Conditional expression of genes and observation of phenotype remain central to biological discovery. Current methods enable either on/off or imprecisely controlled graded gene expression. We developed a 'well-tempered' controller, WTC846, for precisely adjustable, graded, growth condition independent expression of genes in Saccharomyces cerevisiae. Controlled genes are expressed from a strong semisynthetic promoter repressed by the prokaryotic TetR, which also represses its own synthesis; with basal expression abolished by a second, 'zeroing' repressor. The autorepression loop lowers cell-to-cell variation while enabling precise adjustment of protein expression by a chemical inducer. WTC846 allelic strains in which the controller replaced the native promoters recapitulated known null phenotypes (CDC42, TPI1), exhibited novel overexpression phenotypes (IPL1), showed protein dosage-dependent growth rates and morphological phenotypes (CDC28, TOR2, PMA1 and the hitherto uncharacterized PBR1), and enabled cell cycle synchronization (CDC20). WTC846 defines an 'expression clamp' allowing protein dosage to be adjusted by the experimenter across the range of cellular protein abundances, with limited variation around the setpoint.
Since the spectacular demonstration of suppression of nonsense mutations and its application to T4 development (Epstein et al., 1963), means to express genes conditionally to permit observation of the phenotype have remained central to biological experimentation and discovery. During the 20th century, workhorse methods to ensure the presence or absence of gene products have included use of temperature-sensitive (ts) and cold-sensitive (cs) mutations within genes, for example to give insight into ordinality of cell biological events (Hereford and Hartwell, 1974). After the advent of recombinant DNA methods, conditional expression of genes into proteins, for example by derepression of lac promoter derivatives (Goeddel et al., 1979), also found application in biotechnology for production of therapeutics and industrial products (Sochor et al., 2015). In 2021, contemporaneous approaches to conditional expression in wide use include construction of transgenes activated by chimeric activators controlled by promoters whose expression is temporally and spatially restricted to different cell lineages (Brand and Perrimon, 1993), hundreds of approaches based on production of DNA rearrangements by phage-derived site specific recombination (Sauer, 1987), and triggered induction of engineered genes by chimeric transcription regulators with DNA-binding moieties based on derivatives of TetR from Tn10 (Gossen et al., 1995; Garí et al., 1997). Most of these approaches are all-or-none, in the sense that they are not intended to bring about expression of intermediate levels of protein; and the observations they enable are often qualitative.
But it has long been recognized that adjustment of protein dosage can provide additional insight into function that cannot be gained from all-or-none expression. For example, controlled expression of the bacteriophage λ cI and cro gene products was key to understanding how changes in the level of those proteins regulated the phage’s decision to undergo lytic or lysogenic growth (Maurer et al., 1980; Meyer et al., 1980; Meyer and Ptashne, 1980). In S. cerevisiae, contemporaneous means to tune dosage include metabolite induced promoters, such as PGAL1, PMET3, PCUP1 (Maya et al., 2008), in which expression is controlled by growth media composition, and small molecule induced systems, such as the β-estradiol-induced LexA-hER-B112 system (Ottoz et al., 2014). Many of these depend on fusions between eukaryotic and viral activator domains and prokaryotic proteins (Garí et al., 1997; Ottoz et al., 2014; McIsaac et al., 2013; McIsaac et al., 2014) that bind sites on engineered promoters (Brent and Ptashne, 1985). These methods suffer from a number of drawbacks, including basal expression when not induced (Bellí et al., 1998; Garí et al., 1997; Ottoz et al., 2014), deleterious effects on cell growth due to sequestration of cellular components by the activation domain (Gill and Ptashne, 1988) induction of genes in addition to the controlled gene (McIsaac et al., 2013), and high cell-to-cell variation in expression of the controlled genes (Meurer et al., 2017; Elison et al., 2017; Ottoz et al., 2014).
These inducible systems rely on 'activation by recruitment' (Ptashne and Gann, 1997); the activator binds a site on DNA upstream of a yeast gene and recruits general transcription factors and regulators of the Pre-Initiation Complex (PIC). These assemble downstream at the 'core promoter’ and recruit RNA polymerase II to induce transcription (Hahn and Young, 2011). An alternative to inducible activation would be to engineer reversible repression of yeast transcription by prokaryotic repressors (Brent and Ptashne, 1984; Hu and Davidson, 1987; Brown et al., 1987). For TATA-containing promoters, binding of prokaryotic proteins such as LexA and the lac repressor near the TATA sequence can repress transcription (Brent and Ptashne, 1984; Murphy et al., 2007; Wedler and Wambutt, 1995), presumably by interference with the formation of the PIC, transcription initiation, or early elongation. It has long been recognized (Brent, 1985) that prokaryotic repressors likely work through different mechanisms than mechanisms used by repressors native to eukaryotes (Wang et al., 2011; Gaston and Jayaraman, 2003).
