We used a generally applicable strategy to collect and structure the protein interactions of the yeast type II protein phosphatase Ptc1p and its binding partner Nbp2p. The procedure transformed primary unstructured protein interaction data into an ensemble of alternative interaction states. Certain combinations of proteins are allowed in different network configurations. Nbp2p serves as the network hub and brings seven kinases in close contact to Ptc1p. As a consequence, the deletion of NBP2 affects several cellular processes including organelle inheritance and the responses to mating hormone, cell wall stress and high osmolarity; it also impairs the proper execution of the morphogenetic program. Our constraint interaction map provides a basis for understanding a subset of the observed phenotypes and assigns the Ptc1p–Nbp2p module a role in synchronizing the associated kinases during the cell cycle.
Inactivation of proteins that participate in more than one cellular process leads to a variety of apparently unconnected phenotypes. Understanding the molecular cause for each phenotype might reveal how seemingly independent cellular processes are regulated and coordinated in the cell. Genome-wide gene interaction data based on the simultaneous inactivation of more than one gene greatly facilitate this inherently complex analysis because genes with pleiotropic phenotypes often occupy central positions in the corresponding interaction networks (Costanzo et al., 2010; Tong et al., 2004). By assigning physical connections, protein–protein interaction maps provide the necessary complementary information. Interpretation of these maps is usually not straightforward. Genetic interactions can result from complex functional relationships between the investigated pairs of genes and protein interaction maps are generally projections of contacts that occur at different times and places in the cell. To transform protein interaction data into mechanistically meaningful models, it is necessary to resolve these projections into their different interaction planes. We define an interaction plane or state as the sum of all simultaneously occurring contacts. Ideally, these states should be defined by time- and space-resolved in vivo studies. However, these studies are technically demanding and usually not suited for measuring multiple contacts (Maeder et al., 2007). Using the protein pair Ptc1p–Nbp2p of the yeast Saccharomyces cerevisiae as an example and the split-ubiquitin method (Split-Ub) as the experimental tool, we present an alternative approach for defining interaction states. The derived constraint interaction network reduces the number of possible states and thus provides a useful framework for model building and the initiation of more detailed studies.
The yeast type II protein phosphatase Ptc1p and its binding partner Nbp2p both belong to the class of pleiotropic proteins. Loss of either gene leads to a variety of phenotypes including osmo-sensitivity, decreased cell wall integrity, temperature sensitivity, the delay of organelle inheritance, and an increased sensitivity towards rapamycin and other stresses or chemicals (Du et al., 2006; Jin et al., 2009; Lesage et al., 2004; Mapes and Ota, 2004; Ohkuni et al., 2003; Roeder et al., 1998). This multi-functionality correlates with several genetic interactions that places both genes as hubs into similar but non-identical genetic interaction networks (Fiedler et al., 2009; Lesage et al., 2004; Lin et al., 2008; Pan et al., 2006; Tong et al., 2004).
The contribution of Nbp2p–Ptc1p is only well understood for the cells' response to high extracellular osmolarity. This condition will induce a signaling pathway that includes a mitogen-activated protein kinase (MAPK) module consisting of Ste11p (MAPKKK) or alternatively Ssk2p/Ssk22p, Pbs2p (MAPKK) and Hog1p (MAPK) (Hohmann, 2009). Nbp2p recruits Ptc1p to Pbs2p, thus bringing the phosphatase into close proximity to its substrate Hog1p. Dephosphorylation of Hog1p terminates its activity, whereas deletion of PTC1 or NBP2 increases the basal levels of phosphorylated Hog1p and diminishes the cells' ability to rapidly adapt to high osmolarity (Mapes and Ota, 2004). Although Hog1p can be expected to phosphorylate many target proteins, most of the phenotypic characteristics of Δnbp2 or Δptc1 yeast strains cannot be accounted for by the increased basal Hog1p activity alone. For example, the Nbp2p–Ptc1p-dependent delivery of the vacuole and ER to daughter cells depends on the kinases Cla4p and Slt2p, respectively, and the increased sensitivity to cell wall stress was genetically traced to the MAPKK Mkk1p/2p (Bartholomew and Hardy, 2009; Du et al., 2006; Ohkuni et al., 2003).
Yeast possesses the best-investigated protein interactome (Breitkreutz et al., 2010; Gavin et al., 2002; Krogan et al., 2006; Tarassov et al., 2008; Uetz et al., 2000; Yu et al., 2008). However, the richness of phenotypic observations upon deletion of either NBP2 or PTC1 and the wide spectrum of genetic interactions are not matched by a comparably high number of known in vivo binding partners for both proteins. We initiated our study on the multiple roles of Nbp2p–Ptc1p by searching for new binding partners of Ptc1p and Nbp2p. To this end, we used a mating-based interaction assay that allows us to simultaneously test the in vivo interactions of a given protein with a set of 380 selected yeast proteins.
The split-ubiquitin method detects protein–protein interactions in vivo (Johnsson and Varshavsky, 1994). It is based on the ability of the artificially separated N-terminal (Nub) and C-terminal (Cub) halves of ubiquitin to refold into a native-like ubiquitin (Ub). Once coupled to a pair of interacting proteins X and Y, the effective local concentration of Nub and Cub is dramatically increased and their reassociation to Ub is enforced. The concomitant increase in reporter cleavage serves as a measure for the X–Y interaction (Müller and Johnsson, 2008; Paumi et al., 2007; Wittke et al., 1999).
