A Novel Conserved Domain Mediates Dimerization of Protein Kinase D (PKD) Isoforms

Protein kinase D (PKD) isoforms are protein kinase C effectors in signaling pathways regulated by diacylglycerol. Important physiological processes (including secretion, immune responses, motility, and transcription) are placed under diacylglycerol control by the distinctive substrate specificity and subcellular distribution of PKDs. Potentially, broadly co-expressed PKD polypeptides may interact to generate homo- or heteromultimeric regulatory complexes. However, the frequency, molecular basis, regulatory significance, and physiological relevance of stable PKD-PKD interactions are largely unknown. Here, we demonstrate that mammalian PKDs 1–3 and the prototypical Caenorhabditis elegans PKD, DKF-2A, are exclusively (homo- or hetero-) dimers in cell extracts and intact cells. We discovered and characterized a novel, highly conserved N-terminal domain, comprising 92 amino acids, which mediates dimerization of PKD1, PKD2, and PKD3 monomers. A similar domain directs DKF-2A homodimerization. Dimerization occurred independently of properties of the regulatory and kinase domains of PKDs. Disruption of PKD dimerization abrogates secretion of PAUF, a protein carried in small trans-Golgi network-derived vesicles. In addition, disruption of DKF-2A homodimerization in C. elegans intestine impaired and degraded the immune defense of the intact animal against an ingested bacterial pathogen. Finally, dimerization was indispensable for the strong, dominant negative effect of catalytically inactive PKDs. Overall, the structural integrity and function of the novel dimerization domain are essential for PKD-mediated regulation of a key aspect of cell physiology, secretion, and innate immunity in vivo.

substrate/organelle specificity, stability, and functions of PKDs may be markedly affected by formation of complexes containing multiple PKD polypeptides. However, knowledge of the ability of PKDs to oligomerize and functional and regulatory consequences of PKD-PKD interactions is limited and contradictory.
In pioneering studies, Malhotra's group (28) reported that PKD2-PKD3 dimers control constitutive secretory vesicle biogenesis and fission at the TGN of HeLa cells. The Storz lab (29) subsequently observed that a PKD2-PKD3 complex in HeLa and MDA-MB-468 cells regulates F actin-based, directed motility. These investigations suggest heteromultimeric PKDs may play important roles in coupling DAG signals to regulation of key aspects of cell physiology. Conclusions derived from these narrowly focused studies are consistent with the data, but their general applicability is constrained by the following considerations. Only a fraction of HeLa cell D kinases was detected in PKD2-PKD3 complexes (28,29). Activities and functions of heteromeric and non-heteromeric PKDs were not segregated and separately analyzed. Thus, it is not definitively known whether critical PKD activity was restricted to heteromeric complexes. Likewise it is not known whether PKD2-PKD3 complexes are dimers or higher order oligomers or whether PKD polypeptides are normally distributed among monomers and multimers or restricted to a single structural species. The possibility that PKD homo-oligomerization is crucial for function has not been studied. Nothing is known about the ability of broadly expressed PKD1 to self-associate or bind with other PKDs. Importantly, intrinsic structural features that mediate assembly of PKD-PKD complexes have not been elucidated.
In BON neuroendocrine cells, vesicle biogenesis at the TGN and chromogranin A secretion were regulated solely by PKD2 (11). PKD1 or PKD3 depletion had no effect. DT40 B lymphocytes express PKDs 1 and 3. Deletion of either PKD gene had no effect on antigen-or PMA-induced phosphorylation and nucleus to cytoplasm translocation of HDAC5 and HDAC7 (22). Disruption of both PKD genes abrogated HDAC phosphorylation/translocation, but expression of either a PKD1 or PKD3 transgene rescued the PKD null phenotype. Thus, in these studies, hetero-oligomeric PKDs were not required to control TGN vesicle fission or gene transcription.
Embryogenesis proceeded normally in mice globally lacking PKD2 activity (6). PKD3-deficient mice exhibited only a mild skeletal deficit. Thus, Golgi vesicle fission, directed cell migration, and other PKD-modulated processes evidently progressed properly in a vast array of cells in the absence of heteromeric PKD2-PKD3. C. elegans PKDs, DKF-2A and DKF-2B, play critical roles in innate immunity and associative learning (23,24). Because DKF-2A and -2B are expressed in a mutually exclusive manner in intestine and neurons, respectively, hetero-oligomeric kinases are not essential for in vivo regulation in the nematode model.
