In living things, many important behaviors—including animal heartbeats and sleep patterns—happen in cycles. Machines called oscillators, which are found inside cells, control these behaviors. There are many different oscillators and they share some common features, despite involving different genes.

Each oscillator is formed of a set of genes that interact with each other to drive regular cycles lasting seconds, hours, or even months. The oscillators do not need any cues from the environment to maintain these cycles. However, cues such as light or temperature can keep the oscillator synchronized with the environment.

To ask how complex machines like oscillators could evolve, Wildenberg and Murray inserted a gene that makes a fluorescent protein into budding yeast, a single-celled species that does not have an oscillator with a period of 24 hr. These yeast cells were then selectively grown through a few hundred generations to experimentally evolve a yeast strain where the levels of protein fluorescence regularly alternated over 24-hr periods.

Wildenberg and Murray then carried out further experiments to discover the cause of the pattern of protein fluorescence. These revealed that the pattern was due to the yeast cells alternating between forming clumps of multiple cells and living separately. The genes that mutated to create the cycles of cell clumping in the yeast all appear to have unrelated roles.

The 24-hr oscillator that evolved in the yeast has some of the features of the biological oscillators found in nature. It maintains regular cycles even without any cues from the environment and it can control a cell behavior, but the oscillator appears to be unable to accept cues from the environment, a universal property of naturally evolved circadian clocks. Further work to understand how the genes work together in the oscillator will help to better understand how 24-hr oscillators in nature can evolve from genes that lack any 24-hr behavior.

Biological oscillators demand dynamical interactions between multiple components and have evolved many times to drive various behaviors, from heartbeats, on the timescale of seconds, to circadian clocks and seasonal clocks, on the time scale of days and years. The budding yeast, Saccharomyces cerevisiae, lacks homologs to genes that make up the core oscillator or light-sensing pathways of fungal circadian rhythms (Brunner and Kaldi, 2008; Idnurm and Heitman, 2010; Salichos and Rokas, 2010) and does not display any 24-hr behavior capable of sustained autonomous oscillations in a constant environment (Eelderink-Chen et al., 2010; Robertson et al., 2013). As an example of a novel trait, we evolved a 24-hr oscillator in the budding yeast to determine how biologically complex machinery could evolve from proteins and signaling networks that lack any known connection to oscillatory behavior. Because circadian clocks can be entrained by environmental signals and have a temperature-independent period, whereas the behavior we have evolved represents an autonomous oscillator with a period of roughly 24 hr that controls a biological output, we refer to the circuit we have evolved as a diurnal oscillator, referring to its period, rather than the distinction between day and night.

Our initial goal was to use fluorescence activated cell sorting (FACS) and fluorescent reporters to select for diurnal regulation of gene expression. We expressed yellow fluorescent protein (YFP) from the promoter of the FLO1 gene (Verstrepen et al., 2005; Rando and Verstrepen, 2007; Verstrepen and Fink, 2009) and elevated the mutation rate about 100-fold by mutating the proof-reading activity of DNA polymerase ∂ (POL3) (Jin et al., 2005). We used FACS to select for variation in YFP fluorescence over a 24-hr period: each cycle began with selection of the dimmest events, the selected cells were allowed to proliferate, and the population was then selected, 10 hr later; for the brightest events, the selected cells were allowed to proliferate, and the dimmest events were selected, 14 hr later, completing the cycle (Figure 1A). We evolved two parallel populations using two different strengths of selection: the top and bottom 6% (E6) or 15% (E15) of the events detected by FACS. The selected cells proliferated for roughly six divisions before the next selection. Our scheme selects for synchrony rather than entrainment: twice a day, we select the cells with high or low fluorescence, rather than varying an environmental cue and selecting for a response to it. The human analogy would be to go to a large, international airport and select only those travelers who had arrived from the same time zone, rather than taking all newly arrived passengers and using alternating light and dark cycles to entrain their clocks over a period of several days.

