Cell size control in yeast

Abstract

Cell size is an important adaptive trait that influences nearly all aspects of cellular physiology. Despite extensive characterization of the cell cycle regulatory network, the molecular mechanismscoupling growth to division, and thereby controlling cell size, have remained elusive. Recent workin yeast has reinvigorated the size control field and suggested provocative mechanisms forthe distinct functions of setting and sensing cell size. Further examination of size sensing models based on spatial gradients and molecular titration, coupled with elucidation of the pathways responsible for nutrient-modulated target size, may reveal the fundamental principles of eukaryotic cell size control.

Introduction

There is an intimate and often complex relationship between form and function in living organisms[1]. A simple and fundamental aspect of organismal form is size, which is an important determinant of cellular physiology. In unicellular organisms, adjusting the surface to volume ratio may be an adaptation to environmental conditions where surface transport is limiting. Surface-to-volume optimization islikely exhibited in the extensive fossil record of foraminifera, a group of unicellular marine protists. 60% of the variation in mean foraminifera size over the past 400 million years is accounted for by a linear model based solely on atmospheric oxygen concentration[2, 3]. Thus, cell size can be an important selective trait for survival in changing, nutrient-limited environments.

Cell size also affects internal cellular architecture. Not only are the volumes of various organelles proportional to that of the cell[4], DNA content in eukaryotic cells scales linearly with cell size over nearly a million-fold range[5-7]. This is true within as well as between species: diploid yeast cells are about twice as large as haploids [8-10].This constant ratio of DNA to cytoplasm suggests that cell size can adapt to support evolution of DNA content. Conversely, DNA content could adapt to accommodate physiologically-driven changes in cell size. Examples of the latter may be found in the yeast lineage, where a minimal genome could be viewed as an adaptation to small cellular size. With volumes as low as 10’s of femtoliters, yeast are among the smallest eukaryotes and contain relatively few genes, as well as small intergenic regions.For example, whereas metazoan cis-regulatory elements are found thousands of basepairs away from the transcription start site, yeast elements are limited to ~800 basepairs upstream [11]. Thus, evolutionary pressures on cell size may influence mechanisms of transcriptional regulation via selection to maintain the appropriate DNA-to-cytoplasm ratio.

Consistent with the physiological importance of cell size, there is much evidence suggesting that cells have evolved molecular mechanisms for bothmonitoring and controlling size[12]. Cellsof a given species typically vary little about their mean size [13], and this consistency requirestight co-regulation of cellular growth and division [14]. Single-cell studies have provided the most convincing evidence of cell size control. In these studies, variation in cell size at birth is harnessed to infer the presence or absence of cell size control: If size control is present, then cells that are smaller at birthwill grow proportionally more than larger cells in the subsequent cell division cycle[15-17].

This co-regulation of growth and division has been studied in a variety of organisms, but the yeasts Saccharomycescerevisiae and Schizosaccharomycespombe(Figure 1) have been favored model organisms for several reasons:a potentially simpler regulatory architecture, powerful systematic tools for genetic analysis, and simpler cell geometry[18]. More fundamentally, the regulation of growth and division is particularly tractable in yeast. In metazoans, both division and growth are subject to complex extrinsic signaling andmultifarious mutual regulation.In yeast, on the other hand, preventing growth arrests the cell cycle, butgrowth proceeds apace when proliferation is halted[15, 19]. Thus, yeast control size primarily by regulating division in response to growth.This requires detection of the nutrient concentrations that determine growth rate, growth rate itself, or cell size. Here, we review our current understanding of size control mechanisms in yeast, highlighting recent work and open questions.

Figure 1. Yeast models for cell size control.

S. pombe and S. cerevisiae are the preeminent model organisms for cell size control studies. Interestingly, S. cerevisiae and A. gossypii are more closely related than either is to S. pombe. Despite their divergent morphologies, these two yeasts are regulated by similar cell cycle control networks governing the DNA to cytoplasm ratio.

Size and Growth

“the form of an organism is determined by its rate of growth in various directions; hence rate of growth deserves to be studied as a necessary preliminary to the theoretical study of form” D.W. Thompson, On Growth and Form, 1942, P.79 [1].