Overall the frequency, mechanism, regulatory significance, and physiological relevance of PKD-PKD interactions remain largely unknown. To determine whether PKD oligomerization plays a central role in coupling DAG signals to the regulation of cell/tissue physiology, it is necessary to address several pertinent questions: Is widely expressed PKD1 capable of oligomer-ization? Are native PKDs 1-3 predominantly monomers or are D kinases incorporated into hetero-and homodimers or larger multimers? If PKDs oligomerize, what structural features mediate PKD-PKD interactions? Are scaffold proteins essential to guide oligomerization? What are the consequences of disrupting PKD oligomerization in the context of cell physiology and with respect to PKD-mediated regulation of function in vivo ? We demonstrate that human PKDs 1-3 and the C. elegans PKD, DKF-2A, are homo-or heterodimeric in cell extracts and intact cells. We discovered a novel, conserved domain that mediates dimerization of PKD1, 2, and 3 monomers. A similar domain directs homodimerization of DKF-2A. Disruption of PKD dimerization abrogates secretion of PAUF, a protein carried in small TGN-derived vesicles. Disruption of DKF-2A homodimerization in C. elegans intestine impaired the immune defense of the animals against a bacterial pathogen. Finally, dimerization is indispensable for strong, dominant negative effects of catalytically inactive PKDs.

DKF-2A Polypeptides Assemble into Homo-oligomeric Complexes; a Novel Domain Mediates Incorporation of PKD Monomers into Multimers-
The possibility that PKD monomers interact to create homo-oligomeric kinases has not been systematically investigated. C. elegans provides a good model for analysis because the animals selectively express two prototypical PKDs. DKF-2A and DKF-2B accumulate in intestinal epithelial cells and a subset of neurons, respectively (23,24). Thus, DKF-2A and DKF-2B will function either as monomers or homo-oligomers.
Transgenes encoding HA-or FLAG-tagged DKF-2A and DKF-2B were expressed in HEK293 cells. PKD-PKD interactions were assessed by co-immunoprecipitation and Western immunoblot analysis (Fig. 1A). Both the 120-kDa WT DKF-2A and constitutively active DKF-2A EE (A-loop serines 925 and 929 substituted with phosphomimetic Glu) were incorporated into homomultimers. Dimers are not distinguished from higher order complexes in these assays. In contrast, the 98-kDa DKF-2B polypeptide did not homo-oligomerize or bind with DKF-2A or DKF-2A EE.
DKF-2A and DKF-2B have identical regulatory and kinase domains, but amino acid sequences of their N-terminal regions are markedly divergent. Thus, we examined the ability of the unique N-terminal segment of DKF-2A (amino acids 1-319) to mediate oligomerization. DKF-2A truncation mutants lacking amino acids 1-100, 1-199, 1-215, or 1-227 avidly associated with full-length DKF-2A (Fig. 1, B and C). Elimination of 249 or 404 N-terminal residues ablated complex formation (Fig. 1C). These observations place the N-terminal boundary of a novel PKD oligomerization domain (OD) in a segment of DKF-2A encompassing amino acids 227-249. The OD was more precisely mapped by assaying binding properties of internally deleted DKF-2A mutants (Fig. 1, B and D). Excision of amino acids 228 -249 elicited an 80% decrease in complex formation. Binding interactions were suppressed further but still detected in DKF-2A lacking residues 228 -271. Elimination of a 73-residue segment of DKF-2A (⌬228 -300) generated a D kinase that was virtually incapable of engaging in homomultimeric com-  ). After 48 h, the cells were lysed, and PKDs were precipitated from cell extracts by adding anti-HA IgGs and protein G-Sepharose 4B beads. Precipitated and co-precipitated proteins were analyzed by SDS-PAGE and Western blotting, using anti-HA and anti-FLAG IgGs, respectively. The same IgGs detected epitope-tagged PKDs in blots of cell extracts. Tubulin was monitored as a loading control. B, a schematic diagram depicts mutant DKF-2A proteins and DKF-2A domains that were assayed for oligomerization activity. In C-G, transfection, lysis, immunoprecipitation, and Western blot analysis were performed as indicated in A and "Experimental Procedures plexes. The absence of PKD-PKD interactions in a mutant lacking residues 250 -319 directly demonstrated that amino acids located beyond the critical 228 -249 segment were also essential for PKD self-association.
To determine whether DKF-2A regulatory and catalytic domains play auxiliary roles in multimerization, we created transgenes in which (a) the N-terminal segment of DKF-2A (residues 1-319) was fused to GFP and (b) C1a and C1b domains (residues 311-545) were coupled to mCherry red fluorescent protein (Fig. 1B). Although C1a and C1b domains fold and function normally in the absence of surrounding N-and C-terminal segments of signaling proteins (30 -33), the DKF-2A 311-545-mCherry fusion protein failed to bind with either unstimulated or PMA-PKC-activated DKF-2A (Fig. 1E). In contrast, a DKF-2A 1-319-GFP chimera formed complexes with WT PKD and N terminally truncated D kinases (⌬100, ⌬199) that retain the OD (Fig. 1F) Overall, the results indicate that DKF-2A contains a novel domain (OD), encompassing amino acids 228 -319, that mediates assembly of PKD homodimers or higher order multimers. The OD apparently folds and functions independently of the conserved DAG binding and kinase domains of DKF-2A. Multimer formation was not affected by conformational changes caused by mimicking A-loop phosphorylation.
