|Home | About | Journals | Submit | Contact Us | Français|
Cell pairing is central for many processes, including immune defense, neuronal connection, hyphal fusion or sexual reproduction. How does a cell orient towards a partner, especially when faced with multiple choices? Fission yeast Schizosaccharomyces pombe P- and M-cells, which respectively express P- and M-factor pheromones [1, 2], pair during the mating process induced by nitrogen starvation. Engagement of pheromone receptors Map3 and Mam2 [3, 4] with their cognate pheromone ligands leads to activation of the Gα-protein Gpa1 to signal sexual differentiation [3, 5, 6]. Prior to cell pairing, the Cdc42 GTPase, a central regulator of cell polarization, forms dynamic zones of activity at the cell periphery at distinct locations over time . Here, we show that Cdc42-GTP polarization sites contain the M-factor transporter Mam1, the general secretion machinery, which underlies P-factor secretion, and Gpa1, suggesting these are sub-cellular zones of pheromone secretion and signaling. Zone lifetimes scale with pheromone concentration. Computational simulations of pair formation through a fluctuating zone show that the combination of local pheromone release and sensing, short pheromone decay length, and pheromone-dependent zone stabilization leads to efficient pair formation. Consistently, pairing efficiency is reduced in absence of the P-factor protease. Similarly, zone stabilization at reduced pheromone levels, which occurs in absence of the predicted GTPase-activating protein for Ras, leads to reduction in pairing efficiency. We propose that efficient cell pairing relies on a fluctuating local signal emission and perception, which become locked into place through stimulation.
Previous modeling work of pheromone-dependent polarized growth in S. cerevisiae assumed that the cell serves as spherically uniform source of pheromone [8–12]. As the secretion machinery is normally polarized by Cdc42-GTP [13–16], an alternative likely scenario is that the pheromones are released locally at sites of polarization. We used live-cell imaging to probe the possible co-localization of components of the pheromone signaling machinery with dynamic Cdc42-GTP zones during ‘exploration’, i.e. after the last cell division upon nitrogen starvation but prior to polarized growth (also known as ‘shmoo’ formation; see  for review). To label Cdc42-GTP, we used tagged Scd2, a protein that links Cdc42 with its major guanine nucleotide exchange factor, and robustly co-localizes with Cdc42-GTP [7, 18, 19].
The M-factor pheromone is a lipid-modified peptide, exported outside the cell by a dedicated transporter, Mam1 (Figure 1A) [2, 20, 21]. Mam1-GFP signal was weak, but displayed local enrichment at cortical sites often coinciding with Scd2 zones (68 of 74 cells; Figure 1B, S1A–B), in addition to significant internal signal, likely due to endocytic recycling similar to its S. cerevisiae Ste6 homologue [22, 23]. This suggests M-factor is exported not around the entire cell cortex but locally, preferentially at sites of Cdc42 activity. As P-factor is a simple 23aa peptide, processed in the ER and Golgi, and likely secreted through the canonical secretory system (Figure 1A) , we monitored the localization of the secretion machinery by labeling the post-Golgi vesicle-associated Rab11-family GTPase Ypt3 [24, 25]. GFP-Ypt3 showed strong enrichment at Scd2 zones (67 of 75 cells; Figure 1C, S1C–D), similar to our previous description of both Myo52 myosin motor and exocyst complex at these zones . These data indicate that both pheromones are preferentially released at Cdc42-GTP zones.
Engaged pheromone receptors signal through the associated Gα-protein Gpa1 . Because Gpa1 is predicted to be N-terminally myristoylated, we tagged it with mCherry at an internal poorly conserved site, generating Gpa1-mCherrySW, integrated as sole copy at the endogenous gpa1 genomic locus. These cells are fertile, although they exhibit reduced mating efficiencies (30% of cell pairing, n=928 cells), indicating Gpa1-mCherrySW is largely, but not completely functional. Gpa1-mCherrySW fluorescence was weak, but could be detected at dynamic sites at the cell periphery, which often co-localized with Scd2 (45 of 63 cells; Figure 1D, S1E–F). By contrast, pheromone receptors were not exclusively associated with Scd2 zones, displaying instead a broad localization over the entire plasma membrane during exploration, as well as strong internal, likely endomembrane, localization (Figure 1E). This broad peripheral localization is consistent with the ability of cells to perceive a partner and extend a shmoo from any location . During shmoo growth, receptors then became enriched at the shmoo site, as has been previously described in this and other species [26, 27] (Figure 1F). Because pheromone receptors are initially broadly distributed at the membrane, but their associated Gα is enriched at specific sites, we interpret these sites of Gpa1 enrichment as sites of pheromone receptor engagement. The mechanisms of Gpa1 accumulation await future dissection. In summary, these results indicate that release and perception of pheromones occur at discrete cortical sites largely coincident with the dynamic zones of Cdc42 activity observed prior to cell pairing.