Size is the most fundamental aspect of cellular form, and the basis of cell size control is the coupling of growth and division.The growth function f, relates a cell’s current size, V, and cell cycle phase, ϕ, to the rate of size increase, $\frac{dV}{dt}$, so that $\frac{dV}{dt}=f(V,\varphi )$. This relationship defines the requirements of cell size control. More specifically, whether cell growth is linear or exponential has been hotly debated and has important implications for size control[18]. For dV linear growth, cells increase their size at a constant rate C, so that $\frac{dV}{dt}=C$. In this case, specifying the period of the cell cycle specifies mean cell size without requiring a direct link between size and division, because the progeny of large or small cells asymptotically approach the target size over several generations[20]. Linearly growing cells could therefore maintain an average size simply by adjusting the length of the cell cycle.On the other hand, for exponential growth, the rate of growth is proportional to cell size, so that $\frac{dV}{dt}=\alpha V$. In this case, constant cell cycle periods allow growth to a constant multiple of birth size, so that slight fluctuations are not correctedin succeeding generations[21].Therefore, exponentially growing cells must vary cell cycle duration to damp fluctuations. This is accomplished through size-dependent cell cycle progression.

Cell size and growth may be measured by a variety of methods, summarized in Table 1. In budding yeast, an exponential model of growth is supported by both single cell analysis using time lapse microscopy and bulk experiments using radioactive labeling[17, 22, 23].Although it has recently been shown that various cell cycle arrests differentially shift growth rate [24], budding yeast adhere closely to exponential growth in an unperturbed cell cycle. The growth function for fission yeast, however, deviates from exponential (see [18] for a comprehensive review).Fission yeast growth includes two linear segments separated by a rate change point (RCP), followed by a period directly preceding division during which the cell does not increase its length. Despite this difference in growth function, both budding and fission yeasts rely on size control mechanisms to couple growth and division (Figure 2). This may occur because, even though active size controls are not required in linearly growing cells, they may providea more efficient means of controlling size than relying on agrowth rate dependent cell cycle duration.

Table 1.

Methods for measuring cell size

method	measured quantity	measurement principle	advantages/ disadvantages	references
flow cytometry
resistive-pulse sensing (coulter counter)	volume	change in resistance due to volume displacement	fast and accurate/does not track individual cells	[25, 152, 153]
fluorescent flow cytometry and fluorescence activated cell sorting (FACS)	volume and shape	light scatter	high-throughput, can be combined with different fluorescent labels / indirect measurement of size, does not track individual cells	[154, 155]
	protein content	total fluorescence after FITC staining
imaging
bright field	area	volume estimate from area of focal plane	Single cell time course capability, /difficult to accurately identify cell border	[156]
fluorescent proteins	protein content	fluorescence of constitutively expressed fluorophores	Single cell time course capability/ indirect measurement	[17]
interferometry	dry mass	phase shifts caused by biomolecules biomolecules	Single cell time course capability, dry mass determination with subcellular resolution / complicated analysis	[157-159]
microchannelresonators	buoyant mass	Changes in mass affect resonance frequency	Single-cell time course capability/ low throughput, specialized equipment needed	[23, 49]

Figure 2. Size-dependent cell cycle progression in S. pombe and S. cerevisiae.

A. S. pombe cells enter G2 at different sizes following S-phase. They grow in a bilinear fashion and enter mitosis upon reaching a threshold size, so that smaller cells spend more time in G2 than larger cells, as indicated. B.S. cerevisiae daughter cells are born at different sizes and grow exponentially. Smaller cells spend more time in G1 prior to Start than larger cells (as indicated), which partially compensates for initial size variation. Thus, size control, a function of nutrient conditions and growth rate, is exerted at G2-M in S. pombe and within G1 in S. cerevisiae.

Additionally, exponential and linear growth functions different have implications for the effects of cell-to-cell variability and cell size mutants on fitness. Because the instantaneous rate of growth is proportional to cell size during exponential growth, mass doubling time is independent of cell size.Therefore, size control mutations shifting cell size, but not affecting metabolic processes, are not expected to affect population doubling time in exponentially growing cells. This is the case in budding yeast, where size mutants (e.g., whi5Δ) have WT population doubling times [25]. Similarly, although fluctuations may push sizes of individual cells significantly below or above the population mean, this is not expected to affect fitness for exponentially growing cells. Consistent with this argument, budding yeast exhibit much larger size variability than fission yeast, which does not exhibit exponential growth. For haploid cells, the coefficient of variation (standard deviation / mean; CV) of S. pombe cell size at fission is ~0.06, while the CV of S. cerevisiae cell size at budding is ~0.17 [16, 17, 26]. Yet, the noise in budding yeast is still bounded, which may reflect the limits over which growth is exponential. If cells get too large, then a single genome will be unable to support exponential growth of the cytoplasm[27]. Thus, we expect the degree of noise tolerated to be related to the range over which growth is exponential.

