|Home | About | Journals | Submit | Contact Us | Français|
Transient and highly regulated elevations of cytosolic Ca2+ control a variety of cellular processes. Bulk measurements using radioactive Ca2+ and the luminescent sensor aequorin have shown that in response to pheromone, budding yeast cells undergo a rise of cytosolic Ca2+ that is mediated by two import systems composed of the Mid1-Cch1-Ecm7 protein complex and the Fig1 protein. Although this response has been widely studied, there is no treatment of Ca2+ dynamics at the single-cell level. Here, using protein calcium indicators, we show that both vegetative and pheromone-treated yeast cells exhibit discrete and asynchronous Ca2+ bursts. Most bursts reach maximal amplitude in 1–10 s, range between 7 and 30 s, and decay in a way that fits a single-exponential model. In vegetative cells, bursts are scarce but preferentially occur when cells are transitioning G1 and S phases. On pheromone presence, Ca2+ burst occurrence increases dramatically, persisting during cell growth polarization. Pheromone concentration modulates burst frequency in a mechanism that depends on Mid1, Fig1, and a third, unidentified, import system. We also show that the calcineurin-responsive transcription factor Crz1 undergoes nuclear localization bursts during the pheromone response.
Calcium (Ca2+) signals are pervasive in eukaryotic cells, where this divalent cation acts as a messenger that rapidly modifies protein electrostatic charge, shape, and function. Fast and transient elevations of free cytosolic Ca2+ levels control a wide variety of cellular processes and adaptive responses. The versatility of Ca2+ signaling systems is reflected in the very different spatial and temporal distributions that the Ca2+ concentration can display. Some cellular processes, such as Ca2+-triggered exocytosis, are executed in milliseconds within a very localized subcellular environment. Other processes, such as developmental programs and gene transcription control, require longer Ca2+ transients (minutes to hours) that, in multicellular organisms, might even be propagated throughout an entire tissue. This diversity can be captured by live imaging of Ca2+ dynamics, enabling systematic analysis of cell and tissue behavior in response to a changing environment.
In Saccharomyces cerevisiae, cytosolic Ca2+ signals are implicated in the response to several environmental challenges, including osmotic imbalance (Batiza et al., 1996 ; Denis and Cyert, 2002 ), the presence of mating pheromone (Ohsumi and Anraku, 1985 ; Iida et al., 1990 ), protein misfolding (Bonilla et al., 2002 ), high pH (Viladevall et al., 2004 ), and glucose addition under starvation (Nakajima-Shimada et al., 1991 ). There is a vast body of knowledge on the transporters, sensors, and effectors implicated in S. cerevisiae Ca2+ homeostasis (for recent reviews, see Cunningham, 2011 ; Cyert and Philpott, 2013 ). Of note, our understanding of Ca2+ dynamics in yeast relies on bulk monitoring of cellular Ca2+ levels using either radioactive 45Ca2+ or the bioluminescent sensor aequorin. Unlike research on mammalian cells, single-cell monitoring of Ca2+ signals is almost unreported in S. cerevisiae (Cunningham, 2011 ). Here we address this issue by adapting a fluorescent protein Ca2+ sensor to budding yeast and exploring single-cell Ca2+ dynamics during the pheromone response.