A Conserved N-terminal Domain Promotes Homo-and Hetero-oligomerization of Human PKDs-Amino acids comprising the DKF-2A OD aligned with a sequence near the N terminus of PKD2 ( Fig. 2A). The 93-residue PKD2 segment is similar in size to the DKF-2A OD and contains 31 identical amino acids (33%) and 24 highly conservative substitutions (Arg/Lys, Asp/Glu, Leu/Ile, etc.), yielding an overall similarity of 59%. The domain identified in PKD2 is highly conserved in PKDs 1 and 3. For example, the corresponding region of PKD1 is 78% identical and 90% similar to the putative PKD2 OD ( Fig.  2A). We explored the possibility that the conserved domain mediates formation of homo-and/or hetero-oligomeric mammalian PKDs.
A fusion protein containing the C1a and C1b domains of PKD2 but lacking the predicted OD, did not form complexes with PKDs 1-3 (Fig. 2G). This finding and the results presented above suggest that the DAG/PMA binding modules are not involved in PKD oligomerization.
Our studies show that PKDs 1-3 assemble into stable homoor heteromultimeric complexes. A conserved but previously uncharacterized OD consisting of ϳ92 amino acids directs incorporation of PKD monomers into multimers. Multimerization is achieved by interactions between the N termini of PKDs. The recombinant OD, which functions independently and persists in cells, avidly associates with PKDs 1-3. Thus, overexpressed OD could be used to selectively disrupt PKD oligomerization.
PKDs Oligomerize in Intact Cells-It was possible that multi-PKD complexes assembled when previously segregated PKDs intermingled after detergent mediated cell lysis. Thus, we used cell-permeable, irreversible cross-linkers to assay PKD-PKD association within cells. 1,5-Difluoro-2,4-dinitrobenzene (DFDNB) and disuccinimidyl suberate (DSS) are homo-bifunctional compounds that react with Lys ⑀-amino groups. DFDNB has a short spacer arm that enables selective cross-linking of protein subunits, whereas a longer spacer in DSS facilitates cross-linking of both subunits and binding partners.
Transfected cells expressing HA-and FLAG-tagged PKDs were incubated with cross-linker before lysis. Treatment with DFDNB or DSS sharply diminished the content of monomeric DKF-2A (Fig. 3A). Simultaneously, covalently linked, high molecular weight DKF-2A oligomers accumulated. Endogenous PKD2 was also incorporated into high molecular weight complexes (Fig. 3B). Like PKDs, PKC␣ and PKC␦ contain tandem C1a and C1b regulatory domains and a C-terminal kinase module. However, PKCs are monomeric proteins. Levels of PKC␣ and PKC␦ monomers were not affected by incubating cells with DFDNB or DSS (Fig. 3B); neither PKC appeared in a high molecular complex. The results confirm the validity and specificity of the approach.
PKD multimerization was also probed with the reversible, Lys-directed cross-linker dithiobis(succinimidyl propio-nate) (DSP). DSP permeates cells and has a disulfide bond located between the two Lys-reactive moieties. Thus, reduction with ␤-mercaptoethanol will regenerate monomers from crosslinked proteins. Treatment of cells with DSP eliminated DKF-2A monomers and generated multimeric complexes (Fig.  3C, lower panel). Incubation of cell extracts with ␤-mercapto-ethanol, under denaturing conditions, regenerated DKF-2A monomers (Fig. 3C, upper panel). A similar result was obtained for endogenous human PKDs, as illustrated for PKD2 (Fig. 3D). Internal deletion of the OD in the PKD1 ⌬50 -125 mutant dramatically suppressed cross-linker-mediated loss of monomers and simultaneously blocked production of higher order multi-
Overall, the data suggest that most intracellular PKD molecules are in very close proximity (i.e. in complexes) with partner PKDs in the physiological environment of intact cells. Thus, PKDs are multimeric signaling proteins.
DKF-2A and Human PKDs Are Dimers-Gel filtration chromatography and sucrose density gradient sedimentation can provide accurate measurements of the Stokes radius (R s ) and sedimentation coefficient (S 20,w ) of PKDs in minimally manipulated cell extracts (34,35). R s and S 20,w values are then used to calculate the native PKD molecular weight (M r ) via the Svedberg equation (see "Experimental Procedures"). Comparison of native M r with monomer M r (determined from genomic DNA or cDNA databases) reveals the number of monomers in the complex.
