Kel1 is a phosphorylation-regulated noise suppressor of the pheromone signaling pathway

Mechanisms have evolved that allow cells to detect signals and generate an appropriate response. The accuracy of these responses relies on the ability of cells to discriminate between signal and noise. How cells filter noise in signaling pathways is not well understood. We have analyzed noise suppression in the yeast pheromone signaling pathway. By combining synthetic genetic array screening, mass spectrometry and single-cell time-resolved microscopy, we discovered that the poorly characterized protein Kel1 serves as a major noise suppressor of the pathway. At the molecular level, Kel1 suppresses spontaneous activation of the pheromone response by inhibiting membrane recruitment of Ste5 and Far1. Kel1 is regulated by phosphorylation, and only the hypophosphorylated form of Kel1 suppresses signaling, reduces noise and prevents pheromone-associated cell death. Our data indicate that in response to pheromone the MAPKs Fus3 and Kss1 phosphorylate Kel1 to relieve inhibition of the pheromone pathway. Taken together, Kel1 serves as a phospho-regulated suppressor of the pheromone pathway to reduce noise, inhibit spontaneous activation of the pathway, regulate mating efficiency and to prevent pheromone-associated cell death.


Introduction
A crucial aspect of any organism's well-being is the ability of cells to respond to changes in their internal and external environment. Accurate signal-noise discrimination is particularly important during conditions that threaten cellular homeostasis, or when a given signal triggers cellular commitment, such as differentiation. Low levels of noise within a population of cells may be beneficial under certain conditions, by allowing a fraction of cells to survive a dramatic change in environmental conditions (Kaern et al., 2005). However, high noise levels may be detrimental to cellular fitness and has been evolutionarily minimized (Balazsi et al., 2011;Lehner, 2008;Metzger et al., 2015;Wang and Zhang, 2011). Noise has been best studied at the level of gene expression, where it is often referred to as the stochastic variation in the protein expression level of a gene among isogenic cells in a homogenous environment (Raser and O'Shea, 2005;Wang and Zhang, 2011). Gene expression noise can arise from intrinsic and extrinsic variations (Raser and O'Shea, 2005). Intrinsic noise is caused by inherent stochastic events in biochemical processes that can occur at various levels during gene expression, such as transcriptional initiation, mRNA degradation, translational initiation, and protein degradation, as well as during signal transduction (Raser and O'Shea, 2005). Extrinsic noise is caused by differences among cells, either in their local environment or the concentration or activity of any factor that influences gene expression (Raser and O'Shea, 2005;Volfson et al., 2006), such as age, cell cycle stage, metabolic state, and the number and quality of proteins and organelles distributed to the mother and daughter cell during cell division.
The yeast mating pathway is a model system for signal transduction and cellular decision-making (Alvaro and Thorner, 2016;Paliwal et al., 2007). In haploid yeast cells, this pathway detects and transmits a pheromone signal emitted by cells of the opposite mating type to induce a mating response (Fig. 1A). Activation of this pathway results in cell cycle arrest, activation of a transcriptional program and cell wall remodeling to execute the morphological changes required to mate (Alvaro and Thorner, 2016). Given the potential fitness cost associated with inappropriate activation of this pathway (Banderas et al., 2016), pathway activation occurs in normal cells in a switch-like manner with high precision and overall low intrinsic noise (Dixit et al., 2014a;Malleshaiah et al., 2010). The pathway is activated by binding of pheromone to its G protein-coupled target receptor, resulting in dissociation of the heterotrimeric G protein into the α -subunit (Gpa1) and a Gβγ heterodimer Membrane-localized Far1 mediates polarized growth by recruiting the guanine nucleotide exchange factor Cdc24, which stimulates Cdc42 to induce formation of a mating projection (shmoo) (Butty et al., 1998). Cdc42 also activates the PAK-like kinase Ste20, which reinforces activation of the Fus3 MAPK pathway by phosphorylating Ste11 (Pryciak and Huntress, 1998).
