Cross-regulation between CDK and MAPK control cellular fate

Commitment to a new cell cycle is controlled by a number of cellular signals. Mitogen-Activated Protein Kinase pathways, which transduce multiple extracellular cues, have been shown to be interconnected with the cell cycle. Using budding yeast as a model system, we have quantified in hundreds of live single cells the interplay between the MAPK regulating the mating response and the Cyclin-Dependent Kinase controlling cell cycle progression. Different patterns of MAPK activity dynamics could be identified by clustering cells based on their CDK activity, denoting the tight relationship between these two cellular signals. In mating mixtures, we have verified that the interplay between CDK and MAPK activities allows cells to select their fate, preventing them from being blocked in an undesirable cellular program.

Clotet and Posas, 2007; Strickfaden et al., 2007). In this study, we were interested in the 19 interplay between the cell cycle and the mating pathway. 20 Haploid budding yeasts exist in two mating types: MATa and MATa. They produce pheromones 21 (respectively a-factor and a-factor) that can be sensed by a mating partner. Activation of a G-  In order to guarantee that each cell that undergoes fusion possesses a single copy of its genome, 28 active Fus3 phosphorylates the Cyclin Kinase Inhibitor (CKI) Far1 which arrests the cells in G1 29 (Peter et al., 1993;Peter and Herskowitz, 1994). In addition, during division, signaling in the 30 mating pathway is dampened by the action of the Cyclin Dependent Kinase (CDK) Cdc28. 31 Cdc28, the only CDK in S. cerevisiae, associates with the different cyclins to ensure the proper 32 progression through the cell cycle. Inhibition of the mating pathway is made possible by the 33 association between the CDK and the late-G1 cyclins Cln1 and Cln2. The Cln1/2-Cdc28 34 complex has been shown to phosphorylate the scaffold protein Ste5, thereby preventing its 35 recruitment to the plasma membrane and thereby preventing the transduction of the signal from 36 the receptor to the MAPK cascade (Strickfaden et al., 2007). In this study, we have developed a sensitive assay enabling to quantify in parallel MAPK and 46 CDK activity in non-synchronized live single cells using fluorescent biosensors. By exploiting 47 the natural diversity present in the population, we have been able to cluster cells based on their 48 cell cycle position and monitored their MAPK activity pattern. We could confirm the key role of 49 Far1 for the G1 arrest. However, our data suggest that an additional mechanism working in 50 parallel with the Ste5 phosphorylation is required to limit signaling during S-phase. Furthermore, 51 we highlight the importance of the cross-inhibition between MAPK and CDK for cell-fate 52 decision in the mating process. The interplay between these two activities will determine whether 53 cells induce a mating response or commit to a new cell cycle round. 54 7 cluster was identified with cells that cycle briefly in G1 during the time lapse (Through-G1, 3%). 151 No specific behavior in MAPK activity was observed in these cells even when traces are aligned 152 relative to the time of G1 entry (Supplementary Figure 4). However, because CDK plays a major role in controlling the signaling output of the mating 157 pathway, clustering based on the CDK activity pattern allows to group together cells that display 158 similar MAPK activity dynamics. Thus, we observe four types of behaviors: strong and sustained 159 activity for G1 cells, weak and sustained activity in Out-of-G1 cells, strong but transient activity 160 in the G1-exit cluster and delayed activation (depending on the timing of CDK activity drop) 161 present in the G1-entry cells. 162

