Mobility of the gradient tracking machine in mating yeast depends on Bud1 inactivation and actin-independent vesicle delivery

: The mating of budding yeast depends on chemotropism, a fundamental cellular process. Haploid yeast cells of opposite mating type signal their positions to one another through the secretion of mating pheromones. We have proposed a deterministic gradient sensing model that explains how these cells orient toward their mating partners. Using the cell-cycle determined default polarity site (DS), cells assemble a gradient tracking machine (GTM) composed of signaling, polarity, and trafficking proteins. After assembly, the GTM redistributes up the gradient, aligns with the pheromone source, and triggers polarized growth toward the partner. Because strong positive feedback mechanisms drive polarized growth at the DS, it is unclear how the GTM is released for tracking after its assembly is complete. What prevents the GTM from triggering polarized growth at the DS? Here we describe two mechanisms that enable tracking. First, the Ras GTPase Bud1 must be inactivated to release the GTM. Second, actin-independent – but not actin-dependent – vesicle delivery must be targeted upgradient to effect GTM redistribution.

ABSTRACT: The mating of budding yeast depends on chemotropism, a fundamental cellular process. Haploid yeast cells of opposite mating type signal their positions to one another through the secretion of mating pheromones. We have proposed a deterministic gradient sensing model that explains how these cells orient toward their mating partners. Using the cell-cycle determined default polarity site (DS), cells assemble a gradient tracking machine (GTM) composed of signaling, polarity, and trafficking proteins. After assembly, the GTM redistributes up the gradient, aligns with the pheromone source, and triggers polarized growth toward the partner.
Because strong positive feedback mechanisms drive polarized growth at the DS, it is unclear how the GTM is released for tracking after its assembly is complete. What prevents the GTM from triggering polarized growth at the DS? Here we describe two mechanisms that enable tracking. First, the Ras GTPase Bud1 must be inactivated to release the GTM. Second, actinindependent -but not actin-dependent -vesicle delivery must be targeted upgradient to effect GTM redistribution.