Lysates of cells expressing HA-and FLAG-tagged DKF-2A were analyzed by FPLC gel filtration chromatography on a column of Superose 6 and sucrose density gradient centrifugation. The column and gradients were calibrated with purified proteins having previously established R s and S 20,w values (Fig. 4, A and B). Upon gel filtration, DKF-2A appeared in fractions that correspond to elution peaks of 600-kDa globular proteins (Fig.  4C, bottom panel). However, large globular proteins and nonglobular proteins with elongated shapes but much smaller molecular mass can exhibit similar R s values. Introduction of the S 20,w value in the M r calculation corrects for the contribution of protein shape. The measured R s (7.3 nm) and S 20,w (8.8) parameters (Fig. 4, C and D) yielded an estimated M r for DKF-2A that best accommodates two monomers (Table 1). Thus, the C. elegans PKD is a dimer. Using the same approach, R s , S 20,w , and M r values were determined for endogenous PKD1 and PKD2 (Fig. 4, C and D, and Table 1). The hydrodynamic behavior of human PKDs 1 and 2 indicates that these kinases are also dimers. The M r and subunit composition of two internal control proteins, endogenous PKC␦ (monomer) and GAPDH (tetramer), were correctly determined (Fig. 4, A-D, and Table 1), thereby validating the methodology. Based on the preceding results and S max /S 20,w ratios that exceed 1.5 (see "Experimental Procedures" and Table 1), we conclude that DKF-2A and human PKDs 1 and 2 are elongated dimers.
The possibility that repeated transient interactions with scaffolds or modulatory proteins are required to assemble and sustain dimeric PKDs was rigorously evaluated. We constructed recombinant baculovirus that contains cDNA encoding GST-His 6 -DKF-2A under control of the powerful polyhedron promoter. The fusion protein was expressed in infected Sf9 insect cells and initially purified by metal ion affinity chromatography on Ni 2ϩ -NTA-agarose beads (Fig. 5A). Additional purification of doubly tagged DKF-2A was achieved by performing affinity chromatography on GSH-Sepharose 4B. Next, GST and His 6 tags were removed by digestion with thrombin, which cleaves at a unique site that precedes the N terminus of DKF-2A. Finally, the D kinase was further purified and characterized by Superose 6 gel filtration chromatography. Both silver staining (Fig. 5B) and Western immunoblot analysis (Fig. 5C) showed that the   Table 1. Native molecular weights of calibrating proteins (red arrows) are shown to illustrate that PKDs appear to be large oligomers in the absence of information about S 20,W values. Only relevant fractions are shown; V o was collected in fractions 1-8. T indicates the signal obtained from a sample of total protein extract. D, a sample of lysate described in C was fractionated in a sucrose gradient. Fractions were collected and analyzed as described under "Experimental Procedures." Protein peaks are marked with yellow dots. Peaks of calibrating proteins are indicated with red arrows. Experimentally determined S 20,W values are given in Table 1.
Stokes radius of highly purified DKF-2A was similar to that determined for DKF-2A in cell extracts ( Fig. 4 and Table 1). Furthermore, examination of the complete silver-stained SDS-PAGE gel (Fig. 5B) revealed that purified DKF-2A eluted from the FPLC gel filtration column appears to be essentially a single entity. No associated proteins are evident. His 6 -PKD2 was also expressed in Sf9 cells, affinity-purified, and characterized on the Superose 6 column. The R s value determined from the elution peak (Fig. 5D) was similar to that obtained for PKD2 in cell extracts (Table 1). Peak fractions from the Superose 6 column contained only PKD2 polypeptides or minor amounts of large, proteolytically nicked PKD2 fragments. No interacting proteins co-purified with the kinase. Neither DKF-2A nor PKD2 monomers were detected during the analysis. Overall, the results indicate that dimerization is an independent, intrinsic property of nematode and mammalian PKD proteins.
Dimerization Is Required for PKD-mediated Protein Secretion-PKDs regulate the biogenesis and fission of TGN-derived vesicles known as CARTS (36). CARTS selectively incorporate and transport relatively small, secreted proteins to plasma membrane, thereby facilitating constitutive and/or regulated exocytosis. PAUF (pancreatic adenocarcinoma up-regulated factor) is a representative granin family glycoprotein that is incorporated into CARTS at the TGN and then directed to the cell surface (36,37). Secreted PAUF is a multifunctional regulatory protein that stimulates cell growth and motility, increases pancreatic tumor cell invasiveness, triggers cytokine production in immune cells, and

and composition of PKDs and control proteins
The values for R s , S 20,w and K av were determined experimentally as described under "Experimental Procedures" and in the legend for Fig. 4. The M r values were obtained from the Svedberg equation (see Refs. 34 and 35). The number of polypeptides in the native complexes was determined by comparing observed native M r with molecular weights of corresponding monomeric proteins obtained from genomic and cDNA databases. S max was calculated according to Erickson (34). S max /S 20,w Ն 1.5 indicates elongated shape. Determination of the previously established M r and composition of monomeric PKC␦ and tetrameric GAPDH verified the methodology.  potently elicits angiogenesis by activating signal transduction in endothelial cells (37)(38)(39).
