitants and conditions did not yield any protein–ligand co-crystals. Furthermore, soaking experiments with different activator solutions were also unsuccessful owing to the solubility problems of the activators in aqueous buffer. This highlights the importance of developing new molecules that would show improved aqueous solubility, yet would possess high binding affinity. Therefore we performed molecular docking of a library of DAG and phorbol esters using PKCθC1B as the receptor and sequence alignment of PKCθC1B and PKCδC1B to identify the possible activator-binding residues (Figure 2).

Molecular docking

In the phorbol 13-O-acetate-bound PKCδC1B structure (PDB code 1PTQ) [22], Gly²⁵³, Thr²⁴² and Leu²⁵¹ form two, two and one hydrogen bonds respectively with phorbol 13-O-acetate. A sequence alignment of PKCθC1B with PKCδC1B showed that PKCθC1B consists of the same set of residues at the homologous position to the activator-binding residues in PKCδC1B, indicating similar binding could be possible with phorbol 13-O-acetate (Figure 2) [22]. Additionally, in the activator-binding region of PKCθC1B, there were other residues (Tyr²³⁹, Thr²⁴³, Trp²⁵³, Leu²⁵⁵ and Gln²⁵⁸) homologous to PKCδ, which could also play a role in activator binding. Our docking results revealed that the residues we identified from sequence alignment were also capable of forming hydrogen bonds with the docked activators.

Phorbol ester and DAG binding of PKCθC1B and its mutants

To determine the role of the above-mentioned residues in activator binding, Y239A, T243A, W253G, L255G and Q258G mutants were generated and their binding affinities for PDBu (K_d), DOG (K_i) and SAG (K_i) were measured. To study the role of the side chains responsible for ligand binding, the corresponding residues are usually mutated to alanine. However, mutation with glycine is not uncommon in the PKC activator-binding studies [21,28,29,47–51] and in the case of the δC1 domain, the mutational effect caused by either alanine or glycine has been found to be comparable [47]. To compare with the published data for the PKCδ mutants, we made several glycine mutants.

The K_d values for PDBu binding were in the nanomolar range and the mutants showed reduced affinity compared with the wild-type PKCθC1B (Table 2). Representative plots of PDBu binding to wild-type PKCθC1B and mutant L255G are shown in Figure 3. The K_d values of PDBu binding to Y239A, T243A, W253G, L255G and Q258G showed reductions of 8.51-, 1.64-, 7.56-, 11.4- and 1720-fold respectively compared with the wild-type (Figure 3 and Table 2). These data indicated that Q258G had a pronounced effect, whereas T243A has the least effect on PDBu binding. When binding affinities (K_i) of DOG were measured, Y239A, T243A, W253G, L255G and Q258G showed 5.34-, 1.34-, 29.2-, 4.72- and 114-fold reductions respectively, compared with the wild-type (Figure 3 and Table 2). In this case, Q258G showed the highest reduction in binding affinities and T243A showed the lowest effect. Finally, SAG showed a basically similar pattern as did DOG. Compared with the wild-type, Y239A, T243A, W253G, L255G and Q258G showed 4.13-, 0.63-, 31.1-, 8.16- and 260-fold reductions in binding affinities respectively. These data indicated that glutamine was the most critical among all mutated residues tested in the present study. For both DOG and SAG, W253G showed the second greatest effect, but this was not evident for PDBu. The relatively greater importance of a tryptophan residue at this position in other PKC isoforms for DAG binding, as compared with phorbol ester binding, has been described previously [52]. The C1B domain of PKCθ was reported to bind better to vesicles containing SAG than to those containing 1,2-dioleoylglycerol [53]. Consistent with that observation, we observed a 2-fold stronger binding affinity for SAG than for DOG.

Table 2. Binding of PDBu and DOG to PKCθC1B domain variants.

WT, wild-type.