INTRODUCTION
Cellular responses to chemical gradients are important for a wide range of biological phenomena.
The best-known gradient-stimulated outputs are chemotaxis (directed movement) and chemotropism (directed growth). For example, chemotaxis plays a vital role in development, immunity, wound healing, inflammation, and metastasis 1 ; chemotropism is integral to axon guidance, angiogenesis, pollen tube guidance, and fungal life cycles 2, 3 . Although they exhibit different behaviors, chemotactic and chemotropic cells face similar challenges: the responding cell must determine the direction of the gradient source by sensing small chemical concentration differences across its surface, then correctly polarize its cytoskeleton.
The unicellular eukaryote, Saccharomyces cerevisiae, is among the best-studied models of both cell-cycle control and chemotropism 4 . During vegetative growth, haploid yeast cells invariably form new buds adjacent to their last division site, resulting in a characteristic axial budding pattern 5 . Late in the G1 phase of the cell cycle, the Axl2 cortical marker protein recruits the Bud5 guanine nucleotide exchange factor (GEF) to the axial bud site, where it activates its target, the Ras GTPase Bud1 6,7 . Activated Bud1 interacts with the Rho GTPase Cdc42 and its GEF Cdc24, initiating local activation of Cdc42 8 . This positional signal is greatly amplified by two positive feedback loops and the resulting concentrated patch of active Cdc42 triggers the nucleation of actin cables and polarized delivery of secretory vesicles [9][10][11][12] . During the sexual reproduction stage of their lifecycle, haploid yeast cells differentiate into gametes and fuse to form diploid zygotes. Each of the two haploid mating types, MATa and MATα, secretes a peptide pheromone that activates a G-protein-coupled receptor (GPCR) on cells of the opposite type. The pheromone-bound receptor activates its cognate heterotrimeric G protein, causing Gα-GTP to dissociate from Gβγ. Free Gβγ then signals to the nucleus through the Fus3 MAPK cascade, inducing changes in gene expression and cell-cycle arrest in late G1. Gβγ also positions the eventual chemotropic growth site (CS) by linking the receptor to the machinery that nucleates actin cables via the Far1 scaffold protein [13][14][15] . Actin-directed delivery of secretory vesicles to the CS results in the formation of mating projections, commonly known as "shmoos." In mating mixtures, cells find and contact a partner by determining the direction of the most potent pheromone source and polarizing their growth toward it (hereafter referred to as shmooing) 16 .
When cells are treated with isotropic pheromone or are unable to sense a gradient, however, they shmoo adjacent to their last bud site -i.e., at the axial site where they would have budded next if not arrested in G1 17,18 . Hence, the axial bud site is also referred to as the default polarity site (DS) 14,18 .
Like all other chemo-sensing cells, yeast exhibit a remarkable ability to interpret chemoattractant gradients. It has been estimated that a 1% difference in receptor activation across the 5μm diameter of a yeast cell is sufficient to elicit robust orientation 19 . In mating mixtures, yeast cells almost invariably select a single partner, even when surrounded by multiple potential mates. We have recently proposed a deterministic gradient sensing model that explains how mating yeast cells accurately position the CS in response to shallow and dynamic physiological gradients 20,21 . In this model, yeast cells gain their gradient-sensing ability and orient toward their mating partners in four phases. During global internalization, the uniformly distributed receptor and G protein are removed from the plasma membrane (PM). During assembly, mating yeast cells take advantage of the Bud1-positioned DS to assemble the signaling, polarity, and trafficking proteins into a gradient tracking machine (GTM). Assembly of the GTM starts with Far1-Cdc24-Bem1 localization to the DS and ends with the concentration of exocytic and endocytic activities upgradient and downgradient, respectively. During tracking, segregation of exocytosis and endocytosis incrementally redistributes the GTM up the gradient along the PM to the CS. At the CS, the GTM stabilizes when vesicle delivery aligns with the pheromone gradient and the endocytic machinery surrounds the secretion site.
Although our model explains how yeast cells actively track pheromone gradients, it leaves a key question unanswered: How does tracking start? That is, how is the newly assembled GTM released from the DS? The challenge is to understand how the subtle directional information embedded in the extracellular pheromone gradient overrides the strong, feedback-amplified polarity of this intrinsic site. Here, we provide evidence for two mechanisms that explain how the GTM "escapes" from the DS and begins tracking. First, we show that the Bud1 GEF disappears from the PM in cells preparing to mate, whereas the Bud1 GTPase-activating protein (GAP), Bud2, polarizes to the DS and tracks with the receptor. We also demonstrate that tracking requires Bud1 inactivation. Second, we show that the mode of vesicle delivery changes as the GTM transitions from assembly to tracking and from tracking to stabilization. Whereas actindependent vesicle delivery (AD-VD) is active when the GTM is immobile, both before and after tracking, actin-independent vesicle delivery (AI-VD) is necessary and sufficient during tracking.
Our findings suggest that tight regulation of both DS function and the modes of vesicle delivery is essential for yeast gradient sensing.

RFP-Bud2 polarizes to the DS and tracks with the receptor
During the G1 phase of vegetative haploid cells, the Bud1 GTPase is activated adjacent to the cytokinesis site, where it marks the DS as the bud site in the next cell cycle. Active Bud1 also positions shmoo formation at the DS in cells treated with isotropic pheromone. A longstanding question in the study of yeast mating is how shallow pheromone gradients compete with the polarization machinery at a cell's DS to establish a CS aligned with that of its partner. We have shown that G1-arrested yeast cells preparing to mate assemble a GTM composed of signaling, polarity, and trafficking proteins at the DS, which enables them to find the closest mating To test this idea, we engineered MATa cells co-expressing the receptor reporter (Ste2-GFP) as a proxy for the GTM, and either RFP-tagged Bud5, the Bud1 GEF 7 , or RFP-tagged Bud2, the Bud1 GAP 22 . We took time-lapse images of these MATa cells during vegetative growth and in mating mixtures. In vegetative cells, Bud5-RFP localized to the DS after cytokinesis and before bud emergence (Fig. 1A), whereas RFP-Bud2 concentrated at the bud neck but was not detectable at the PM between cytokinesis and bud emergence (Fig. 1B). These observations are consistent with those reported by Park et al. 22 and Marston et al. 7 . Conversely, in mating yeast, Bud5-RFP gradually disappeared from the mother-daughter neck and was never detectable at the DS or elsewhere on the PM (Fig. 1C), whereas RFP-Bud2 polarized to the DS after cytokinesis and before the receptor (Fig. 1D). The polarized RFP-Bud2 and receptor tracked together along the PM to the CS before shmoo formation and fusion. These data suggest that Bud1 is inactivated in mating cells before tracking begins.