To create a Nub fusion array, a cassette consisting of the PCUP1-Nub-HA was placed upstream of 380 yeast open reading frames (ORFs) by homologous recombination. This ORF collection is enriched in genes that participate in the establishment of cell polarity, including genes involved in stress response, cytokinesis, cell wall integrity, cell cycle control, TOR signaling, exit of mitosis and polarized secretion (supplementary material Table S1). To measure interactions between this array of Nub fusions and any Cub fusion protein, the Cub-RUra3p module (CRU) was integrated via recombination downstream of the ORF of the corresponding gene of interest (Wittke et al., 1999). Coexpression of both fusions in one cell could be achieved in diploids by mating yeast strains expressing the Nub and the Cub fusions of opposite mating types. In diploids where the Nub and Cub fusion proteins interact, the Ura3p reporter is released from Cub and concomitantly degraded. Because fluoroorotic acid (5-FOA) is converted by Ura3p into toxic 5-fluoro-uracil derivatives, cells bearing interacting fusion proteins can be distinguished from those bearing non-interacting fusion proteins by their ability to grow on medium containing 5-FOA (Wittke et al., 1999).
We first screened for interaction partners of Ptc1p and Nbp2p by introducing the respective CRU fusion into the Nub array independently four times using a robot-based mating strategy. Interactions between Nub and Cub fusion proteins were scored when the yeast quadruple grew on medium containing 5-FOA (Wittke et al., 1999). After subtracting promiscuously interacting Nub fusions, the kinases Gin4p, Kcc4p, Pbs2p and Bck1p were identified as binding partners of both Ptc1p and Nbp2p (Fig. 1A,B; Table 1). Nbp2p was detected as additional ligand of Ptc1CRU, whereas Ptc1p and Bem4p were identified as specific interaction partners of Nbp2CRU. Both screens independently confirm the interaction between Nbp2p and Ptc1p (Mapes and Ota, 2004).
The prevalence of protein kinases among the binding partners of Nbp2p and Ptc1p is a second striking feature of the analysis. To further explore this observation, we tested a set of 31 protein kinases as CRU fusions against the Nub array. All kinases are either known or suspected to participate in the establishment of cell polarity or in the response to cell stresses. We also included CRU fusions of Nap1p, Bem4p and Msg5p in our analysis. Nap1p was found by two-hybrid analysis to bind to Nbp2p, Bem4p was discovered by our Nub array as a binding partner of Nbp2p, and Msg5p is a phosphatase also involved in the cellular stress response and hormone adaptation, yet with a different and much smaller set of genetic interactions than Ptc1p (Doi et al., 1994; Ohkuni et al., 2003). These Split-Ub interactions are summarized in Table 1. Using the concurrent interaction with Nbp2p and Ptc1p as stringent criteria, the proteins Bck1p, Bem4p, Gin4p, Kcc4p, Nap1p, Pbs2p, Cla4p, Ste20p and Hsl1p were selected as specific binding partners of the Nbp2p–Ptc1p module from the tested set. The analysis thus expands the list of Nbp2p–Ptc1p ligands obtained in the first round of screening by the kinases Hsl1p, Ste20p, Cla4p and the proteins Nap1p and Bem4p. In the following sections we describe a generally applicable strategy to elucidate which of the Nbp2p–Ptc1p interactions can occur simultaneously, which exclude each other, and which depend on each other.
Dissection of Nbp2p binding sites
Nbp2p can be divided into a central SH3 domain (residues 113–170) and two stretches of unknown structure (residues 1–112 and 171–236; Fig. 1C). We constructed two C-terminal truncations of Nbp2p that are either missing the C-terminal region (Nbp21–173p) or half of the SH3 domain as well as the C-terminal region (Nbp21–140p). Both truncations were tested as CRU fusions against the interaction partners of Nbp2p. Although the measured interaction between Nub-Nap1p and Nbp2 CRU was initially classified as a false positive, we also included Nub-Nap1p in this assay, because its interaction with Nbp2p was confirmed by the reverse Split-Ub configuration and two-hybrid analysis (Table 1; supplementary material Table S2) (Ohkuni et al., 2003). Deletion of the C-terminal domain of Nbp2p abolished the interactions with Nap1p, Gin4p and Kcc4p. Further truncation of part of the SH3 domain removed the interactions with Bem4p, Bck1p and Pbs2p, leaving Ptc1p as sole interaction partner of the N-terminal Nbp2p domain (Fig. 1D) (Mapes and Ota, 2004). The outcome specifies three seemingly independent regions within Nbp2p, each binding a different set of proteins.
The binding partners of the SH3 domain
Bck1p is the MAPKKK of the cell wall integrity pathway (CWIP) and activates the MAPKKs Mkk1p and Mkk2p (Mkk1p/2p) by phosphorylation, which in turn stimulate the MAPK Slt2p (Lee and Levin, 1992; Levin, 2005). Bck1p interacts with Slt2p, Mkk1p, Nbp2p–Ptc1p and also weakly with Spa2p (Table 1, Fig. 2A). These multiple interactions suggest a scaffold function of Bck1p for the MAP kinase module of the CWIP. The measured interaction with Spa2p suggests that Bck1p also localizes to sites of polar growth (van Drogen and Peter, 2002). Indeed, the natively expressed Bck1-GFP can be detected at the tips of small buds (Fig. 2B). However, the most prominent staining is seen in tubular structures that form nets throughout the cell. The colocalization with a mitochondrial marker protein confirmed that these Bck1p-positive structures are mitochondria, a localization already observed for Bck1p in presence of toxic farnesol concentrations (Fig. 2B) (Fairn et al., 2007).
Bck1p can be structurally separated into an N-terminal region of 1174 residues (regulatory domain, RD) and the C-terminal kinase domain (KD) (Fig. 2A). We compared a subset of the protein interactions of full-length Bck1p with the interactions of a bck1 allele missing the kinase domain (BCK1ΔK). The interaction strengths of Spa2p, Mkk1p and Slt2p with Bck1CRU were drastically reduced upon removing KD, whereas binding to Nbp2p and Ptc1p was unaffected by this deletion (Fig. 2A). The interaction between Bck1p, Bck1ΔKp and Nbp2p was confirmed by precipitating C-terminally Myc-tagged versions of Bck1p and Bck1ΔKp from yeast extracts with a GST fusion of Nbp2p expressed in E. coli. This binding was specific, because GST alone did not precipitate Bck1–Myc or Bck1ΔK–Myc (supplementary material Fig. S1).
