Conserved Modular Domains Team up to Latch-open Active Protein Kinase Cα*

Background: Protein kinase C (PKC) is a nodal regulator of cell signaling. Results: Multiple interactions between conserved regulatory domains in PKC synergistically stabilize a nanomolar affinity homodimer critical for cellular function. Conclusion: Homodimerization regulates the equilibrium between the auto-inhibited and active states of PKCα. Significance: Multiplexed interactions between modular domains dictate PKC function. Signaling proteins comprised of modular domains have evolved along with multicellularity as a method to facilitate increasing intracellular bandwidth. The effects of intramolecular interactions between modular domains within the context of native proteins have been largely unexplored. Here we examine intra- and intermolecular interactions in the multidomain signaling protein, protein kinase Cα (PKCα). We identify three interactions between two activated PKC molecules that synergistically stabilize a nanomolar affinity homodimer. Disruption of the homodimer results in a loss of PKC-mediated ERK1/2 phosphorylation, whereas disruption of the auto-inhibited state promotes the homodimer and prolongs PKC membrane localization. These observations support a novel regulatory mechanism wherein homodimerization dictates the equilibrium between the auto-inhibited and active states of PKC by sequestering auto-inhibitory interactions. Our findings underscore the physiological importance of context-dependent modular domain interactions in cell signaling.

The use of modular protein domains has emerged as a prominent feature of increasing phylogenetic complexity (1,2). Linking modular domains within a single protein allows complex regulation while conserving the sequence and structure of the individual domains, especially for those that have catalytic activity (3). For instance, spatiotemporal control of signaling proteins is often achieved by stringing together a conserved catalytic domain with one or more regulatory modules. These modules can play multiple roles including masking the catalytic site to inhibit basal activity (auto-inhibition), releasing autoinhibition through conformational changes triggered by second messenger stimuli, and facilitating translocation to subcel-lular compartments through binding secondary messengers or scaffolding proteins. Each additional module in a signaling protein provides a combinatorial enhancement to its regulation and cellular function (4). The protein-context independent structure and cellular function of individual modules have been extensively researched using biophysical approaches such as x-ray crystallography and NMR spectroscopy (5). However, coordination of interactions between domains in intact proteins remains unexplored primarily due to the reliance on reductionist structural and biochemical approaches. As a corollary, our current structural understanding of modular signaling proteins does not adequately address the versatility of their cellular function (6 -8). In this study we overcome this limitation with a technique we previously developed to systematically modulate interactions between individual protein domains in the context of the entire molecule (9,10).
Generally, stark perturbations are required to study intramolecular interactions in proteins. For instance, to measure the strength of an intramolecular interaction between two domains it is typical to split them into two individual peptides (11). An alternative approach is to truncate or mutagenize a proposed interaction interface and look at a functional consequence associated with the loss of the interaction (12)(13)(14). In the first approach the strength of an interaction can be underestimated given that when the two interacting domains are in the same polypeptide, they will likely have a high local concentration that can affect the on-rate of the interaction. In the second approach, it is often difficult to anticipate whether a change in function is directly or tangentially effected by the truncation or mutagenesis.
To address the technical shortcoming in studying intramolecular domain-domain interactions, we previously developed and applied a novel genetic method termed Systematic Protein Affinity Strength Modulation (SPASM) (10). This technique utilizes an ER/K motif that has unusual biophysical properties (15). The ER/K motif primarily adopts an extended ␣-helical secondary structure in solution with an end-to-end distance of 10 -30 nm, depending on the length of the motif (16). However, it is prone to stochastic breaks in helicity that allow the N and C termini of the helix to come in close proximity of each other at a low but predictable frequency (10). Capitalizing on this phe-nomenon, a genetically encoded FRET pair as well as interacting polypeptides can be fused to either end of the ER/K motif to build a single polypeptide (xpolypeptide-FRETdonor-ER/K linker-FRETacceptor-ypolypeptide) (9,17). At low (nM) concentrations, the interaction between the polypeptides will report a binary high FRET interacting state or low FRET noninteracting state that will be averaged out in ensemble measurements (10). It has been empirically determined that the on-rate for interactions in this system are dictated by the frequency that the ␣-helix stochastically brings its ends in close proximity, whereas the off-rate is dependent on the interaction between the polypeptides attached to the end of the helix (10). Additionally, the frequency at which the ends of the ␣-helix come in close proximity is a function of the length of the ␣-helix; longer helices have a lower frequency of end-to-end proximity (10). By incorporating different length ER/K motifs between interacting polypeptides, the on-rate can be modulated such that the apparent concentrations can be varied from 100 nM to 10 M (10). Finally, the SPASM module is a transposable genetic element that can be incorporated into naturally occurring linkers between modular domains in most proteins. This technology allows intramolecular interactions to be perturbed in systematic ways such that domain-domain interactions can be directly observed and characterized (9). Here we utilize this technology to explore domain-domain interactions in the multidomain protein kinase C (PKC).
The PKC family exemplifies the use of modular architecture in cellular function. PKC is composed of a conserved catalytic domain (CD) 3 fused to multiple, modular regulatory domains (RDs: pseudo-substrate, C1a, C1b, and C2) that are sensitive to different small molecule/second messenger stimuli (␣ isoform: C1a-diacylglycerol (DAG) and phorbol ester (PMA); C1b-DAG, PMA; C2-Ca 2ϩ ). In the absence of stimuli, the RDs cooperatively mask the catalytic site, whereas conformational changes facilitated by effector binding to these domains facilitate interactions with cellular membranes and scaffolding proteins (6). Although numerous studies have dissected the structure-function relationship of the individual regulatory modules, interactions between modules remain largely unexplored. Hence, the current model of PKC activation presents it as a binary switch that is basally turned off through a RD-CD interaction and is switched on when effectors bind and release the RD from the CD (18,19). As a glimpse beyond this binary view, a high affinity intermolecular interaction between the effector-bound C1 and RD (12 nM) has been previously reported (11). Furthermore, PKC activation has been proposed to trigger homodimerization (20) with a consequent influence on catalytic activity (11,20). However, these intermolecular interactions have not been incorporated into current models of PKC activation, presumably due to the lack of mechanistic or structural details.
Here, we demonstrate that PKC␣ readily homodimerizes upon stimulation with activating effectors in vitro. We generated and characterized a unimolecular FRET sensor that is both functional as PKC␣ and reports on effector-induced dimerization in vitro and in CHO cells. We then used multiple FRET sensors, some including the ER/K linker, to dissect domaindomain interactions in the activated state of PKC␣ to reveal that dimerization stems from several weak interactions involving the C1, C2, and catalytic domains that together contribute to a nanomolar affinity interaction. We then address the functional significance of PKC␣ dimerization through four separate vignettes. 1) We found that the in vitro specific activity of PKC␣ is more sensitive to dimerization than it is to modulation of the auto-inhibitory interactions. 2) We observed homodimerization in the basal state upon mutagenesis of the turn motif priming phosphorylation site (Thr 3 Ala), a finding with implications in PKC maturation. 3) We revealed that PKC function in cells can be modulated in the presence of peptides designed to destabilize the dimer state. 4) We observed that a commonly used PKC inhibitor bisindoylmaleimide I (BimI) can alter PKC␣ localization by destabilizing basal, auto-inhibitory interactions. We conclude with a simple model in which dimerization allows PKC to overcome a high degree of basal auto-inhibition by latching open the effector stimulated kinase.