Figure 1 with 1 supplement see all

Download asset Open asset

After 30 days of selection, the intensity distributions resulting from the selections for high and low fluorescence differed. To analyze the behavior of the evolved populations in more detail, we cultured them from stocks frozen that had been prepared at the end of the original selection. When they are first thawed, these populations are asynchronous, presumably because the manipulations associated with freezing and thawing disrupt the synchrony between the oscillations of different individuals. We therefore subjected these cultures to the same, twice daily, selection used for the evolution: the dimmest events are collected, cells are allowed to proliferate for 10 hr, the population distribution of fluorescence is recorded by FACS, and the brightest events are collected. After 14 hr of proliferation, distributions are recorded, and the dimmest events are collected to complete the cycle (Figure 1B). This protocol is a form of selection synchrony: it synchronizes a population by selecting those members that are at the peaks and troughs of a cycle of fluorescence intensity, rather than by applying a stimulus that alters the phase of the oscillations until all members of the population have the same phase (induction synchrony or entrainment). After synchronization, both evolved populations showed a cyclical change in the fluorescence distribution, with the E6 population giving larger shifts than the less strongly selected E15 (Figure 1C). It takes more cycles of synchronization of E15 to achieve the full amplitude of daily oscillation than it does for E6, which is largely synchronized within 24 hr of selection.

Our starting strain also contained red fluorescent protein (RFP) under the control of the ACT1 (actin) promoter, allowing us to ask if RFP fluorescence, which had not been selected on, also cycled. To our surprise, the distribution of RFP also cycled (Figure 1—figure supplement 1A), prompting us to examine the cells at different phases of the cycle. Just before their scheduled selection for low fluorescence, the population was predominantly single-celled, and just before their selection for high fluorescence, most cells were in multicellular clumps (Figure 1D). The fluorescence per cell is constant over 24 hr, for both YFP and RFP, and thus the cyclical variation in intensity seen by FACS is entirely due to changes in cell association (Figure 1—figure supplement 1B). Because we see similar fluctuations in the fluorescence from P_ACT1-RFP as we do from P_FLO1-YFP, we believe that we would have selected for oscillations in cell aggregation with any promoter that drove the expression of a fluorescent protein to levels that were sufficient for our FACS-based selection.

The cyclical behavior of our evolved populations could be due to interactions between two or more different genotypes. To eliminate this possibility, we isolated and analyzed fifteen clones from each population after subjecting them to FACS-based synchronization: all 30 clones produced strong oscillations in YFP and RFP distributions, and we analyzed the clone that showed the strongest oscillations (E6-1) in detail (Figure 1C,D). Except for the free-run experiments, described later in the paper, all other experiments examined populations that were subject to FACS-based selection twice a day with a 10-hr interval between the selection for the dimmest and brightest events and a 14-hr interval between the selection for the brightest and dimmest events.

The cycle is not a response to daily variations in environmental factors such as temperature. If populations were responding to environmental fluctuations, they should eventually synchronize in response to these fluctuations. To look for this behavior, we subdivided a single population of clone E6-1 and subjected it to two different treatments: synchronizing it by FACS selection, as described above, or subjecting it to identical manipulations, except that we collected all the events that passed through the FACS, without regard to their fluorescence intensity. The first population oscillated, showing primarily single cells just before the selection for low YFP fluorescence, and primarily clumps just before the selection for high YFP fluorescence, whereas the unsynchronized population showed very similar distributions at these two times (Figure 1—figure supplement 1C, D). We conclude that we have evolved a cyclical change in cell association with a period of about 24 hr, and cells that show this oscillation can be synchronized by selecting on their level of fluorescence as a proxy for the number of cells in a clump.