In addition to influencing the selective pressures on size control systems, growth functions may themselves be selected for. In particular, the growth functions of fission and budding yeasts may reflect the physiological requirements of symmetric and asymmetric division patterns. Upon nutrient limitation, budding yeast will produce daughter cells less than 20% of the mother cell size [15]. This asymmetric division may select for growth functions that are efficient over a larger range in cell sizes, such as exponential growth. In turn, efficient growth over a large size rangelessensthe pressure tohaveprecisesize control. Therefore, we expect to see the degree of cell-to-cell variation tolerated in size control increase with the degree of asymmetry in division size, illustrating the potential interdependence of the growth function, division pattern, and size control.

Perfect and imperfect size thresholds

Despite different growth functions, size control in budding and fission yeasts is broadly similar. As indicated in early physiological studies,both yeasts implement continuous monitoring of a size-dependent signaland restrict cell cycle progression at specific stages in a size-dependent manner. In S. pombe, evidence for a size requirement for division came from experiments that tracked the growth and division of cells following synchronization by S-phase arrest [19]. Cells continued to lengthen during the arrest, so that they became much larger than normal cells. Upon release, the progeny of these cells returned to normal size after only a few division cycles, because each cycle took only about 60% of the normal time, preventing mass doubling before division. Cell division cycles returned to their normal durationonce the cells had regained their normal size distribution. These shortened cycles indicated that the abnormally large arrested cells had surpassed a size threshold, manifested primarily at the G2-M transition [28, 29]. Later work uncovered an additional size requirement at the G1-S transition, which is revealed by adverse growth conditions or mutations reducing cell size[30-34]. This secondary size checkpoint has been little-studied, so we do not discuss it here.

In S. cerevisiae, early single-cell studies showed that small cells spend a longer time in G1, which allows them to grow more than initially larger cells[15]. These experiments suggesteda size threshold at the G1-S transition inS. cerevisiae, rather than at the G2-M transition, as in S. pombe. Additionally, the asymmetric division pattern of budding yeast has implications for size control. Mother cells fulfilled the size requirement when they first became mothers and bud at regular intervals[35].Interestingly, even small mothers do not exhibit size control and proceed through G1 rapidly, compared to similarly-sized daughter cells [36].This difference in mother-daughter size control programs was later traced to the asymmetric distribution of the daughter-specific transcription factors Ace2 and Ash1,which regulate transcription of the G1 cyclinCLN3[37-41]. Thus, S. cerevisiae size control acts primarily on the G1-S transition in daughter cells.

Similar to the secondary size control at G1-S in fission yeast, there is evidence for additional size control during budding yeast S/G2/M. This was initially suggested by the finding that halting bud growth by depolymerizing actin induces a G2/M delay in cells with buds smaller than a threshold size[42]. This effect depends on homologues of the key regulatory proteins for mitotic entry in fission yeast (see below).A subsequent study challenged the causal role of sizeby showing that stopping bud growth through disruption of vesicular transportdoes not cause a G2/M delay [43]. This study suggested that actin depolymerization engages the morphogenesis checkpoint [44], and that the apparent size dependence is in fact due to differences in G2/M progression. Nevertheless, single-cell studies show a weak dependence of G2/M duration on size at budding (S. DiTalia, personal communication), suggesting the need for further investigation.

It is important to note that, wherever it occurs in the cell cycle, size control can be more complex than a simple threshold. Typically, cell cycle control has been viewed in terms of ‘sizers’ and ‘timers’. Sizers require that cells pass a size threshold, while timers require that cells wait a fixed amount of time independent of size[45].However, these concepts are insufficient to describe all size-dependent cell cycle progression. The degree of size control can be inferred by plotting a metric of cell growth within a given interval versus the size of the cell upon entering the interval (Figure 3A). If the slope of a curve fit to the single cell data is −1, the amount of growth in the interval will exactly compensate the difference in initial sizes. If the slope is 0, growth within the interval is uncorrelated with the entrance size and there is no size-dependent control, i.e., a timer. Imperfect size control will yield intermediate values. Haploid fission yeast cells exhibit a perfect sizer[16], whereas earlyfrog embryos exhibit a perfect timer[46]. Budding yeast G1-S control has significant but imperfect size control, as exhibited by a slope of −0.7 (Figure 3B) [17]. Thus, eukaryotic cells exhibit distinct degrees of cell size control.

Figure 3. Single-cell size control assay as applied to S. cerevisiae.