	PKCθ C1B WT	PKCθ C1B Y239A	PKCθ C1B T243A	PKCθ C1B W253G	PKCθ C1B L255G	PKCθ C1B Q258G
PDBu Kd (nM)*	0.39 ± 0.2	3.32 ± 0.34	0.64 ± 0.06	2.95 ± 0.19	4.44 ± 0.56	670 ± 150
DOG Ki (nM)*	16.1 ± 0.7	86 ± 15	21.5 ± 1.8	470 ± 69	76 ± 24	1840 ± 320
Ki/Kd	41	26	34	159	17	3
SAG Ki (nM)*	8.7 ± 2.0	35.1 ± 5.1	5.5 ± 1.5	271 ± 22	71 ± 10	2260 ± 330
Ki/Kd	22	11	9	92	16	3

^*Values represent the means ± S.E.M. of triplicate independent experiments.

Figure 3. Determination of K_d and K_i.

(A and B) Determination of K_d. Saturation plots for the binding of [³H]PDBu to (A) PKCθC1B and (B) PKCθC1B L255G. (C and D) Determination of K_i. Plots for competitive inhibition of the binding of [³H]PDBu by DOG to (C) PKCθC1B and (D) PKCθC1B L255G. The results are the means ° S.E.M. of the triplicate points in a single experiment. Where the error bars are not visible they are smaller than the symbol. The experiments illustrated are representative of triplicate experiments. The mean values for K_d and K_i are presented in Table 2.

To determine the effect of composition of anionic PS on activator binding, we determined radioactive PDBu binding by PKCθC1B and its mutants in the presence of varying proportions of PS in PS/PC mixtures while maintaining a fixed concentration of total phospholipid.

Effect of the phospholipid composition on [³H]PDBu binding

Binding of [³H]PDBu to PKCθC1B showed a marked dependence on the presence of PS, with essentially no binding observed in the presence of 100% PC (Figure 4). Approximately 30% binding was observed at 5% PS and maximal binding was attained with PS percentages between 75% and 100%. The T243A mutant showed almost the same dependence on PS as did the wild-type. On the other hand, mutants W253G and L255G showed appreciably greater dependence on PS for binding than did the wild-type, with 25% PS required, yielding the same level of binding observed in the wild-type at 5% PS. Finally, for both L255G and Y239A, maximal binding was obtained at 50–75% PS, with a decrease at 100%, whereas for the wild-type binding was maximal and constant between 75 and 100% PS.

Figure 4. Effect of PS on phorbol ester binding.

PDBu binding to PKCθC1B and its mutants was determined as a function of the percentage of PS in PS/PC phospholipid mixtures and was expressed relative to binding at 100% PS. The results are the means ± S.E.M. of three independent experiments.

As analysed by surface plasmon resonance [21], the W253G and L255G mutants of PKCθ showed significant reduction in the binding affinity (6- and 14-fold respectively) for DiC₁₈ in lipid vesicles [POPC (1-palmitoyl-2-oleoyl phosphatidylcholine)/POPS(1-palmitoyl-2-oleoyl phosphatidylserine)/DiC₁₈ = 69:30:1, by vol.]. This reduction is more than that of the corresponding homologous residues W181G and L183G (1.7- and 1.6-fold respectively) in C1A. Taken together, our data clearly indicate that the concentration of PS strongly affects the activator-binding affinity and the extent of this variation is different for different mutants.

The effect of PS on the activator binding for the homologous sites of PKCδ revealed that Leu²⁵⁰ and Leu²⁵⁵ of PKCδC1B strongly influenced the binding of PDBu [54]. The L250D and L255D mutations completely abolished the strong binding of PDBu in the presence or absence of PS. Also, the L250R and L255K mutations caused dramatic 340- and 250-fold reductions respectively in PDBu binding in the presence of lipid. Only a modest reduction in the PDBu binding (6.6- and 2.9-fold respectively) was observed in the absence of lipid. Mutation of L250K or W252K had a 3-fold effect compared with the wild-type in the presence of PS, whereas their binding affinity remained almost the same in the absence of the PS. At 50% PS, where PKCδC1B showed activity of approximately 100%, its W252K mutant showed approximately 55% of the maximum activity [54]. In the present study we found that, at 50% PS, PKCθC1B and its mutants Y239A, T243A, W253G, L255G and Q258G showed 90.5%, 130.7%, 98.2%, 59.3%, 92.8% and 85.7% of the level of PDBu binding observed at 100% PS respectively (Figure 4). The highest reduction observed for the W253A indicated that the Trp²⁵³ was critical for its interaction with membranes and subsequently the PDBu binding.