Mammalian Cell Culture-CHO Flp-in cells (Invitrogen) were cultured and incubated in DMEM supplemented with 10% FBS, 4.5 g/liter D-glucose, 1% glutamax, 20 mM HEPES, pH 7.5, and 1% antibiotic-antimycotic (Invitrogen) at 37°C and 5% CO 2 . Constructs in the pcDNA vector were transiently transfected into CHO cells using FuGENE HD (Promega) and harvested after 48 h. CHO Flp-in cells stably expressing mCerulean (mCer)-PKC␣-mCitrine (mCit)-FLAG constructs were generated according to the manufacturer's protocol with cells maintained in 600 g/ml hygromycin. For live cell imaging, cells were trypsinized, and 2 ϫ 10 5 cells were re-plated onto 35-mm glass-bottom dishes (MatTek Corp.) coated with 1 g of fibronectin (Sigma) and allowed to adhere for 2 h. The plates were washed with 2 ml of HEPES-buffered saline with 0.2% dextrose, and 500 l of this medium was added to the cells before imaging. Cells were imaged at 40ϫ or 60ϫ with a Nikon TiE microscope equipped with an Evolve 512 ϫ 512 EM-CCD camera (Photometrics), a mercury arc lamp, and the appropriate band pass filters. For FRET images, a Dual-View filter (Photometrics) was used for simultaneous acquisition of mCerulean and mCitrine peak fluorescence. After image acquisition of untreated condition (pre), cells were perfused with HBS containing 10 M PMA and 900 M CaCl 2 or 50 M LPA at room temperature. For live cell fluorometer experiments (Figs. 2a and 8c), 1 confluent 10-cm plate was trypsinized, washed with 10 ml of DMEM with 10% FBS to inactivate the trypsin, and washed twice with 5 ml of HBS with 0.2% dextrose before being resuspended in 1 ml. This is similar to methods previously reported (17). All conditions involving myristoylated peptides were incubated with 20 M peptide for 15 min before imaging or fluorometer experiments. For translocation imaging experi-ments 1.5 M BimI was added to cells 2-4 min before image acquisition.
Fluorometer Data Acquisition-A Fluoromax-4 fluorometer (Horiba Scientific) was used to obtain FRET spectra. Samples were excited at 430 nm (8-nm band pass), and the emission was recorded from 450 to 650 nm in 1-nm intervals (4-nm bandpass). For unimolecular sensors, 20 -80 nM protein was used in vitro, and for bi-molecular sensors, 40 nM mCerulean and 160 nM mCitrine constructs were used unless otherwise noted. All in vitro assays were performed in a buffer containing 20 mM HEPES, 5 mM MgCl 2 , 500 M EGTA, 1 mM DTT, 5 g/ml aprotinin, and 5 g/ml leupeptin and were mixed in tubes precoated with 0.1 mg/ml BSA. Final concentrations of 1.5 mM CaCl 2 , 3.2 M PMA, 10 M BimI, or 50 g/ml DAG and 25 g/ml PS were added to the sample where indicated. All conditions were allowed to equilibrate for Ͼ30 min at 30°C after the reaction was fully mixed. The figures of FRET depict mean and S.E. from three independent spectra obtained using the same batch of recombinant protein and the same reagent preparations on the same day and are representative of results observed from three independent measurements. For experiments monitoring MANT-ADP, binding samples were prepared in the same manner but obtained spectra with direct excitation of tryptophan at 290 nm (8-nm band pass), and emission was recorded from 300 to 550 nm at 1-nm intervals (4-nm band pass) or direct excitation of MANT at 340 nm (4-nm band pass) from 400 to 550 nm at 1-nm intervals (2-nm band pass).
Kinase Assay-Kinase assays were performed using the Kinase-Glo Max Luminescence assay kit (Promega) according to the manufacturer's protocol. Synthetic liposomes were made from Brain Polar Lipid Extract (Porcine) (Avanti; high in PS) mixed with 2% w/w 1,3-diolein (Sigma) using established procedures (21). Briefly, the liposomes were suspended in chloroform for storage and dried under nitrogen and stored under vacuum for 18 h. The mixture was resuspended to a concentration of 12.5 mg/ml in HEPES buffer and underwent at least three freeze-thaw cycles in liquid nitrogen then hot water. The mixture was then extruded (Avanti) using a 0.08-m Nuclepore Track-Etch Membrane (Whatman) and diluted into a 20ϫ reaction concentration of 1.6 mg/ml. Histone IIIs (Sigma), bovine myelin-binding protein (MBP; Sigma), or MBP residues 4 -14 (Santa Cruz Biotechnology, Inc) at final concentrations of 40 M were used as substrates unless otherwise noted. The reaction was initiated with 40 M ATP (Sigma) and were briefly mixed and incubated with mild shaking at 30°C from 2 to 10 min. For all conditions the ATP consumption is calculated against a matched control without kinase. ATP calibration curves were obtained under identical conditions (without kinase). End-point luminescence was measured in white, 96-well plates (Thermo Scientific) on a Synergy HT (Biotek).
Cross-linking Assay-A cross-linking assay was performed as previously reported (20). Briefly, 200 -400 nM PKC was preincubated in a HEPES buffer containing 2% glycerol and 0.02% Triton X-100 and the indicated effectors. A 50ϫ stock of BMH (1.25 mM) in DMSO was added to the preincubated solution for 2-5 min then quenched with 5ϫ SDS-PAGE loading dye with 20 mM DTT. All blots shown are representative of at least three independent experiments. It should be noted that the level of cross-linking observed is highly dependent on the concentration of BMH resulting in variability between experiments. To circumvent this limitation inherent in cross-linking assays, the extent of cross-linking is always evaluated with respect to a matched control in each experiment.
In Vitro Membrane Partitioning Assay-This assay was performed as previously described (21). Liposomes were prepared as in the kinase assay except the liposomes were extruded with an equal volume of 340 mM sucrose buffer. Purified protein was clarified at 80,000 rpm, and the clarified protein was incubated in the indicated conditions. The reaction mixture was spun at 80,000 rpm, the supernatant was taken off (soluble fraction), and the remaining pellet was suspended in an equal volume of 1ϫ SDS-PAGE loading dye (pellet fraction).