Interestingly, similarly to RD, the isolated KD of Bck1p did not bind to Mkk1p, Slt2p or Spa2p either (our unpublished observation). As these findings suggest cooperation between the two domains, we performed an intramolecular Split-Ub assay (Raquet et al., 2001). The assay revealed that the N- and C-termini are in closer proximity to each other in the full-length Bck1p than in Bck1ΔKp, suggesting that the KD of Bck1p folds back onto its RD and that both regions of the molecule collaborate for achieving the strong binding to Mkk1p–Slt2p (supplementary material Fig. S1).
As the SH3 domain of Nbp2p (SH3Nbp2) mediates the interaction with Bck1p, we searched the sequence of RD for the presence of a SH3Nbp2 interaction consensus motif as defined by a phage selection procedure for SH3-containing yeast proteins (Fig. 1) (Tong et al., 2002; Tonikian et al., 2009). A very good match was found between the residues 804 and 812 of Bck1p (PKREAPKPP) (Fig. 2A). Accordingly, we measured the interactions of Nub-Ptc1p and Nub-Nbp2 with two C-terminally truncated Bck1p derivatives Bck11–824CRU and Bck11–801CRU, which either contained or lacked this motif, respectively (Fig. 2A). The Split-Ub interaction assay clearly defined the stretch between residues 802 and 824 as the Nbp2p interaction site of Bck1p (Fig. 2C). This conclusion was confirmed by a pull-down assay using GST–Nbp2p and Myc-tagged Bck11–824p and Bck11–801p (Fig. 2D).
The simultaneous loss of binding to Nbp2p and Ptc1p upon deletion of the PKREAPKPP motif in Bck1p, together with our analysis of the Nbp2p truncations, suggests that Nbp2p mediates the interaction between Ptc1p and Bck1p (Fig. 2C). We integrated in a Δnbp2 strain the Nub-module upstream of PTC1 and the Cub-module downstream of the BCK1 ORF. Comparison of the growth on 5-FOA with a corresponding yeast strain harboring the native copy of NBP2 revealed that the interaction between Bck1p and Ptc1p vanishes in the absence of Nbp2p (Fig. 2E). Importantly, the deletion of NBP2 does not measurably change the expression level of Nub-Ptc1p (Fig. 2E).
Cla4p and Ste20p are p21-activated kinases (PAKs) that are stimulated by GTP-bound Cdc42p. Ste20p activates one branch of the osmolarity MAP kinase pathway and alternatively, the MAPK module that responds to extracellular mating hormone. Cla4p is involved in the correct assembly of the septin cytoskeleton during the early phase of bud growth (Keniry and Sprague, 2003). Besides their individual binding partners, the Cub fusions of Cla4p and Ste20p are both shown by the Nub array to interact with Nbp2p, Ptc1p and Bem1p. The binding sites for both PAKs on Nbp2p could not be mapped, because the corresponding Nub fusions did not reveal an interaction with Nbp2CRU (Fig. 1B). However, both kinases harbor short proline-rich stretches that resemble the interaction motif of Bck1p and Pbs2p for SH3Nbp2 (Fig. 3A,F). We created two equivalent C-terminal deletion mutants for both kinases that either contain (Ste201–487p, Cla41–471p) or lack this motif (Ste201–473p, Cla41–448p). Strains harboring Ste20CRU, Cla4CRU or the CRU fusions of their C-terminal truncations were individually mated against strains expressing Nub-Nbp2p and Nub-Bem1p and tested for interactions. The binding of Nbp2p to either of the two PAK kinases depends on the PxRxxPxxP motif because Ste201–473CRU and Cla41–448CRU were unable to bind to Nbp2p (Fig. 3B). This finding points to SH3Nbp2 as the interaction site for both PAKs. Interestingly, our data revealed that Ste20p and Cla4p differ in respect to binding to Bem1p. Confirming a previous study, we show that Nbp2p and Bem1p dock onto the same site in Ste20p because both interactions are simultaneously lost upon removal of the PxRxxPxxP motif (Fig. 3A) (Winters and Pryciak, 2005). However, Cla41–471p displayed a clear and specific interaction with Bem1p that remained even in the absence of the PxRxxPxxP motif. This indicates a separate, more N-terminally located Bem1p binding site. To decipher whether Bem1p binds the Cla4p PxRxxPxxP motif in addition to this N-terminal site, we tested a short Cla4p fragment of 101 residues (Cla4370–471p) covering this motif. This isolated polypeptide stretch is sufficient for binding to Nbp2p, but not for binding to Bem1p (Fig. 3B).
Binding of Ptc1p to Ste20p depends on Nbp2p because the interaction between Nub-Ptc1p and Ste20p is lost upon removal of Nbp2p (Fig. 3C). Cla4CRU did not yield sufficient Ura3p activity for performing a Split-Ub assay in a Δnbp2 strain (our unpublished observation). However, the concomitant loss of Nbp2p and Ptc1p binding upon deletion of the PxxP motif in Cla41–448p presents strong evidence that Nbp2p also mediates the binding of Ptc1p to Cla4p (Fig. 3D). In support of the Split-Ub-derived conclusions Myc-tagged versions of Ste20p, Cla4p and Cla41–471p were specifically precipitated from yeast extracts with GST–Nbp2p. Cla41–448p lacking the PxxP motif also failed to bind to Nbp2p in this assay (Fig. 3E).
Alignment of the SH3-interacting motifs of Ste20p, Cla4p, Bck1p and Pbs2p indicates a positively charged residue at position seven within the previously described consensus sequence as an important feature for SH3Nbp2 binding (Fig. 3F). We exchanged in Cla4385–471CRU K459 for S and in separate experiments I462 for S, and P461 for A. Fig. 3F clearly shows that the P461A exchange interferes with the strong binding of the motif to Nbp2p and that the K459S exchange reduces its affinity, whereas the exchange at position 462 has no measurable influence on its binding.