A.Size control leads to a negative correlation between cell size at birth and growth in G1. Live-cell imaging techniques allow cells to be tracked from birth to budding, which is concomitant with DNA synthesis.B. Plotting the logarithm of size at birth vs. relative growth in G1 quantifies the efficiency of G1 size control. A slope of −1 would indicate perfect size control, whereas as slope of 0 indicates a lack of size control. Wild-type S.cerevisiae exhibit imperfect but significant size control, as indicated by the −0.7 slope relating birth size to growth in G1.C. Deletion of the cell cycle control gene WHI5 greatly diminishes the efficiency of size control, as indicated by the reduced slope of −0.3. Data adapted from ref. [17].

The intermediate degree of size control exhibited by budding yeast correlates with imperfect size compensation through a single cell cycle. In other words, it will typically take more than one cycle to damp size fluctuations. Additionally, the imperfect size control means that the ‘size threshold’ for the G1-S transition is itself size-dependent and that, on average, cells that are smaller when they enter G1 will bud at smaller sizes than their larger brethren. This cannot be accounted for by a simple combination of sizer and timer modules. A full account of the molecular mechanism underlying budding yeast size control will be required to understand itscomplex size-dependence. Nevertheless, “shifting the size threshold” up or down may be interpreted as shifting the size control curve (Figure 3) right or left to generate larger or smaller cells at budding, respectively, for the same input birth size. Although common in the literature, G1 progression kinetics in budding yeast are imperfectly characterized by a single size-threshold and require a curve for accurate description.

Size-dependent cell cycle progression suggests that cells somehow sense their own size. A discussion of how cells sense their own size requires first a clarification of what we mean by size. At its most general, “size” conflates cellular mass and volume, which are often correlated. This correlation is imperfect, however, as cell density varies through the cell cycle significantly inS. pombe, and more subtly inS. cerevisiae[47-49]. Total protein or ribosome content may also represent relevant additional, correlated size metrics, whichmightbe sensed by distinct molecular mechanisms. Twosize-sensing mechanisms have emerged as favored models: a protein synthesis rate-based mechanism in budding yeast and a gradient mechanism that directly detects cell length in fission yeast.

Geometric size sensing in S. pombe

Fission yeast are rod-shaped and grow highly asymmetrically. That fission yeast maintain a constant width and increase only in length has been exploited in size control studies, as it allows measurement of a single dimension, length, to indicate three-dimensional volume [47]. Intriguingly,S. pombe cells might exploit their own well-defined geometry by using a spatial gradientto sense cell length[50, 51]. The principle issimple: a mitotic activator is localized to a region in the middle of the cell, and an inhibitor of the activator is arranged in a gradient originating from the cell poles. Thus, as the cell grows, the amount of inhibitor interacting with the activator decreases,leading to an increase inmitotic cyclin activity thatdrives mitosis.

The first gene specifically identified as a size controller was Wee1, which restrains CDK activity[28] by phosphorylating Tyr 15 (Figure4A).Removal of this specific phosphate by Cdc25 is rate-limiting for mitotic entry [52-55]. The spatial size sensor model operates through the Wee1-inhibitory kinases, Cdr1 and Cdr2 [56-60], which are themselvesinhibited by Pom1, a kinase identified through its pleiotropic effects on cell morphology and growth [61]. While Pom1 is localized in a spatial gradient emanating from the cell poles [61, 62], Cdr1 and Cdr2 are localized in cortical nodes in the middle of the cell[63].As cells grow, their poles move apart, reducing medial Pom1 concentration. This leads to reduced Pom1-dependent Cdr1/2-inhibition, potentially yielding a gradient-based sizer mechanism [50, 51]. The Pom1 gradient arises due to Tea4 dependent localization to the tips followed by autophosphorylation-induced membrane dissociation [64]. In support of the gradient model, mislocalization of Pom1 lengthens cells [50, 51].

Figure 4. Models of size-dependent cell cycle regulatory networks.

Activators of cell cycle progression are colored green; inhibitors are colored red. A. In S. pombe, size control acts at the G2-M transition. The mitotic inhibitory kinase Pom1, localized to a membrane gradient originating at the cell poles, inhibits the pro-mitotic kinase Cdr2, which is localized in cortical nodes at the midcell. Cdr2, in turn, inhibits the Wee1 kinase, which inhibits CDK1. As pombe cells grow in length, the medial concentration of Pom1 decreases, allowing activation of Cdc13-CDK1 and entry into mitosis. B. In S. cerevisiae, size control acts at the G1-S transition. CDK1 in complex with the upstream G1 cyclin Cln3, whose activity is growth-dependent, activates the G1-S transcription factor SBF by direct phosphorylation and by inactivating the transcriptional inhibitor Whi5. This activates a transcriptional positive feedback loop that leads to a switch-like entry into S-phase. Importantly, Cln3 acts on promoter-bound SBF-Whi5 complexes, suggesting that Cln3 actiivty may be titrated against promoter-bound SBF.