Western Blotting-Lysates were prepared from stably expressing CHO cells generated using the Flip-in system (Invitrogen). Cells were plated in 6-well dishes (NEST) to reach 80 -90% confluence within 24 h and were subsequently serum-starved for an additional 24 h. Cells were treated with the indicated peptides (20 M) 15 min before the addition of 1-oleoyl lysophosphatidic acid (LPA, Cayman Chemicals, 10 M). Cells were incubated with LPA for 15 min before lysis on ice with ice-cold buffer (1% Triton X-100, 250 mM NaCl, 50 mM Tris pH 6.8, 4 mM MgCl 2 , 2 mM EDTA, 1:100 phosphatase inhibitor mixture II (Sigma), and containing 10 g/ml aprotinin, 10 g/ml leupeptin, and 1 mM PMSF). Lysates were syringed with a 26-gauge needle and clarified by spinning at 16,000 ϫ g (10 min, 4°C). Total lysate protein was assayed either by Bradford assay or by total ERK1/2 blotting (see below) with no significant difference observed between the two methods. Lysates were separated on 10% polyacrylamide, SDS gels before being transferred to PVDF membranes for 3 h at 300 mA. Blots were blocked with 2% BSA, TBS ϩ 0.1% Tween (TBST) for 1 h at room temperature. Primary phospho ERK1/2 antibody (4376S, Cell Signaling) was used at a concentration of 1:5000 in 2% BSA/TBST and incubated overnight at 4°C. Blots were washed with TBST (3 ϫ 10 min) before the addition of secondary antibody (goat anti-rabbit (Jackson ImmunoResearch Laboratories, Inc.) 1:10,000 in 1% milk/TBST). Blots were washed again with TBST (3 ϫ 10 min) and developed using Immobilon Western chemiluminescent HRP substrate (Millipore). Blots were imaged using a ChemiDoc-it imaging system (UVP). Blots were stripped with Restore PLUS Western stripping buffer (Thermo Scientific) and reblotted with total ERK1/2 antibody (4695S, Cell Signaling; 1:10,000, 2% BSA/TBST). Blots were exposed to secondary as above and developed under the same conditions. All blots shown are representative of at least three independent experiments. Cross-linked PKC was probed (Fig. 1f) using a PKC␣ antibody (sc-8393, Santa Cruz Biotechnology, 1:10,000 in 2% BSA/TBST).
Live Cell FRET Analysis-Images were analyzed using custom software in Matlab (Mathworks Inc.). Briefly, for images of LPA-induced cells with constant excitation over the second-tomin time scales, images were subjected to a background threshold/subtraction followed by correction for photobleaching using a maximum threshold. FRET ratios were computed as the pixel-by-pixel ratio between mCitrine and mCerulean intensities across the time series. For images of PMA-induced cells in which excitation and image acquisition occurred for 200 -500 ms every 30 s for 15-20 min, the straight pixel-by-pixel ratio was calculated, as little photobleaching effects were observed. Deconvolution was performed using AutoQuant X3 (MediaCybernetics).
Dissociation Constant of Dimer Formation-The equilibrium between mCer-PKC␣-mCit (F, fluorescent) and PKC␣-FLAG (D, dark) can be represented by the following five equations, where F t and D t are the total concentrations of mCer-PKC␣-mCit and PKC␣-FLAG, respectively.
Assuming the three interactions are equivalent and interchangeable, knowledge of the dissociation constant (K D ) is necessary and sufficient to explicitly calculate the partitioning between the species. A custom Matlab substitution solver was used to numerically calculate each of the five species (F, D, F⅐F, D⅐F, D⅐D) for a range of K D values. Knowledge of the individual species allows us to estimate the K D by evaluating the best fit (least squares) to the titration curve of measured FRET ratio (R obs ) as a function of PKC␣-FLAG concentration (D t ). As previously reported, R obs can be expressed as a linear combination of the FRET ratio for the dimer (R F⅐F ), monomer (R F ), and interspecies dimer (R D⅐F ), R F⅐F is set to the observed FRET ratio at saturating concentrations (Ͼ100 nM) of mCer-PKC␣-mCit, R D⅐F is set to the observed FRET ratio at high D t /F t ratios (Ͼ20), and R F is set to the basal FRET level in the absence of effectors. With these constraints and K D as the only free parameter, we find that a K D Ͻ 5 nM provides an indistinguishable fit to the experimental measurements. This competitive binding measurement with interchangeable, equivalent species can only provide an upper bound to the K D rather than a precise measurement. Lower concentrations of the mCer-PKC␣-mCit can be used to improve this estimate; however, they suffer from low signal-tonoise in our fluorometer measurements.

RESULTS
To investigate the role of intramolecular interactions in the regulation of PKC␣, several distinct FRET reporters were used throughout the study. These are composed of discrete domains of PKC␣ fused to ER/K linkers, and the GFP-derived mCit and mCer FRET pair (Fig. 1a) tethered with (Gly-Ser-Gly) 2-4 linkers to allow for rotational freedom. Basally within a cell or in the presence of EGTA in vitro, the RDs (includes N-to-C-terminal pseudosubstrate, C1a, C1b, and C2 domains) of PKC are known to engage in cis inhibitory intramolecular interactions with the CD (includes the kinase domain and the C-terminal extension termed the C-tail) (18,22). Furthermore, the small molecules Ca 2ϩ , DAG/ PMA, and BimI have been established to interact primarily with the C2, C1, and catalytic domains, respectively (6). f, a representative immunoblot using an anti-PKC␣ antibody resolves single BMH cross-linked bands of equivalent mass but differing intensities after incubation of PKC␣-FLAG with indicated effectors as in c. g, increasing concentrations of PKC␣ result in a constant amount of cross-linked protein despite increased PKC binding to a fixed density of liposomes. Top, representative SDS-PAGE of cross-linked PKC␣ for a fixed concentration of liposome (ϳ80 g/ml PS rich brain polar lipid extract (Porcine) mixed with 2% w/w DAG) as a function of increasing concentrations of PKC␣-FLAG. All samples were treated with 25 M BMH for 3 min. The final well includes C619A mutant PKC␣-FLAG under matched conditions. Bottom, PKC␣-FLAG was incubated with or without the sucrose loaded liposomes used in the cross-linking assay (top) and fractionated by ultracentrifugation; the supernatant or pellet was subsequently separated by SDS-PAGE. Saturation of the liposomes with PKC␣-FLAG occurred at a concentration between 800 and 1600 nM. PKC␣-FLAG (C619A) does not cross-link under the same conditions as WT but demonstrates a comparable ability to associate with liposomes.