Bem4p is unique among the binding partners of the SH3Nbp2 domain in not being a protein kinase and not displaying any PxxP motif in its sequence. The Bem4CRU fusion revealed that, in addition to Nbp2p and Ptc1p, the two Rho-type GTPases Cdc42p and Rho1p area also binding partners (Fig. 4A). The latter two proteins have been previously identified as interaction partners by a systematic two-hybrid analysis (Drees et al., 2001). The Nbp2p–Bem4p interaction could be verified by a GST–Nbp2p pull-down (Fig. 4C). Our observation that Bem4CRU did not bind to Nub-Ptc1p in a Δnbp2 deletion strain confirms Nbp2p as the mediator of this interaction (Fig. 4B).
The binding partners of the C-terminal Nbp2p domain
The septin-associated kinases Gin4p, Hsl1p and Kcc4p all interact with Nbp2p and Ptc1p (Fig. 5). Kcc4p shows in addition strong interactions with the septins Cdc11p and Cdc3p, the cyclin-dependent kinase Pho85p, and a less prominent interaction with Bud6p and Bni5p (Table 1) (Fig. 5B). Bud6p is a member of the polarisome and belongs to the family of actin-organizing proteins, Bni5p is a septin-interacting protein (Amberg et al., 1997; Lee et al., 2002). Hsl1p interacts, among others, with the two septin-associated kinases Gin4p and Kcc4p (Table 1).
Using the corresponding wild-type and Δnbp2 strain, we could demonstrate for all three septin-associated kinases the dependency on Nbp2p for their interaction with Ptc1p (Fig. 5C). The separately expressed N-terminal kinase domain of Gin4p still bound to Nbp2p, thus ruling out the PxxP stretch within its C-terminal sequence as a binding site (Fig. 5A). This is in accordance with our finding that the C-terminal region of Nbp2p and not its SH3 domain mediates the interaction with Gin4p (Fig. 1D).
Nap1p is a histone chaperone that controls aspects of the cell cycle, bud morphogenesis and the septin cytoskeleton (Mortensen et al., 2002; Ohkuni et al., 2003). The Nub array yielded besides the known interaction partners Nbp2p, Gin4p, and Kcc4p, the proteins Hsl1p, Ptc1p and Tco89p as novel binding partners of Nap1p. Tco89p is a member of the TOR complex 1 and thus links the Ptc1p–Nbp2p–Nap1p interaction state to the TOR pathway (Reinke et al., 2004). To gain further insight into this pathway, we tested Tco89CRU against our Nub array and found in addition to Tor1p as the known binding partner, Tor2p, Tco89p, Nbp2p and Ptc1p as novel binding partners of Tco89p (Table 1) (Reinke et al., 2004).
The C-terminal region of Nbp2p contains an acidic stretch of 40 residues at its end. Deletion of this stretch removed binding to Kcc4p, Gin4p and Nap1p (our unpublished observation). This finding indicates either a common interaction site for the independent association of Kcc4p, Gin4p and Nap1p with Nbp2p, or one protein serving as an adaptor for the others. Nap1CRU was shown in this study to bind to Gin4p and Kcc4p (Table 1). In addition, Nap1p was found by coprecipitation with Gin4p and/or Kcc4p in a complex that did not contain Nbp2p (Gavin et al., 2002; Mortensen et al., 2002). We thus tested the idea that Nap1p is the adaptor by measuring the interactions between Nub-Nbp2p and Kcc4CRU, Gin4CRU or Hsl1CRU in the presence or absence of NAP1 (Fig. 5D). All three kinases lost proximity to Nbp2p in the absence of Nap1p, whereas the SH3Nbp2p-domain-interacting kinases Ste20p and Pbs2p retained their affinity for Nbp2 in the Δnap1 strain. Myc-tagged Gin4p and Myc-tagged Pbs2p could be precipitated with bacterially expressed Nbp2p from extracts of cells containing Nap1p. The presence of Pbs2p and the absence of Gin4p in the Nbp2p-bound fraction of extracts from Δnap1 cells supports our conclusions from the Split-Ub analysis and confirms Nap1p as the link between Nbp2p–Ptc1p and the septin-associated kinases Gin4p, Kcc4p and Hsl1p (Fig. 5E and supplementary material Fig. S3).
Correlating single states of the Nbp2p–Ptc1p network with kinase activities and phenotypic observations
Fig. 8 summarizes the dissection of the Nbp2p–Ptc1p protein interaction network. We define Nbp2p as its hub and an interaction state of this network as the simultaneous binding of a subset of its members to Nbp2p. Can we correlate each of the eight Nbp2p–Ptc1p kinase states with a certain activity and phenotype? In a first approach, we compared the effects of overexpressing the kinases of the Nbp2p–Ptc1p network and some of their known downstream kinases in wild-type and Δnbp2 cells. The overexpression of Slt2p, Hog1p, Kcc4p and Ste20p specifically abolished the growth of Δnbp2 cells, indicating that Nbp2p–Ptc1p negatively regulates these pathways (Fig. 6A). Overexpression of Cla4p, Bck1p and Gin4p led to a strong growth reduction in both Δnbp2 and wild-type cells, whereas the overexpression of the kinases unrelated to the Nbp2p–Ptc1p network did not display any significant growth effects (Fig. 6A). To more directly demonstrate that the toxicity of overexpression in Δnbp2 cells was due to the unconstrained activities of the respective kinases, we constructed a cla4 allele where the critical prolines within the Nbp2p binding site at positions 458, and 461 were replaced by alanines (cla4AA). Mild overexpression of cla4AA reduced the growth rate when compared with cells expressing the wild-type allele under identical conditions (Fig. 6B). By contrast, the overexpression of both alleles is equally toxic in Δnbp2 cells, suggesting that their diverse effects in wild-type cells is due to their different affinities to Nbp2p (Fig. 3F, Fig. 6B).