Tracking is defective in mating cells expressing Bud1 G12V
The absence of the Bud1 GEF (Bud5) from the DS along with the localization of the Bud1 GAP (Bud2) to the GTM during both assembly at the DS and redistribution suggested to us that Bud1 inactivation is required for gradient tracking. To test this, we imaged the receptor reporter in mating MATa BUD1 cells expressing a constitutively active form of Bud1, Bud1 G12V , from a centromeric plasmid (hereafter, BUD1 G12V /BUD1 cells) (Fig. 2) 23 . The G12V amino acid substitution blocks the GTPase activity of Bud1, thereby locking it in the active state. Cells forced to express Bud1 G12V in the absence of Bud1 are not viable. Consistent with our hypothesis, about 30% of the BUD1 G12V /BUD1 cells ignored MATα cells with which they were in direct contact, a behavior we see one tenth as often in wild-type (WT) mating mixtures. These BUD1 G12V /BUD1 cells either continued to bud ( Fig. 2A) or failed to redistribute the polarized receptor toward the potential partner (Fig. 2B). Of the BUD1 G12V /BUD1 cells that successfully formed zygotes, significantly fewer exhibited gradient tracking behavior as compared to control cells: They either fused with partners positioned near their DS or at the presumptive distal bud site ( Fig. 2C) 24,25 . In the distal-mating class of cells, the receptor polarized directly to the fusion site rather than tracking from the DS toward the mating partner -a phenotype we call jumping 21 .
Given that Bud1 partially rescues DS function in cells co-expressing Bud1 G12V , as evidenced by the viability of BUD1 G12V /BUD1 cells, these data support our hypothesis that Bud1 must be inactivated to allow GTM tracking.

The receptor polarizes to the DS but fails to track in bem1 ΔCPX cells
The polarization of RFP-Bud2 to the DS before the receptor in mating cells suggested that Bud1 is inactivated early during GTM assembly. Miller 26 . In mating cells, the earliest event in GTM assembly detected thus far is the localization of Far1 to the DS, presumably in complex with Cdc24 20 . We postulated that Far1-Cdc24 recruitment to the assembly site depends on the reported interaction of Cdc24 with Bem1 [13][14][15] , and further, that Bud1 GDP initiates GTM assembly by recruiting Bem1 to the DS. To test this conjecture, we took time-lapse images of MATa cells expressing Bem1 ΔCPX and the receptor reporter in mating mixtures. In the absence of Bud1 GDP -Bem1 interaction, we expected a failure to recruit Far1-Cdc24, and therefore, no receptor polarization to the DS. Surprisingly, the receptor polarized to the DS in bem1 ΔCPX cells just as well as in the WT cells (Fig. 3). However, bem1 ΔCPX cells showed no evidence of gradient-sensing: their polarized receptor crescents did not track toward potential mating partners, and they invariably shmooed and mated at the DS. These observations suggest that the interaction between inactive Bud1 and Bem1 is not required for GTM assembly, but that the Bem1 CPX domain is required for tracking.