Several feedback loops have been identified in the pheromone signaling pathway that improve transmission of information and that help switch off the pathway. An example of positive feedback is the increased expression of FUS3 that occurs upon activation of Ste12 by Fus3 (Roberts et al., 2000). However, Fus3 also provides negative feedback to dampen the response and help switch off the mating response (Yu et al., 2008). This is physiologically relevant, because hyperactivation of the mating pathway, or an attempt to mate in absence of a partner, can lead to cell death (Severin and Hyman, 2002;Zhang et al., 2006), which can be caused by weakening of the cell wall during formation of the mating projection (Iida et al., 1990;van Drogen et al., 2019). A major negative feedback mechanism consists of Fus3mediated phosphorylation of Ste5, which attenuates signaling output (Choudhury et al., 2018;Repetto et al., 2018). Another example of Fus3-mediated negative feedback occurs through phosphorylation of Sst2 (Yu et al., 2008). Sst2 is a Regulator of G protein Signaling (RGS) and inhibits G-protein signaling by acting as a GTPase Activating Protein (GAP) for Gpa1 (Apanovitch et al., 1998), and it mediates desensitization of cells after pheromone treatment (Dohlman et al., 1995). Interestingly, Sst2 has also been shown to function as a noise suppressor (Dixit et al., 2014a). The dominant source of variation in the pheromone pathway is thought to be extrinsic noise (Colman-Lerner et al., 2005), and mutant cells either lacking SST2 or expressing a mutant form of Gpa1 that is resistant to the GAP activity of Sst2 show increased levels of noise both in the transcription and in the morphogenesis branches of the pathway (Dixit et al., 2014a). Noise suppression is thought to be required for proper gradient detection and morphogenesis, and one potential physiological consequence of elevated noise in mutant cells is reduced mating efficiency (Dixit et al., 2014a).
Despite the fact that pheromone signaling has been studied intensively during the past decades, there are still several genes with poorly characterized functions in the pathway. One such gene is KEL1, which encodes a 131 kDa protein consisting of a KELCH propeller in the N-terminal region and three coiled-coil domains in the C-terminal region (Fig. 1B). Kel1 was first identified in a screen for genes whose overexpression relieved the mating defect caused by activated alleles of PKC1 (Philips and Herskowitz, 1998). kel1Δ mutants are elongated and heterogeneous in shape and have a defect in cell fusion. Kel1 localizes to sites of polarized growth and forms a ternary complex with the cell fusion regulator Fus2 and activated Cdc42 during mating. Kel1 also interacts with formins to regulate the assembly of actin cables (Gould et al., 2014). These findings suggest that Kel1 may serve as a hub during the pheromone response, but how and to which extent Kel1 controls the pheromone response remains unknown.
Here, we characterized the function of Kel1 in the pheromone response. We present data showing that Kel1 is regulated by phosphorylation, suppresses spontaneous activation of the pheromone pathway, has a major role in filtering noise in the pheromone pathway, and that it prevents pheromone-induced cell death.

Kel1 is important for the pheromone response
During the course of our experiments, we serendipitously observed that approximately 35% of kel1Δ mutant cells died upon treatment with pheromone ( Fig. 1C-E). Careful inspection of the data revealed that also a small but significant number of wild-type (WT) cells died after pheromone treatment (Fig. 1E), consistent with previous findings that mating is a potentially lethal affair (Severin and Hyman, 2002). The importance of Kel1 in preventing cell death was reflected in pheromone halo assays, where we observed a 30% cell-density reduction in the kel1Δ strain compared with the WT strain inside the halo, where the concentration of pheromone is highest (Fig. 1F, G). Although there was no significant change in the area of the halo, which is typically observed in mutants with defects in the mating pathway, we did notice a sharper interface of the halo (Fig. 1F, G), consistent with a role for Kel1 in the adaptive response to pheromone. Light microscopy revealed not only that kel1Δ mutants had significant morphological defects during vegetative growth, such as elongated buds and misshapen cells (Suppl. Fig. S1A, B), but more importantly, they were also 15 times more prone than WT cells to initiate the development of more than one shmoo during pheromone treatment (Fig. 1H). This indicates that kel1Δ mutants fail to commit to the original projection, resulting in abandonment of shmoos.
Together, these data indicate that Kel1 prevents pheromone-associated cell death.