A physical connection between Nbp2p–Ptc1p and the MAPKKK Bck1p explains the previously observed link between Nbp2p–Ptc1p and the CWIP (Ohkuni et al., 2003). In particular, it was shown that the level of phosphorylated Slt2p, the MAPK downstream of Bck1p, is increased in Δnbp2 cells (Du et al., 2006). The increased sensitivity of Δnbp2 cells to the cell wall stressor caspofungin confirms these observations (Fig. 6C).
Our data imply that Nbp2p–Ptc1p negatively regulates the mating response of yeast cells via the their association with Ste20p (Fig. 3B,C; Fig. 6A). We thus tested the sensitivities of wild-type and Δnbp2 cells against the α-factor mating hormone (Fig. 6D). Cultures of Δnbp2 cells showed a significantly wider halo upon treatment with α-factor than was seen in wild-type cells, indicating a higher level of activation when the intimate connection between the phosphatase Ptc1p and Ste20p is dissolved (Fig. 6D).
Overexpression of Cla4p as well as deletion of PTC1 delays the delivery of vacuolar vesicles to the daughter cell (Bartholomew and Hardy, 2009; Jin et al., 2009). Our proof of an Nbp2p-mediated molecular link between Ptc1p and Cla4p provides a satisfying explanation for both observations. Transport of vacuoles, peroxisomes and mitochondria depends on Myo2p (Pruyne et al., 2004). To find out whether the Nbp2p–Ptc1p–Cla4p state controls any of these additional Myo2p-dependent transport processes, we integrated the inducible PMET17 promoter upstream of Cla4p and followed the segregation of fluorescently labeled mitochondria and peroxisomes under inducing conditions (Fig. 6E, supplementary material Fig. S2). Buds of cells overexpressing Cla4p received both organelles on average at later stages of their growth. The effect of the NBP2 deletion on the kinetics of organelle segregation was less pronounced, but especially at 37°C, clearly visible (Fig. 6F).
Cells overexpressing Kcc4p or Gin4p form elongated buds (Okuzaki et al., 2003). As overexpression of Kcc4p is lethal in Δnbp2 cells, we closer inspected the morphology of Δnbp2 cells. A fraction of these cells displayed more elongated buds and smaller bud necks than the wild-type cells (Fig. 7A) (our unpublished observations). Because Gin4p, Hsl1p and Kcc4p control aspects of the septin cytoskeleton, we chose to inspect septin organization in Δnbp2 cells (Barral et al., 1999). The GFP-coupled septin Cdc11p (Cdc11–GFP) highlighted abnormal septin structures in a fraction of those Δnbp2 cells that form elongated buds. We occasionally observed dots of fluorescence in the tip of the elongated buds (Fig. 7Bp), bars of fluorescence shifted from the bud neck closer to the center of mother and daughter cells (Fig. 7Bj,p), and septin bars with a seemingly perpendicular orientation to the division plane (Fig. 7Bl,n).
The fraction of multi-budded cells is increased in Δnbp2 cells. Live-cell imaging of Cdc11–GFP revealed in ~10% of the Δnbp2 cells, small buds emerging from points where transient Cdc11p fluorescence had been visualized, before it reappeared at a different location to mark the formation a new bud. First buds often remained as non-growing ‘ghosts’ during the rest of the cell cycle. This process of aborted bud formation can be observed two or three times before a stable bud is finally formed (Fig. 7C).
Using the increased temperature and caspofungin sensitivity as well as the elongated bud morphology of Δnbp2 cells as criteria, we evaluated the functionality of nbp2 alleles that either lack the C-terminal tail (Nbp21–173p) or a functional SH3 domain, in addition to the C-terminal sequence (Nbp21–140p). Nbp21–173p could complement all three phenotypes. Cells expressing Nbp21–140p displayed the same bud morphology and caspofungin sensitivity as Δnbp2 cells, but still grew at 38°C, albeit less well than wild-type cells (supplementary material Fig. S4).
The Split-Ub assay detects direct protein–protein interactions in their natural cellular environment as well as the assembly of subunits in larger protein complexes, even if the subunits are not in direct physical contact. Both features of the system were prerequisites for its application in this study (Eckert and Johnsson, 2003; Müller and Johnsson, 2008; Wittke et al., 2000).
By selecting the proteins according to their locations and functional connections, we created an array of Nub fusion proteins whose composition is dedicated for investigating processes associated with polarized growth in yeast. We applied the Nub array in a series of experiments that aimed to understand the multiple roles of the phosphatase Ptc1p and its adaptor protein Nbp2p. Both proteins were jointly associated with widely different aspects of yeast cell biology. Yet the mechanisms by which these proteins affect so many divergent pathways were enigmatic.
Using the Nub array, we confirmed the interaction between Nbp2p–Ptc1p and Pbs2p, the MAPKK of the high osmolarity pathway, and were able to discover new physical links between Nbp2p–Ptc1p and the kinases Bck1p, Kcc4p, Gin4p, Hsl1p, Ste20p and Cla4p. Additionally, we identified new interaction partners of yeast protein kinases (Table 1). Further work is certainly needed to evaluate the physiological significance of these interactions, but we suggest that our detection of several heteromeric kinase pairs indicates new regulatory circuits in the yeast (Fig. 8E).
The models in Fig. 8 summarize key features of the obtained constraint interaction network. The adaptor protein Nbp2p mediates all interactions between Ptc1p and the different kinases. Ste20p, Bck1p, Cla4p and Pbs2p compete with similar PxxP motifs for the same site on Nbp2p. As a result of its missing PxxP motif, we propose that Bem4p binds to a different site on SH3Nbp2. Gin4p, Kcc4p and Hsl1p bind indirectly via Nap1p to a region C-terminal to the SH3 domain of Nbp2p. The strong sequence identity between the two kinase domains suggests that Kcc4p and Gin4p compete for binding to Nap1p. We could further show that Hsl1p binds to both Kcc4p and Gin4p, and very weakly to Nap1p. These observations imply that Hsl1p is connected to Nap1p and Nbp2p–Ptc1p via either Kcc4p or Gin4p. Thus each of the SH3Nbp2 binding kinases can be part of a distinct interaction state that includes the polarity protein Bem4p, the phosphatase Ptc1p and the septin-associated kinases Gin4p, Kcc4p and Hsl1p.