PKC␣ Homodimerizes upon Stimulation with Effectors in
Vitro-We engineered an initial unimolecular FRET sensor (mCer-PKC␣-mCit-FLAG; each element is listed from the N to the C termini and dashes consist of a (GSG) 2-4 linker) capitalizing on the design scheme of a previously reported PKC␦ sensor that elicited changes in FRET upon activation both in vitro and in cells (23). This PKC␣ flanking sensor (50 nM) showed enhanced FRET upon activation with Ca 2ϩ and liposomes containing DAG and PS ( Fig. 1b; see "Experimental Procedures"). To address the possibility that the observed increase in FRET is primarily driven by confined localization on the lipid surface, the effect of Ca 2ϩ and PMA, which is a soluble effector (24), was measured with similar results (Fig. 1b). The observed increase in FRET can arise from conformational changes within the PKC molecule or interactions between PKCs. GSG linkers inserted between PKC␣ and the two fluorophores (mCer and mCit) were designed to make the FRET readout sensitive to changes in distance rather than orientation of the two fluorophores. The current model of PKC␣ activation anticipates an increase in Nto C-terminal distance (18) and, therefore, is in contrast with the observed increase in FRET. To address whether intermolecular interactions contributed significantly to the observed FRET response, we generated a second, bimolecular FRET reporter system (mCer-PKC␣-FLAG and PKC␣-mCit-FLAG). Upon the addition of Ca 2ϩ and liposomes containing DAG/PS, Ca 2ϩ and PMA, or Ca 2ϩ alone a significant increase in FRET was observed compared with the sensors in their basal state, consistent with an intermolecular interaction (Fig. 1c). The observed increases in FRET with the bimolecular FRET reporter are significantly higher than that due to cross-excitation of the acceptor (mCit). The effects of cross-excitation were determined by summing the spectra of samples with only mCer-PKC␣ (light blue) and only PKC␣-mCit (yellow) ( Fig. 1c; combined spectrum, dashed line). To assess what aspects of these observed increases in mCit/mCer occur from non-intended effects on the fluorophores and/or clustering of the proteins, we generated an additional bimolecular FRET reporter system containing the C2 domain of PKC␣ with mCit or mCer (residues 185-292). At equal molarity of each fluorophore, the ratio of mCit/mCer was assessed for both of the bimolecular FRET reporter systems basally, with Ca 2ϩ and DAG/PS or with Ca 2ϩ and PMA (Fig. 1d). Basally, both reporters display equivalent low/negligible levels of FRET, but PKC effectors increase FRET for the sensors containing full-length PKC␣ compared with the controls with only the C2 domain. Of note, the C2 domain sensors displayed a substantial increase in FRET in the condition with liposomes, which could be explained by a specific interaction between C2 domains in the presence of both of its cofactors Ca 2ϩ and PS or from the local confinement of the sensors in a PS-and DAG-rich microdomain that forms in liposomes in the presence of Ca 2ϩ (25).
As an orthogonal approach to FRET, a cross-linking assay was used. It has been previously reported that incubation of activated PKC␣ with a Cys-Cys cross-linker (BMH; 1.3-nm spacer arm) elicits an SDS-PAGE gel shift consistent with dimer formation (20). This result was reproducible with enrichment of cross-linked protein on Coomassie-stained SDS-PAGE gels between 170 and 250 kDa accompanied by loss of monomer (75 kDa) after effector stimulation of PKC␣-FLAG (Fig. 1e). Although a consistent, single, higher molecular mass band between 170 and 250 kDa was always observed in experiments, the relative partitioning between monomer and cross-linked protein was variable between protein preparations and experimental repeats. Hence, cross-linking results were only compared within each experiment with matched controls. To probe for additional bands on the SDS-PAGE gel after cross-linking with increased sensitivity of detection, gels were transferred to a PVDF membrane and probed with a PKC␣ antibody (Fig. 1f). Increasing levels of a single cross-linked band were observed with Ca 2ϩ , Ca 2ϩ and PMA, and Ca 2ϩ and DAG/PS (Fig. 1f). The band is consistent with PKC␣-FLAG running between the expected molecular weight of dimeric and trimeric PKC␣. A recent report used a similar Cys-Cys cross-linking assay (26) for the PKC-related protein (PRK) kinase and observed a homodimer running above its expected molecular weight. Based on this previous observation and the presence of a single higher molecular weight band, PKC␣ likely homodimerizes rather than forms a higher order oligomer.
The extent of cross-linking is independent of the concentration of PKC␣-FLAG above concentrations of 100 nM (Fig. 1g). This observation is not consistent with clustering or crowding of PKC on membranes wherein increasing PKC concentrations with a constant concentration of lipids should increase crosslinked product (Fig. 1g). Importantly, an in vitro membrane localization assay shows that PKC␣ does not saturate the fixed concentration of lipids until Ͼ800 nM PKC␣ is added (Fig. 1g, bottom  panel). As an additional control, a C619A mutant of PKC␣-FLAG does not cross-link despite its ability to co-sediment with liposomes as well as the WT protein. Homodimerization of active PKC␣ has been previously suggested (11,20,27,28) and together our in vitro observations, including the increase in FRET upon activation of the initial unimolecular sensor, are consistent with the formation of homodimer.
Functional Characterization of PKC Sensors-The addition of fluorophores and other modifications to PKC␣ did not compromise the basic enzymatic and processing functions of the WT protein. Several characteristics of PKC␣ including a single band on SDS-PAGE matching its expected molecular mass, priming phosphorylations at the hydrophobic motif (Ser(P)-657) and turn motif (Thr(P)-638) (29), and PKC effector-sensitive kinase activity support this observation (Fig. 2, a-d) (30). Of note we found that three commonly used PKC phosphorylation substrates, histone IIIs, MBP, or a peptide corresponding to residues 4 -14 of MBP (MBP pep) at equal molarity elicited dramatically (roughly 1 order of magnitude) different ATPase activities of PKC␣ (Fig. 2d). We found that PKC␣-FLAG and mCer-PKC␣-mCit-FLAG behaved interchangeably in the ATPase assays (Fig. 2e). However, the PKC␣-⌬C-tail-mCit-FLAG sensor used in Fig. 6, a and c, lacks the priming phosphorylations and, as previously reported (31), has no detectable ATPase activity. To assess if this sensor retains any of its native tertiary structure in the catalytic domain and as such can provide useful information on the binding interface, we adopted an assay that has commonly been used to measure nucleotide binding in other ATPases, including in protein kinase A (32), but has to our knowledge not been tested with PKCs. MANT-labeled ADP is postulated to be a resonance energy acceptor from fluorescing tryptophans in the ATPase domain only if it binds to the catalytic site (32). Enhanced MANT-ADP excitation (450 nm) was detected upon the addition of activated PKC␣-FLAG, which could be quenched with an excess of unlabeled ADP (Fig. 2f). A titration of MANT-ADP was fit to a one site-specific binding function (Prism) to establish a binding affinity of 20.6 Ϯ 5.6 M (Fig. 2g), similar to previous reports of ADP binding affinities of 5.5 M for the ribosomal protein s6 kinase (S6K) (33) and 9.0 M for protein kinase A (32). Despite the loss of catalytic activity upon deletion of the C-tail, PKC␣-⌬Ctail-mCit-FLAG retained its ability to bind MANT-ADP (Fig. 2h), suggesting retention of its native tertiary structure. This is not surprising when considering homologous kinases such as PDK1 that fold properly without having a C-tail and are capable of kinase activity after trans binding of the C-tail of another AGC kinase (34). Taken together, these assays provide a toolbox for assessing the functionality of PKC␣-derived sensors. Estimate of PKC␣ Homodimer Binding Affinity and Effects on Specific Activity-The critical concentration at which dimerization of PKC␣ is observed is beyond the limit of both the bi-molecular FRET and cross-linking assays used in Fig. 1. Hence, a bimolecular FRET competition assay was utilized to estimate the binding affinity of a homodimeric PKC complex. The increased FRET for the mCer-PKC␣-mCit-FLAG with the addition of Ca 2ϩ and PMA or the addition of Ca 2ϩ and PS/DAG can be competed off by the addition of PKC␣-FLAG (Fig. 3a). By titrating PKC␣-FLAG into a mixture of PS/DAG-activated mCer-PKC␣-mCit-FLAG, the FRET was systematically quenched to below basal levels (Fig. 3b). The decrease in FRET does not stem from saturation of the liposomes, as the total PKC␣ concentration is well below the saturation limit determined in Fig. 1g. The competitive FRET inhibition was fit to an equilibrium partitioning between the three species (see "Experimental Procedures") to yield an upper bound on the bimolecular dissociation constant (K D ) of 5 nM. Of note, it was previously observed that the specific activity of PKC␣ increased in vitro at concentrations above 4 nM (11). To investigate the sensitivity of specific activity of mCer-PKC␣-mCit-FLAG to the dimeric state, a fixed concentration of sensor was mixed with increasing concentrations of PS/DAG liposomes. Across a range of 80 -300 g/ml liposomes, a modest decrease in FRET and a correlated decrease in specific activity was observed ( Fig. 3c; R 2 ϭ 0.91). These results are consistent with a previous report that systematically demonstrated that the specific activity of PKC␣ is sensitive to its dilution across lipid vesicles (35). Higher liposomal concentrations did not allow for robust FRET measurements in our assays due to high light scattering from the lipids.