It is important to note that the present study using PKCθC1B was performed with GST-cleaved C1B constructs, whereas those reported for PKCδC1B were done with proteins in which the GST had not been cleaved. In any case, our results emphasize for PKCθC1B, as for other C1 domains, that measured binding represents the ternary complex of ligand, C1 domain and phospholipid, with all three elements contributing to the energetics of binding. Differences in the composition of various cellular phospholipids may thus make important contributions to ligand selectivity.

After measuring the binding affinities of phorbol ester and DAG to PKCθC1B and its mutants and the dependence on PS of their binding, we determined the effect of the mutations on the membrane translocation of the full-length kinase.

Effect of mutation in the C1B domain on membrane translocation

Translocation was measured separately in the presence of 50 nM PMA and 250 nM DOG. Activator-induced membrane translocation was quantified by the decrease of PKCθ and its mutants in the cytoplasmic pool (or the increase of the membrane pool). In the presence of 50 nM PMA, translocation was observed both for the wild-type PKCθ and for all mutants. Y239A, T243A, W253G, L255G and Q258G showed 17.1%, 12.5%, 14.8%, 19.4% and 21.6% reductions respectively in membrane translocation compared with the wild-type (Figure 5A). In the presence of 250 nM DOG, Y239A, T243A, W253G, L255G and Q258G showed 12.0%, 8.98%, 5.9%, 15.4% and 15.9% reduction in membrane translocation respectively compared with the wild-type (Figure 5B).

Figure 5. Effect of mutations on the membrane translocation properties of PKCθ and its C1B domain mutants.

Distribution between cytosol and membrane of PKCθ and its mutants following treatment for 2 h with (A) 50 nM PMA, and (B) 250 nM DOG. Full-length proteins were expressed in HEK-293 cells. The cells were treated with either PMA or DOG, harvested and fractionated. The cytosolic (C) and membrane (M) fractions were analysed by Western blotting. Each data point represents the means ± S.D. of three measurements. *P < 0.05.

These results indicated that the residues mutated were important for the membrane translocation of the full-length kinase. Among the residues, Gln²⁵⁸ was found to be the most important, because mutation of Gln²⁵⁸ showed significant reduction in membrane translocation in the presence of both PMA and DOG. However, the mutational effect depended on the activator type, as evident in the comparison of PMA with DOG in full-length mutant translocation. The percentage reduction of the mutant translocation was somewhat greater in the presence of PMA compared with DOG.

Translocation of GFP–PKCθ wild-type and mutants in living LNCaP cells in response to PMA

Using the GFP-tagged PKCθ wild-type and mutants, we examined the sensitivity, kinetics and cellular localization of translocation in living cells in response to PMA. The constructs were expressed in LNCaP human prostate cancer cells because this cell line spreads well on the substratum, permitting good visualization, and has been extensively used for the analysis of PKC translocation. Representative images are shown in Figure 6. The images from all of the experiments were then quantified in terms of the signal/intensity ratio for membrane/cytoplasm (Table 3). At 10 nM PMA, negligible response was observed for all of the constructs (Table 3). At 100 nM PMA, the GFP–PKCθ wild-type showed marked translocation to the plasma membrane within 2 min (Figure 6). For T243A, translocation had begun within 2 min and was largely complete by 5 min. By 20 min, plasma membrane translocation was largely complete for all constructs, except Q258G, which showed little response. At 1000 nM PMA, the same pattern was evident, but with more extensive and faster translocation, and partial translocation of Q258G could now be seen at the later times. Although the constructs differed in their sensitivity and kinetics of response to PMA-induced translocation, in all cases translocation was to the plasma membrane.

Figure 6. The translocation pattern of the GFP-tagged PKCθ wild-type and its mutants in living LNCaP cells after PMA treatment.

Cells expressing the wild-type or mutant GFP–PKCθ constructs were treated with the 1000 nM (A) or 100 nM (B) PMA. The translocation pattern was examined by confocal microscopy as a function of time. Each panel represents images typical of three to four independent experiments.

Table 3. Quantification of membrane/cytoplasmic intensity ratios in confocal images.