However, the pleiotropy of Pom1 perturbations raises some doubt as to the specificity of its role in size control.Among other functions, Pom1 is required for proper positioning of the division plane[62, 65]. Furthermore, the ability of the Pom1 gradient itself to accomplish size control has been insufficiently tested. For instance, the Pom1 concentration on the membrane decays exponentially from the cell pole with a length scale of~1.5μm [66]. Since pre-mitotic S. pombe cells have an average length of more than 10 μm, this finding suggests that the medial concentration of Pom1 will drop to extremely low levels well before mitotic entry. This is particularly relevant incdc11-119mutants deficient in septation, which yield multinucleate cells containing a DNA to cytoplasm ratio similar to WT (F. Neumann personal communication). In addition, size scaling with ploidy suggests that non-spatially based mechanisms also operate in fission yeast [8].

Prior to the discovery of the Pom1 gradient mechanism, S. pombe size control studiesanticipated non-spatial mechanisms.It was suggested that accumulation of Cdc13might drive the transition[67], but most studies focused on possible Wee1 or Cdc25-mediated mechanisms. While regulation of Wee1 by Cdr2 and Cdr1kinases has been associated primarily with nutrient modulation of cell size [59, 68, 69], size sensing was more commonly ascribed to Cdc25. An early study suggested that Cdc25 (rather than Cdc13) induced mitosis at a critical size through an accumulation mechanism [70], and the finding that Cdc25 levels increase during cell cycle arrest strengthened the link between cell growth and Cdc25 synthesis[71, 72]. Finally, Cdc25 levels are disproportionately sensitive to translation initiation rate, which is mediated by uORFs and additional 5′ regulatory sequences[73]. Although these studies are far from conclusive, they suggest that a Cdc25-dependent translational sizer may act in concert with aPom1-Cdr2-Wee1 spatial sensor. Furthermore, the maintenance of a consistent, albeit greater, division length in cells with both key phosphosites mutated in Cdc2 indicates that other pathways must be involved in mitotic size control [74]. Certainly, taken together, these data suggest multiple size sensing mechanisms operating in fission yeast.

Synthesis-ratebased size sensing in S. cerevisiae

A gradient-based geometric size sensing mechanism in budding yeast cells would require the formation of linear surface gradients in these nearly spherical cells, or cytoplasmic gradients with length scales of a few microns. Although both mechanisms are feasible, as they have been observed in bacterial cells[75-77], there is strong evidence against a geometric size sensor operating in budding yeast. In particular, cell size increases linearly with ploidy over a six-fold range despite significant differences in geometrical characteristics such as cell wall curvature, surface-to-volume ratio, and eccentricity[10, 17]. It is therefore unlikely that geometric parameters other than size are responsible for initiating cell cycle progression. Furthermore, the filamentous fungusAshbyagossypii presents an interesting case in comparison to budding yeast, because it exhibits similar G1-S control, despite its distinct geometry [78](Figure 1).G1 control mutants that affect cell size in budding yeast, such as whi5 and cln3deletionmutants that make cells smaller and larger respectively, have a similar effect on the amount of cytoplasm per nucleus in A. gossypii (A. Gladfelter personal communication).This argues that the size control, as manifested in the DNA to cytoplasm ratio,is unlikely to be operating on geometric considerationsin species more closely related to budding yeast.

A mechanism that has no geometric requirement and could maintain the DNA to cytoplasm ratio for cells of any shape is a protein synthesis rate based sensor.Such a mechanism could explain the similar effects of mutations inAshbyaand buddingyeast despite gross shape differences. At its most basic, this class of models posits that a division-promoting protein is synthesized at a rate proportional to overall protein synthesis andthat the cell enters the cell cycle when that factor reaches a threshold level. Such models were proposed early on to underlie the size threshold for division in bacteria and metazoans [79-81], as well as in S. cerevisiae and S. pombe[29, 82, 83]. Consistent with this model, reducing translation rate by incubation with low doses of cycloheximide extends G1 and increases the threshold cell sizein S. cerevisiae[84-87]. Interestingly, although similar results were observed in S. pombe treated with pulses of cycloheximide[88], extended incubation of S. pombe cells in low levels of cycloheximide does not affect size at division, providing evidence against a synthesis rate mechanism in fission yeast (F. Navarro, personal communication).