To probe the logic behind the process of cell-cell pairing, we designed a simple numerical simulation mimicking an experimental two-dimensional field of cells of opposite mating types in a mating reaction. We implemented Cdc42-GTP zone (simply referred to as “zone” below) dynamics in each cell with a positional bias around cell poles, as measured experimentally . The simulation is based on two simple assumptions. The first is that a zone lifetime scales with the local opposite-type pheromone concentration (Figure 2A). This assumption is validated by experimental observations: Heterothallic M-cells lacking the P-factor protease Sxa2 exposed to homogeneous low-level P-factor exhibit dynamic zones, with only high-level P-factor promoting polarized growth at a single site . Furthermore, zone lifetimes increase with pheromone concentration ranging from 0.01 to 1μM (Figure 2B). We note that S. cerevisiae polarity site motility is also constrained by higher pheromone levels [28–30]. The second assumption is that cells paired (i.e. with zones facing each other) for > 100 min have grown and fused together, and are thus taken out of the reaction (Figure 2C). Indeed, previous examination had shown that cells do not polarize towards fused pairs and readily re-orient if the target cell fuses with another partner . The simulation relies on two parameters. The first, b, defines the linear response of the zone lifetime to the opposite type pheromone. For simplicity, we assume that this value is identical in the two distinct mating types. The second, λ, is the characteristic decay length of the pheromone signal [8, 9], which likely results from the action of secreted proteases that restrict it near the source [31–33].
Starting from a field of exploring cells (Figure 2D), the simulations lead to a pattern of cell pairing (Figure 2E) that varied in each realization due to the randomness of the exploration process (Movie S1). To monitor the success of the simulations, we quantified the fraction of paired cells over time and found the value of parameter b that gives the most efficient cell pairing for a given decay length λ (Figure 2F). The selected optimal value of b separates a high sensitivity regime where exploratory zones stabilize along unproductive orientations and a low sensitivity regime where cells spend most time exploring (Figure S2A–C; Movies S2 and S3).
In agreement with the local enrichment of pheromone secretion machineries at Cdc42-GTP zones, we found that local release of pheromones at the polarized patch (model 1) yielded more efficient pairing than global pheromone release (model 2) (Figure 2F). Consistently, low pheromone decay length allowed more efficient cell pairing (Figure 2F). Both models make the assumption that the pheromone signal is perceived at the site of the polarized patch, in agreement with local Gpa1 enrichment. As expected, decoupling pheromone perception from the polarized patch in the simulation (model 3) nearly blocked cell pairing (Figure 2F). Thus, the optimal conditions for efficient cell pairing in the simulation are local pheromone release and perception, as observed above, and pheromone decay length of order 1 μm or less. These results remained valid when considering configurations of cells that do not touch (Figure S2D–E), nonlinear response to pheromone concentration (Figure S2F–G) and cells entering the exploratory phase over an extended time interval (Figure S2H–I). We also built a more detailed model accounting for the excluded cell volume in pheromone distributions. This detailed model recapitulated the more efficient cell pairing observed upon local pheromone release and pheromone decay length of order 1μm or less as in model 1 above (see supplementary text; Figure S2J–L). Because this model was very computationally expensive, we used the simple model below.
To compare simulations to experiments, nitrogen-starved P and M cells were placed in a monolayer on a thin agarose pad at a density similar to that used in the simulations above (about 19’000 cells per mm2) and the fraction of paired cells was monitored over time. Up to 60% of cells became engaged in a pair and fused to form a zygote within 15h of starvation, with similar kinetics to that observed in the simulation with local pheromone release (Figure 2G–H).
To better define the efficiency of pairing, we focused on isolated small groups of cells, in which cells of distinct mating types were labeled with distinct fluorescence (Figure 3A). This small group analysis is interesting, because the low number of cells permits exact counting of all possible outcomes, including the optimal pairing configurations with the largest number of pairs. Remarkably, 77% of the cells mated, close to the maximum predictable efficiency of 83% that would have been obtained if the maximum number of pairs had been formed in each configuration (Figure 3A–B, where a different configuration could have yielded one more pair in each of experimental panels f and i). In silico arrangements of cells in small groups of identical distribution as those analyzed experimentally revealed comparable kinetics and pairing efficiency of 72 ± 3 % (mean ± stdev), with fluctuations due to the different partner choices occurring on each run of the simulation with local pheromone release (Figure 3A–B, where a different configuration could have yielded one more pair in each of simulation panels f and d; Movie S4). By contrast, we calculated that if each cell in the groups of Figure 3 made a single random choice among its possible partners, only 52 ± 2 % of the cells would mate, with several cells making an unreciprocated choice. If the cells with unreciprocated choices are allowed to make a second, third, or more choices, this increases the fraction of successful pairs to 74 ± 1.4 %, close to the realized value. Thus, the large fraction of pairs in simulations is realized through transient cell engagement followed by either simulated fusion or bond breakage. The same process can also be seen in time lapse images of Scd2 zones of mating cells (Figure S3; ).