To analyze the oscillations in cell association, we followed the behavior of E6-1 cells grown at low population density using two different labels. The first, cerulean fluorescent protein (CFP) was expressed from a constitutive promoter (P_ACT1) to distinguish two genetically identical subclones: one expressed CFP (CFP⁺) and the other did not (CFP⁻). The second was introduced by labeling cells, just after FACS selection, by covalently linking Oregon Green 488-X, a yellow fluorescent molecule, to their cell walls (Hoch et al., 2005). Because a daughter’s cell wall is entirely new, the original cells retain the yellow label and their daughters are unlabeled (Barral et al., 2000). We followed these cultures as they proliferated and eventually switched to the other phase of the morphological oscillation (Figure 2A,B and Figure 2—figure supplement 1A). Single cells that were selected for by sorting for low YFP fluorescence and covalently labeled (T = 0 hr), proliferated to become clumps composed of a single yellow cell surrounded with non-yellow cells that had the same CFP expression state as the original single cell. The absence of clumps containing more than one yellow cell or both CFP⁺ and CFP⁻ cells in the same clump shows that cells emerging from the single-cell phase build lineage-based clumps. Consistent with this interpretation, we saw synchronous oscillations in cell association even at very high dilution (∼10³ cells/ml). Multicellular clumps, which were covalently labeled at the time of peak aggregation gave rise to single cells by a combination of two mechanisms: fragmenting to produce smaller clumps and proliferating to produce single-celled offspring. Smaller clumps consisting of yellow cells and newly born non-yellow cells were observed 3 hr after (T = 13 hr) the peak aggregation time (10 hr). By 18 hr, the population was a mixture of medium-sized and smaller clumps and single cells, and by 24 hr non-yellow, single cells, and small clumps were the bulk of the population, representing the completion of one cycle. When unsynchronized CFP⁺ and CFP⁻ subclones were sonicated to break up all the clumps and cultured together for 90 min, clumps that re-associated were composed of a mixture of CFP⁺ and CFP⁻ cells showing that clumps form by cells sticking together rather than failing to separate after division (Figure 2—figure supplement 1B). These results show that cells oscillate between two states over a 24-hr period: single cells produce offspring that immediately stick to their mothers to assemble a lineage-based clump, which later produces a population of single cells and small clumps as a result of clumps fragmenting and producing single cells which divide to produce a mixture of small clumps and single cells (Figure 2C).

Figure 2 with 1 supplement see all

Download asset Open asset

Because we typically sort populations twice every 24 hr, their behavior could reflect a damped oscillation, driven by selection, rather than a true oscillator, which should continue to cycle, even in the absence of selection. We therefore asked if our synchronized populations could free run: continue to oscillate in the absence of both entraining signals and periodic selection for cells in particular phases of the oscillation. We performed two experiments: one on populations and another to follow lineages initiating from single cells. In the population experiment, E6-1 was subjected to three different conditions. (1) Unsynchronized: cells that had been passaged without earlier synchronization were passed through the FACS machine and diluted twice every 24 hr, with intervals of 10 and 14 hr between the two dilutions, but all cells that passed through the FACS machine were collected, regardless of their fluorescence level and then diluted. (2) Synchronized: a population was synchronized by selection for 3 days with selection at alternating 10 and 14 hr intervals. The culture was synchronized by FACS for another 3 days, by FACS selection and dilution twice every 24 hr, with intervals of 10 and 14 hr between the two dilutions. (3) Free run: a population was synchronized by selection for 3 days at 10:14-hr intervals. After 3 days of synchronization (marked by a dotted line), this culture was not subject to FACs but was simply diluted twice every 24 hr, with intervals of 10 and 14 hr between the two dilutions for an additional 3 days. The initial, synchronized population used to start the free run was the same one that was used to start the population with continuing synchronization. Thus, all three conditions were diluted with intervals of 10 and 14 hr between the two daily dilutions and maintained at a density that allowed continual exponential proliferation. The free run populations continued to show detectable oscillations, which were undetectable in a control population that had not been synchronized (Unsynch) (Figure 3A). We next monitored how individual lineages oscillate without any prior FACS-based synchronization. A single cell was grown in a well of a 96-well plate and imaged every 3.5 hr. After 24 hr, the resulting population was collected and diluted back to single objects, regardless of clump size, into three separate new wells and followed every 3.5 hr for another 24 hr. This procedure was repeated one more time for each lineage (i.e., 3 cells from each of 3 sub-lineages) for a third day, as shown by the cartoon in Figure 3B. This protocol dilutes cells once every 24 hr to keep cells in exponential growth throughout the experiment. Cell lineages produced sustained oscillations for 3 days (Figure 3C) with many lineages remaining in phase with each other. In some lineages, their phase shifted within a 24-hr period, but the overall period, over 3 cycles was approximately 24 hr. These results suggest that we have evolved an autonomous oscillator capable of sustained oscillations in the absence of an environmental stimulus.