Kel1 activity is regulated by phosphorylation
High-throughput experiments have detected at least 50 phosphorylated amino acids in Kel1, suggesting that Kel1 may be regulated by certain signaling pathways (Albuquerque et al., 2008;Bodenmiller et al., 2010;Gnad et al., 2009;Helbig et al., 2010;Holt et al., 2009;Li et al., 2007;Smolka et al., 2007). Indeed, λ-phosphatase treatment of immunoprecipitated Kel1 resulted in increased mobility on SDS-PAGE, confirming that Kel1 exists as a phosphoprotein in vivo ( Fig. 2A). It should be noted that bulk Kel1 phosphorylation did not appear to be altered much by pheromone treatment ( Fig. 2A), even though it has previously been shown that phosphorylation of certain residues in Kel1 is increased after treatment with pheromone (Li et al., 2007). Likely, only a few phosphorylation sites in Kel1 respond to pheromone treatment, and their effect on the mobility of Kel1 on SDS-PAGE may be obscured by the other phosphorylation sites.
A substantial number of putative phosphorylation sites in Kel1 consists of [S/T-P] sites (Albuquerque et al., 2008;Bodenmiller et al., 2010;Gnad et al., 2009;Helbig et al., 2010;Holt et al., 2009;Li et al., 2007;Smolka et al., 2007), which are typically targeted by proline-directed kinases such as CDKs and MAPKs. One such kinase is the pheromoneactivated MAPK Fus3, and several proline-directed phosphorylation sites in Kel1 have been detected specifically during pheromone treatment (Li et al., 2007). We hypothesized that proline-directed phosphorylation might regulate Kel1. Therefore, we first constructed two mutant alleles in which all [S/T-P] sites were substituted with either alanine or aspartate residues (kel1-ala and kel1-asp, respectively, see Methods for details). These alleles were fused to a C-terminal FLAG tag and integrated into the KEL1 locus under control of the endogenous KEL1 promoter. Phos-tag gel electrophoresis of immunoprecipitated proteins revealed that Kel1-ala migrated faster than wild-type Kel1, confirming that at least some of these [S/T-P] sites are phosphorylated in vivo (Fig. 2B). Although follow-up experiments are required to tease out exactly which specific [S/T-P] sites are regulated by pheromone, these data show that Kel1 is phosphorylated on one or several [S/T-P] sites in vivo.
Next, we analyzed the morphology of kel1-ala and kel1-asp mutant cells. We noticed that during vegetative growth the bud shape of WT cells was slightly oblong, whereas the kel1-ala buds were almost perfectly spherical (Fig. 2C, D). kel1-asp buds were significantly more elongated than WT buds, although not as much as those of kel1Δ cells (Fig. 2C, D), and both kel1-asp and kel1Δ mutant cells frequently showed aberrant morphology (Fig. 2E). More importantly, upon pheromone treatment more than half of the kel1-ala mutant cells failed to form a shmoo (Fig. 2C, F), and those cells that did form a shmoo always generated a single, highly spherical shmoo (see below). In contrast, kel1-asp mutant cells readily formed shmoos and were more likely to form multiple shmoos after pheromone treatment than WT cells (Fig.   2F). Thus, the phenotype of kel1-asp cells resembles that of kel1Δ mutant cells, albeit more modest. Pheromone-associated cell death was also significantly higher in kel1-asp mutants, whereas in kel1-ala mutants it was significantly lower than in WT cells (Fig. 2G, H). This was reflected in the pheromone halo assay, which showed that the cell-growth interphase was sharper in the kel1-asp strain and more diffuse in the kel1-ala mutant compared to the WT strain ( Fig. 2I, J). Finally, we analyzed the ability of the kel1 mutants to mate with WT cells of the opposite mating type, which revealed that the mating capacity of kel1Δ and kel1-ala mutants was reduced significantly (Fig. 2K).
Taken together, these results indicate that phosphorylation of Kel1 is important for regulation of the mating process. Given that kel1-ala mutants have a phenotype opposite to that of phospho-mimicking kel1-asp mutants, and that the phenotype of kel1-asp cells generally resembles that of kel1Δ cells (although milder), we conclude that (i) hypophosphorylated Kel1 suppresses the pheromone response, and (ii) phosphorylation of certain [S/T-P] sites relieves the inhibitory effect of Kel1 on the pheromone response.