S. cerevisiae has two sexes or mating types, MATa and MATα. Haploid cells secrete a mating type–specific pheromone that is sensed by cells of the complementary mating type. Pheromone binding to its cognate receptor triggers a canonical mitogen-activated protein (MAP) kinase signaling cascade that leads to cell cycle arrest, transcriptional reprogramming, and polarized growth toward the mating partner. Pioneering work using 45Ca2+ showed that haploid yeast cells treated with mating pheromone take up extracellular Ca2+ in a dose-dependent manner (Ohsumi and Anraku, 1985 ). The 45Ca2+ uptake begins 30–40 min after pheromone addition and is sustained over an additional 60–80 min, being coincident with active polarized growth and formation of a cell shape called shmoo (Ohsumi and Anraku, 1985 ). Bulk Ca2+ monitoring using the bioluminescent sensor aequorin showed a similar timing in cytosolic Ca2+ level changes during the pheromone response (Nakajima-Shimada et al., 1991 ; Muller et al., 2003 ). Pheromone-induced Ca2+ uptake depends on at least two plasma membrane transporting systems, the high- and low-affinity calcium import systems (HACS and LACS, respectively). HACS depends on the simultaneous presence of three plasma membrane proteins: Cch1, a homologue of the catalytic α-subunit of voltage-gated Ca2+ channels (Fischer et al., 1997 ; Paidhungat and Garrett, 1997 ; Muller et al., 2001 ); Mid1, an integral membrane glycoprotein that physically interacts with Cch1 (Iida et al., 1994 ; Locke et al., 2000 ); and Ecm7, a predicted tetraspanning membrane protein with similarity to the catalytic γ-subunit of voltage-gated Ca2+ channels (Martin et al., 2011 ). Cch1, Mid1, and Ecm7 proteins are constitutively expressed, but HACS activity is very low in vegetative growing cells, presumably due to negative feedback regulation mediated by the Ca2+/calmodulin-activated phosphatase calcineurin (CN; Locke et al., 2000 ; Muller et al., 2001 ). LACS requires the presence of Fig1, a pheromone-induced plasma membrane protein. Like Ecm7, Fig1 has four predicted transmembrane domains with structural resemblance to the transmembrane domains of claudins and the γ-subunit of voltage-gated Ca2+ channels (Muller et al., 2003 ). Here we show that the increase of cytosolic Ca2+ during response to the pheromone is not continuous but is composed of a series of discrete Ca2+ bursts that are more frequent as pheromone concentration increases. This pheromone-dependent increase of Ca2+ burst frequency depends on HACS, LACS, and a third, unidentified Ca2+ transport system. In parallel with the cytosolic Ca2+ bursts, the CN-responsive transcription factor Crz1 undergoes nuclear localization bursts in response to pheromone.
To monitor cytosolic Ca2+ ([Ca2+]cyt) signals in S. cerevisiae, we adapted different versions of the green fluorescent protein (GFP)–calmodulin fusion protein (GCaMP) Ca2+ indicators originally developed to follow neuronal activity. We built strains bearing Ca2+ indicator genes GCaMP3 (Tian et al., 2009 ) and GCaMP6f (Chen et al., 2013 ) stably inserted in a single copy in the URA3 locus (Materials and Methods). Growth rate measurements in different conditions indicated that neither GCaMP3 nor GCaMP6f interfered with S. cerevisiae cell growth in standard culture conditions (Materials and Methods). In vivo responses for both Ca2+ reporters were evaluated by live fluorescence microscopy monitoring of cells subjected to calcium stress. Inspection of short (10–13 min) time-lapse movies revealed that cells growing in standard synthetic medium show low levels of cytosolic fluorescence and sporadically exhibit transient increases of fluorescence. These observations are consistent with previously reported bulk measurements of [Ca2+]cyt levels showing that mid logarithmic–growing cells incubated in standard medium keep low [Ca2+]cyt and low CN activity and do not actively accumulate extracellular calcium (Ohsumi and Anraku, 1985 ; Stathopoulos and Cyert, 1997 ; Muller et al., 2003 ). However, soon after the addition of calcium, numerous cells showed high cytosolic fluorescence levels, strongly suggesting that both indicators are functional in S. cerevisiae (Figure 1A). Cell segmentation of time-lapse images and quantitation of normalized fluorescence levels (F/F0; see Materials and Methods) confirmed that both GCaMP3 and GCaMP6f are responsive to Ca2+ stress in S. cerevisiae, as suspected (Figure 1B). Single-cell F/F0 traces were then used to compare the performance of both GCaMP variants in terms of registered maximum amplitudes and signal-to-noise ratios (SNRs; Figure 1, C and D). In contrast with tests in cultured neurons and animal models, in which GCaMP6f clearly outperformed GCaMP3 (Chen et al., 2013 ), the two indicators exhibited similar performances in budding yeast. Both GCaMP3 and GCaMP6f outperformed yellow Cameleon YC2.12, a Förster resonance energy transfer–based calcium sensor that exhibited very low SNR in S. cerevisiae cells (Cai et al., 2008 ). Unlike the calcium dye fura-2, which, after loading rapidly, accumulates in yeast vacuoles (Iida et al., 1990 ), GCaMP3 and GCaMP6f are stable cytosolic proteins enabling long-term monitoring of [Ca2+]cyt dynamics. Given that GCaMP3 and GCaMP6f behaved similarly, we decided to use GCaMP6f for the experiments because it has been further optimized for fast Ca2+ imaging (Chen et al., 2013 ).