Receptor tracking is defective in exo70 ΔdC cells
We have shown that the receptor tracks normally in bud1Δ cells 20 . Therefore, the inability of the receptor to track in bem1 ΔCPX cells must be because the CPX domain of Bem1 provides a critical tracking function independent of Bud1, and not because tracking depends on the interaction of Bud1 GDP with Bem1. We have also shown that tracking correlates with the concentration of the vesicle delivery marker, Sec3, to the upgradient side of the GTM 20 . Notably, Bem1 has been reported to direct vesicle delivery independent of actin cables through the interaction of its Phox homology domain with Exo70 27 .
Sec3 and Exo70 serve as partially redundant "pioneer proteins" for the vesicle-tethering exocyst complex -they position and catalyze assembly of the complex at discrete locations on the PM 28,29 . Although Sec3 and Exo70 are transported to the PM along with the other components of the complex by Myo2 on actin cables, they are unique in their direct recruitment to the PM independent of F-actin: Sec3 binds to Cdc42 and Rho1 while Exo70 binds to Bem1 [28][29][30][31] . The C-domain of Exo70 is essential for Exo70-Bem1 interaction and its deletion (denoted ∆dC) diminishes AI-VD without affecting AD-VD 30 .
To determine whether the failure of the GTM to track in bem1 ΔCPX cells is due to the loss of Bem1-directed AI-VD, we took time-lapse images of MATa exo70 ΔdC cells expressing Ste2-GFP in mating mixtures (Fig. 4A). Consistent with our hypothesis, about 40% of the exo70 ΔdC cells ignored MATα cells with which they were in direct contact, a behavior we see ten times less often in WT mating mixtures (Fig. 4B). In most of these cells, the receptor polarized to the DS but did not track toward proximal partners; consequently, such cells shmooed at the DS ( Fig. 4A and B). Of the exo70 ΔdC cells that successfully formed zygotes, a significantly larger fraction mated at their DS or at the presumptive distal bud site as compared to the control cells, while a significantly smaller fraction exhibited gradient tracking and chemotropic mating (Fig. 4C).
Given that AI-VD is partially maintained by Sec3 in exo70 ΔdC cells 30 , these data suggest that AI-VD is required for tracking.

Markers for AD-VD do not track with the receptor but polarize strongly at the eventual chemotropic site in mating cells
It has been proposed that the movement of the polarity complex along the cell cortex during yeast gradient sensing is driven by vesicles delivered to the PM on micro actin filaments 32,33 . To determine whether AD-VD is associated with gradient tracking in mating cells, we engineered MATa cells co-expressing the receptor reporter and RFP-tagged Myo2, a marker for vesicles delivered to the PM on actin cables 34 ; RFP-tagged Abp1, a marker for actin patches 35 ; or RFPtagged Abp140, a marker for actin cables 36 . We took time-lapse images of these MATa cells from cytokinesis to fusion as they mated with MATα cells ( 5C and 6F). This is consistent with our conclusion, based on time-lapse imaging of Sla1-RFP, that receptor-driven endocytosis is maximal behind the peak of tracking receptor and surrounds the eventual CS 20 . Together, these observations suggest that AD-VD does not contribute to GTM tracking, but is operative during GTM assembly at the DS and stabilization at the eventual CS.

The receptor tracks in myo2-16 cells mated with WT MATα cells at restrictive temperature
To determine whether AD-VD is required for tracking, we engineered MATa cells expressing Ste2-GFP and the temperature-sensitive allele myo2-16 37 . Myo2, the type V myosin motor protein in yeast, docks post-Golgi vesicles to actin cables and carries them to sites of actin-dependent polarized secretion on the PM 34 . At the restrictive temperature of 33 ºC, the myo2-16 mutant protein cannot dock post-Golgi vesicles to actin cables but has no effect on actin-cable assembly 37  global gradient-sensing mechanism that competes with the DS, a mobile GTM is assembled at the DS, which then incrementally redistributes toward the mating partner. However, concentrating the key polarity and secretory proteins at the DS presents its own problems: What prevents polarized growth at that site? And how is the GTM released for tracking after its assembly is complete? Here we describe two mechanisms that enable tracking (Fig. 8). First, the Ras GTPase Bud1 must be inactivated to allow GTM release. Second, actin-independent -but not actin-dependent -vesicle delivery must be targeted upgradient to drive GTM redistribution.