Kel1 acts downstream of Sst2 during the mating response
To identify pathways associated with phosphorylation of Kel1 we performed a genome-wide synthetic genetic array (SGA) screen (Baryshnikova et al., 2010;Tong et al., 1 0 2001b). Since kel1-asp mutants have a loss-of-function phenotype that generally resembles that of the kel1Δ strain, but weaker, we decided to use only the kel1Δ and kel1-ala strains as query mutants. These mutants were crossed into the library of non-essential knock-out genes as well as the Decreased Abundance by mRNA Perturbation (DAmP) library (Breslow et al., 2008;Giaever et al., 2002). For all genetic interactions we calculated the genetic interaction score (see Methods). We focused on a subset of genes that showed differential genetic interactions with the kel1Δ and kel1-ala mutations, using a strict cut-off in which the difference between kel1-ala and kel1Δ scores deviated more than three standard deviations from the mean (Fig. 3A, Suppl. Fig. S1C, and Suppl. Table S1). To find patterns in this gene list, we studied the reported phenotypes for these genes using information gathered from the Saccharomyces Genome Database. The most highly overrepresented phenotype was "Resistance to enzymatic treatment" (Fig. 3B). Mutations in genes associated with this phenotype often lead to cell wall defects, resulting in increased sensitivity to cell wall-lytic enzymes. This is consistent with our finding that kel1Δ mutants frequently undergo cell death during pheromone-induced cell wall remodeling. The second most overrepresented phenotype was "Mating response". This is likely an underestimate, because many pheromone-associated genes are not easily identified in SGA screens due to the fact that mutations in this pathway lead to severe mating defects, and efficient mating is essential for the genome-wide crosses that form the basis of SGA screening.
We were intrigued by one mating response-associated gene in particular, i.e. SST2, which encodes a Regulator of G protein Signaling (RGS) that suppresses the pheromone response (Dohlman et al., 1996). Loss of SST2 substantially reduced the fitness of kel1Δ mutants, but did not appear to have a strong effect on the fitness of kel1-ala mutants (Fig.   3A). Interestingly, even in absence of pheromone approximately 8% of the cells in sst2Δ 1 1 the SGA screen. Pheromone treatment strongly increased cell death of kel1Δ mutants, and although cell death in sst2Δ cultures was similar to WT (Fig. 3D), it was significantly higher in sst2Δ kel1Δ double mutants than in either single mutant (Fig. 3C, D).
We next studied the cellular morphology of the mutants in presence and absence of pheromone. 15% of vegetatively growing sst2Δ cells showed aberrant morphology, and the buds of sst2Δ mutants were generally more elongated than those of WT cells, but to a lesser extent than kel1Δ mutants ( Fig. 3E-G). Pheromone treatment resulted in a small but significant increase in the number of sst2Δ cells with multiple projections (Fig. 3H). The morphology of cells and buds of the sst2Δ kel1Δ double mutant resembled that of the kel1Δ single mutant (Fig. 3F, G), and pheromone treatment of sst2Δ kel1Δ double mutants did not further increase the number of cells with multiple projections compared with the kel1Δ single mutant (Fig. 3E, H). However, the kel1-ala mutation completely suppressed the morphological defects and inhibited shmoo formation of sst2Δ mutants ( Fig. 3E-H).
Taken together, these data show that Kel1 and Sst2 regulate cell morphogenesis both during vegetative growth and in the presence of pheromone, with Kel1 having a dominant role, and that Kel1 functions downstream of Sst2 in this process.