Having successfully adapted two GCaMP indicators to S. cerevisiae, we proceeded to characterize single-cell [Ca2+]cyt dynamics during the pheromone response. Aequorin luminescence monitoring of [Ca2+]cyt indicated that, after a lag of 40 min, pheromone-treated cells experienced a slow increase of [Ca2+]cyt that peaked after 30 min and then needed an additional ~60 min to decline to basal levels (Muller et al., 2003 ). Because calcium-stress results indicated that [Ca2+]cyt fluctuations were on the order of seconds to minutes (Figure 1), we first imaged pheromone-treated cells at high speeds (e.g., 5 frames/s). MATa GCaMP6f cells were grown to mid log phase and treated with α-factor, and then, at different time points (10–80 min after pheromone addition), cells were registered over 10-min periods. Again, we observed that control cells had very low fluorescence and that a few cells exhibited transient increases in the fluorescence signal. We also observed that those cells that were imaged within the first 30 min of α-factor treatment exhibited low fluorescence levels, which was consistent with the previously reported 40-min lag before [Ca2+]cyt increase. In contrast, movies taken after 30 min of α-factor treatment revealed very active [Ca2+]cyt dynamics: yeast cells underwent fast and transient increases, or bursts, of [Ca2+]cyt (Figure 2A). Data analysis showed that pheromone-induced [Ca2+]cyt bursts are heterogeneous and sometimes overlapping and have different morphologies and lifespans (Figure 2, A and B). Most bursts have lifespans <30 s (70% of 90 registered bursts) and exhibit abrupt increase, reaching maximum levels in <4s (Figure 2, B and C). We also observed that the majority of [Ca2+]cyt bursts (70% of 225 registered bursts) exhibit a decay behavior that fits with a single-exponential function (Supplemental Figure S1). Because the decay times derived from these fits are more than one order of magnitude larger than the reported unbinding time of Ca2+ from GcaMP6f (Chen et al., 2013 ), we propose that the observed fluorescence decays faithfully reflect the underlying reduction of the free cytosolic Ca2+ concentration. Thus the observed single-exponential decay suggests that the cytosolic calcium concentrations reached during the pheromone response do not saturate the prevalent cytosolic calcium–removing systems.
We next addressed how mitotically active cells experience [Ca2+]cyt dynamics and how the dose of pheromone affects this during cell growth polarization. For this, we cultured MATa GCaMP6f bar1Δ cells to mid log phase, treated them with different concentrations of α-factor, and then monitored them over 100 min. We chose this imaging range because, in our experimental conditions, it was long enough to accommodate a whole cell cycle in nontreated cells and shmoo formation in pheromone-treated cells (Supplemental Movie S2). Because we did not detect [Ca2+]cyt bursts with lifespans <7 s (Figure 2B), we used a 0.4 frame/s sampling rate to avoid phototoxicity without losing signal detection capacity. We quantified two basic features of [Ca2+]cyt bursts—amplitude (F/F0) and duration—and also monitored the number of bursts per cell over 100 min (frequency; Figure 3). We found that 40% of mitotically active cells transit a complete cell cycle without experiencing noticeable changes in [Ca2+]cyt. The remaining 60% of cells experienced at least one burst during the 100-min registration (Figure 3A). This raised the question of whether [Ca2+]cyt bursts were characteristic of a specific phase of the cell cycle. Morphological categorization of cells at the moment of burst occurrence indicated that mitotically active yeast cells experience bursts at any point of their cell cycle. However, when we quantified the number of bursts per cell cycle phase and compared the resulting distribution with the average time the monitored cells invest in each phase, we realized that G1 phase bursts were clearly overrepresented (Figure 3B). A more detailed probabilistic analysis indicated that this observed bias cannot be accounted for by a model in which calcium bursts occur at equal rates regardless of the cell phase and that they preferentially arise when cells are transitioning either G1 or S phase (Materials and Methods).
Burst morphologies of mitotically active cells were characterized by low amplitudes (F/F0 ≤ 1 in 92% of total burst population; Supplemental Figure S3) and short lifespans (<20 s in 87% of total burst population; Supplemental Figure S3). These results indicated that although [Ca2+]cyt is maintained at low levels during normal cell growth in standard conditions, there are discrete [Ca2+]cyt bursts during the cell cycle. These observed bursts might be implicated in fine-tuning of different cell cycle regulators that are CN effectors, such as Hcm1 (Arsenault et al., 2015 ), Elm1 (Goldman et al., 2014 ), and Rga2 (Ly and Cyert, 2017).