Membrane/cytoplasmic intensity ratios were calculated from the confocal images of the multiple experiments. Values are the means ± S.E.M. of at least three images for 1000 nM and 100 nM PMA-treated samples and at least two images for the 10 nM PMA treated samples. WT, wild-type.

Treatment	Time (min)	PKCθ WT	PKCθ Y239A	PKCθ T243A	PKCθ W253G	PKCθ L255G	PKCθ Q258G
100 nM PMA	5	14.5 ± 2.2	3.8 ± 1.8	10.3 ± 1.9	11.4 ± 3.6	4.8 ± 1.5	1.2 ± 0.1
	10	11.3 ± 2.1	4.2 ± 1.6	9.9 ± 1.5	12.7 ± 3.9	6.0 ± 1.2	2.0 ± 0.3
	20	12.1 ± 2.2	5 ± 0.4	7.4 ± 1.5	13.2 ± 5.0	7.2 ± 1.9	2.1 ± 0.4
100 nM PMA	5	8.1 ± 1.2	1.7 ± 0.4	3.6 ± 1.1	1.9 ± 0.4	1.3 ± 0.1	1.0 ± 0.0
	10	6.5 ± 0.7	2.3 ± 0.8	3.7 ± 0.8	3.0 ± 0.7	1.6 ± 0.2	1.1 ± 0.0
	20	5.4 ± 0.5	2.3 ± 0.5	3.1 ± 0.6	2.6 ± 0.6	2.3 ± 0.4	1.1 ± 0.0
10 nM PMA	5	1.0 ± 0.1	1.1 ± 0.0	1.0 ± 0.0	1.0 ± 0.1	1.1 ± 0.1	1.2 ± 0.1
	10	1.0 ± 0.1	1.1 ± 0.1	1.3 ± 0.2	1.1 ± 0.1	1.1 ± 0.1	1.1 ± 0.1
	20	1.2 ± 0.4	1.0 ± 0.0	1.7 ± 0.1	1.2 ± 0.2	1.1 ± 0.0	1.1 ± 0.0

DISCUSSION

PKCθ is unique among all the PKCs, because of its predominant expression in T-cells. It is the only PKC isoform which is uniquely targeted to the immunological synapse in T-cells and involved in the T-cell activation process [55,56]. We undertook this study to identify the activator-binding residues of the PKCθC1B subdomain and to accurately define the shape of the binding cleft with the future goal of designing PKCθ-specific molecules. A previous study by Melowic et al. [21] demonstrated a dominant role of the C1B subdomain in the membrane binding and activation process of PKCθ over CIA, its other (twin) cysteine-rich subdomain.

To identify the activator-binding residues, we first determined the crystal structure of the C1B domain at 1.63 Å resolution and used this structure to dock several known phorbol ester and DAG analogues to identify the possible residues in the protein that interact with these activators. Furthermore, by comparison with the activator-binding residues of PKCδC1B, for which the activator-bound structure is known [22], we identified five possible residues that could affect the activator binding.

Since the binding of DAG and phorbol ester to PKC showed differences in its dependence on structural features of the C1 domains [22,29,57–59], we used both phorbol ester and DAG for the present study. Our results indicated that, for PKCθC1B and its mutants, phorbol ester had higher binding affinity than DAG. Although to different extents, all the mutants showed reduced binding affinities for both phorbol ester and DAG as compared with the wild-type PKCθC1B. Out of the five residues that were tested, Gln²⁵⁸, a conserved residue among the PKC isoforms, showed maximum influence both for binding by the phorbol ester and DAG and for membrane translocation in response to these ligands.

A careful inspection of the PKCθC1B structure revealed that Gln²⁵⁸ stabilizes the activator-binding pocket by bridging the left and the right activator-binding clefts. The backbone amine nitrogen of Gln²⁵⁸ could form a hydrogen bond (3.61 Å) with the backbone carbonyl oxygen of Tyr²³⁹ and the side-chain nitrogen (NE2) of Gln²⁵⁸ could form another hydrogen bond (2.87 Å) with the backbone carbonyl oxygen of Gly²⁵³. Mutation of Gln²⁵⁸ with glycine resulted in the disruption of this bridge, leading to the highest reduction of binding affinity and membrane translocation.