Kel1 physically interacts with pheromone pathway components and may be phosphorylated by Fus3 and Kss1
We wished to better understand how Kel1 is regulated during the pheromone response.
We immunopurified Kel1 from untreated and pheromone-treated cells and identified interaction partners using mass-spectrometry (Tables S2 and S3). We identified a large number of proteins previously known to physically interact with Kel1, thus validating the approach (Suppl. Fig. S1D). We looked for proteins that differentially interacted with Kel1 depending on pheromone treatment, and observed a significant enrichment of proteins with functions in the mating response ( Fig. 4A

Kel1 prevents spontaneous recruitment of Ste5 and Far1 to inhibit formation of mating projections in absence of pheromone
Vegetatively growing kel1Δ mutant cells often form elongated structures that resemble shmoos. To determine whether these structures are indeed spontaneous shmoos we monitored membrane localization of Ste5 and Far1, which are markers for formation of mating projections (Nern and Arkowitz, 1999). Strikingly, even in absence of pheromone, we observed that NeonGreen (mNG)-tagged Ste5 and Far1 were recruited to patches at the cell cortex both in kel1Δ mutants and in kel1-asp mutant cells (Fig. 5A, B). Such spontaneous membrane recruitment of Ste5 and Far1 was not observed in WT cells or in kel1-ala mutants.
Furthermore, whereas pheromone treatment resulted in localization of Ste5-mNG and Far-mNG at the cortex of WT cells, recruitment of these proteins was undetectable in the kel1-ala mutant ( Fig. 5A-C). These findings show that the polarized structures that are observed in vegetatively growing kel1Δ and kel1-asp mutants are indeed spontaneous shmoos, and suggest that Kel1 suppresses spontaneous signaling through the pheromone pathway.
Therefore, we analyzed the mRNA levels of two genes known to be upregulated by the pheromone pathway, i.e. FUS3 and STE2 (Oehlen et al., 1996). Although there was no significant difference in FUS3 and STE2 mRNA levels between populations of WT cells and any of the kel1 mutants after pheromone treatment, kel1Δ mutant populations showed a significant increase in mRNA levels in absence of pheromone (Fig. 5D), confirming that Kel1 suppresses spontaneous activity of the pheromone pathway.
If Kel1 indeed inhibits spontaneous signaling through the pheromone pathway by preventing the accumulation of Ste5 at the cell cortex, then deletion of STE5 should reverse the phenotype of kel1Δ and kel-asp mutants. We found that the buds of vegetatively growing ste5Δ mutants were significantly more spherical than those of WT cells (Fig. 5E, F). More importantly, deletion of STE5 also modestly but significantly reduced the bud phenotype of

Hypophosphorylated Kel1 cooperates with Sst2 to prevent spontaneous pheromone signaling and to dampen the pheromone response
Spontaneous activation of the pheromone pathway has been observed previously and was found to originate downstream of the receptor at the level of the G protein (Siekhaus and Drubin, 2003), which we confirmed (Suppl. Fig. S1E). Mechanisms have evolved that suppress spontaneous signaling; for instance, RGS proteins like Sst2 inhibit unscheduled G protein signaling (Siekhaus and Drubin, 2003). Given our findings that Kel1 acts downstream of Sst2 but at/upstream of Ste5, and that the kel1-ala mutant suppresses the morphological defects observed in sst2Δ mutant cells, we hypothesized that Kel1 might cooperate with Sst2 to inhibit spontaneous signaling. We made use of a dual reporter system based on expression of GFP under control of the pheromone pathway-sensitive FUS1 promoter and mCherry under control of the pheromone pathway-independent ADH1 promoter; the ratio of GFP/mCherry is a measure of signaling pathway activity at single-cell resolution (Dixit et al., 2014a). Interestingly, single-cell microscopy revealed that in absence of pheromone the GFP/mCherry ratio was significantly higher in kel1Δ mutants than in WT cells ("0 min" in