163
In order to compare the results obtained with the SKARS, we performed a similar analysis with 164 two different assays that report on mating pathway activity. Using a fluorescently tagged Kss1 165 (Kss1-mScarlett), the relocation of the MAPK from the nucleus to the cytoplasm upon 166 pheromone stimulus has been monitored in a strain carrying Whi5-mCitrine. This change in 167 cellular compartments of Kss1 has been shown to be a consequence of the disassembly of a 168 complex formed between Dig1/Dig2, Ste12 and Kss1, upon phosphorylation by Fus3 and/or 169  Figure 6D), compared to the sharper dynamics of the SKARS reporter in the 174 same cell cycle stage ( Figure 2D). In addition, when comparing the Kss1 relocation of G1 and 175 G1-entry cells relative to the time of pheromone addition (Supplementary Figure 6E), we see that 176 they are almost identical. This suggests that cells late in the division process have fully recovered 177 their signaling ability. 178 The second assay is based on the dPSTR (dynamic Protein Synthesis Translocation Reporter) al., 2016). This promoter drives the expression of a small peptide that promotes the relocation of 181 a fluorescent protein in the nucleus of the cell. We use the previously published pAGA1-dPSTR 182 to monitor the dynamics of mating gene induction. AGA1 has been shown to be strongly induced 183 by a-factor in a MAPK dependent manner (Roy et al., 1991;Oehlen et al., 1996;Aymoz et al., 184 2018). Clustering of the gene expression data based on cell-cycle stage demonstrates a strong 185 expression in G1 and G1-entry clusters, while clusters for Out-of-G1 and G1-exit cells display an 186 attenuated and delayed response (Supplementary Figure 7). 187 The global outcome that can be obtained by comparing these three types of reporters is that full 188 signal competence is observed in the G1 cluster, while Out-of-G1 cells have a reduced ability to 189 signal. In addition, cells exiting G1 will only transiently activate the MAPK. Interestingly, this 190 transient activation is not sufficient to drive gene expression as protein production in the G1-exit 191 cluster is delayed by 20 to 30 minutes. 192

193
The comparison between the SKARS, the Kss1 relocation and the pAGA1-dPSTR also 194 highlights a discrepancy for the G1-entry cluster. While the two latter assays display a signaling 195 ability that is comparable to the one present in G1 cells for the G1-entry cluster, the SKARS 196 measurements suggest that the MAPK activity is only recovered when CDK activity has 197 dropped, as cells enter in G1. 198 The influence of the cell cycle on the nuclear enrichment of the SKARS could potentially 199 explain this observation. Indeed, a lower enrichment of the corrector can be observed in Out-of-200 G1 cells compared to G1 cells (Supplementary Figure 8A and B). In the G1-entry cluster, this 201 results in a slow transition from a low to a high nuclear to cytoplasmic ratio (N/C) of the 202 corrector around the time of entry into G1. In signaling dead cells, the exact same behavior is 203 observed for the functional sensor (Supplementary Figure 1B). However, the dynamic of this 204 transition is strikingly different from the dynamic of MAPK activity measured upon G1-entry. 205 While the corrector slowly rises from -10 to +10 min, the MAPK activity shift takes place 206 between +5 and +10 min after the CDK activity drop. Thus, the fact that these two events are not 207 synchronous, strongly suggests that the sharp increase in MAPK activity is not an artifact from a 208 reduced sensitivity during the division of the cells.  Figure 8D  212 and E). If the G1-entry behavior was an artifact due to a poor enrichment of the sensor, we would 213 expect that cells with a relatively high nuclear enrichment would not display this behavior or at 214 least to a lower extent. On the contrary, we observe that G1-entry cells which keep a high N/C of 215 the corrector throughout the time-lapse display a stronger change in MAPK activity 10 minutes 216 after G1 entry. 217 In agreement with the Kss1 and dPSTR assays, previous works (Oehlen and Cross, 1994; less, it remains difficult to estimate their relative influence on the signaling outcome. Our data 237 suggest that at low a-factor concentrations, negative feedbacks are prevalent and contribute to feedbacks stabilize the system in a high activity state for a long time. Similar experiments were 240 performed in BAR1+ cells, where the presence of the pheromone protease adds another layer of 241 regulation to the system and leads to a faster decline in signaling activity at all concentrations but 242 the highest one, which remains sustained over the course of the time lapse experiment 243 (Supplementary Figure 9). 244 It is well-established that treatment with pheromone prevents cells from entering a new cell cycle 245 round (Hartwell et al., 1974), therefore one expects a difference in the proportion of G1 cells at 246 the end of the experiment between a-factor treated and mock treated cells ( Figure 3D).  We have shown above that the efficiency of the cell cycle arrest depends on the level of MAPK 291 activity. We next want to verify how the CDK activity influence the MAPK signal transduction. 292 In budding yeast, a single Cyclin Dependent Kinase, Cdc28, associates with the different cyclins 293 throughout the cell cycle to orchestrate the division of the cell. While Cdc28 is an essential 294 protein, it is none-the-less possible to acutely inhibit its kinase activity by using an analog 295 sensitive allele (cdc28-as (Bishop et al., 2000)). The inhibitor NAPP1 was added to the cells 6 296 minutes after the pheromone stimulus. Upon NAPP1 treatment, Out-of-G1 cells show a rapid MAPK activity, that is low during division, increases to a level comparable to the one observed 299 in cells stimulated in G1 ( Figure 5A). 300 Additionally, this experiment allows to verify that the transient MAPK activation observed in the 301 G1-exit cluster is shaped by the rising CDK activity. Cells with an increasing CDK activity after 302 pheromone stimulus were clustered. In the DMSO control experiment, this sub-population 303 displays the expected fast activation of the MAPK upon a-factor addition followed by a decay to 304 basal activity as CDK activity rises. In the inhibitor treated cells, the MAPK activity decay is 305 blocked upon NAPP1 addition and the MAPK signal rises to reach full activity ( Figure 5B). 306 These experiments demonstrate that the activity of Cdc28 directly and quickly regulates the level 307 of MAPK activity present in the cell. 308