Figure 3

Download asset Open asset

Because there could be multiple oscillating lineages, each with a different genetic basis, in our evolved population, we analyzed the behavior of four individual clones from the E6 population, focusing on the clone E6-1. We used bulk segregant analysis (Yvert et al., 2003; Segre et al., 2006; Birkeland et al., 2010; Koschwanez et al., 2013) to find candidates for the mutations that caused E6-1 to oscillate and verified them by engineering them into the ancestral, un-evolved strain. We began by crossing the evolved clone to its ancestor, putting the resulting diploid through meiosis, selecting for 24-hr oscillations on the resulting spores to isolate a pool of haploid progeny that showed robust cycling, and sequencing this pool's genomic DNA at high coverage. The full set of mutations needed to cause cycling should be present in the vast majority of the selected progeny, whereas the allele frequency of the neutral mutations that have accumulated during evolution should be ∼50% (Birkeland et al., 2010; Koschwanez et al., 2013). We classified eight mutations that were present in ≥95% of the selected cells as putative causal mutations (Figure 4A). We recreated the evolved oscillator by engineering these eight mutations into a wild type laboratory strain. After synchronization, the oscillations of the recreated strain are indistinguishable from those of E6-1 (Figure 4B) and produce a similar free run behavior as E6-1 both as populations (Figure 4C) and as single lineages (Figure 4D).

Figure 4

Download asset Open asset

We performed three tests to investigate the role of the different mutations in producing the evolved phenotype: deleting the mutated gene, replacing it with a wild-type allele, and investigating selection for or against the mutant allele when we selected for different levels of cell association. We investigated the role of removing genes because two of the causative mutations were stop codons, suggesting that other causative mutations might also inactivate the genes in which they occurred. Each causative mutation could be individually replaced by a gene deletion without destroying the oscillation (Figure 5A). For each of these single gene deletions, the corresponding wild-type version of the same gene was added back and tested for oscillations. Surprisingly, only two genes substantially altered the oscillations: SIR4 made E6-1 single-celled and YJL070C produced constitutive clumps (Figure 5B). To determine how the mutations whose wild-type expression had no effect were contributing to the oscillator, pools of spores from a diploid heterozygous for the eight mutations were selected for four ‘sub-traits’ of the oscillator: batch culture was used to select for exponential growth and FACS to select for constitutive clumps, single cells, and stochastic clumping. After 5 days of selection, the eight mutations were PCR amplified and Sanger sequenced from genomic DNA of the surviving pool of spores to estimate the frequency of the ancestral and evolved allele (Koschwanez et al., 2013) (Figure 5C). In three independent replicate experiments, sir4-100 was selected for in all conditions, scw11-E261* was selected for in constitutive clumps, and three mutations, vps5-S520L, pet127-D650N, and whi2-R127* were positively selected for in single cells and selected against in clumping cells. Different combinations of HBN1, IDS2, and YJL070C were selected for between replicate experiments in both clumping and single cell sub-traits, making it hard to assess their significance. Because of this variability, and the absence of a phenotype when HBN1 or IDS2 were either deleted or replaced with wild-type genes, it is unclear if these genes are required for the evolved oscillations, and if they are, how they contribute to them.

Figure 5 with 1 supplement see all

Download asset Open asset

Figure 5—source data 1

Gene expression changes in clone E6-1, relative to its ancestor.

https://doi.org/10.7554/eLife.04875.010

Download elife-04875-fig5-data1-v2.txt

The nature of the mutated genes does not produce simple hypotheses for the mechanism of the oscillator: they have a wide range of molecular functions and no pair of them has a previously described functional relationship. In search of other clues, we measured gene expression to look for changes in E6-1. We performed two comparisons. The first was between synchronous and asynchronous populations. We isolated RNA at the same times from three cultures: the peak periods of single cells and multicellular clumps for a synchronized E6-1 culture, an unsynchronized E6-1 culture passaged through the FACS machine, but without sorting, and an ancestral population subjected to the synchronization protocol. We compared each population to the other two and failed to find a significant excess of genes that appeared to cycle in the evolved, synchronized population.