The principal intracellular pool of PAUF consists of two partially processed polypeptides with apparent M r values of 22 and 24 kDa. Mature, secreted PAUF has an apparent M r of ϳ28 kDa, but it is visualized as a diffuse band (on Western blots) because of variations in length and composition of its N-linked oligosaccharide and other modifications. Because CARTS biogenesis and PAUF secretion are PKD-dependent (36), this secretion system provides a good model for analysis of the physiological relevance of D kinase dimerization.
PAUF precursors are absent or low in various lanes of the Fig.  6 (A, D, and E), where PKD activity is impaired. Observation of limited PAUF precursor accumulation in these instances is consistent with reports from the laboratories of Seufferlein (11), Ricci (12), and Evers (41). They demonstrated that PKDs control not only signal-dependent secretion of PAUF, chromogranin A (a PAUF-related glycoprotein), insulin, and neurotensin but also biogenesis of specialized secretory vesicles that transport these secreted proteins from the TGN to the cell surface. When PKD-mediated vesicle biogenesis was blocked, both secretion and intracellular levels of secretory proteins (and their precursors) declined coordinately and precipitously. Secretory protein synthesis was not inhibited. Instead, secretory protein precursors that were not properly packaged and processed in the absence of vesicles were rapidly degraded. This current model suggests the following scenario. When a strong dimerization disruptor or PKD KD mutant is expressed in cells (Fig. 6, A-E), PAUF precursors are synthesized, but few CARTS vesicles are properly generated. Consequently, PAUF precursors that are not packaged in CARTS are degraded, whereas secretion is simultaneously inhibited.
To directly determine whether the results obtained from transient transfection experiments (Fig. 6, A-F) were affected by variations in PAUF precursor levels or stability, we generated and studied cloned HEK293 cells that contain a stably integrated, abundantly expressed PAUF transgene. Fig. 6G shows PAUF precursor levels were elevated and relatively constant under all conditions tested. PAUF was not released into the medium under basal conditions. However, stimulation of endogenous PKDs with PMA elicited robust PAUF secretion. Again, expression of the PKD2 OD or dominant negative PKD1 KD sharply suppressed PMA-stimulated secretion (Fig. 6G). Thus, dimerization of endogenous PKDs is required for signaldependent, D kinase-mediated regulation of PAUF exocytosis. Observations on PAUF secretion presented in Fig. 6G are in excellent agreement with results obtained from transient transfections depicted in Fig. 6 (A-F).
Overall, the data indicate that native PKD dimers are essential for regulation of CARTS-mediated secretion. Thus, the newly described PKD dimerization domain is indispensable for proper control of a key physiological process. Dimerization is also essential for potent inhibitory effects exerted by catalytically inactive PKDs.
Impaired Dimerization Impedes PKD-catalyzed Phosphorylation of a Substrate Effector-Because PAUF secretion is an indirect assay, we also analyzed the role of dimerization in phosphorylation of a bona fide PKD substrate effector, phosphatidylinositol 4-kinase III␤ (PI4KIII␤) (13). The cells were cotransfected with plasmids encoding GFP-PI4KIII␤ and either FLAG-PKD2 or FLAG-PKD2 mutants containing defective dimerization domains. GFP-PI4KIII␤ was isolated from cell extracts by immunoprecipitation with IgGs directed against GFP. Phosphorylation of PI4KIII␤ was monitored by using anti-phospho-(Ser/Thr) PKD substrate antibody as described under "Experimental Procedures" and Ref. 13. PMA promoted phosphorylation of GFP-PI4KIII␤ in cells expressing wild type PKD2 (Fig. 6H). In contrast, little or no phosphorylation of GFP-PI4KIII␤ was observed in cells expressing dimerization defective PKD2 mutants (PKD2 ⌬66 or ⌬138). These data demonstrate that dimerization plays a central role in coupling upstream signals to PKD-mediated phosphorylation of a representative target effector protein. To assess the relevance of C. elegans DKF-2A dimerization to substrate phosphorylation, we exploited our observation that the nematode D kinase robustly phosphorylates PI4KIII␤ in transfected HEK293 cells. Consequently, the cells were cotransfected with transgenes encoding GFP-PI4KIII␤ and either wild type HA-DKF-2A or constitutively active HA-DKF-2A EE. PMA elicited strong phosphorylation of GFP-PI4KIII␤ by WT DKF-2A (Fig. 6I). Disruption of DKF-2A dimerization via overexpression of DKF-2A 1-319-mCherry sharply reduced PI4KIII␤ phosphorylation. As expected, HA-DKF-2A EE efficiently phosphorylated PI4KIII␤ in the absence of stimulus (Fig. 6J). However, constitutive phosphorylation was also strongly inhibited when dimerization of DKF-2A EE was impaired. These observations show that dimerization of the nematode PKD, like mammalian PKDs, is essential for optimal phosphorylation of a target effector protein in the context of intact cells.