In pheromone-treated cells, we observed that an α-factor concentration as low as 1 nM triggers a small but noticeable increase in [Ca2+]cyt burst frequency, whereas an intermediate dose (5 nM), near the α-factor receptor Ste2 dissociation constant, is sufficient to produce a fivefold increase in the average number of burst occurrences (Figure 3C). High pheromone concentrations (10, 100 nM) elicit a maximum response in which 88% of the cells undergo between 5 and 15 bursts. Within the registered time frame (100 min), we did not find a periodic occurrence of bursts, but we did observe a lag phase (see later discussion). Unlike this clear increase in frequency, [Ca2+]cyt burst morphologies suffered minor changes all over the dose–response curve. Burst amplitude distributions moved toward higher values with α-factor addition. For example, 23% of bursts in nontreated cells had F/F0 > 1, whereas this percentage increased to 40% in 100 nM α-factor–treated cells (Supplemental Figure S3). Similarly, α-factor addition triggered the appearance of more bursts with longer lifespans: whereas 26% of bursts in nontreated cells lasted >20 s, in 100 nM α-factor–treated cells, this proportion rose to 45% (Supplemental Figure S3). However, for both parameters, we did not observe a clear dose–response behavior as we did for burst frequencies (compare insets in Figure 3C and Supplemental Figure S3). Integration of [Ca2+]cyt traces of 100 nM α-factor–treated cells confirmed our observations of a 30-min lag phase before burst frequency increase (Figure 3D) and are consistent with bulk measurements using radioactive calcium or aequorin (Ohsumi and Anraku, 1985 ; Muller et al., 2001 ). In accordance with those authors’ observations, we found that [Ca2+]cyt burst frequency increase is coincident with shmoo formation active growth polarization at all α-factor concentrations tested (Figure 3D).
Frequency modulation (FM) of Ca2+ bursts is widely used by mammalian cells to encode different input signals, which are then decoded by signaling molecules and transcription factors to execute appropriate cellular responses (Dolmetsch et al., 1998 ; Kupzig et al., 2005 ; Thurley et al., 2014 ). The α-factor–burst frequency dose–response curve (Figure 3C, inset) matches the equilibrium binding curve of α-factor to Ste2 and the α-factor–transcriptional induction dose–response curve, indicative of the phenomenon known as dose–response alignment, common to many other signaling systems (Yi et al., 2003 ; Colman-Lerner et al., 2005 ). FM of intracellular signals has been reported in budding yeast, in which different stresses modulate the frequency of nuclear translocation of transcription factors Msn2 (Jacquet et al., 2003 ; Petrenko et al., 2013 ; Lin et al., 2015 ), Mig1 (Lin et al., 2015 ), and CN-responsive Crz1 (Cai et al., 2008 ). Remarkably, we observed that during the pheromone response, Crz1 exhibits bursts of nuclear localization (Supplemental Figure S3). We do not know whether these observed Crz1 bursts correlate with [Ca2+]cyt bursts.
In the absence of stress, Crz1 locates in the cytosol, but high extracellular Ca2+ stress triggers a series of unsynchronized Crz1 nuclear localization bursts (Cai et al., 2008 ). Extracellular Ca2+ concentration modulates the frequency of Crz1 nuclear localization bursts, which, in turn, promote transcriptional bursts of Crz1-dependent downstream genes. FM of Crz1 nuclear localization enables coordinated control of multiple target genes, keeping their relative expression constant over a wide dynamic range of stressor input (Cai et al., 2008 ). A similar scenario is proposed for stress-induced Msn2 nuclear localization bursts, for which FM elicits a more robust transcriptional response than does amplitude modulation (Hao and O’Shea, 2012 ; Petrenko et al., 2013 ). It is assumed that during mating, the strength of pheromone signaling increases up to the culminating point of cell–cell fusion (Brizzio et al., 1996 ). Thus, as mating progresses, frequency-modulated regulation of Ca2+ bursts might result in coordinated control of Ca2+ effectors in both partners. Yeast cells down-regulate the pheromone signal over time, resuming mitotic growth when a partner is not found, even in the presence of pheromone (Moore, 1984 ). Multiple negative feedback mechanisms prevent sustained pheromone activation because it causes cell death (Zhang et al., 2006 ). The Ca2+/CN signaling down-regulates the pheromone response by inhibiting Ste12-dependent gene expression and promoting internalization of the α-factor receptor Ste2 (Cyert et al., 1991 ; Alvaro et al., 2014 ; Goldman et al., 2014 ). Ly and Cyert (2017) found that at high pheromone concentrations, CN dephosphorylates Rga2, leading to Cdc42 signaling reduction and pheromone response down-regulation. Thus pheromone-induced modification of Ca2+ burst frequencies might help yeast cells to evaluate and decide whether to quit or pursue the mating program.