The second comparison was to examine the mean expression of genes in the evolved and the ancestral populations. There were 155 genes whose expression differed between the ancestral and evolved populations (Figure 5D and Figure 5—source data 1). The most striking of these were genes involved in sexual behavior. Haploid-specific genes (hsg), which are normally expressed in haploids and repressed in diploids, were repressed in the evolved clone; 8 of the 10 most strongly repressed genes were haploid-specific. This phenotype is due to the causal mutation that inactivates SIR4, a gene required for transcriptional silencing of sub-telomeric genes, and the mating genes at the two silenced loci, HMLα and HMRa (Rine and Herskowitz, 1987; Herskowitz, 1988; Pillus and Rine, 1989). In sir4 cells, the expression of a and α information from these two loci mimics a diploid cell and represses haploid-specific genes. Consistent with the effect of expressing SIR4 in E6-1, deleting HMLα from the evolved clone, which removes the only source of α information restores the haploid pattern of gene expression and eliminates cell clumping (Figure 5—figure supplement 1A,B).

We investigated the evolutionary trajectory of the E6 population by asking when the causative mutations found in the E6-1 clone appeared and how widespread they were by the end of our evolution. Each mutation was PCR amplified and Sanger sequenced at different time points over the course of the evolution to determine its frequency to an accuracy of roughly 10%. None of the mutations in E6-1 reached a frequency above 0.4, two were undetectable in the overall population, and the remaining six were first detected between 12 and 30 days of evolution (Figure 6A). The low allele frequencies at the final time point of our evolution have two possible interpretations: the evolved population contains a single oscillating lineage, which has yet to eliminate a substantial fraction of non-oscillating lineages, or it is a mixture of genetically distinct oscillator lineages. All the clones from the evolved populations cycled suggesting that the populations have multiple lineages. We therefore performed bulk segregant analysis on three additional clones from the strongly selected population (E6-2,3,4) (Figure 6B and Figure 6—source data 1A). Our results show that two cycling populations have evolved entirely independent, and each of these has split into two subpopulations. E6-2 contained 6 of the 8 causative mutations found in E6-1 and had three additional putative causative mutations. E6-3 and E6-4 share three mutations, but have no mutations in common with the lineage that produced E6-1 and E6-2. Analyzing the entire open reading frame of WHI2, the one gene that was mutated in all 4 sequenced clones, suggests that there are at least five independently evolved oscillators in the E6 and E15 populations, since one population (E6) contains 2 mutant and one wild-type allele, and the other contains one mutant and one wild-type allele (Figure 6—source data 1B).

Figure 6

Download asset Open asset

Figure 6—source data 1

Amino acid change of candidate causal mutations.

(A) Table of the candidate causal mutations identified in clones E6-1,2,3,4 and the resulting amino acid change. (B) The entire WHI2 gene was PCR-amplified, Sanger sequenced, and aligned to the wild type sequence of WHI2 to identify mutations.

https://doi.org/10.7554/eLife.04875.013

Download elife-04875-fig6-data1-v2.eps

We asked whether the oscillations in other lineages also depend on the silencing of haploid specific genes. We measured mRNA levels of the two genes in HMLα, one of which, HMLα2, inhibits haploid-specific gene expression, and three representative haploid-specific genes in eight other clones derived from our two evolved populations (E6 and E15). All nine clones have increased expression of HMLα and reduced expression of haploid-specific genes when compared to the ancestor (Figure 6C). Furthermore, deleting HMLα in these clones resulted in a loss of multicellular aggregates in all but one clone (Figure 6D, E6-5), even though only one of the clones (E6-2) has the sir4 allele found in E6-1.

Do the oscillations that we detect have any relationship to other oscillations that have been detected in budding yeast? Fluctuations in metabolic activity can be driven by temperature oscillations with a 24-hr period but rapidly die out at constant temperature (Eelderink-Chen et al., 2010), and a particular regime of growth to high density can induce prolonged metabolic oscillations whose period is 4–5 hr (Tu et al., 2005). We think it is unlikely that either of these cycles are connected to the behavior we observe: the 24-hr cycles do not free-run and the shorter cycles have the wrong period and can only be seen at cell densities higher than those that we used. We cannot, however, exclude the possibility that the behavior we see is the result of coupling changes in cell–cell adhesion to a previously unknown circadian clock in budding yeast.