Protein synthesis rate based models rely on asizer protein, whose abundanceincreases proportionally to the protein production rate.Such a sizer protein should exhibit three key properties. First, the sizer should be a dose-dependent activator of cell cycle progression; that is, it must berate limiting and not merely a necessary component for cell cycle progression. Second, its expression must be nearly constitutive during the growth period, so that it consistently indicates overall protein synthesisrate. Finally, the sizer should be unstable, so that its activity reflects current protein production rate rather than total synthesis[89-91]. The protein in S. cerevisiae that exhibits these three properties, and is thus the prime candidate for a sizer protein, is the G1 cyclin Cln3. CLN3 is a dose-dependent activator of the G1-S-transition [92-94], and unlike other cyclins, its expression oscillates only weakly through the cell cycle[40, 94-98]. Finally, Cln3 is a highly unstable protein, with a half-life of less than five minutes [92-96]. Although the properties of Cln3, when examined using bulk assays, suggest that it acts as a synthesis ratesizer, its instability has prevented its analysis in live cells. A quantitative analysis isstill required to demonstrate that, as predicted by the model, total amounts ofthe Cln3 protein correlate with size and determine precise timing of the G1-Stransition.

Although Cln3 fits the classical description of a sizer protein, there is a basic problem facing the Cln3 synthesis ratesizer model: given that the total amount of Cln3 is proportional to protein synthesis rate, which is proportional to size in exponentially growing budding yeast cells, the correlation between ribosome content and volume will render the concentration of Cln3 roughly constant. How then does the cell produce a signal whose activity increases with cell size, when the key activator remains at constant concentration? This problem is inherent to all synthesisrate sizer models and suggests that cells normalize sizerprotein abundance against a standard that does not change with cell growth.

Early formulations of this model proposed genomic DNA (or specific sites thereon, such as replication origins) as a non-growing standard [79, 83]. In budding yeast, the nucleus was proposed as a proxy for the genome, as its volume was long assumed to depend on DNA content [99, 100]. This mechanism is supported by the fact that Cln3’s primary targets are the transcription factor SBF and its inhibitor, Whi5[101, 102], and by Cln3’s active nuclear localization[103, 104]. However, the nucleus has recently been shown to grow in G1 and maintain a nearly constant nuclear to cytoplasmic volume ratio in both S. cerevisiaeand S. pombe[8, 9]. Thus, it is necessary to consider how Cln3 abundance might be compared directly to cellular DNA content. The recent finding that Cln3 is present at SBF binding sites in the promoter of a key downstream target (CLN2) suggests a model in which Cln3 activity is titrated against the fixed number of SBF binding sites (called SCBs) in the genome[105]. Evidence for this model is currently limited to the finding that transformation of yeast with a high-copy plasmid containing several SCBs increases cell size at budding in a Cln3 and Whi5-dependent manner [105]. This promising model requires further study.

Positive feedback at the basis of the budding yeast threshold response

Althoughthe Cln3-synthesis-rate model might provide a signal whose strength increases smoothly in proportion to cell size, it does not obviously provide a threshold mechanism. A threshold in G1 was first identified by Lee Hartwell in colleagues, who defined Start as the point of commitment to the mitotic cell cycle [106]. When to exposed to mating pheromone, pre-Start cells immediately arrest their cell division cycle, whereas post-Start cells divide once more. Positive feedback loops have been shown to underlie threshold responsesin bothcell cycle and developmental transitions[107-109]. Indeed, in budding yeast, the threshold for Start is provided by a G1 cyclin positive feedback loop. The SBF inhibitor Whi5 is the rate-limiting target of Cln3 in G1, whose phosphorylation by Cln3-CDK initiates its dissociation from SBF and export from the nucleus [101, 102, 110]. Partial relief of SBF inhibition results in the transcriptional activation of the two downstream G1 cyclins, Cln1 and Cln2, completing a positive feedback loop (Figure 4b)[111-114]. Activating the Cln1/2 positive feedback loop ensures rapid export of the remainderof the nuclear Whi5 and commits the cell to division[110]. It has been proposed that ultrasensitivity in Cln3 activity is a result of cytoplasmic retention mechanisms [115, 116] or a uORF limiting translation rate [117].However, these models are inconsistent with slow and steady export of Whi5 from the nucleus in cln1Δcln2Δ cells lacking the cyclin positive feedback loop, but containing endogenous levels of Cln3[114].Importantly, progression through G1 becomes size-independent after Whi5 export[17], indicating that flipping the positive feedback-switch converts the gradually increasing size-dependent signal into a threshold response thatmarks the end of the size control program.