Kel1 suppresses noise in the pheromone signaling pathway
We noticed that there existed considerable cell-to-cell variability in the pheromone response in the population of kel1Δ mutants, which suggests that Kel1 may suppress noise in the mating pathway. To investigate this further, we measured the overall signaling noise as the coefficient of variation (CV) of the median population response before and after pheromone treatment. The dispersion in WT signaling was between 40-50% both before and after pheromone treatment, but increased slightly in the dynamic range of the pheromone response as previously reported (Fig. 7A, B and Suppl. Interestingly, the signal dispersion in untreated cell populations was higher in kel1Δ mutants than in WT cells, but strongly reduced in kel1-ala mutants (Fig. 7A, B and Suppl. Fig. S3B).
As expected, signal dispersion was also slightly increased in the sst2Δ mutant, although less so than in the kel1Δ strain, while expression of kel1-ala reduced the variability of the sst2Δ mutant to that of the WT strain. In contrast, after 180 minutes of pheromone treatment we could no longer observe the increase in overall noise that we detected in populations of kel1Δ and sst2Δ mutants in absence of pheromone (Fig. 7B). This is likely a consequence of the elevated signaling and increases in average GFP expression in these mutants, which causes an expected decrease in CV for overdispersed distributions ( Taken together, these results show that Kel1 suppresses noise and limits spontaneous signal transmission in the absence of pheromone.

Discussion
High levels of noise in signaling pathways can result in a substantial fitness cost, and therefore filtering mechanisms have evolved to suppress noise. Here, we identify Kel1 as a major noise suppressor in the pheromone pathway. Importantly, Kel1 is regulated by Our working model is shown in Suppl. Figure S2. In absence of pheromone, hypophosphorylated Kel1 prevents recruitment of Ste5 and Far1 to the cell membrane (Suppl. Expression of non-phosphorylatable Kel1-ala blunts pheromone-induced recruitment of Ste5 (Suppl. Fig. S2E), suppresses noise in the pathway and reduces pheromone-induced cell death, limits formation of mating projections and results in low mating efficiency.
The kel1-asp mutant partially phenocopied the kel1Δ mutation, which is consistent with the idea that, at least in terms of the pheromone response, hypophosphorylated Kel1 suppresses the pathway, whereas phosphorylation of Kel1 results in its inactivation. It should be mentioned that the effects of the kel1-asp mutant were generally milder than the kel1Δ mutation, which could at least partially be due to the fact that phosphomimicking residues often do not fully replicate the effect of phosphorylation (Paleologou et al., 2008). Some of the other limitations of our study that remain to be addressed are to determine the exact phosphorylation sites in Kel1 that respond to pheromone, identify the phosphatase that targets Kel1, and unravel the molecular mechanism by which Kel1 controls activity of the mating pathway. We presently also do not know how phosphorylation regulates the function of Kel1 at the molecular level. Kel1 consists of a Kelch propeller and coiled coil domains, which are known to mediate multimerization and protein-protein interactions. Given that Kel1 can physically interact with Ste2, Sst2, Far1, Ste5, Fus3, we speculate that Kel1 may form a phospho-dependent platform that integrates signals to regulate the pheromone pathway.
Future studies will focus on unraveling the molecular mechanism by which Kel1 regulates membrane recruitment of Ste5 and Far1.
In addition to [S/T-P] sites, Kel1 appears to be phosphorylated on many other sites, suggesting that several other signaling pathways may converge onto Kel1. Interestingly, it was recently shown that in response to mechanical stress Pkc1 prevents lysis of pheromonetreated cells by phosphorylating Ste5, resulting in dispersal of Ste5 from the site of polarized growth (van Drogen et al., 2019). We found that Kel1 also suppresses lysis during pheromone treatment and that it regulates Ste5 localization. Moreover, Kel1 was first identified in an overexpression screen for genes that overcome the fusion defect of cells expressing activated Pkc1 (Philips and Herskowitz, 1998). It will be interesting to investigate the relationship between Pkc1, Kel1 and Ste5 at the molecular level.
Computer simulations and experimental studies have revealed mechanisms that provide robustness to the mating response (Chen et al., 2016;Howell et al., 2012), which is defined here as the persistence of a system's behavior under conditions of uncertainty.
Robustness is critical for efficient yeast mating and one way cells improve robustness is by reducing the sensitivity to pheromone, which is in part mediated by Sst2 (Chen et al., 2016).
We found that the transcriptional response to pheromone in individual kel1Δ mutant cells fluctuated considerably over time, whereas kel1-ala mutants showed less fluctuations compared to WT cells. This suggests that, whereas phosphorylation of Kel1 promotes activity of the pathway, hypophosphorylated Kel1 may provide robustness to the system. The phenotype of the kel1Δ mutant and the kel1-ala mutant masked the increase in signaling noise in sst2Δ mutants, indicating that proper control of Ste5 may be the dominant limiting factor for signal transmission fidelity and amplitude. We speculate that the balance between phosphorylated and a hypophosphorylated Kel1 is important for fine-tuning the output of the pathway by controlling Ste5. Clearly, more studies are needed to more fully understand how Kel1 regulates pheromone signaling.
In conclusion, we have shown that Kel1 is an important regulator of the mating pathway with a major function in noise suppression. 1