Morphogenetic processes are realized by the integration of different cellular activities. Each of the kinases regulated by Nbp2p–Ptc1p participate in at least one of these activities. Pbs2p controls the production of force for cellular growth by adjusting the osmotic imbalance (Hohmann, 2009). Bck1p regulates the enforcement and the weakening of the cell wall at strategic places during growth (Levin, 2005). The PAKs and the septin-associated kinases control the assembly of the septin ring, a structure that is crucial for cell division and growth in yeast (McMurray and Thorner, 2009). Our results, as well as the experiments by others, suggest a crucial role for Ptc1p and Nbp2p in synchronizing these activities. Owing to its interaction-induced proximity, Ptc1p might dephosphorylate either the Nbp2-bound kinases directly or their downstream targets to reduce the activities of the corresponding pathways (Fig. 8A,B) (Du et al., 2006; Jin et al., 2009; Lesage et al., 2004; Mapes and Ota, 2004; Ohkuni et al., 2003; Roeder et al., 1998). Thus, disruption of the link between Ptc1p and Nbp2p would activate all seven kinases at once. Severing only the link between the SH3 domain and its targets would selectively activate the pathways controlled by Cla4p, Ste20p, Bck1p and Pbs2p, whereas the dissociation of Nbp2p from Nap1p would increase the activities of the septin-associated kinases Gin4p, Kcc4p and Hsl1p.
Synchronization of organelle inheritance is one example where the concerted activation of the SH3Nbp2-bound kinases are likely to have a role. Overexpression of Cla4p inhibits the movement of mitochondria, vacuole and peroxisomes, whereas overexpression of Slt2p delays ER inheritance (Fig. 6E) (Bartholomew and Hardy, 2009; Du et al., 2006). Deletion of Ptc1p or Nbp2p leads to similar phenotypes (Fig. 6F) (Du et al., 2006; Jin et al., 2009; Roeder et al., 1998). Interrupting the link between Nbp2p and the PxxP kinases might therefore act as switch to slow down organelle migration to the daughter cell at the end of each cell cycle. Loss of Ptc1p results in the mislocalization of the motor proteins responsible for the movements of organelles as well as the stability of adaptor proteins linking organelles to their respective motors (Bartholomew and Hardy, 2009; Jin et al., 2009). Rather than signalling via organelle-specific Ptc1p substrates, we propose a general influence of Ptc1p on these processes via associations with the kinases Bck1p, Ste20p and Cla4p. Localization of Bck1p on the mitochondria is counterintuitive to its proposed role in the control of ER inheritance (Fig. 2B). However, because the mitochondria and the ER are physically linked, it might reflect a crosstalk between both organelles that guarantees their coordinated and robust inheritance (Kornmann et al., 2009).
Demonstrating the coexistence of all members in the same complex is the missing step in the analysis of some of the postulated interaction states. For example, we were able to prove the Ptc1p–Nbp2p–Nap1p–Gin4p interaction state by showing the interaction between its most distant members, Ptc1p and Gin4p, in addition to the physical links between the consecutive neighbors of this postulated chain. Nap1p activates Gin4p during mitosis (Mortensen et al., 2002). The docking of Nap1p to Nbp2p–Ptc1p might thus reduce the activity of Kcc4p–Gin4p and restrict the action of this complex to this phase of the cell cycle. Unexpectedly, deletion of the C-terminal binding site of Nbp2p for Nap1p has no obvious morphogenetic consequences (supplementary material Fig. S4). This finding might indicate that the interaction state is not entirely understood. Indeed, Tjandra and colleagues discovered an epistatic relationship between CLA4, NAP1 and GIN4 that implied a Nap1p-dependent phosphorylation and activation of Gin4p by Cla4p (Tjandra et al., 1998). Our data support the existence of Cla4p–Nbp2p–Nap1p–Gin4p interaction state (Fig. 8A). Accordingly, interrupting the binding between Nap1p and Nbp2p simultaneously abolishes the positive influence of Cla4p and the negative influence of Ptc1p on the activity of Gin4p–Kcc4p. A Cla4p–Nbp2p–Nap1p–Gin4p interaction state might thus explain the lack of obvious phenotypes of cells missing the contact region of Nbp2p for Nap1p (supplementary material Fig. S4).
The observation that Cdc42p is required for the activation of Cla4p and indirectly of Gin4p extends the Cla4p–Gin4p state by the link between Cla4p and Bem1p (Fig. 8A) (Tjandra et al., 1998). Bem1p recruits the GEF Cdc24p and GTP bound Cdc42p to Cla4p and Ste20p to activate their kinase functions (Bose et al., 2001). Our dissection of the Nbp2p binding sites revealed a subtle difference in the regulation of the two PAKs. Both have a very similar Nbp2p binding motif located N-terminal to the kinase domain in the center of the molecule. We could show that Nbp2p and Bem1p use their SH3 domains to bind the same site of Ste20p, whereas Cla4p provides a distinct and more N-terminally located binding site exclusively for Bem1p (Fig. 3B) (Winters and Pryciak, 2005). The additional Bem1p binding site raises the molecular heterogeneity of Cla4p from the two states Bemp1–Cla4p and Cla4p–Nbp2p to a third state Bem1p–Cla4p–Nbp2p that might be required for overcoming the negative influence of Ptc1p on the activation of Gin4p in the Cla4p–Gin4p interaction state (Fig. 8A,C). A competition between Nbp2p and Bem1p excludes Bem1p from the Ste20p–Gin4p state and thereby eliminates any positive influence of Ste20p on the activity of Gin4p (Fig. 8A). Indeed, it was shown that Ste20p does not activate Gin4p (Tjandra et al., 1998).