We next asked how the known calcium import systems HACS and LACS affect amplitude, duration, and frequency of pheromone-induced [Ca2+]cyt bursts. We observed that in the absence of extracellular calcium, pheromone treatment is unable to induce an increase in [Ca2+]cyt burst frequency (Supplemental Figure S3), indicating that this response is mainly caused by extracellular calcium import and not by release from intracellular stores. To address this question, we monitored MATa mutant cells impaired for HACS (mid1Δ), LACS (fig1Δ), or both (mid1Δ fig1Δ). [Ca2+]cyt traces of mitotically active mutant cells exhibited no differences from those of wild type growing in the same conditions: most cells remained with low basal levels of GCaMP6f fluorescence and exhibited few low-amplitude bursts (Figure 4, Supplemental Figure S4, and Supplemental Table S5). On α-factor treatment, however, the three mutants exhibited a statistically significant increase in [Ca2+]cyt burst frequency (Figure 4, A–C, and Supplemental Table S5) but a very limited one compared with wild-type cells (Figure 4D). These results indicate that in response to pheromone, both influx systems are active. Moreover, analyzing the cumulative distributions of [Ca2+]cyt burst frequencies, we found no differences between single and double mutants, suggesting that in our tested conditions, HACS and LACS function nonadditively (Figure 4D and Supplemental Table S5). The fact that the mid1Δ fig1Δ double mutant is able to increase [Ca2+]cyt burst frequency also suggests that either there is a third import system, like the proposed glucose-induced calcium flux system (Groppi et al., 2011 ), or there is a compensatory release from internal calcium stores. How does pheromone response trigger [Ca2+]cyt bursts? Although the mechanisms of HACS or LACS activation have not been defined, it has been suggested that Cch1 is a mechanosensitive channel (Kanzaki et al., 1999 ). Thus cell growth polarization and subsequent plasma membrane stretch may activate Cch1.
Statistical analysis of the cumulative distributions of [Ca2+]cyt burst amplitudes and lifespans showed that in both tested conditions, mid1Δ cells underwent bursts with higher amplitudes than did wild-type, fig1Δ, and fig1Δmid1Δ cells (Figure 5, A and B, and Supplemental Table S5). In contrast, lower amplitudes characterized fig1Δ cell bursts in the absence of pheromone (Figure 5B). Like mid1Δ cells, fig1Δ mid1Δ double mutants showed bursts but with higher amplitudes in response to pheromone (Figure 5A). Although burst lifespans seem to be different for vegetative growing fig1Δ and mid1Δ cells (Figure 5D), the Kolmogorov–Smirnov (KS) test does not reject the hypothesis that lifespans of all strains belong to the same distribution (Supplemental Table S5). On pheromone treatment, mid1Δ cells showed bursts with higher lifespans, whereas no differences were detected for the other three strains according to the KS test (Figure 5C and Supplemental Table S5). In short, these results indicated that HACS-impaired cells (mid1Δ) experienced bursts with higher amplitudes and lifespans in response to α-factor. Of interest, whereas Fig1 localizes at the shmoo tip during growth polarization, Cch1 is broadly distributed throughout the plasma membrane and is excluded from the shmoo tip (Yoshimura et al., 2004 ; Aguilar et al., 2007 ). Therefore it is expected that LACS- and HACS-mediated Ca2+ entries are spatially segregated, perhaps reflecting signaling over different subsets of Ca2+ effectors.
Live monitoring of [Ca2+]cyt in fungi at the single-cell level has been hampered by the lack of sensitive, stable, and high-SNR sensors. Our results indicate that GCaMP sensors can be used to obtain detailed information on Ca2+ dynamics in S. cerevisiae. GCaMP sensors are being used in filamentous fungi (Nick Read, personal communication), where it is well established that Ca2+ signaling is required for hyphal growth and pathogenicity (Munoz et al., 2015 ). In Candida albicans, Ca2+ uptake is required for tropic growth of hyphae (Brand et al., 2007 ). It has been suggested that localized Ca2+ uptake may provide positional information needed to define the site for polarized growth via a mechanism that requires Cch1, Cdc42, and its guanine exchange factor, Cdc24 (Brand et al., 2014 ). Monitoring Ca2+ dynamics in single cells might reveal the presence of cytosolic Ca2+ gradients in different fungi species and their relevance for normal and pathogenic physiology.