In conclusion, though we cannot exclude that pheromone concentration may also bias the site of zone assembly, our simulations suggest that efficient cell pairing is achieved through local pheromone release and sensing at dynamic zones that become stabilized by increased pheromone levels.
One prediction of the model is that optimal pairing occurs for pheromone decay lengths of order 1 μm or less. To test this prediction, we examined the ability of sxa2Δ M-cells to pair with wildtype P-cells. Sxa2 is a secreted protease that cleaves P-factor, and thus likely contributes to shaping the pheromone landscape around cells [31–33], similar to the proposed role of Bar1 protease in S. cerevisiae [8, 10]. In our 2D assay, the pairing efficiency of h− sxa2Δ x h+ wt crosses was reduced (Figure 4A, S4D), though interestingly not as dramatically as when assessed in 3D on agar plates ( and data not shown). Importantly, increasing the pheromone decay length of only a single partner, as in the wt x sxa2Δ crosses, also led to reduced pairing efficiency in simulations (Figure 4B).
The stronger phenotype of sxa2Δ mutants observed in 3D is in agreement with the idea that shaping of the pheromone gradients by the protease is particularly important when a large number of close partners each produce P-factor. To partly mimic this situation in 2D and further test the importance of pheromone concentration for cell-cell pairing, we added excess synthetic P-factor to fields of wt x wt or sxa2Δ x wt crosses, with the aim to further “confuse” the cell as to the origin of the pheromone gradient [34, 35]. Addition of 10μg/ml synthetic P-factor led to a partial decrease in pairing efficiency in wt crosses, but to a strong impairment in sxa2Δ x wt crosses, in which the excess P-factor cannot be degraded, and resulted in the majority of cells growing unproductive, unpaired projections (Figure 4A, S4D). In simulations, we mimicked the predicted increased zone lifetime due to the homogeneous higher P-factor concentration by increasing the overall patch lifetime, τ0, in M cells (while keeping bτ0 constant to maintain the same level of sensitivity; see supplemental text). This led to a decrease in pairing efficiency, which appeared additive rather than synergistic with the effect of increased P-factor decay length (Figure 4B). This is likely because the fate of the added synthetic pheromone is different in wildtype, where it is progressively degraded, than in sxa2Δ mutants, where its concentration remains high throughout the experiment. Additional increase of τ0 in the simulation indeed further lowered pairing efficiency, better mimicking the wt x sxa2Δ experimental situation (Figure 4B). We conclude that experiments and simulations are in agreement that short pheromone decay lengths are critical for efficient cell pairing.
We then sought to experimentally modify the internal cell mechanisms determining zone lifetime and its dependence to pheromone concentration, thus varying the equivalent of parameters τ0 and b in the model. The molecular mechanisms underlying oscillations of Cdc42-GTP zones in response to pheromone are not well defined, but the small GTPase Ras1 is a likely important factor . Indeed, Ras activity requires and in turn promotes pheromone signaling [36–38]. Ras was also proposed to be a positive regulator of Scd1, the major guanine nucleotide exchange factor activating Cdc42 . Consistent with this idea, we found that ras1Δ cells failed to recruit Scd1 to the cell cortex during mating and formed broad zones of Scd2 polarization that did not dynamically explore the cell cortex (Figure S4A–B). N-terminally tagged GFP-Ras1 also accumulated at sites of Scd2 dynamic exploration, though it was also present at other cortical regions (Figure S4C). Thus, Ras1 is a likely positive regulator of local Cdc42 activation during exploratory polarization.
A zone lifetime is likely modulated by negative signals promoting zone disassembly. Indeed, deletion of Gap1, the predicted GTPase activating protein (GAP) for Ras1 [39, 40], caused an important increase in zone lifetime: heterothallic gap1Δ M-cells lacking the P-factor protease Sxa2 exposed to homogeneous P-factor (0.01 to 1μM) displayed dynamic Scd2 zones, but these exhibited significantly longer lifetimes than gap1+ controls exposed to the same pheromone concentrations (Figure 4C). Some of these cells also lysed, as previously reported . Thus, Ras inactivation promotes polarity zone disassembly.