FRET Increase from mCer-PKC␣-mCit-FLAG Is Observed upon PMA and LPA Activation in Live Cells-The flanking FRET sensor (mCer-PKC␣-mCit-FLAG) was stably expressed in CHO cells to ϳ8-fold above endogenous PKC␣ levels (data not shown). The sensor basally resided in the cytosol and, consistent with previous studies using (N-terminal) eGFP-tagged PKC␣, translocated to the plasma membrane after stimulation with PMA (Fig. 4, a and c), effectively mimicking the canonical response for PKC␣ (36). The basolateral membrane of the cell was imaged to provide a larger surface area for integration of FRET levels. The mean pixel-by-pixel FRET ratio increased steadily for the first 3 min post PMA stimulation and remained elevated through 15 min of imaging ( Fig. 4b; black). As a control, DMSO treatment did not induce translocation or significant changes in FRET (Fig. 4b). As an orthogonal approach to measure FRET (17), the FRET response to the addition of PMA was determined using a live cell suspension in a spectrofluorometer (Fig. 4b, green). A substantial increase in FRET was observed after PMA addition, with a peak at 4 min, after a gradual decrease to basal levels within 8 min (Fig. 4b, green). Together, our measurements suggest that cellular activation of mCer-PKC␣-mCit-FLAG by PMA causes an increase in FRET, but cell physiology plays a role in the persistence of the FRET response past the initial activation. A typical time-lapse of the accumulation of fluorescence at the basolateral membrane post PMA stimulation is shown (Fig. 4c) as well as a deconvolved Z-stack of the same cells at the 16-min time point (Fig. 4d). Most of the fluorescence accumulates at the plasma membrane, although a small fraction accumulates in the cytosol and nucleus consistent with results reported in NIH 3T3 cells (37). CHO cells stably expressing the sensor protein elicit a characteristic downstream PMA response by phosphorylating ERK1/2 (Fig. 4e). The PMA-dependent ERK1/2 phosphorylation persists Ͼ30 min after stimulation (Fig. 4e). The PMAstimulated ERK1/2 phosphorylation is dependent on PKC activity, as it is abolished by the PKC-specific inhibitor BimI (1.5 M) (Fig. 4f).
PMA activates PKC directly by binding its C1 domains (38). As a physiological stimulus, LPA was used to stimulate PKC as previously demonstrated (39). Unlike PMA, LPA elicits the characteristic membrane translocation response in only ϳ50% of the cells, consistent with previous reports (40). A transient PKC translocation to the plasma membrane was observed after LPA stimulation (Fig. 5a). Membrane translocation was accompanied by a transient increase in FRET for mCer-PKC␣-mCit-FLAG (Fig. 5, a and b). The sensor protein also induced a biphasic ERK1/2 phosphorylation response with a dramatic increase at 2 min post stimulation and return to basal levels at 4 min followed by sustained increase at 6 and 8 min (Fig. 5c). The persistent ERK1/2 phosphorylation 15 min post LPA treatment was inhibited by pretreatment with the PKC inhibitor BimI (Fig. 5d). We speculate that the biphasic ERK1/2 response to LPA is caused by bifurcation in LPA-induced PKC signaling with reconvergence of the signals at the downstream ERK1/2 at different time scales. Together, the LPA and PMA responses are consistent with mCer-PKC␣-mCit-FLAG reporting an activated dimeric state for PKC␣, consistent with our in vitro observations. However, other factors including clustering of the sensor at the plasma membrane or conformational changes independent of homodimerization may also contribute to the observed cellular FRET response after PKC activation.
Elements in Both the RD and CD Are Critical for Homodimerization-To identify the domains contributing to homodimerization in vitro, several domain truncations were made into the bimolecular FRET reporter system (mCer-PKC␣-FLAG ϩ PKC␣-mCit-FLAG). It was observed that deletion of either both of the C1 domains or the conserved C-terminal extension of the AGC kinase domain (C-tail (V5) domain) completely abolished the effector induced increase in FRET, whereas the deletion of the C2 domain retained a partial response (Fig. 6a). The involvement of the C1 and C2 domains in homodimerization is consistent with a previous report of a high affinity interaction between C1 and RD after activation (11). To probe for individual interactions between RD and CD (RD-RD, RD-CD, CD-CD), bi-molecular FRET pairs containing only the RD or CD were engineered. Although these confirm an enhanced RD-RD interaction (Fig. 6b), no appreciable interaction involving the CD (RD-CD, CD-CD) were detected upon activation even at concentrations much higher than the K D (150 nM). These latter observations are in contrast with the effects of C-tail truncations in the context of full-length PKC␣ (Fig. 6a). The C-tail-truncated catalytic domains retain their ability to bind nucleotide, attesting to the retention of a folded catalytic site (Fig. 2h). The C-tail is a conserved region of the kinase domain that is required to stabilize the active state of the kinase and is critical for hetero-dimerization of almost all AGC kinases (41). An alternative interpretation for the lack of FRET response for the C-tail deletion sensor is the displacement of the C-terminal mCit fluorphore such that no FRET is detected despite the occurrence of homodimerization. To test this possibility, a second bimolecular FRET pair (mCer-PKC␣-FLAG ϩ mCit-PKC␣-FLAG) was tested. The increase in FRET upon activation was found to be independent of which termini the fluorophores are positioned on PKC (Fig. 6c). The C-tail involvement in homodimerization was confirmed by an almost complete loss of FRET after deletion of only one C-tail in either of the bimolecular FRET pairs (mCer-PKC␣ ϩ PKC␣-⌬C-tail-mCit; mCer-PKC␣-⌬C-tail ϩ mCit-PKC␣) (Fig. 6c). Together, these observations suggest that intermolecular interactions involving both the RD and CD contribute to homodimerization.