Interestingly, under our binding conditions, W253G caused only a 4-fold greater loss of affinity for DOG than it did for PDBu. In the case of the C1B domain of PKCβI/II and PKCα under different lipid conditions, the presence of a tryptophan residue in place of tyrosine (Figure 7) greatly influenced binding of DAG, but not that of phorbol ester [52,60].

Figure 7. Sequence comparison of the C1 domains.

The Trp²⁵³ in PKCθ and its homologous residues in other C1 domains are boxed.

A comparison of the crystal structures of PKCδC1B and PKCθC1B revealed remarkable similarity of the overall structure. PKCθC1B has a high sequence homology (80%) with PKCδC1B (PDB code 1PTQ) with an rmsd of 0.66, indicating the similarity between the structures. Superimposition of the PKCθC1B structure with the PKCδC1B showed that the activator-binding pocket opening of PKCθC1B is slightly narrower than is that of the PKCδC1B. Another prominent difference is that, in the activator-binding surface, Met²³⁹ in PKCδC1B is replaced by Lys²⁴⁰ in PKCθC1B (Supplementary Figure S3 at http://www.biochemj.org/bj/451/bj4510033add.htm). This suggests that the affinity for the PS-rich membrane would be different for PKCθC1B and PKCδC1B.

Although the orientations for most of the residues of both the structures are almost identical, the orientation of Trp²⁵³ is totally opposite in these two structures (Figure 8). In a representation of the structure where the activator-binding loops are positioned towards the membrane, Trp²⁵³ is found to be orientated towards the membrane in PKCθC1B, whereas the homologous Trp²⁵² of PKCδC1B is orientated downwards away from the membrane. Because of this particular orientation of Trp²⁵³ in PKCθC1B, its interactions with nearby residues are distinctly different from PKCδC1B. In PKCδC1B the indole ring of homologous Trp²⁵² is found to be orientated in parallel to the imidazole ring of His²⁷⁰ at a distance of 4.13 Å, possibly involved in jr-stacking interactions [61–63]. In the PKCθC1B, however, similar interactions between the Trp²⁵³ and His²⁷⁰ are absent because these two rings are far apart (>10 Å) from each other (Figure 8). The pK_a of the imidazole ring nitrogen being ~6.0 means it would exist mostly in its neutral form at the physiological pH. Furthermore, since this histidine residue co-ordinates with one of the Zn²⁺ atoms, its possibility of existing as a cation is completely ruled out. The ε-NH₂ of Lys²⁷¹ in PKCδ, which is protonated at physiological pH and located at a distance of 5.12 Å from the imidazole ring of His²⁶⁷ and at a distance of 5.49 Å from the indole ring of Trp²⁵², is most likely to be involved in cation-it interactions with both the rings, for this type of interaction occurs within a distance of 6 Å [63–65]. In PKCθC1B, however, this lysine residue is replaced by Arg²⁷², which is involved in cation-π interactions with His²⁷⁰ at a distance of 3.81 Å. Furthermore, upward orientation of Trp²⁵³ at the rim of the activator-binding pocket of PKCθC1B may also exert additional hydrophobic interactions, with the activator facilitating its binding to the membrane.

Although the Trp²⁵³ side chain is not directly orientated inside the activator-binding loop like the homologous Trp⁵⁸⁸ in Munc13.1, which occludes the DAG-binding site [66], our binding data for W253G showed 8-, 29- and 31-fold reductions in the binding of PDBu, DOG and SAG respectively by the mutant as compared with the wild-type. Although binding data of the DiC₁₈ and the purified full-length W253G showed a 6-fold reduction in binding affinity in SPR analysis in a previous study [21], the results from the present study also showed reduced membrane translocation for both PDBu and DiC₈. It is important to emphasize, in any case, that the crystallographic analysis was for the isolated C1B domain, whereas ligand binding is measured in the presence of phospholipid. The orientations of Trp²⁵³ will undoubtedly be influenced by the presence of lipid and the effect of the W253G mutation will reflect, among other factors, its impact on the C1B-membrane interaction. Furthermore, comparison with other known C1 domain structures reveals that β2-chimerin [67] also has the homologous tryptophan residue in an orientation opposite to that of PKCθC1B.