Yeast strains and growth conditions
S. cerevisiae strains were grown in at 30°C until mid-log phase in standard yeast extract peptone dextrose (YPD) medium or in synthetic medium supplemented with relevant amino acids. Strains were derived directly from the S288c strains BY4741 (Brachmann et al., 1998) and RDKY3032 (Flores-Rozas and Kolodner, 1998) using either standard gene-replacement methods or intercrossing. See Suppl. Table S3 for strains.
To construct kel1-ala and kel1-asp mutants, KEL1 was fist replaced with the URA3 gene.

Plasmids
For a list of plasmids see Suppl.  Table S5), which also include HindIII and BamHI sites for the tag replacement.

Pheromone treatment
Cells were treated with 15mg/L of alpha actor (custom synthesized by GenScript) for 2 hrs unless otherwise indicated.

Mating efficiency assay
To evaluate the matting efficiency 1ml of mid-log-phase MATa and MATα cells carrying complementary markers were mixed an incubated at 30°C in absence of agitation for 4 hours.
100μl of five consecutive serial dilutions (1:10) of the crosses were plated on YPD plates and diploid selective media (YPD supplemented with G418 200μg/ml (Sigma) and nourseothricin 100μg/ml (WERNER)). After 2 days of incubation at 30°C the colonies of the plates were counted using the Colony Counter mobile application (Promega). See Suppl. Table S3 for strains.

FITC staining
Cells were harvested, washed with PBS and stained for 10 min in the dark with 1mg/ml FITC (Sigma-Aldrich) in carbonate buffer (0.1M, pH 9.5). Cells were washed 3 times with PBS and visualized with the microscope.

Flow cytometry
GFP and RFP-expressing cells were harvested and fixed with 4% paraformaldehyde for 15 min at room temperature. Paraformaldehyde was quenched with Glycine 0.5M for 15 min at room temperature. Cells were washed 2 times with PBS, resuspended in PBS and sonicated for 10 seconds at 30% amplitude. Flow cytometry was performed using an LSR II Flow Cytometer (BD Biosciences). Data analysis and plotting was performed in R.

Mass spectrometry
Cells untreated or treated with α-factor for 2 hrs. were collected, washed with cold PBS and lysed in lysis buffer (100mM Tris-HCl pH8, 150mM NaCl, 1mM EDTA, 1% Triton X-100, 10% Glycerol, 2mM DTT, protease and phosphatase inhibitor cocktails). After centrifugation, supernatants were immunoprecipitated by mixing 800 µg of cell extract with 20 µl of anti- DTT, protease and phosphatase inhibitor cocktails) and lysed by vortexing with glass beads followed by centrifugation to remove cell debris. Equal amounts of proteins were used for immunoprecipitation using magnetic beads conjugated covalently to relevant antibodies. After extensive washing with lysis buffer, coimmunoprecipitated proteins were resolved by SDS-PAGE and analyzed by Western-blotting with the indicated antibodies.
For λ phosphatase treatments, cells were harvested and lysed as described above, magnetic beads were washed 2 times with lysis buffer without phosphatase inhibitors. Samples were treated with 800 units of λ-phosphatase (NEB) for 30 min after which the enzyme was heatinactivated (5 min at 95ºC) and the proteins resolved by SDS-PAGE and analyzed by Western-blotting.
Briefly, cells were harvested and washed with ice-cold 20% TCA (Sigma-Aldrich) and resuspended in 20% TCA. Cells were lysed by vortexing with glass beads followed by centrifugation (10 min. at 13.000 rpm). Precipitated proteins were washed with ethanol 70%.
Equal amounts of sample were loaded onto 4.5% acrylamide (BioRad) 20μM Phos-tag (Supersep, Wako chemicals) gels, followed by electrophoresis for 6 hours at 150 V. Gels were washed three times for 10 min with transfer buffer containing 5 mM EDTA and three times for 10 min with transfer buffer without containing EDTA. The gel was transferred to a membrane and protein phosphorylation was analyzed by Western blotting as described above.
Protein densitometric quantifications were performed using FIJI software (Schindelin et al.,