SPASM Technique Demonstrates That the RD and CD Interact in the Dimeric
State-In the context of the intact PKC protein, all domains are tethered in close spatial proximity, which increases their local effective concentration. Hence, weak interactions between domains (RD-CD, CD-CD) that are nonetheless essential for homodimerization may not be detected in bimolecular assays. To address this limitation, SPASM modules were spliced into different locations of full-length PKC␣. The SPASM module consists of a FRET pair (mCit/mCer) flanking an ER/K linker, which controls the effective concentration of the interaction between proteins fused to its ends (10). These modules are designed to (a) artificially increase the local concentration of domains as compared with untethered domains while (b) using the FRET readout to resolve effector-stimulated changes in interactions between domains. Given that the bimolecular RD-RD interaction is enhanced after PKC activation (11), it can serve to nucleate additional RD-CD and CD-CD interactions. Thus, a SPASM module was spliced in between the RD and CD to test for an intermolecular RD-CD or CD-CD interaction that can stabilize a high affinity dimer. To avoid potential complications from intramolecular interactions, a 30-nm ER/K linker was used (low effective concentration of the intramolecular interaction (10)), and the system was engineered to report only on bimolecular interactions (FRET donor placed adjacent to either the RD or CD; FRET acceptor adjacent to the CD of a different sensor). Both sensor pairs showed a comparable increase in FRET upon activation providing direct evidence for a RD-CD interaction, with the potential for a CD-CD interaction in the homodimeric state (Fig. 6d). Next, a 10-nm SPASM module was inserted in three separate locations in the PKC gene (1, between the pseudosubstrate (PS) and C1a domain; 2, between the C1b and C2 domains; 3, between the C2 and CD). Although the PS is known to interact with the CD basally (42), this interaction did not change after activation (type 1; Fig. 6e). In contrast, the C1 domains display a marked enhancement in interaction with either the C2 or CD after activation (type 2), whereas collectively the RDs interact very prominently with the CD after activation (type 3; Fig. 6e). Using the type 3 configuration to directly report on the interaction between the RD and CD, deletions of C1a, C1b, C2, or the C-terminal 15 residues of the C-tail were made. Surprisingly, the activated RD-CD interaction remained partially intact after the deletion of any of the three regulatory domains, but shortening of the C terminus of the C-tail completely disrupted this interaction (Fig. 6f). Together, these data suggest that all three regulatory domains (C1a, C1b, and C2) contribute to the dimer interface, but the C-tail is absolutely critical for the interaction. Given that the C-tail is essential for the active conformation of the kinase domain (43), these data cannot distinguish between a direct role for the C-tail as part of the dimerization interface and its indirect contribution through the global conformation of the kinase domain.
To probe for precise interfaces involved in homodimerization, we sought to identify Cys residues that were being crosslinked by BMH, which requires two Cys residues to be within 1.3 nm. The only Cys residues within PKC␣ are found in the Cys-rich C1a and C1b domains and within the CD. Activationspecific cross-linking was still detected after deletion of either the C1a or C1b, yet a single point mutation in the C-tail (C619A) completely abolished the cross-linking response (Fig.  6g). Interestingly, the corresponding residue in PKC␤II (Cys-622), which is found in the active-site tether region, is buried within an interaction interface with the C1b domain in a recent crystal structure (18). This raises the possibility that the interaction between the C1b and c-tail captured in the crystal structure could be an interaction utilized in dimerization. Regardless, our data propose a new conformation for activated PKC wherein the CD, C1a, C1b, and the C2 domains all contribute to stabilizing a homodimer (Fig. 6h). It was reported from two separate groups that truncation of the C-terminal 10 residues in the C-tail (31) or alanine mutagenesis of regions from the C1a and C2 domains (44) completely abolished activity of PKC␣. A potential explanation for these observations is the disruption of homodimerization and its consequent effects on PKC catalytic activity. To probe the function of a homodimeric conformational state, we report on four independent vignettes concerning different aspects of PKC␣ function. Each provides some insight and highlight how our mechanistic understanding of PKC␣ function is enhanced when considering this additional state. acceptor are on the same sensor but separated by an ER/K linker. In the absence of an interaction, the 10-nm ER/K linker separates domains beyond FRET distance. An increase in FRET indicates enhanced interaction between domains at either end of the ER/K linker. a, C1 and C-tail domains are essential for PKC␣ homodimerization; mCer-PKC␣ ϩ PKC␣-mCit with no truncation (WT) or ⌬C1, ⌬C2, ⌬C-tail in both proteins. b, a bimolecular RD-RD interaction, but not an RD-CD interaction or CD-CD interaction is detected after activation. RD-CD, mCer-CD ϩ mCit-RD; RD-RD, mCit-RD ϩ mCer-RD; CD-CD, mCer-CD ϩ mCit-CD (50 nM donor sensor, 100 nM acceptor sensor). c, the C-tail in the catalytic domain is essential for homodimerization. 1, mCer-PKC␣ ϩ mCit-PKC␣; 2, mCer-PKC␣ ϩ PKC␣-⌬C-tail-mCit; 3, mCer-PKC␣-⌬C-tail ϩ mCit-PKC␣. d, both the RD and CD domains reside at the dimerization interface. 1, RD-mCer-30-nm ERK-CD ϩ RD-30 nM ERK-mCit-CD; 2, RD-30-nM ERK-mCer-CD ϩ RD-30-nm ERK-mCit-CD. e, enhanced RD-CD interaction in the PKC homodimer. 1, PS-mCer-10-nm ERK-mCit-C1-C2-CD; 2, PS-C1-mCer-10-nm ERK-mCit-C2-CD; 3, RD-mCer-10-nm ERK-mCit-CD. f, C1a, C1b, and C2 domains partially contribute to dimerization, whereas the C-terminal 15 residues of the C-tail are essential. Shown is change in FRET upon activation of type 3 configuration (as in e), with truncation of the C1a, C1b, C2 domains or the C-terminal 15 residues of the C-tail. g, proximity of the C-tail to the homodimer interface. Shown is a representative Coomassiestained gel of PKC␣-FLAG protein under basal or activated (Ca 2ϩ ϩDAG/PS) conditions. Data for full-length PKC␣ (WT), domain deletions (⌬C1a, ⌬C1b), and mutant (C619A) proteins are shown. Note the complete loss of detectable dimerization upon mutagenesis of a Cys-619 in the C-tail region of the CD. The BMH cross-linker has a maximal reach of 1.4 nm. For reference the C1b domain is about 3 ϫ 2 ϫ 2 nm (18). h, model for dimer formation based on data presented in a-g. I, RDs interact with each other upon effector stimulation to nucleate homodimerization. II, RD-CD interaction in trans facilitated by the C-tail. III, proposed conformation for the PKC homodimer. a-f, error bars, S.E. n Ն 3.

Equilibrium between Closed, Open, and Dimerized
Conformations of PKC␣-In the absence of effectors PKC␣ predominantly populates a "closed" conformation with the CD autoinhibited by the RD (18). Splicing a SPASM module between the RD and CD should decrease the effective concentration of the RD for the CD in proportion to the length of the linker.