Disruption of PKD Dimerization Compromises C. elegans Innate Immunity-C. elegans feeds on microbes and normally lives ϳ3 weeks on a diet of Escherichia coli OP50, a rich source of nutrients. When C. elegans persistently ingests Pseudomonas aeruginosa 14 (PA14), this pathogenic bacterium colonizes the intestine and causes toxicity, which kills the nematodes in 4 -5 days. If PA14 and OP50 are simultaneously available, C. elegans will produce a complex array of immune effectors that cooperate with pathogen avoidance and foraging behaviors to protect animals against transient infection, restore health, and reinitiate reproduction. The C. elegans innate immune system, which is located principally in intestinal epithelial cells, defends against PA14 by synthesizing and secreting antimicrobial proteins and lectins and increasing the concentration of detoxifying enzymes and proteins that sustain the intestinal lining.
PA14 infection triggers DKF-2A activation in intestine (23). DKF-2A (in concert with p38 MAP kinase) suppresses PA14 toxicity by promoting induction of ϳ85 mRNAs encoding protective, anti-microbial proteins. C. elegans lacking DKF-2A (dkf-2 null) is hypersensitive to killing by PA14. Hypersensitivity is rescued by targeting expression of a DKF-2A-GFP transgene to intestinal cells in dkf-2 null animals (23). Moreover, resistance to PA14 increased markedly (super-resistance) in both WT and dkf-2 null animals when a DKF-2A transgene was modestly overexpressed in intestine. Thus, we investigated the physiological significance of PKD dimerization in vivo by assaying PA14 sensitivity of transgenic animals in which (a) dimerization of endogenous DKF-2A is disrupted or (b) a DKF-2A mutant lacking a dimerization domain is expressed in intestine of dkf-2 null animals.
Synchronized animals are grown on OP50 and then transferred to plates seeded with PA14. Because PA14 is the sole food source, all animals eventually succumb to infection. However, the rate of PA14-mediated killing will be fast in populations of hypersensitive animals with a compromised immune system and relatively slow in animals with WT or enhanced immune systems. Thus, consequences of impaired PKD dimerization on a critical function, immune defense against a life-threatening pathogen, can be elucidated by determining PA14 killing kinetics (survival curves).

Discussion
Current knowledge of the occurrence, molecular basis, and biological significance of PKD multimers is extremely limited and somewhat contradictory. We addressed these issues by determining the oligomerization status of PKDs and elucidating relationships between PKD dimerization and physiological functions. We discovered and characterized a novel domain that mediates dimerization of PKDs. Dimerization is essential for PKD-mediated secretion of PAUF, PKD-dependent activation of the C. elegans innate immune system and potent, dominant negative effects of catalytically inactive PKD mutants.
Virtually nothing was previously known about assembly and functions of homo-oligomeric PKDs. Consequently, initial studies focused on DKF-2A, which regulates innate immunity in the absence of other PKD isoforms. Co-immunoprecipitation of DKF-2A polypeptides labeled with different epitope tags revealed robust homo-oligomerization. Chemical cross-linking showed that DKF-2A monomers were closely apposed within intact cells. The highly efficient conversion of DKF-2A monomers to oligomers by cross-linkers contrasted starkly with results obtained for PKCs ␣ and ␦. Like PKDs, PKCs have N-terminal DAG binding sites and a C-terminal kinase domain, but they remained monomeric at all cross-linker concentrations. The results also indicate that PKD oligomers detected in cell extracts were not generated artifactually after detergent-based lysis, e.g. by liberating segregated PKD monomers from binding sites on different organelles.
Characterization of DKF-2A in cell extracts by gel filtration and sucrose density gradient sedimentation yielded an estimated M r value corresponding to a homodimer. Stokes radii of highly purified DKF-2A and DKF-2A in extracts are similar, indicating that stable, high affinity binding between two DKF-2A monomers is an intrinsic property of the PKD. Overall, DKF-2A is preponderantly (if not exclusively) dimeric.
Deletion mutagenesis disclosed that a domain composed of 92 amino acids (residues 228 -319) that precedes the C1a DAG binding site mediates DKF-2A dimerization. A DKF-2A 1-319-GFP fusion protein accumulated in transfected cells and avidly bound DKF-2A and DKF-2A mutants that retain amino acids 228 -319. Robust formation of these complexes shows that the newly discovered domain folds and functions independently of conserved, regulatory, and catalytic regions of PKDs.