How similar is the diurnal oscillator we evolved to naturally occurring circadian clocks? All circadian clocks share four features: (1) a limit-cycle, cell autonomous cellular oscillator that continues to oscillate in the absence of environmental cues (free run); (2) the oscillator maintains a roughly 24-hr period over a range of temperatures (temperature compensation); (3) the period of the internal oscillator can be synchronized, or entrained, to the phase of a daily environmental cue (e.g., light, temperature); and (4) the ability to control cellular and organismal behavior (Young and Kay, 2001; Tauber et al., 2004; Bell-Pedersen et al., 2005; Rosbash, 2009). Our results show the yeast oscillator we evolved has two of these four behaviors: its roughly 24-hr period is maintained in the absence of environmental cues or selection on the output of the oscillator (Figures 2B, 3A, 3C, 4C, and 4D), and it regulates a cellular behavior, the ability of cells to form clumps. Further investigations are required to determine if the yeast oscillator uses the combination of positive and negative feedback loops that form the core of circadian clocks and could be evolved to be temperature-compensated and entrained by environmental signals.

Our experiment shows that mutating genes in multiple, unrelated biochemical networks can produce evolutionary novelty and reveals that identifying adaptive mutations is often insufficient to explain the mechanism of these new phenotypes. The ability of multiple combinations of mutations to produce the same complex phenotype leads to two inferences. First, we know relatively little about the circuitry of many existing molecular pathways and evolving novel traits may help illuminate previously unknown connections between pathways. Second, that multiple mutations are capable of producing unexpected changes in cellular physiology by connecting previously unrelated pathways to produce new circuits. The few mutations that are shared between different oscillating lineages suggest that there are multiple, functionally different ways that such rearrangements can lead to the same phenotype.

The common loss of silencing despite little commonality in other mutated genes between evolved clones has two implications. First, it suggests that wild yeast may regulate their clumpiness during the diploid phase, which accounts for most of their life cycle, and that our evolved populations have exploited this regulation to respond to our selection for oscillatory behavior. Second, the independent evolution of the same trait appears to have a mixture of common and diverse features. All nine of the populations we examined have switched from the haploid- to the diploid-specific pattern of gene expression and all four lineages analyzed in detail have mutated WHI2. The repeated occurrence of these genetic changes suggests that they greatly increase the number of subsequent mutations that could lead to the appearance of an oscillator. Whether the later mutations act on a single pathway or multiple, different pathways will require the detailed dissection of the oscillator's mechanism. It would be interesting to determine whether a selection for oscillation in diploids, where the inactivation of genes requires two independent genetic changes, would take a similar trajectory and produce the same mechanism that we have uncovered in haploids.

The inconsistency among bulk segregant analysis, gene deletions and wild type reversions illustrates the difficulty in identifying the contribution of different mutations to a complex, evolved phenotype. We can nevertheless make inferences about the molecular underpinnings of evolutionary novelties. Four of the six mutations that are strongly identified as causal mutations are loss of function mutations. The mutations in SCW11 and WHI2 are nonsense mutations, whereas those in SIR4 and YJL070C can be replaced by gene deletions, but not by restoration of the wild-type gene. The preponderance of loss of function mutants suggests that sophisticated functions can be produced primarily by inactivating genes, rather than modifying or increasing their activities. Whether similar paths are followed in nature is likely to depend on the strength and unidirectionality of selection and the extent of antagonistic pleiotropy: the contrast between the benefits of a mutation in one environment and its costs in others. We speculate that short bursts of intense selection, such as those that occur when a species occupies a new and unoccupied niche, could lead to adaptation that is primarily driven by gene inactivation, whose short-term benefits exceeds its costs. These costs could later be reduced either by suppressor mutations that would minimize the harmful effects of loss of function mutations, while maintaining the evolved phenotype, or even by reversion of the original mutation, once the evolved phenotype had been stabilized by further mutations. Further investigation will be required to determine how often loss of function mutations participate in a variety of evolutionary processes, ranging from cancer to the formation of new species and forms of biological organization.