"Prying" open the RD-CD interaction with SPASM modules containing increasing-length ER/K linkers systematically reduces the basal level of FRET (Fig. 7a) and increases the basal level of activity ( Fig. 7b; 10 nm, ϳ22/12%; 30 nm, ϳ37/22% maximum activity, compared with ϳ3/3% for WT PKC␣ (histone IIIs/MBP pep)). Of note, prying open the RD-CD interaction alone is not sufficient to induce dimerization in the absence of effectors. Yet the presence of effectors increases FRET regardless of ER/K linker length (Fig. 7a). Given the importance of RD-CD interactions for dimer formation (Fig. 6), it is not surprising that increasing ER/K linker length also disrupts dimer formation (Fig. 7a). Despite the different levels of FRET with 10-and 30-nm ER/K linkers after activation, they both have specific activity similar to that of the CD in the absence of the RD (TEV cleaved) ( Fig. 7b; 10 nm, ϳ88/86%, 30 nm, ϳ94/95% maximum activity (histone IIIs/MBP pep)). Following TEV treatment the RD should partition to the lipid vesicles leaving the CD completely uninhibited. It was observed that in the presence of effectors both the 10 and 30 nm ER/K linkers demonstrate similar activities despite different levels of FRET. Taken together these results are consistent with a dynamic equilibrium for PKC␣ between a closed (basal autoinhibited: medium/low FRET), an open (no auto-inhibition: low FRET), and a dimeric conformation (no auto-inhibition: high FRET) with equivalent specific activities in the open and dimer forms (Fig. 7c). In the proposed model for mature PKC␣, the specific activity is regulated by only one parameter, namely, the ratio of non-auto-inhibited catalytic domains to the total number of catalytic domains in a given reaction.

Loss of the Turn Motif Priming Phosphorylation Causes Basal
Dimerization-We next aimed to bridge our global structural model of PKC to high resolution x-ray crystallographic data. A recent partial crystal structure of PKC␤II suggests a direct binding interaction between its C1b and C-tail domains (18). To investigate the possibility that this interaction is similar to the one utilized in homodimerization of PKC␣, individual point mutations in PKC␣ (K621A, D357A, H633G, T638A) were engineered in an attempt to disrupt the C1b-C-tail interaction. The residues are mapped onto the crystal structure of PKC␤II (orange surface rendering) as well as Cys-619 (yellow surface rendering) for visual reference (Fig. 8a). Each mutation engineering into the bimolecular FRET reporters elicited modest effects, except for T638A, which counter intuitively increased the FRET response under basal and effector-stimulated conditions (Fig. 8, b and c). Cross-linking analysis of the T638A mutant confirmed the basal formation of a dimeric species (Fig.  8d). Residue Thr-638 is an essential priming phosphorylation site (turn motif (TM)) in PKC␣ that is essential for full catalytic activity after activation (45). The T638A mutant has ϳ36/22% (histone/MBP pep) of WT PKC␣ activity after activation with Ca 2ϩ ϩ PS/DAG (data not shown), similar to previous reports (45). The Thr 3 Ala mutant is commonly incorporated in the TM site to mimic the loss of TM site phosphorylation (45)(46)(47). Using such an approach for multiple AGC kinases including PKC, it has been broadly suggested that TM phosphorylation functions to anchor the C-terminal portion of the C-tail, containing the hydrophobic motif, to the catalytic domain (47). Based on this model, it might be anticipated that basal homodimerization of PKC␣-T638A is the result of an indirect destabilization of the interaction of the catalytic domain with the phosphorylated hydrophobic motif site on the C-tail. To assess this possibility, phosphonull mutation of the hydrophobic motif site (S657A) was incorporated into the FRET sensors. In contrast with the proposed model, basal homodimerization induced by T638A is independent of the S657A mutation (Fig. 8e). The mechanism by which T638A induces basal dimerization is informed by an alignment of two crystal structures of the catalytic domain of PKC␤II ( Fig. 8f; Ref. 48). The region just N-terminal of the TM site (AST region; Ref. 43) undergoes a dramatic conformational change when bound to C1b (cyan), yet in both structures the C-tail converges at the phosphorylated TM residue that appears tightly anchored to the catalytic domain by electrostatic interactions with four basic residues (red) (Fig. 8f). We postulate that disruption of the anchoring phosphorylated TM site allows the AST region of the C-tail to adopt a conformation that can induce homodimerization in the absence of effectors. Independent of this interpretation, the finding that an unphosphorylated TM can basally homodimerize may provide a mechanism by which PKC␣, unlike most AGC kinases, is autophosphorylated at its three priming sites (6,49) and deserves further investigation.

Disruption of PMA/LPA-induced High FRET State in Cells
Modulates PKC-specific Phosphorylation of ERK1/2-To test the functional significance of dimerization in cells, one peptide each from the RD and CD was selected to specifically disrupt (relative to a matched scrambled peptide; all myristoylated) prospective PKC␣ dimerization interfaces. Although the effects of the peptides on FRET were modest in vitro, they were specific and concentration-dependent across two different sensor systems (Fig. 9, a and b). Addi-tionally, the RD peptide did not disrupt liposome binding in sedimentation assays (Fig. 9c). When treating cells with the myristoylated RD peptide, the PMA-induced high FRET state of mCer-PKC␣-mCit-FLAG was disrupted in live cell suspensions ( Fig. 9d; fluorometer measurements). Both the RD and CD peptides had a significant and sustained dampening effect on the PMA-induced increase in FRET of the sensor compared with their respective scrambled peptide controls ( Fig. 9e; microscope measurements). Additionally, neither peptide had a noticeable effect on the rate or relative degree of basolateral membrane translocation of the sensor in CHO cells after PMA treatment ( Fig. 9f; data not shown). Next, the effect the peptides had on the PMA-and LPAinduced phosphorylation of ERK1/2 was assessed by pretreatment of cells for 15 min before stimulation with PMA or LPA. The native RD peptide specifically lowered the level of ERK1/2 phosphorylation after PMA (Fig. 9g) or LPA treatment (Fig. 9h) relative to the scrambled control. When the cells were stimulated with serum, which is expected to activate ERK1/2 through both PKC-dependent and independent pathways (50), both the RD peptide and the known PKC inhibitor BimI had comparable partial effects (Fig. 9i). Last, the CD peptide was able to specifically disrupt LPA-induced ERK1/2 phosphorylation compared with its scrambled control (Fig. 9j). Together, these results are consistent with a model in which disruption of PKC␣ homodimerization is capable of inhibiting PKC-mediated downstream ERK1/2 phosphorylation.

Destabilizing the Closed Conformation Provides Stable PKC␣
Translocation to the Membrane-The PKC inhibitor GF109203x is known to stabilize the translocation of PKC␣ after effector stimulation (46). Stensman et al. (46) hypothesized that the mechanism of GF109203x-influenced translocation sensitivity is due to a destabilization of the basal RD-CD interaction. Additionally, several analogues of this inhibitor, including BimI, are known to induce conformational changes in the CD (51). Using a PKC␣ sensor with a SPASM module inserted between the RD and CD, we directly detect a reduction in the strength of the interaction between RD and CD when incubated with BimI under basal conditions (Fig. 10a). Based on the three-state model (Fig. 7c), destabilization of the closed conformation should drive the equilibrium toward dimer formation after effector stimulation. Accord-ingly, we find increased dimerization of PKC␣ in the presence of BimI (Fig. 10, b and c). We hypothesize that both the open and dimeric conformations of PKC␣ can engage with the plasma membrane. Consistent with this concept, LPA stimulation of cells stably expressing the flanking sensor (mCer-PKC␣-mCit), substantially increased the fraction of cells with stable localization of the sensor at the membrane in the presence of BimI (Fig. 9, d and e). Our findings provide direct evidence that a small molecule that can destabilize the closed conformation can facilitate both dimerization and stabilize membrane localization of PKC␣ in the presence of effectors. This opens the possibility of other small molecules or interacting proteins influencing localization of PKC by influencing the stability of either the closed or dimer states. As a corollary, PKC translocation is not necessarily an appropriate proxy for PKC activation in live cells as demonstrated by the decoupling of activity and translocation with the peptide inhibitors ( Fig. 9) and BimI (Fig. 10, d and e).