The bud-positioning function of the DS must be inactivated to release the GTM
During the GTM assembly process in cells preparing to mate, we found that the Bud1 GEF became undetectable, whereas the Bud1 GAP polarized to the DS. Subsequently, the Bud1 GAP tracked with the receptor to the CS. These results suggest that Bud1 is inactivated during GTM assembly and that it remains inactive throughout tracking. Is Bud1 inactivation required for tracking to begin? Even when co-expressed with WT Bud1, which is necessary for viability, GTP-locked Bud1 conferred a severe defect in tracking: A large fraction of BUD1 G12V /BUD1 cells either continued to bud or shmooed at the DS. Taken together with our previous finding that bud1Δ cells assemble multiple GTMs at random positions on the PM -a maladaptive phenotype that occasionally results in bud1Δ cells fusing with multiple partners and forming heterokaryons rather than zygotes 20 -we conclude that Bud1 plays an important albeit transitory role in yeast mating. It is initially required to promote assembly of a single GTM at a specific point in the cell cycle (late G1) and at a specific cortical position (the DS). It must then be inactivated and remain inactive to permit tracking. In this view, the Bud1 GAP travels with the GTM to ensure that Bud1 stays off. Without this protection, stochastic activation of Bud1 could trigger positive feedback amplification of the Cdc42-Cdc24-Bem1 loop, leading to local nucleation of actin cables [8][9][10][11][12] . This would likely cause the GTM to stall before aligning with the gradient source, as we observed in BUD1 G12V /BUD1 cells.

Actin-independent vesicle delivery is essential for GTM tracking
Based on the localization of Sec3, a pioneer component of the exocyst complex, we previously concluded that vesicle delivery is involved in all three phases of gradient sensing after global internalization 20 . We showed that Sec3-RFP polarizes to the DS during GTM assembly, concentrates to the upgradient side of the GTM during tracking, and sharply peaks in the center of the GTM after stabilization. The observations reported in this paper suggest a relationship between the mode of vesicle delivery and the phase of gradient sensing: Vesicles are delivered to the PM by both actin-cable-dependent and actin-cable-independent mechanisms during GTM assembly and stabilization, but exclusively by an actin-cable-independent mechanism when the GTM is tracking. We infer that the high rate of vesicle delivery along actin cables 28 , which is needed to assemble the GTM and polarize growth at the CS, must be turned off to permit gradient tracking. In other words, there is a tradeoff between faster vesicle delivery and GTM mobility.
In principle, actin-cable-directed secretion could be the primary determinant of GTM behavior or the stability/mobility of the GTM could be the primary determinant of when and where actin cables are nucleated. We favor the latter possibility, "GTM first, cables second." Following pheromone-induced global internalization of the receptor and G protein, during the 10-15 minute GTM assembly phase, receptor and G protein gradually increase at the DS 20 . We found that AD-VD markers were detectable at the DS in about half of the cells examined and that blocking AD-VD prevented GTM assembly in about half of the cells examined as well. The simplest way to explain why some cells need F-actin and Myo2 to complete GTM assembly and others do not is variability in how much receptor and G protein remain at the DS after global internalization. Another observation that supports the "GTM first, cables second" view is that we never detected the AD-VD markers at the CS before the GTM began to stabilize.
Our conclusion that AI-VD is both necessary and sufficient to enable tracking is based on the following observations. First, AD-VD is dispensable during tracking and chemotropic mating. Second, the direct interaction between Bem1 and Exo70, which localizes Exo70 (and thus the exocyst complex) to secretion sites on the PM 27 , is critical for tracking. We Therefore, we propose that Gβγ recruits Exo70 upgradient within the GTM through their mutual interaction with Far1-Cdc24/Bem1 and thereby biases AI-VD toward the gradient source ( Fig. 8). It will be interesting to determine whether the pheromone-induced phosphorylation of Gβ enhances the affinity of Gβγ for Far1-Cdc24/Bem1, as we have recently shown that phosphorylated Gβ concentrates on the leading side of the GTM and is a directional cue 21 .
Why does the mode of vesicle secretion change as mating cells progress from GTM assembly to tracking and from tracking to stabilization? Robust polarized growth such as bud and shmoo formation requires rapid, focused, and stable vesicle delivery along actin cables. In contrast, gradient tracking requires vesicle delivery at a rate that does not result in polarized growth, as well as dynamic positioning of the secretory site in response to the pheromone gradient. AI-VD positioned by a heterotrimeric G protein whose local activity and concentration directly reflect that of the pheromone receptor is likely a faster and more flexible way to effect tracking than a mechanism that depends on the polymerization and depolymerization of actin cables.
What controls the transition from AD-VD at a fixed site during GTM assembly to AI-VDpowered GTM tracking, and back to AD-VD at a fixed site after GTM stabilization? Our results suggest that Bud1 must be inactivated to enable tracking. Because Bud1 positions Far1-Cdc24/Bem1-Cdc42 to nucleate actin cables during bud emergence, it likely plays the same role during GTM assembly. As active Bud1 disappears, the Far1-Cdc24/Bem1 complex is freed to interact with Gβγ and Exo70. In this view, Bud1 inactivation is the switch that turns off AD-VD and allows Gβγ-positioned AI-VD to predominate during the transition from assembly to tracking. Our data also indicate that AD-VD markers become detectable again only after the GTM reaches the CS and stabilizes. This observation suggests that actin cables cannot be nucleated if the polarity complexes are mobile. In this view, the positional stability of the GTM determines where and when actin cables will form, thus aligning AD-VD with the pheromone source.