GPD1 promoter was obtained from PYM-N14 (Janke et al., 2004 ) and amplified by PCR using primers BamH1GPDp-F (CGT AGG ATC CGA GCT CAG TTT ATC ATT) and Xho1GPDp-R (GTC ACT CGA GTC TAG AAT CCG TCG AAA) adding BamHI and XhoI sites and cloned into plasmid pRS306N from the EUROSCARF collection (Taxis and Knop, 2006 ). This plasmid was digested with SacI, and the 988–base pair fragment containing GPD1-promoter followed by ADH1-terminator was ligated into SacI-digested pRS306K, an integrative S. cerevisiae plasmid, at URA3 loci with the dominant marker kanMX4 (Taxis and Knop, 2006 ). The resulting vector was called pRS306K-GPD1p-ADH1t-a.
To generate the final vector, pCMV-GCaMP3 and pCMV-GCaMP6f mammalian expression vectors obtained from AddGene (Cambridge MA) were used as templates to perform RFcloning reactions (van den Ent and Lowe, 2006 ) designed to precisely insert GCaMP3 or GCaMP6f open reading frames in GPD1p-ADH1t in the yeast integrative vector pRS306K-GPD1p-ADH1t-a. These final vectors were verified by sequencing and subsequently linearized for yeast transformation. The Crz1-GFP tagging strain was generated by PCR and homologous recombination using the plasmid pFA6a-link-yEGFP-CaURA3 (Sheff and Thorn, 2004 ).
A complete list of strains used in each experiment is given in Supplemental Table S5. All strains were in the W303 background (leu2-3, 112 trp1-1 can1-100 ura3-1 ade2-1 his3-11, 15). Cells were grown in either rich medium YPD (1% yeast extract, 2% Bacto-peptone, and 2% glucose) or synthetic defined medium (SD) containing 0.67% yeast nitrogen base without amino acids (YNB; Difco), 2% glucose, and the appropriate drop-out mixture of amino acids according to supplier’s instructions (CSM; Sunrise Science Products, San Diego, CA). Synthetic growth media lacking calcium were prepared similarly using calcium-free yeast nitrogen base reagents (YNB-CaCl2; Sunrise Science Products). In general, yeast strains were grown in 10 ml of rich medium YPD to an OD600 = 0.8–1. For all microscopy experiments, yeast cells were cultured for a minimum of 18 h in SD and supplemented with 80 ng/ml adenine (Sigma- Aldrich, St. Louis, MO) at 30°C to OD = 0.2–0.4. For imaging, 200 μl of cell culture was incubated in concanavalin-A (Sigma-Aldrich)–coated chambers for 10 min (#1.5 glass; MatTek, Ashland, MA). Finally, SD medium was replaced by a fresh one, and cells were imaged. For calcium-free treatments, cells were grown overnight in SD and then transferred to chambers and washed three times with SD-Ca2+ before use. The α-factor mating pheromone was obtained from Sigma-Aldrich.
Time-lapse imaging was performed on a Leica DMI6000,TCS-SP5 with a resonant scanner (8000-Hz scanning frequency) using an HCX PL APO 63/1.4-060 CS oil objective and equipped with an automatically programmable XY stage (Mark & Find; Leica) in an environment chamber preequilibrated to 30°C. Images were acquired using an argon ion laser at wavelength of 488 nm with 30% laser power, a pinhole set to 3 airy units, photomultiplier tube range of 495–589 nm, 784 gain, and –0.2 offset. Image formats of 512 × 512– and 282 × 282–nm pixel size were used for all acquired movies. For calcium stress, the sampling rate was 1 frame/10 s. For high-speed imaging of pheromone-treated cells, it was 5 frames/s, and for long-term monitoring, it was 0.4 frames/s. Measurements of mitotic cell growth and cell growth polarization during the pheromone response indicated that all defined imaging conditions were innocuous for both processes (unpublished data).