Increasing Cln-CDK effects S-phase entry by activating the related heterodimerictranscription factors SBF (Swi4-Swi6) and MBF (Mbp1-Swi6) to promote the coherent transcription of more than 200 genes [111, 114, 118-120]. SBF is activated both via the inhibition of Whi5 and through a Whi5-independent mechanism [101, 114]. MBF is also activated by G1-cyclin CDK activity, possibly through the Swi6 subunit shared with SBF [121]. Despite co-regulation, the transcription of the regulon through the mitotic cell cycle is temporally organized: the G1 cyclinCLN1 is the earliest activated gene (CLN2 is also relatively early) implying that G1 cyclin positive feedback is initiated prior to genome-wide changes to the transcriptional program [122]. The G1 cyclin promoters are therefore important determinants of the CDK-activity threshold required to trigger the positive feedback loop. Thus, werequire a better understanding of the molecular mechanisms that determine the susceptibility of G1-Sregulon promoters to Cln3-CDK activity in order to understand the molecular basis ofcell size regulation.

Environmental modulation of cell size

For any given environment, yeast maintain a consistent cell size by monitoring growth during the cell cycle and restricting proliferation in a size-dependent manner. However, it is well known that cell size can change dramatically as a function of extracellular conditions. Thus,the size control program is modulated according to nutrient conditions and growth rate.Typically, cells growing quickly on rich media divide at a larger size than cells growing slowly under poor nutrient conditions.

Modulation of size controlis particularly important for synthesis rate-based mechanisms.Since protein synthesis rates for slow growing cells of a given mass are lower, these cellswould be expected torequire significant increases in cell size to overcome a static threshold[90]. However, the opposite is generally the case: though there is no linear relationship between the growth rate and the mean size of the population,[25, 123]slow growth on poor nutrients tends to reduce cell size[123]. Intriguingly, a change in the available nutrients immediatelytriggers an adjustment in sizecontrol. For example, when S. cerevisiae is shifted from ethanol (slow growth) to glucose (fast growth) containing media the budding index (% cells in S/G2/M) temporarily decreases as cells rapidly adjust to a larger target size[87, 124-126]. Upon a nutrient downshift, such asa shift from glutamate to proline[30, 68, 127], or sudden glucose starvation, S. pombe immediately undergoes several rounds of division without growth phases, to attain a much smaller average cell size [128, 129].These observations have led to the conclusionthat, in addition to a size sensing mechanism, there must be a mechanism that rapidlymodulatesthe target size in response to nutrient availability[90, 130].

Several important proteins involved in metabolism and growth have also been implicated in size control. Specifically, the highly conserved kinases, target-of-rapamycin (TOR) and protein kinase A (PKA) respond to nutrient availability by sensing external cues through nutrient receptors[131], as well as internal cues such as glycolytic activity or amino acid concentrations (reviewed in [132-135]). The downstream targets of PKA and TOR include metabolic pathways and biosynthetic genes thatpromote growth. Several mutants of both the PKA and TOR pathways have been implicated in yeast size control [25][136]. As Yanagida et al[136]and Shiozaki[137]have recently provided detailed reviews on size setting in S. pombe, which highlight the importance of TOR, we will focus our discussion on summarizing recent findings in S. cerevisiae.

When screening the S. cerevisiae deletion collection for genes determining cell size, Jorgensen et al. identified SFP1 and SCH9, previously poorly characterized components of the TOR and PKA pathways [25]. Sfp1 and Sch9 activity is both modulated by nutrients andactivates ribosome production through the ribosome biosynthesis (Ribi) and ribosomal protein (RP) genes[86]. Jorgensen et al.(2004) proposed a model where ribosome biosynthesis rates set the size threshold, while current ribosome activity, i.e.,protein synthesisrate, enables passing that threshold.

Over the last decade, the pathways that connect nutrient sensing, Sfp1 and Sch9 activity, and ribosome biosynthesis have been investigated, and a more complete picture is beginning to emerge (reviewed in [134, 138]). Briefly, Sfp1 and Sch9 are both directly phosphorylated by the TOR complex 1, with crosstalk to the PKA pathway [139, 140]. Sfp1 activity is regulated by localization to the nucleus. Under unfavorable conditions such as nutrient limitation or chemical stress, Sfp1 is retained in the cytoplasm byMrs6, a component of the secretory pathway[138, 141]. In the nucleus,Sfp1 activates the transcription of the Ribi and RP genes, likely by controlling the localization of the transcription factors Fhl1 and Ifh1 [86, 142]. In turn, Sch9 activates Ribi, RP and rRNA gene expression by inactivating the transcriptional repressors Stb3, Dot6 and Tod6 [143].