309
One mechanism of inhibition of the mating pathway by the CDK has been shown to consist in Cln1/2 on Ste5 has been mutated (Bhaduri and Pryciak, 2011)) ( Figure 6A). We monitored the 318 response of the mating pathway for these three alleles with the SKARS, the relocation of Kss1 319 and the pAGA1-dPSTR at 10 and 100 nM a-factor ( Figure 6B, C and D). The wild-type STE5 320 behaved similarly to the WT parental strains, demonstrating the functionality of the 321 complementation. As expected, signaling activity in G1 cells is minimally influenced by either 322 mutation in Ste5 because the CDK is inactive. In the G1-exit cluster, the inhibition following the 323 transient activation of the pathway is less pronounced in the two Ste5 mutants than for the wild-324 type allele. Generally, the behavior is more pronounced at 10nM than at 100nM a-factor and the 325 Ste58A mutant displays a weaker inhibition by the CDK than the Ste5CND mutant. 326 However, our data suggest that an additional mechanism might contribute to limit the signal 329 transduction of the mating pathway in the early stage of the division process. This phenomenon 330 is best observed at low concentrations of pheromone where the activity of the MAPK cascade is 331 weaker. This behavior has not been detected previously, probably because most experiments 332 have been performed at saturating levels of pheromone. However, in mating mixture where 333 pheromone concentration is low, this mechanism could contribute to the cell fate decision. In 334 order to identify other potential targets of the CDK, we have tested various alleles of Ste20 335 because it has often been suggested as a potential target for this regulation (Oehlen and Cross, shown to regulate this process. Our data indicate that other mechanisms, that remain to be 346 identified, contribute to the repression of MAPK activity in dividing cells. 347

MAPK activity in mating 348
After studying the response of cells stimulated by exogenous pheromone, we next wanted to 349 understand how the cell cycle and the mating pathway were regulated during mating and how 350 this cross-inhibition allowed an efficient cell-fate selection in these physiological conditions. In 351 order to achieve this, we have imaged the MATa cells bearing the SKARS, the corrector and the 352 Whi5 marker in presence of a MATa partner expressing constitutively a cytosolic CFP at high 353 levels. In Figure 7A, thumbnails of such an experiment are displayed. The fusion events (arrow 354 heads) can be detected by observing a sudden increase in CFP signal in the MATa cells. In the 355 frame preceding the two fusion events displayed in Figure 7A, we observe that the MATa cells of The dynamic measurements of the mating process allowed us to monitor how cells reach this 359 state. Using our automated image analysis pipeline, we have detected more than one hundred 360 fusion events and curated them manually to remove any artifacts due to a mis-segmentation of 361 the cells. The single cell traces of these events have been computationally aligned relative to the 362 fusion time. Time zero corresponds to the last frame before the increase in CFP intensity, 363 because it is the last time point where MAPK activity can be reliably quantified ( Figure 7B). As 364 a reference, the median MAPK activity of cells imaged under the same conditions but in absence 365 of a mating partner is plotted. In the fusing cell, a gradual increase in MAPK activity starts 40 to 366 60 minutes prior to fusion. In parallel, we observe that the median CDK activity is low for all 367 time-points. However, the 75 percentile stabilizes to a low value 40 minutes prior to fusion, 368 denoting that a fraction of the population enters in G1 within the hour preceding the fusion. 369 On the CDK side, Cdc28 kinase activity is directly regulating the signal flow in the MAPK 406 pathway. Blocking Cdc28 activity relieves this inhibition allowing recovering full MAPK 407 activity, within minutes after addition of the chemical inhibitor ( Figure 5). This suggests a very 408 direct mechanism of action on the mating pathway. The primary candidate for this process has 409 been Ste5. However, our quantitative measurements speak for an additional mechanism 410 detectable mostly at low pheromone concentrations. Other potential candidate targets of Cdc28 411 could be the G-protein and the receptor, or other proteins in the MAPK signal transduction 412

cascades. 413
The experiments performed with mating mixtures also illustrate that the interplay between the 414