Synthetic genetic array (SGA)
The SGA query strain Y8205 (can1::STE2pr-Sp_his5 lyp1::STE3pr-LEU2 his3Δ1 leu2Δ0 ura3Δ0; kind gift from C. Boone, University of Toronto, Canada) either harboring the kel1Δ or the kel1-ala mutation was crossed with a collection of deletions of non-essential genes (YKO) and with a collection of mutants with reduced mRNA levels of essential yeast genes (DAmP) according to (Tong et al., 2001a). Briefly, mutations of interest were linked to natMX selectable marker, while mutations in the collections to kanMX. After mating and sporulation, the spores were transferred to medium which enables growth of MATa meiotic progeny. In a last step, double mutants were obtained after selection on medium containing kanamycin and nourseothricin. Each cross was done in quadruplicate on 1536-format plates.
All double mutants were grown at 30 o C and 37 o C and imaged after 2 days. Image analysis and scoring were done with SGAtools (Wagih et al., 2013).
The phenotype enrichment score (Fig. 3B) was computed in R using the bc3net library (de Matos Simoes et al., 2012) and it was calculated over genes annotated with phenotypes in invasive growth, response to pheromone, pheromone sensitivity, mating projection morphology, cell shape, endomembrane system morphology, cell wall morphology and cytoskeleton morphology. All annotations were downloaded from the Yeast Phenotype Ontology database (https://www.yeastgenome.org/ontology/phenotype/ypo).

Author contributions
Designed research: NG and JME. Performed research: NG, PC, SMO, JE. Analyzed data: NG, JME, ANA, PC, SMO, JE. Wrote the paper: NG, ANA, JME.     The analysis was performed as described in Figure 2D.   and kel1-asp mutants. Cell death was visualized by FITC staining as described in Figure 1D.

Acknowledgments
WT image was taken from an independent experiment. I, Quantification of cell death of the experiment shown in (H). Analysis was performed as described in Figure 1E, ** 0.001<p<0.01. The analysis was performed as described in (C). *** p<0.001; ** 0.001<p<0.01; * 0.01<p<0.05.    A  b  r  i  g  h  t  m  o  n  o  m  e  r  i  c  g  r  e  e  n  f  l  u  o  r  e  s  c  e  n  t  p  r  o  t  e  i  n  d  e  r  i  v  e  d  f  r  o  m  B  r  a  n  c  h  i  o  s  t  o  m  a  l  a  n  c  e  o  l  a  t  u  m  .   N  a  t  M  e  t  h  o  d  s   .  1  0  :  4  0  7  -4  0  9  .  S  i  e  k  h  a  u  s  ,  D  .  E  .  ,  a  n  d  D  .  G  .  D  r  u  b  i  n  .  2  0  0  3  .  S  p  o  n  t  a  n  e  o  u  s  r  e  c  e  p  t  o  r  -i  n  d  e  p  e  n  d  e  n  t  h  e  t  e  r  o  t  r  i  m  e  r  i  c  G  -p  r  o  t  e  i  n  s  i  g  n  a  l  l  i  n  g  i  n  a  n  R  G  S  m  u  t  a  n