For short-term (10–13 min) movies, cells were manually segmented. For long-term (100 min) movies, single cells were automatically segmented from transmitted light images using CellStat (Kvarnstrom et al., 2008 ). For mitotically active cell analysis, cells were randomly selected using CellStat, and only those cells that were segmented throughout the whole time lapse were kept for further analysis. Because we used asynchronous cultures, for α-factor treatment experiments, after CellStat segmentation, we selected cells that at t = 0 were morphologically compatible with G1 phase (unbudded cells). In all cases, we excluded from further analysis selected cells that evidenced loss of integrity upon movie inspection.
Connected regions on the image Al(tk) corresponding to single cells (each one identified by the subscript l) were determined as a function of time tk using segmented images.
We computed the normalized calcium signal for each cell l at time tk as
where g(xi, yj, tk) is the registered fluorescence at time tk from the pixels located at (xi, yj) and Np(Al(tk)) is the number of pixels in the region Al(tk) of the image associated with the l th cell at time tk. We computed the background noise for each cell BGI(tk) as
where is an empty region contiguous to the analyzed cell. The number of pixels is similar to the number of pixels of Np(Al(tk)). We then subtracted the background noise BGI(tk) from the normalized fluorescence to define
To correct for photobleaching, we took the sequence for each cell , with T the duration of the experiment, and extracted from it the values hl(tk) associated with easily recognizable fluorescence bursts. We then fitted this pruned sequence by an exponential:
Finally, we defined a corrected expression for photobleaching-normalized fluorescence as
To account for differences in GCaMP expression or cell size between cells, which result in different fluorescence basal levels, we computed the basal level associated with each cell as
where N* is the total number of ti* times of the sequence without easily recognizable fluorescence bursts. We finally computed the increase in the corrected-for-photobleaching normalized fluorescence for each cell as a function of time as
We assumed that this quantity was a good reporter of the changes in cytoplasmic Ca2+ and can be compared between cells despite their differences in dye load or size. It is thus the quantity that we used for the analysis.
To set a criterion that distinguishes Ca2+ bursts from spurious (random) changes in basal fluorescence, we applied the following iterative procedure. The procedure first takes, for each cell l, the pruned sequence with the times ti* as defined before. It then computes for each time ti* of the sequence the mean µi,l and SD σi,l of over the 250 times ti* preceding ti*:
From the element 251 to ti*, the protocol sweeps over the 250 elements of the sequence before ti* and computes . If this ratio is >3 for some ti*, the corresponding element is excluded from the pruned sequence. The newly excluded time is included in a new set, tburst. The procedure is repeated, iteratively sweeping back and forth (using the 250 times that follow ti*) in time until no more elements departing from the mean by >3 SDs are found. The procedure is applied subsequently to all the times ti* in the latest pruned sequence. Finally, we look for sets of at least three consecutive times ti* in tburst and associate the corresponding elements with a Ca2+ burst.
Because newly born cells spend more time in G1 phase than mother cells and because complete registration of cell cycles of mother cells was more frequent in our 100-min time-lapse movies, we restricted this analysis to mother cells (114 cells over three independent experiments) that experienced one or more Ca2+ bursts. To check how meaningful it was to pool the data from the three experiments together, we compared the distributions of phase duration obtained for each experiment. A KS test did not reject the hypothesis that the data from any two of the experiments were drawn from the same distribution (confidence level of 0.05). Therefore we decided to pool the data coming from the 114 cells together. In principle, calcium bursts can occur at any time along each cell phase. Thus, to distinguish situations with different numbers of bursts during the time course of one cycle, we calculated probability densities per unit time. To this end, we coarse-grained the data set using the time duration dt of the longest typical Ca2+ burst (~22 s). This duration was much smaller than the mean times ~500, 1750, and 2000 s that each cell invests in each of the three phases that we considered for the analysis—G1, S, and G2/M, respectively; it was also much smaller than the mean interburst time interval (~2050 s). We then looked at the time course of the fluorescence, F/F0(t), of each selected cell and assigned a 1 or a 0, respectively, to each time interval of duration dt if F/F0(t) either showed or did not show a burst in the corresponding time interval. Given the relationship between dt and the mean interburst time, we expected each interval of size dt to contain at most one burst. It may occur that a burst starts during a particular interval and ends during the following one. In such cases, we assigned a 1 to the interval that contained most of the burst duration and a 0 to the other one. If the burst was equally distributed between the two consecutive intervals, we assigned the 1 to either one of them at random. In this way, we digitized the data set, and the computation of the probability of burst occurrence was simply done by counting bursts (i.e., differences in burst duration were not considered for the computation). Thus we calculated the probability that a burst occurs during any time interval of duration dt as
where T is the duration of the experiment, Nc is the number of analyzed cells (mothers with at least one Ca2+ burst that are visualized during the whole experimental time course), and Nb is the total number of bursts displayed by the whole set of analyzed cells during the experiment.