Another important aspect of the Ptc1p–Nbp2p–Bck1p–Mkk1p–Slt2p state also demands further clarification: the interaction data clearly show that Ptc1p is linked via Nbp2p to the MAPKKK Bck1p, whereas complementary genetic experiments suggest that Ptc1p dephosphorylates Mkk1p/2p (Ohkuni et al., 2003). In analogy to the osmolarity pathway, we thus postulate that Nbp2p brings Ptc1p into close proximity to its substrate Mkk1p/2p. By deactivating Mkk1p/2p, Slt2p stays hypo-phosphorylated and inactive (Fig. 8D). Although we detected each of the consecutive interactions in this chain, experimental evidence for the postulated presence of Ptc1p and Mkk1p in one complex is still missing.
Materials and Methods
Construction of the Nub array, Cub fusion genes and other molecular manipulations
Using primers composed of sequences upstream and downstream of the start codon of the respective gene and sequences annealing with the kanMX6-PCUP1-Nub-HA cassette, a PCR product was obtained that could be integrated in front of or within the gene of interest via homologous recombination. Successful recombination was verified in Geneticin-resistant strains by a diagnostic PCR using primers annealing with the Nub sequence and sequences 300–600 bp downstream of the integration site within the respective gene. A minority of genes was assembled as full-length Nub gene fusions behind the kanMX6-PCUP1-Nub-HA on a centromeric pRS vector (Sikorski and Hieter, 1989).
Full-length CubRURA3 (CRU) fusion genes expressed from their native promoters as well as their C-terminally truncated versions were obtained and their successful recombination verified essentially as described (Wittke et al., 1999). The identical PCR products used for the construction of the CRU gene fusions was placed in front of the GFP sequence or a sequence coding for the 9×Myc-epitope in a pRS304 or 306 recombinative vector to create the full-length and natively expressed GFP and Myc fusion genes (Sikorski and Hieter, 1989). Ectopic expression of fragments of the genes as CRU fusions was achieved by inserting the respective sequences between the PMet17 promoter and the CRU module on a centromeric pRS313 or 315 vector (Sikorski and Hieter, 1989).
Exchange of the native promoter of a certain gene in the yeast genome by the PMET17 promoter was achieved by recombination of a PCR fragment containing the natNT2 resistance gene followed by the PMET17 promoter sequence. The 5′ and 3′ ends of the PCR fragments were capped by 45–47 bp sequences identical to the 5′ upstream sequence or to the sequence immediately following the start codon of the target gene respectively. Recombination was verified by a diagnostic PCR using primers annealing in the PMET17 promoter or within the sequence of the ORF of the gene respectively (Janke et al., 2004).
NBP2 and NAP1 were replaced by the natNT2 resistance gene or the CmLEU2 gene in the strains JD53 (MATα) and JD47 (MATa) through a one-step gene deletion procedure to create the Δnbp2 and Δnap1 strains. Deletions were verified by diagnostic PCR of the strains using primer combinations annealing outside of the ORF of the gene and primer combinations annealing within the marker gene and outside of the ORF of the genes respectively (Janke et al., 2004; Knop et al., 1999). The GST-NBP2 fusion gene was obtained by placing the ORF of NBP2 behind the E. coli GST sequence on the pGEX-6P-1 vector (GE Healthcare, Freiburg, Germany). Mutations were introduced in the corresponding genes or fragments thereof by using primers annealing at the respective site but carrying a mismatch to change the codon by the ensuing PCR.
Split-ubiquitin assay was carried out with 383 individual MATα strains each expressing a certain Nub fusion gene. They were printed on a YPD-Geneticin plate (Gen+) and replicated as quadruplicates onto YPD medium by a rotor robot system (Singer Instruments, Watchet, UK). A MATa strain expressing a certain CRU fusion gene was then mated with the Nub strains by printing it onto the 1532 positions of the array. After overnight incubation at 30°C the mated cells were transferred onto His−, Gen+ medium to select for the presence of the Nub and Cub fusion genes. After incubation for 5 days at 30°C the yeast strains were transferred to His− Gen+ medium or His− Gen+ medium containing 0.1% FOA supplemented with either no or 50 mM or 100 mM additional copper. Growth of the cells at 30°C was scored each day for up to 7 days. Promiscuous Nub fusion proteins were defined by their interaction pattern with a set of 79 CRU fusions including a subset of the fusions shown in this work and CRUs whose interactions will be shown elsewhere. Nub fusions that showed interactions with more than 20% of the baits of the test-set were considered to be constitutively false positives and not included in the list of specific interaction partners (supplementary material Table S2). Interactions with Nub-Sso1p and/or Nub-Snc1p defined CRU fusions as physically associated with the plasma membrane or the late secretory system, respectively (Wittke et al., 1999). In this case Nub fusions known to be predominantly associated with the plasma membrane were considered as potential false positives and not counted as binding partners. Interaction with Nub-Ubc6p and/or Nub-Sec62p defined CRU fusions as residents of the ER membrane, and Nub fusions known to be predominantly associated with the ER membrane were considered as potential false positives and not counted as binding partners (Wittke et al., 1999). Measurement of interactions between individual Nub and Cub fusion proteins by spotting yeast cells expressing both fusions onto 5-FOA and SD Ura-containing medium was essentially as described (Wittke et al., 1999; Eckert and Johnsson, 2003).
Microscopy and live-cell imaging
Differential interference contrast (DIC) and fluorescence microscopy were performed using the Delta Vision System (Applied Precision, Issaquah, WA). The System is equipped with the Olympus IX71 microscope, with a 100× Plan-Apo/1.4 objective lens and a charge-coupled-device (CCD) camera from Photometrics. Images were acquired and analysed with the softWoRx software of the Delta Vision System, and processed using Adobe Photoshop. Fluorescent proteins were visualized using a live cell filter set (Chroma Technology) using the EGFP (λex470, λem525) and mCherry filter (λex572, λem632) respectively. Yeast strains were grown overnight in liquid selective medium, diluted the next day, and grown to mid-log phase at the indicated temperatures. Cells were fixed after washing once in selective medium by incubation in 4% paraformaldehyde for 10 minutes at room temperature. Cells were washed twice in 1× PBS before analysis.