PKC, a Conservative and Dynamic
Kinase-The negative consequences of unregulated PKC activity have been apparent since its discovery as the receptor for tumor-producing phorbol esters (38). In the subsequent decades it has been observed that the dysregulation of PKC function is tied with several different pathologies (52)(53)(54)(55). Concurrently, multiple distinct mechanisms of auto-inhibition of PKC have been described (18,42,56,57). Collectively these findings portray PKC as a conservative signaling protein that is only capable of functioning under a limited ensemble of conditions. In contrast, an increasing body of evidence positions PKC as a lynchpin in intracellular communication in diverse signaling environments (58). The mechanisms allowing PKC to transition from the tightly regulated autoinhibited state to a catalytically active state in a dynamic manner and in the context of multiple signaling inputs have been largely unexplored. In this study we find that homodimerization is an additional regulatory mechanism exploited by PKC␣ to overcome intramolecular auto-inhibition.
Three-state Model-Our data confirm previous reports that find a positive correlation between PKC concentration or dimerization and specific activity (11,20,35) (Fig. 3c; data not shown). A possible explanation is that intermolecular interactions in the homodimer enhance the catalytic activity of the kinase domain. However, the fully activated homodimeric PKC is less active (86%) than just the catalytic domain (Fig. 7). Furthermore, varying the fraction of activated PKC in the monomer and dimer states with different length ER/K helices between the RD and CD did not affect specific activity in the presence of effectors (Fig. 7). Orthogonally, it was observed that prying open the basal auto-inhibitory interactions using ER/K helices only modestly increased specific activity (Fig. 7). These observations lead us to propose a model wherein dimerization acts as a "mass action sink" by sequestering the regulatory domains away from their autoinhibitory interactions. In our proposed three-state model, PKC is in dynamic equilibrium between the auto-inhibited state that has no catalytic activity (0%), the monomeric and non-auto-inhibited (open) state that has full catalytic activity (100%), and the homodimeric state that also has full catalytic activity (100%). In Fig. 11 we use this model to demonstrate how dimerization can influence the specific activity of PKC. Of particular note in this three-state model there are now two largely independent ways in which PKC could be activated; through disruption of the auto-inhibitory interactions (K open ) or through stabilization of the homodimer (K dimer ). Likely, canonical PKC effectors, such as those used in this study, will modulate both equilibrium constants. We postulate that for the nuanced regulation observed in cells, interactions between PKC and small molecules or proteins may differentially influence these two parameters.
Peptides and Small Molecules Can Modulate PKC Function through Two Distinct Mechanisms-In this study we demonstrate the feasibility and physiological significance of peptide and small molecule modulation of intramolecular interactions in PKC. Two peptides were identified and characterized that selectively disrupt the homodimeric state (K dimer ) of PKC ( Fig.  9, a-f). When introduced into a cellular system, both peptides specifically and dramatically reduced the PKC-driven downstream phosphorylation of ERK1/2 ( Fig. 9, g--j). Conversely, we observed that the small molecule PKC inhibitor BimI disrupted the interaction between the auto-inhibited and open states (K open ; Fig. 10a), which in turn leads to an increase in the homodimeric state (Fig. 10, b and c). When introduced into a cellular system, BimI was able to alter the translocation response of PKC post LPA stimulation (Fig. 10, d and e). These results demonstrate potential approaches for PKC intervention as well as provide precedence and detailed mechanism for noncanonical PKC regulation that could be utilized in physiological contexts.
Dimeric State Versus Higher Order Oligomerization-Although the results presented here are consistent with homodimerization of PKC␣, they do not rule out the formation of higher order oligomers. The FRET measurements do not differentiate between homodimers and higher order oligomers. However, the cross-linking studies show a single dominant higher molecular weight band that is consistent with a dimer. Given that BMH cross-linking is limited to the close proximity (1.3 nm) of two cysteines, it is possible that higher order oligomers are not efficiently cross-linked, resulting in dimerization as the predominant species detected using this method. With the absence of evidence for higher order oligomerization, we have presented concepts along the lines of the simplest oligomeric state (dimer). Notably, the concept of dimerization sequestering the regulatory domains away from the auto-inhibited state is compatible with higher order oligomerization. In regard to higher order oligomerization, it has been previously reported that PKC as well as certain PKC phosphorylation substrates can drive phase separation on lipid membranes (25). Additionally, the potent PKC activator protamine sulfate has been shown to aggregate PKC in an active state in the absence of additional effectors in vitro (59). Further studies are necessary to probe the possibility of higher order PKC oligomerization and its physiological relevance.
Domain-Domain Interactions in Other AGC Kinases-Interactions between modular domains are known to broadly regulate AGC kinases. However, none of them have been proposed to utilize the specific mechanism described here. The AGC kinase PDK1 is known to heterodimerize with several AGC kinases including PKCs, AKT, PKN, and S6K through interactions involving the PIF binding pocket on the PDK1 catalytic domain and the hydrophobic motif found on the C-tail of the other kinase (34). This interaction has been shown to dramatically enhance PDK1 catalytic activity and cellular function (34). In contrast, homodimerization facilitated by interactions between the catalytic and PH domains has been speculated to down-regulate PDK1 activity in cells (60). In the protein kinase C-related protein 2 (PRK2) an inhibitory homodimerization occurs between the N-terminal regulatory domains (including a C2 domain) and the catalytic domains (26). The AGC Rhoassociated kinase I and II (ROCKI and -II) are also observed forming a homodimer between regulatory domains and the C-tail (61). Although dimerization of ROCKI does not appear to grossly stimulate or inhibit kinase activity, it does alter enzyme kinetics by lowering the K m of ATP by 60 -80-fold (61). Homodimeric truncations of ROCKI and -II and subsequently MRCK␤, have been crystallized demonstrating the involvement of the C-tail at the intermolecular interface (62)(63)(64). In PKC an inhibitory intramolecular interaction between the C1 domain and PIF-pocket on the catalytic domain has been previously reported (14). Together these studies suggest that domain-domain interactions in either an intermolecular or intramolecular context are common regulatory mechanisms in AGC kinases. Furthermore, they hint at the recurrence of particular domain-domain interactions such as between the C1 and AGC kinase C-tail domains. Detailed studies in related kinases are necessary to further our understanding of AGC kinase regulation. However, as we report here, the domains involved in auto-inhibitory interactions in PKC␣ also engage under activated conditions. The conformational flexibility apparent in multidomain interactions belies traditional structural approaches including x-ray crystallography and NMR and also complicates insight gained from truncation and site-directed mutagenesis. As such, we propose that the SPASM technique can bridge the gap between our knowledge of modular domain structure and multiplexed cellular function.