Despite our increased understanding of the role of Sfp1 and Sch9 signaling in ribosome synthesis, there has been very little progress in determining how this contributes to size control. In fact, aspects of Jorgensen and Tyers’s model have been challenged by several studies: Bernstein et al.(2007) showed that disturbing the maturation of ribosomes, without inhibiting their transcription, delays Start in a Whi5-dependent manner and leads to bigger cells [144]. Another study addressed the pleiotropic effects of SFP1 deletion and showed that Sfp1 also directly targets metabolism [126]. Notably, the overexpression of Msr6, which leads to retention of Sfp1 in the cytoplasm, was shown to downregulate glycolysis[141]. Thus, Sfp1 may modulate cell size in a way that is independent of ribosome synthesis. Additionally, nutrient dependent signaling may also act on ribosome biosynthesis in a Sfp1/Sch9-independent way, e.g. possibly through transcriptional inhibitors such as Crf1 [145]. A full understanding of the roles of Sfp1 and Sch9 at Start will likely require the identification of a molecular link between these two proteins and the core cell cycle regulatory network.

Additionally, there are multiple TOR- and PKA-independent pathways in nutrient signaling that may also play a role in size control[133].For example, there is strong evidence for glucose metabolism regulating several cell cycle genes independently of PKA and TOR [146]. In this context, recent findings from Cai and Tu[147, 148] appear especially interesting: the concentration of acetyl-CoA (the endproduct of glycolysis) directly modulates the acetylation of regulatory proteins, which in turn modify histone acetylation and thereby enhance transcription of many genes required for growth and proliferation. Another interesting link between metabolism, growth and cell cycle was proposed by Futcher[149], whosuggestedthat the accumulation of storage carbohydrates could be an important determinant of Start and cell size under nutrient limited growth conditions.

Although the specific connection remains unknown, metabolic regulation must interact with cell cycle control at the G1-S transition. Studies shifting cells from poor to rich mediaindicate that size resetting is downstream of Cln3.Several studies demonstrated thatcln3 deleted cells still show nutrient modulation of size[124, 125, 150]. Baroni et al. [124] and Tokiwa et al. [150] showed that upon either glucose or cyclic adenylate-monophosphate (cAMP) addition, transcription of the G1 cyclinCLN1(and to a lesser extend CLN2) is inhibited, whereasCLN3transcription is unaffected. Flick et al [125] later mapped this inhibition to the CLN1 promoter. Inhibition of downstream G1 cyclin transcription was not only found during transient size resetting, but also played a role during steady-state growth on glucose.

Since size resetting acts downstream ofCLN3, we expect to find it acting on an element of the G1-S positive feedback loop. How is the activity of SBF modulated independently of CLN3? One likely candidate is the rate-limiting transcriptional inhibitorWhi5. While long-term steady state adjustments could simply be due to transcriptional regulation of Whi5, this would not account for immediate responses to glucose addition. We can thus speculate that modifications of Whi5 or of the Whi5-SBF interaction are responsible for rapid changes in target size. However, studies from Tyers and coworkers have shownthat size adjustment is maintained in cells lacking either or both CLN3 and WHI5[86, 102]. This implies that either a Cln3/Whi5-independent mechanism or multiple interactions set the target size. We therefore anticipate the discovery of novel transcriptional regulators that interact with SBF to fine tune the CDK requirement for passage through Start in response to nutrient conditions.

Concluding remarks

The many years of work on cell size control in budding and fission yeasts have revealed the basic principles underlying the maintenance of a consistent cell size in these organisms: critical cell cycle transitionsare delayed in a size-dependent manner, while the target size is modulated in response to environmental conditions. The protein regulatory networks that effect size sensing and cell cycle control have been elucidated in both budding and fission yeasts, but understanding the mechanisms underlying nutrient modulation of target size remains a major challenge. For size sensing, the remaining challenge is to determine how the various regulatory components work together to generate a size-dependent signal. In budding yeast, a protein synthesis rate model, based on titrating sizer molecules against genomic binding sites, shows promise.In fission yeast, a gradient-based spatial sensormay play an important role. The close homology between budding yeast and metazoan G1-Sregulation, and between fission yeast and metazoan G2-Mregulation [151], suggestsconservation of the regulatory principles underlying size control. Thus, understanding size control mechanisms in yeast will help us answer the fundamental question of how eukaryotic cells sense and set their own size.