Successful analysis of dimerization of the prototypical C. elegans PKD provided guidance and an experimental framework for investigations on oligomerization of mammalian PKDs. Co-immunoprecipitation analysis, chemical cross-linking, mutagenesis, and hydrodynamic studies revealed that PKDs 1-3 form stable heterodimers and homodimers. This may diversify PKD functions. For instance, cells expressing all PKD isoforms can generate three homodimers and three heterodimers. The six distinct PKDs could potentially interact differentially with upstream PKCs or downstream substrates, exhibit qualitatively or quantitatively different patterns of intracellular localization, have different susceptibilities to degradation, etc. A highly conserved but previously unappreciated domain (59% similar to amino acids 228 -319 in DKF-2A) directs incorporation of monomeric PKD1, PKD2, and PKD3 polypeptides into stable homo-or heterodimeric complexes. The 92-residue domain precedes the C1a module in PKD polypeptides, indicating that dimerization depends on protein-protein interactions occurring near the N termini of D kinases. FLAG or mCherry fusion proteins that contain the PKD dimerization domain alone avidly ligated PKDs 1-3 or selfassociated. The data suggest that the dimerization domain is both necessary and sufficient to establish and sustain the quaternary structure of PKDs.
Homodimeric and heterodimeric complexes were generated by various combinations of WT, PMA-activated WT (data not shown), and constitutively active PKD EE isoforms. The results suggest that the dimerization domain acts independently and is not influenced by configuration of the A-loop, DAG/PMA binding, or catalytic activity. In contrast, constitutively active PKD1 EE ⌬50 -125, which lacks a dimerization region, was neither susceptible to chemical cross-linking nor co-immunoprecipitation with WT PKDs or full-length PKD1 EE. Dimerization was also essential for physiological effects of PKDs (see below).
Cross-linking experiments and determinations of Stokes radius, S 20 , w and native molecular weight values revealed that endogenous PKD1 and PKD2, like recombinant D kinases 1-3 and C. elegans DKF-2A, are exclusively dimers. Previous studies by the Malhotra group (28) and Storz group (29) documented the formation of endogenous PKD2-PKD3 complexes. In addition, disruption of endogenous PKD dimerization, caused by expressing a PKD2 1-146-mCherry transgene in HEK 293 cells, strongly suppressed PMA-induced PAUF secretion. Collectively, these observations show that the newly characterized, conserved dimerization domain mediates stable formation of endogenous PKD homodimers and heterodimers. Dimerization of endogenous PKDs is essential for D kinasemediated regulation of a key cell function, secretion from the TGN.
Our studies provide a molecular basis for the generation of PKD2-PKD3 complexes that affect secretion and motility in HeLa cells (28,29): high affinity binding between conserved dimerization domains that are embedded in each isoform. PKD-PKD association appears to be limited to dimer formation; no monomers were detected, and small amounts of high molecular weight PKD observed by gel filtration appeared to be nonspecific aggregates. No evidence was obtained for trimers, tetramers, or higher order PKD complexes. A report indicating that nearly all tissues in PKD2-deficient or PKD3-defective mice develop and function normally (6) argues against an essential requirement for PKD2-PKD3 heterodimers in broadly regulating TGN fission or F-actin-based motility in vivo. Only PKD2 was required for constitutive and regulated production of TGN-derived vesicles that mediate chromogranin A secretion from BON endocrine cells (11). PKD1 or PKD3, presumably acting as monomers or homo-oligomers, can independently regulate phosphorylation and nucleus to cytoplasm translocation of HDAC7 in B cells (22). The preceding observations raise a central question: is dimerization critical for PKD-mediated regulation of key physiological processes?
Disruption of endogenous DKF-2A dimerization in C. elegans intestinal cells markedly impaired immune responses that defend against an ingested pathogenic bacterium PA14. Further, targeted intestinal expression of a dimerization-deficient DKF-2A variant, which has normal DAG binding and kinase domains, failed to rescue PA14 hypersensitivity of dkf-2 null C. elegans. In contrast, modest overexpression of WT DKF-2A dimers in intestine confers super-resistance to PA14 in dkf-2 null or WT animals (23) (Fig. 7). Thus, homodimerization is essential for a critical PKD-regulated function, activation of the immune system in intact animals.
In transfected cells, PKD1 EE promoted PAUF secretion, which is mediated by TGN-derived vesicles designated CARTS (36). PAUF secretion was abrogated when the PKD1 EE dimerization domain was deleted. PMA/PKC-induced activation of endogenous PKDs in HEK293 cells triggered robust PAUF secretion. However, abundant expression of a PKD dimerization domain fusion protein (PKD2 1-146-mCherry), which forms stable complexes with full-length PKDs 1-3, potently inhibited PMA-stimulated PAUF secretion. The phenotype caused by competitively interfering with dimerization of endogenous PKDs resembled suppression of PAUF secretion elicited by a dominant negative PKD mutant. Classically, a single dominant negative PKD isoform, which contains a mutation that sharply diminishes ATP binding at the catalytic site, is sufficient to suppress actions of all intracellular PKDs. The ability of dominant negative PKD1 to suppress PAUF exocytosis was neutralized when dimerization was impaired by a deletion mutation (PKD1 KD ⌬50 -125). Loss of dimerization capacity may also explain why some N-terminal mutations compromise intracellular targeting of PKDs (33). Overall, dimerization is critical for both dominant negative PKD-mediated ablation and WT PKD-mediated up-regulation of an important physiological process, secretion.