For time-lapse microscopy exponentially grown cells were diluted to a density of 3×106 cells/ml and filled into the cell inlet well of a microfluidic plate (CellASIC Corporation, San Leandro, CA). Temperature was held constant at 30°C using a Delta Vision System supplied temperature chamber.
Plasmids pYC142-mtRFPff, expressing the mitochondrial targeting domain of the ATP9 of N. crassa fused to a fast-folding DsRed, and pJR233, expressing GFP coupled to the peroxisomal targeting sequence SKL, were transformed into wild-type, Δnbp2 and PMETCLA4 strains for measuring the cellular inheritance of mitochondria and peroxisomes (Kondo-Okamoto et al., 2006). The obtained strains were grown to mid-log phase at 30°C or 37°C in SD medium supplemented with or without methionine. The widths of mother and daughter cells were measured and the presence of the organelles in the daughter was determined. After computing the ratio of mother to daughter cell width, each cell was classified into one of three bud size categories: 0–0.3 (small bud); 0.3–0.5 (middle bud); 0.5–1 (large bud).
Preparation of yeast cell extracts
For pull-down experiments, cell cultures were grown to an OD600 of 1.5, pelleted, washed once in ice-cold water, and then transferred into liquid nitrogen. The cell pellets were ground in liquid nitrogen using a mortar. The cell powder was collected in protein extraction buffer (50 mM HEPES pH 7.5, 150 mM NaCl, 1 mM EDTA) containing 1× protease inhibitor cocktail (Roche Diagnostics, Mannheim, Germany), 1 mM dithiothreitol (DTT) and 1 mM phenylmethylsulfonyl fluoride (PMSF). 0.1–0.2% Triton X-100 was added, incubated for 10–20 minutes on ice, and extracts were clarified by centrifugation.
GST pull-down analysis
E. coli cells (BL21, Amersham, Freiburg, Germany) expressing GST or GST–Nbp2p were grown at 37°C to an OD600 of 0.6, and the expression of GST fusion proteins were induced by 1 mM IPTG. The cells were shaken for additional 3 hours at 30°C, harvested by centrifugation, and resuspended in lysis buffer (150 mM NaCl, 5 mM Imidazol, 50 mM KH2PO4 pH 8.0) containing 1× protease inhibitor cocktail, 0.4 mM PMSF and 1 mg/ml lysozyme (Sigma). After incubation for 10 minutes at 4°C, cells were sonicated, and then treated with 0.01 mg/ml DNaseI in 25 mM MgCl2 and 1% Triton X-100 for 10 minutes on ice. The lysates were clarified by centrifugation at 18,000 r.p.m. for 30 minutes at 4°C, and either stored at −80°C or immediately incubated with lysis buffer equilibrated Glutathione–Sepharose 4B beads (GE Healthcare, Freiburg, Germany) for 1 hour at 4°C under rotation. Finally, bound material was washed three times in 1× PBS.
GST or GST–Nbp2p Sepharose slurries were incubated with 1 ml of yeast cell extract containing the Myc-tagged proteins for 3 hours at 4°C under rotation. The slurry was poured over a spin column and washed three times with three volumes of protein extraction buffer, and twice with two volumes of 1× PBS. The bound material was eluted with 1× PBS containing 10 mM gluthatione and analyzed by SDS-PAGE and immunoblotting.
Proteins were separated by a 12.5% SDS-PAGE and transferred onto nitrocellulose membrane (Bio-Rad, Munich, Germany), using a semi-dry blotting system (GE Healthcare, Freiburg, Germany). The quality of the transfer was checked by Ponceau S staining of the membrane. Blots were incubated for 1 hour in blocking solution (5% non-fat dry milk in 1× TBST) and treated with a mouse monoclonal anti-Myc (HISS Diagnostics, Freiburg, Germany) or anti-HA antibody (Covance, Freiburg, Germany) for 2 hours at room temperature or overnight at 4°C. After washing the membranes three times in 1× TBST, the membranes were incubated with goat anti-mouse horseradish peroxidase-conjugated secondary antibody (Bio-Rad) for 2 hours at room temperature. Bound secondary antibodies were visualized using the enhanced chemiluminescence detection system (Roche Diagnostics, Mannheim, Germany).
Pheromone halo assay
10 μl of a saturated culture of NBP2 (JD47) or Δnbp2 (AHY140) cells were added to 4 ml of 0.5% (w/v) agar, mixed gently and poured directly onto a YPD plate. Filter paper disks (5 mm diameter) containing 1 μg α-factor (Carl Roth, Karlsruhe, Germany) were placed on the solid surface and the diameters of the halos were measured after 2 days at 30°C.
NBP2 and Δnbp2 cells harbouring BG1805-PGAL1-X vectors (X=BCK1, MKK1, SLT2, NBP2, PTC1, CLA4, SKM1, STE20, GIN4, KCC4, KIC1, STE7, FUS3, HOG1 and KSS1) (EUROSCARF, Frankfurt, Germany) or empty pRS316 vector were grown in liquid selective medium containing 2% glucose (SD) to an OD600 of 1 and spotted in tenfold dilutions on SD-agar plates supplemented with either 2% galactose to induce or 2% glucose to repress the expression of the respective genes.
The work was supported by a DFG Research Grant JO 187/5-1 to N.J. We thank Andreas Wilbers for his technical support, Prof. Dr Janet M. Shaw for plasmid pYC142-mtRFPff and Dr Judith Müller for comments on the manuscript. Caspofungin was a gift from MSD Sharp & Dohme (Haar, Germany).