1. 1. Ares M

   (2012) Isolation of total RNA from yeast cell cultures

   Cold Spring Harbor Protocols 2012:1082–1086.

   https://doi.org/10.1101/pdb.prot071456

   • Google Scholar
2. 1. Barral Y
   2. Mermall V
   3. Mooseker MS
   4. Snyder M

   (2000) Compartmentalization of the cell cortex by septins is required for maintenance of cell polarity in yeast

   Molecular Cell 5:841–851.

   https://doi.org/10.1016/S1097-2765(00)80324-X

   • Google Scholar
3. 1. Bell-Pedersen D
   2. Cassone VM
   3. Earnest DJ
   4. Golden SS
   5. Hardin PE
   6. Thomas TL
   7. Zoran MJ

   (2005) Circadian rhythms from multiple oscillators: lessons from diverse organisms

   Nature Reviews. Genetics 6:544–556.

   https://doi.org/10.1038/nrg1633

   • Google Scholar
4. 1. Birkeland SR
   2. Jin N
   3. Ozdemir AC
   4. Lyons RH Jnr
   5. Weisman LS
   6. Wilson TE

   (2010) Discovery of mutations in Saccharomyces cerevisiae by pooled linkage analysis and whole-genome sequencing

   Genetics 186:1127–1137.

   https://doi.org/10.1534/genetics.110.123232

   • Google Scholar
5. 1. Brunner M
   2. Káldi K

   (2008) Interlocked feedback loops of the circadian clock of Neurospora crassa

   Molecular Microbiology 68:255–262.

   https://doi.org/10.1111/j.1365-2958.2008.06148.x

   • Google Scholar
6. 1. Eelderink-Chen Z
   2. Mazzotta G
   3. Sturre M
   4. Bosman J
   5. Roenneberg T
   6. Merrow M

   (2010) A circadian clock in Saccharomyces cerevisiae

   Proceedings of the National Academy of Sciences of USA 107:2043–2047.

   https://doi.org/10.1073/pnas.0907902107

   • Google Scholar
7. 1. Herskowitz I

   (1988)

   Life cycle of the budding yeast Saccharomyces cerevisiae

   Microbiological Reviews 52:536–553.

   • Google Scholar
8. 1. Hoch HC
   2. Galvani CD
   3. Szarowski DH
   4. Turner JN

   (2005) Two new fluorescent dyes applicable for visualization of fungal cell walls

   Mycologia 97:580–588.

   https://doi.org/10.3852/mycologia.97.3.580

   • Google Scholar
9. 1. Idnurm A
   2. Heitman J

   (2010) Ferrochelatase is a conserved downstream target of the blue light-sensing White collar complex in fungi

   Microbiology 156:2393–2407.

   https://doi.org/10.1099/mic.0.039222-0

   • Google Scholar
10. 1. Jin YH
    2. Garg P
    3. Stith CM
    4. Al-Refai H
    5. Sterling JF
    6. Murray LJ
    7. Kunkel TA
    8. Resnick MA
    9. Burgers PM
    10. Gordenin DA

    (2005) The multiple biological roles of the 3'-->5' exonuclease of Saccharomyces cerevisiae DNA polymerase delta require switching between the polymerase and exonuclease domains

    Molecular and Cellular Biology 25:461–471.

    https://doi.org/10.1128/MCB.25.1.461-471.2005

    • Google Scholar
11. 1. Koschwanez JH
    2. Foster KR
    3. Murray AW

    (2013) Improved use of a public good selects for the evolution of undifferentiated multicellularity

    eLife 2:e00367.

    https://doi.org/10.7554/eLife.00367

    • Google Scholar
12. 1. Picot J
    2. Guerin CL
    3. Le Van Kim C
    4. Boulanger CM

    (2012) Flow cytometry: retrospective, fundamentals and recent instrumentation

    Cytotechnology 64:109–130.

    https://doi.org/10.1007/s10616-011-9415-0

    • Google Scholar
13. 1. Pillus L
    2. Rine J

    (1989) Epigenetic inheritance of transcriptional states in S. cerevisiae

    Cell 59:637–647.

    https://doi.org/10.1016/0092-8674(89)90009-3

    • Google Scholar
14. 1. Rando OJ
    2. Verstrepen KJ

    (2007) Timescales of genetic and epigenetic inheritance

    Cell 128:655–668.

    https://doi.org/10.1016/j.cell.2007.01.023

    • Google Scholar
15. 1. Rine J
    2. Herskowitz I

    (1987)

    Four genes responsible for a position effect on expression from HML and HMR in Saccharomyces cerevisiae

    Genetics 116:9–22.