In the same way that we assigned 0’s and 1’s to each analyzed cell at a particular time interval of duration dt, we assigned a symbol G1, S, or G2/M, depending on which of these three phases the cell was in. We identified each phase according to morphological characterization of transmitted-light time-lapse movies as follows: unbudded cells were categorized as G1-phase cells. The S phase was defined as the lapse between the first evidence of a nascent bud and the point where the bud/mother cell diameter ratio was <0.5. The G2/M phase was defined as the lapse between the end of S phase and the point where the common bud–mother cell longitudinal axis was broken, evidencing physical separation of daughters from mother cells. When any of these criteria was not easily distinguished, the cell was removed from the analysis. We then computed the probability that a cell is at a given phase Pph (ph = G1, S, G2/M) as the mean over all selected cells of the fraction of time each selected cell spends at the given phase:
where the subscript l identifies any of the Nc analyzed cells and Nph, l is the number of time intervals of duration dt during which the l th cell remains in the phase ph. We assumed that Pph is also the (mean) fraction of time a cell spends at phase ph during each cell cycle:
where λph is the mean duration of phase ph during a cell cycle and λ is the mean cell cycle duration. We computed the duration of each phase for each cell following each selected cell during one cycle. We then computed λph as the mean over these phase durations. We computed the joint probability that, during a time interval of duration dt a cell is at a given phase and displays a burst as
where Nb, ph, l is the total number of burst displayed by the l th cell while being at the phase ph. Clearly, the total number of bursts displayed by the l th cell, Nb, l, satisfies
where the sum is performed over the three phases. We computed the conditional probability that a cell is at a given phase ph at a particular time interval of duration dt, given that at that time interval the cell displays a burst, as
This is the quantity that we compared against Pph in the text.
We computed the conditional probability that a cell has a burst during a time interval of duration dt, given that it is at a particular phase ph during that time interval, as
Now each cell l spends a total time tph,l in each phase ph given by
during the time course of the experiment, so that, on average, the cell spends a time
in phase ph during the experiment. The duration of each of the three performed independent experiments was slightly larger than the mean cell cycle duration λ. Thus we relate the mean phase duration λph with the mean time tph a cell spends in phase ph during the time course of the experiment by
Inserting this expression in Eq. (16), we obtain
from which we compute the probability per unit time that a burst occurs while the cell is at a given phase ph as
These are the rates that are reported in Figure 3A.
We tested whether the observed bias of burst occurrence was compatible with a model in which calcium bursts occur at equal rates regardless of the cell phase. For this, we performed a chi-square goodness-of-fit test, considering the null hypothesis that the mean number of bursts during a phase was solely proportional to the mean phase duration. Pooling the data from three independent experiments (total number of bursts, 168), a comparison with the mean number of bursts per phase observed experimentally gave a chi-square value (8.92) that allows us to reject the null hypothesis at the 0.02 significance level. We then conclude that calcium bursts occur at higher rates when cells are transitioning either the G1 or S phase.
We thank Agustina Olivera-Couto for advice on microscopy, Tabaré de los Campos for help with preliminary image analysis, and Alberto Muñoz and Nick Read for invaluable advice on using GCaMP sensors. We thank Martha Cyert and Nina Ly for helpful discussions and critical reading of the manuscript. This work was supported by an ANII-Caldeyro Barcia Fellowship, Uruguay (PDNAC201217826 to N.C.), the International Centre for Genetic Engineering and Biotechnology (Grant CRP/URU11-01 to P.S.A.), the Fondo de Convergencia Estructural Mercosur (Grant COF 03/11 to the Institut Pasteur de Montevideo), the Secretaría de Investigación Universidad de Buenos Aires (Grant 20020130100480BA to S.P.D.), and the Fondo para la Investigación Científica y Tecnológica, Argentina (Grants PICT 2013-1301 to S.P.D. and PICT-2014-3034 to P.S.A.). N.T. is a fellow from Consejo Nacional de Investigaciones Científicas y Tecnológicas, Argentina.
This article was published online ahead of print in MBoC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E16-07-0481) on December 28, 2016.