The gap1Δ mutant represents a condition in which cells exhibit an apparent increase in the parameters τ0 and b. As predicted in simulations with increased τ0 and b for one cell type, wt x gap1Δ formed pairs with significantly reduced efficiency, with gap1Δ cells stabilizing a growth axis at unreciprocated locations (Figure 4D–E, S4D). The stronger phenotype observed experimentally may stem from the formation of these unpaired projections, as well as from the lysis of some gap1Δ cells. Indeed, modification of the simulation to remove cells with zones stable for >200min better mimicked the experimental situation (Figure 4E). Thus, reducing the pheromone concentration-dependency for zone stabilization prevents transient engagements and locks polarity zones in inappropriate locations. We conclude that an optimal zone lifetime, stabilized only by high concentrations of pheromone, is required for efficient cell-cell pairing.
We propose that local pheromone signal release and perception underlies flexible cell-cell communication for efficient pair formation. Our experimental data indicate that exploratory polarity sites represent discrete zones of localized pheromone signal release and sensing, and our computational approach demonstrates that fluctuating local signal emission and perception, which becomes locked into place through stimulation, serves for optimal pairing of cells during yeast mating. The fluctuating nature of the zones of signal release and perception is critical to this process, as demonstrated by the inefficient pairing of gap1Δ cells, in which polarity zones are stabilized at significantly lower pheromone concentrations. The short decay length of the pheromone gradient highlighted in the simulations is consistent with the short range of fission yeast cell mating and validated by the lower pairing efficiency of sxa2Δ cells. We note that we observed similar reduction in pair formation with P-cells lacking Sxa1, a predicted M-factor protease . A prior calculation suggested that a very high concentration of diffuse proteases would be needed to restrict the pheromone concentration profile to scales of order one cell diameter . While some proteases may not be freely diffusing, we note that an increase in nonlinear sensitivity to pheromone concentration (parameter n) gives a similar trend to that of decreasing pheromone decay length (compare Figure 3B to Figure S2F, G). Thus nonlinear sensitivity to pheromone concentration could be an additional mechanism that cells use to effectively achieve local sensing. We also note that we assumed spatially uniform protease activity to highlight the effects of local pheromone secretion and decay length, however non-uniform protease activity over scales comparable to the cell size may also contribute to the mating kinetics .
The pheromone gradient sensing mechanism proposed here has features of sensing by temporal averaging , since exploratory zones localize in regions of higher pheromone concentration on average. More than just contributing to gradient sensing, fluctuations in the position of the zone also lead to different transient cell engagements. This mechanism of partner switching that would not be possible through local zone wandering [28–30] allows the cell population to test different conflicting mating configurations (reminiscent of the configurations of frustrated physical systems with quenched disorder ). This non-deterministic conceptual pairing framework may be valid beyond yeast sex, for instance for the formation of connections in filamentous fungal mycelia  or for activity-dependent stabilization of neuronal connections .
S. pombe strains used are listed in Table S1. Standard S. pombe media and genetic manipulations were used . For all mating experiment cells were grown in MSL ± N (minimal sporulation medium with or without nitrogen), as described [7, 46]. For construction of gpa1-mCherrysw-kanMX4 a mCherry fusion was inserted after amino acid S132 at endogenous genomic locus. GFP-Ypt3 was expressed from plasmid (pREP41-GFP-Ypt3) under the control of nmt41 promoter, as described . Further details are provided in the supplementary file.
Images in Figures 2, ,3,3, ,4,4, S3 and S4A, B, D were acquired on a DeltaVision epifluorescence system. The DeltaVision platform (Applied Precision) was composed of a customized Olympus IX-71 inverted microscope and a Plan Apo 60×/1.42 NA (for DIC) or a U-Plan Apo 100×/1.4 NA oil objective (for fluorescence), a CoolSNAP HQ2 camera (Photometrics), and an Insight SSI 7 color combined unit illuminator. To limit photobleaching, images were captured by OAI (optical axis integration).
Peripheral kymographs were constructed in ImageJ v1.46 by drawing a 3 pixel-wide line around the cell cortex.
All imaging was performed at room temperature (20–22°C). Figures were assembled with Adobe Photoshop CS5 and Adobe Illustrator CS5.
This work was supported by a Swiss National Science Foundation (SNF) research grant (31003A_155944) and a Marie Curie Initial Training Network (FUNGIBRAIN) to SGM, NIH grant R01GM098430 to DV, and an SNF International Short Visit grant to SGM and DV. DH was supported by the National Science Foundation Research Experience for Undergraduates site grant PHY-1359195 at Lehigh University.
Author ContributionsLM, FOB, VV and SGM performed experiments and analyzed data. BK, DH and DV developed and performed numerical simulations. SGM and DV wrote the manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.