    • Google Scholar
16. 1. Robertson JB
    2. Davis CR
    3. Johnson CH

    (2013) Visible light alters yeast metabolic rhythms by inhibiting respiration

    Proceedings of the National Academy of Sciences of USA 110:21130–21135.

    https://doi.org/10.1073/pnas.1313369110

    • Google Scholar
17. 1. Rosbash M

    (2009) The implications of multiple circadian clock origins

    PLOS Biology 7:e62.

    https://doi.org/10.1371/journal.pbio.1000062

    • Google Scholar
18. 1. Salichos L
    2. Rokas A

    (2010) The diversity and evolution of circadian clock proteins in fungi

    Mycologia 102:269–278.

    https://doi.org/10.3852/09-073

    • Google Scholar
19. 1. Segrè AV
    2. Murray AW
    3. Leu JY

    (2006)

    High-resolution mutation mapping reveals parallel experimental evolution in yeast

    PLOS Biology 4:e256.

    • Google Scholar
20. 1. Tauber E
    2. Last KS
    3. Olive PJ
    4. Kyriacou CP

    (2004) Clock gene evolution and functional divergence

    Journal of Biological Rhythms 19:445–458.

    https://doi.org/10.1177/0748730404268775

    • Google Scholar
21. 1. Tu BP
    2. Kudlicki A
    3. Rowicka M
    4. McKnight SL

    (2005) Logic of the yeast metabolic cycle: temporal compartmentalization of cellular processes

    Science 310:1152–1158.

    https://doi.org/10.1126/science.1120499

    • Google Scholar
22. 1. Verstrepen KJ
    2. Fink GR

    (2009) Genetic and epigenetic mechanisms underlying cell-surface variability in protozoa and fungi

    Annual Review of Genetics 43:1–24.

    https://doi.org/10.1146/annurev-genet-102108-134156

    • Google Scholar
23. 1. Verstrepen KJ
    2. Jansen A
    3. Lewitter F
    4. Fink GR

    (2005) Intragenic tandem repeats generate functional variability

    Nature Genetics 37:986–990.

    https://doi.org/10.1038/ng1618

    • Google Scholar
24. 1. Young MW
    2. Kay SA

    (2001) Time zones: a comparative genetics of circadian clocks

    Nature Reviews Genetics 2:702–715.

    https://doi.org/10.1038/35088576

    • Google Scholar
25. 1. Yvert G
    2. Brem RB
    3. Whittle J
    4. Akey JM
    5. Foss E
    6. Smith EN
    7. Mackelprang R
    8. Kruglyak L

    (2003) Trans-acting regulatory variation in Saccharomyces cerevisiae and the role of transcription factors

    Nature Genetics 35:57–64.

    https://doi.org/10.1038/ng1222

    • Google Scholar

Author details

1. Gregg A Wildenberg

   1. Faculty of Arts and Sciences Center for Systems Biology, Harvard University, Cambridge, United States
   2. Department of Molecular and Cellular Biology, Harvard University, Cambridge, United States

   Contribution

   GAW, Conception and design, Acquisition of data, Analysis and interpretation of data, Drafting or revising the article

   For correspondence

   greggwildenberg@gmail.com

   Competing interests

   The authors declare that no competing interests exist.

2. Andrew W Murray

   1. Faculty of Arts and Sciences Center for Systems Biology, Harvard University, Cambridge, United States
   2. Department of Molecular and Cellular Biology, Harvard University, Cambridge, United States

   Contribution

   AWM, Conception and design, Analysis and interpretation of data, Drafting or revising the article

   Competing interests

   The authors declare that no competing interests exist.

• 19

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.