Intrinsic polarity may be integral to the differentiation of many cell types
Depending on environmental conditions, haploid yeast cells choose one of three distinct fates late in the G1 phase of the cell cycle: When well nourished, they commit to mitosis and begin to polarize the growth of a daughter cell, or bud, at the DS, concomitant with the initiation of S phase; in mating mixtures, they assemble the GTM at the DS preparatory to locating and fusing with a proximal partner; when starved, they form long, chained projections called pseudohyphae, likely at the DS, which are thought to be used to forage for nutrients 38 . In addition to being determined at a unique point in the cell cycle (late G1), we infer that yeast cell fate is regulated at a unique cortical site (the DS). The importance of DS regulation during budding is well documented 39 . The work we have presented here, and previously 20 , demonstrates the essential role of DS regulation during mating. Other studies have shown that yeast cells cannot initiate pseudohyphal growth in the absence of Bud1, Bud2, or Bud5, suggesting that DS regulation is essential for cellular differentiation in response to starvation 40 . Thus, in S. cerevisiae, the specific structure generated at the intrinsically determined polarity site depends on the environmental input.
Like S. cerevisiae, most cell types in higher eukaryotes begin to differentiate when their progenitor cells complete a division cycle, and like the DS in yeast, polarity sites on the PM are associated with cell division 41 . Moreover, the basic components and systems required to generate cell polarity and direct vesicle secretion are highly conserved across the eukaryota. Historically, cells were thought to polarize in response to environmental cues as they differentiated 41 .
For example, actomyosin flows generate mechanical constraints that result in the establishment of polarity, which subsequently determines cell fate in M. musculus and C. elegans 42,43 ; during inner-outer lineage differentiation at the eight-cell stage of mice embryonic development, F-actin and polarity-related proteins such as PKC, PARs, and Ezrin are gradually translocated from the division plane to the apical cortex of the outer cells before fate determination 44 ; and, human pluripotent stem cells autonomously develop polarity before differentiating 45 . It remains to be seen whether these and other differentiating cell types use division-marked polarity site(s) to assemble protein complexes that enable them to respond to environmental cues. If so, it will be interesting to determine whether these protein complexes relocate to environmentally determined positions, how the pre-existing polarity sites are regulated, and whether different modes of vesicle delivery are involved in these processes.

MATERIALS AND METHODS
Molecular and microbiological techniques. Standard methods were used for microbial culture and molecular manipulation, performed as described previously [46][47][48] .