NLS-NLS-CFP into the second MCS of pED92 (AatII-SphI). Plasmids were transformed in yeast 449
of W303 background expressing the Hta2-iRFP (yED136) or the Hta2-tdiRFP (yED152). Whi5 450 was tagged with mCitrine using pGTT-mCitrine plasmid . Kss1 was tagged 451 with mScarlet (pGTL-mScarlet). pAGA1 induction was monitored by integrating the pAGA1-452 dPSTR R in the URA3 locus (Aymoz et al., 2018). For each cell, the average nuclear intensity in the fluorescent channel corresponding to the 508 SKARS, the Corrector and Whi5-mCitrine were divided by the average intensity in the 509 cytoplasm for every time point (ratio N/C). To quantify the MAPK activity, the Adjusted ratio 510 for each individual cell was obtained by dividing the ratio N/C of the Corrector by the ratio N/C 511 of the SKARS. The CDK activity of each cell was defined as the inverse of the Whi5-mCitrine 512 N/C ratio. The initial CDK activity was defined as the average of three time-points before the 513 stimulus (3-6 minutes before a-factor stimulus). All these quantities are unitless numbers, since 514 they are ratios between fluorescence intensities obtained from the microscope camera. We used 515 the symbol [-] to represent this lack of units. 516

517
When plotting a histogram using the Whi5 ratio N/C (Nuclear intensity over Cytoplamsic 518 intensity) of all cells at all time points, we identified two populations. To properly distinguish 519 between ratios corresponding to nuclear Whi5 versus ratios corresponding to cytoplasmic Whi5, 520 we first separated ratio values using a fixed threshold (threshold=2). However, this value 521 overestimated the limit of what could be considered as nuclear Whi5. We then identified cells 522 with ratios below this threshold at all time points and replotted a histogram of all Whi5 ratio N/C 523 values from this sub-population. This method enables to enrich the sample with low N/C ratio. 524 This second histogram presented a clearer overview of the values corresponding to cytoplasmic 525 Whi5. We then calculated the derivative of the histogram and identified the position of fifth-526 lowest derivative value. The corresponding Whi5 ratio N/C was used as a final threshold to 527 separate nuclear to cytoplasmic Whi5 values. This procedure enables to correct for slight 528 differences between experiments mostly due to the chosen experimental settings (96 well-plate 529 or pad experiments). Globally, the procedure adjusted the threshold between 2 and 1.7. If the 530 Whi5 ratio N/C was above the threshold, corresponding to a low CDK activity value (C/N) and 531 thus the cell was considered in G1. On the opposite, values below the threshold indicate that the 532 cell was Out-of-G1. G1 cells sub-population was defined by single cell traces which remained 533 above the threshold for the entire duration of the time lapse. Conversely, Out-of-G1 cells cluster 534 was defined by traces that stay below this threshold. The other single cell traces are scanned for 535 G1-entry and G1-exit events. If a pattern of two values below the threshold followed by two 536 values above the threshold was found, we considered that the cell belonged to the G1-entry 537 cluster. The same strategy was used to identify the G1-exit sub-population, using a pattern of two 538 values above the threshold followed by two values below the threshold. Cells in the G1-entry 539 cluster were defined as having a single event of G1-entry. Cells in the G1-exit cluster were 540 selected as having only one event of G1-exit. Finally, the Through-G1 cluster is made of cells 541 that display a G1-entry event followed by a G1-exit event. Because the duration of the time-lapse 542 experiments is much shorter than the cell cycle of the yeast (45 min vs. 90 min), we considered 543 that the number of events cannot exceed two. Cells which do not follow this criterion were 544 The authors have declared that no competing interests exist. 573          Ste5 + ?