Our binding data revealed that there were differences in the binding affinities of the mutants of PKCδ and PKCθ, particularly for W253G and L255G. In PKCθC1B, W253G and L255G showed 8- and 11-fold reductions respectively in their affinities for PDBu as compared with the wild-type. On the other hand, in PKCδC1B, the homologous mutants W252G and L254G showed 31- and > 1000-fold reductions, respectively. For Gln²⁵⁸ and Thr²⁴³, mutation caused similar reductions of PDBu affinity for both PKCδ and PKCθ. For the Tyr²³⁹ site, the Y239A in PKCθC1B showed a 9-fold reduction; the homologous Y238G in PKCδC1B showed a 60-fold reduction in the binding affinity (Supplementary Table S1 at http://www.biochemj.org/bj/451/bj4510033add.htm). This is not surprising, since each mutation can cause a change in overall lipophilicity of the domain. Correspondingly, we found that the mutants differed in their PS dependence for phorbol ester binding. Whether the difference at the Tyr²³⁹ site between PKCθC1B and PKCδC1B is because of the different substitutions, alanine versus glycine respectively, or is a true reflection of the difference between PKCδ and θ is a matter of further investigation. Our modelling and molecular dynamics studies on Y238A and Y238G, however, showed that there was no significant structural change between the two mutants and both of them form four hydrogen bonds with PDBu. That the same residues in the two different isoforms with similar structures showed differential ligand-binding affinities indicated that there were cumulative effects, including the other nearby residues in determining the binding affinity, along with possible differences in their influence on membrane binding.

What are the key determinants for a C1 domain to be phorbol ester/DAG-responsive? The current concept is that PKC C1 domains anchor to the membranes as a ternary system including C1 domains, C1 activators and the anionic phospholipids present in the membrane, which bind to the positively charged residues present above the rim of the activator-binding cleft [28,29,68–70]. Alteration of any component in this ternary system, therefore, is expected to affect the ligand-binding affinity. For example, in the atypical PKCζ, several arginine residues lining the activator-binding pocket are responsible for its unresponsiveness towards the activator [71]. It has been suggested that these arginine residues reduce access of ligands to the binding cleft and change the electrostatic profile of the C1 domain surface, without altering the basic structure of the binding cleft. In Vav1C1, the presence of Glu⁹, Glu¹⁰, Thr¹¹, Thr²⁴ and Tyr²⁶ along the rim of the binding cleft reduced the overall lipophilicity of the rim, impairing the membrane association and thereby preventing the formation of the ternary complex [45]. In Munc13.1, occlusion of Trp⁵⁸⁸ in the activator-binding pocket reduces its affinity for the activators [66]. Our results of the present study emphasize that the structure and orientation of the residues are also important in determining the responsiveness of a C1 domain towards its activator.

In conclusion, our 1.63 Å structure of PKCθC1B revealed close similarity between the PKCδ and PKCθ C1B structures, except for the orientation of the Trp²⁵³ residue, which plays a critical role in membrane binding. Out of the five residues, Gln²⁵⁸ showed the most prominent effect on both activator binding and PKC translocation, with each site having its own contribution to the activator-binding affinity of the C1B subdomain.

Although the present study has identified several unique structural features in the activator-binding domain of PKCθC1B, further studies will be required to understand the mechanism by which PKCθ is specifically targeted to the immunological synapse of the T-cells.

Abbreviations used:

AP-1

activator protein 1

APC

antigen-presenting cell

DAG

diacylglycerol

DOG

sn-1,2-dioctanoylglycerol

DMEM

Dulbecco’s modified Eagle’s medium

EGFP

enhanced green fluorescent protein

FBS

fetal bovine serum

GFP

green fluorescent protein

GST

glutathione transferase

HEK

human embryonic kidney

HRP

horseradish peroxidase

MALDI–TOF-MS

matrix-assisted laser-desorption ionization–time-of-flight MS

NF-κB

nuclear factor κB

PC

L-α-phosphatidylcholine

PDBu

phorbol 12,13-dibutyrate

PKC

protein kinase C

PS

L-α-phosphatidylserine

rmsd

root mean square deviation

SAG

sn-1-stearoyl 2-arachidonyl glycerol

TB

terrific broth

TCR

T-cell receptor