The discovery of the dimerization domain and characterization of the dimeric structure of PKDs provide platforms for further analysis. It is of interest to determine whether homoand heterodimers are assembled according to intrinsic binding affinities and PKD isoform concentrations or through regulated processes controlled by signal transduction. Signal-induced activation of PKDs could potentially elicit isoform-specific changes in proportions of homodimers versus various heterodimers, whereas total PKD dimer concentration remains constant. Future investigations on this topic may yield new insights into mechanisms of D kinase regulation. Site-directed mutagenesis and structural elucidation of the dimerization domain should reveal individual amino acids and features of secondary and tertiary structure that govern formation of complexes between monomeric PKD polypeptides. The possibility that the dimerization domain affects intracellular localization of PKDs or binding with regulatory proteins merits exploration. Biochemical, moleculars, and genetic approaches could be used to separate and characterize various PKD dimers, thereby determining their redundant and unique properties and providing clues regarding which isoforms regulate specific physiological processes. The dimerization domain appears to occur only in PKDs. Thus, development of peptido-mimetic compounds that selectively disrupt dimerization or co-associate activating or inhibitory modulators with PKD monomers might be useful for pharmacological intervention in certain diseases, such as tumors in which the epithelial to mesenchymal transition, proliferation, and/or metastasis are driven by altered activities of PKDs (42)(43)(44)(45)(46).

Experimental Procedures
Cell Culture-Human HEK293 cells were grown as previously described (47). Cells (6-cm plates) were transfected with recombinant plasmid DNA (3 g) complexed with polyethylenimine (12 g) (48). Cloned cell lines stably expressing transgenes were selected with 1 mg/ml G418 (49). Effects of WT and mutant PKDs on secretory vesicle biogenesis and export from the TGN were assayed by monitoring both intracellular accumulation and secretion of PAUF, a granin family glycoprotein (36,37). Growth medium was removed 40 -48 h post-transfection, and cells were incubated in Opti-MEM I (serum-free) medium (Life Technologies). After 5 h, the medium was collected, and cells were lysed. Aliquots of medium and detergentsoluble cell proteins were assayed for PAUF content by Western immunoblot analysis.
Transgene Construction-cDNAs encoding WT or mutant PKDs 1-3, DKF-2A, DKF-2B, and PAUF were amplified and tagged with restriction sites by the polymerase chain reaction. Product DNA was cloned into a modified pCDNA3.1 expression plasmid (Invitrogen), which appends an N-terminal HA or FLAG tag to the protein. C-terminal Myc was added to PAUF. To express fusion proteins tagged with C-terminal GFP, cDNAs were cloned into the pEGFP-N1 vector (Clontech). In some constructs, GFP was replaced by mCherry. Site-directed and internal deletion mutagenesis were performed with the Q5 mutagenesis kit (New England BioLabs) according to the manufacturer's protocols. Full-length PI4KIII␤ cDNA was obtained from the DNASU plasmid repository at Arizona State University. The cDNA insert was provided in the pLP-EGFP-C1 vector, which appends a GFP tag at the N terminus of PI4KIII␤.
Cell Lysis and Western Immunoblot Analysis-Cells (6-cm plate) were lysed on ice in 0.3 ml of 25 mM Tris-HCl, pH 7.4, containing 0.15 M NaCl, 0.2 mM EGTA, 1 mM dithiothreitol, 0.8% Triton X-100, 2% glycerol, and cocktails of protease (Roche) and phosphatase (Sigma) inhibitors. Lysates were sonicated for 5 s and clarified by centrifugation at 44,000 ϫ g for 15 min at 4°C. Proteins were size-fractionated by denaturing electrophoresis and transferred to a polyvinylidene difluoride membrane (50,51). The blots were probed with primary IgGs immunoblotting. Subsequent purification and characterization of PKDs were achieved by gel filtration on Superose 6 (as described above) after elution buffer was exchanged for GF buffer on a spin desalting column (Thermo-Zeba). GST and His 6 tags were removed from DKF-2A by incubating the fusion protein with thrombin (Millipore).
Pathogen-mediated Killing-P. aeruginosa (PA14) was seeded on agar plates containing C. elegans growth medium. A group of 120 synchronized L4 animals was transferred from normal food (E. coli OP50) to PA14 pathogen plates. The plates were incubated at 25°C, and the number of living worms was determined at 8-h intervals. Immobile worms unresponsive to touch were scored as dead.
Author Contributions-C. S. R. and C. A. R. conceived and designed the studies, analyzed and interpreted results, and wrote the paper.