We envisioned that an ideal conditional expression system to support genetic and quantitative experimentation would: (1) function in all growth media, (2) be inducible by an exogenous small molecule with minimal other effects on the cell, (3) manifest no basal expression of the controlled gene in absence of inducer, allowing generation of null phenotypes, (4) enable a very large range of precisely adjustable expression, and (5) drive very high maximum expression, allowing generation of overexpression phenotypes. Moreover, since differences in global ability to express genes into proteins (Colman-Lerner et al., 2005) lead to differences in allelic penetrance and expressivity (Burnaevskiy et al., 2019), the ideal controller should (6) exhibit low cell-to-cell variation at any set output, facilitating detection of phenotypes that depend on thresholds of protein dosage, and other inferences of single-cell behaviors from population responses.
Here, we describe the development of a prokaryotic repressor-based transcriptional controller of gene expression, Well-tempered Controller846 (WTC846), that fulfils the criteria outlined above. This development had three main stages. We first engineered a powerful eukaryotic promoter that is repressed by the prokaryotic repressor TetR and induced by the chemical tetracycline and its analogue Anhydrotetracycline (aTc), to use as the promoter of the controlled gene. Next, we used instances of this promoter to construct a configuration of genetic elements that show low cell-to-cell variation in expression of the controlled gene, by creating an autorepression loop in which TetR repressed its own synthesis. Third and last, we abolished basal expression of the controlled gene in the absence of the inducer, by engineering a weakly expressed 'zeroing' repressor, a chimera between TetR and an active yeast repressor Tup1. With WTC846, adjusting the extracellular concentration of aTc can precisely set the expression level of the controlled gene in different growth media, over time and over cell cycle stage. The gene is then 'expression clamped' with low cell-to-cell variation at a certain protein dosage, which can range from undetectable to greater abundance than wild type. We showed that strains carrying WTC846 allelic forms of essential genes recapitulated known knockout phenotypes, and one demonstrated a novel overexpression phenotype. We constructed strains bearing WTC846 alleles of genes involved in size control, growth rate, and cell cycle state and showed that these allowed precise experimental control of these fundamental aspects of cell physiology. We expect that WTC846 alleles will find use in biological engineering and in discovery research, in assessment of phenotypes now incompletely penetrant due to cell-to-cell variation of the causative gene, in hypothesis directed cell biological research, and in genome-wide studies such as gene-by-gene epistasis screens.
Our goal was to engineer efficient repression of eukaryotic transcription by a bacterial repressor. We started with a strong (Ho et al., 2018), well-characterized, constitutive, and endogenous yeast promoter. This promoter, PTDH3, has three key Transcription Factor (TF) binding sites, one for Rap1 and two for Gcr1 (Yagi et al., 1994; Kuroda et al., 1994) in its Upstream Activating Region (UAS), and a TATA sequence at which PIC assembles on the core promoter (Figure 1A). Based on earlier work, we knew that binding of prokaryotic repressors to sites flanking the TATA sequence of PTDH3 repressed activity of this promoter (Wedler and Wambutt, 1995), presumably by interfering with PIC formation, transcription initiation, or early elongation. We therefore placed well characterized, 15 bp long TetR-binding sites (tetO1) (Bertram and Hillen, 2008) immediately upstream and downstream of the PTDH3 TATA sequence to create P2tet. To determine whether repressor binding could also block function in the UAS, we placed a single tetO1 directly upstream of each Rap1 and Gcr1-binding site to create P3tet. We also combined the operators in these constructs to generate P5tet (Figure 1B). We integrated a single copy (Gnügge et al., 2016) of constructs bearing these promoters directing the synthesis of the fluorescent protein Citrine into the LEU2 locus (Griesbeck et al., 2001).
We compared the Citrine fluorescence signal (measured by flow cytometry at wavelengths 515–545 nm) from these promoters to quantify their activity. We compared the strains Y2551[P2tet], Y2564[P3tet], and Y2566[P5tet] with an otherwise-isogenic strain in which Citrine was expressed from native PTDH3 (Y2683). This fluorescence signal measures Citrine expression, but also includes autofluorescent background from the yeast cells. We quantified this background by using the otherwise-isogenic parent strain Y70. Measured in this way, P2tet had 76%, P3tet 69%, and P5tet 51% of PTDH3 activity (Figure 1B). To assess repressibility of these promoters, we compared Citrine expression in these strains with expression in otherwise-isogenic strains in which a genomically integrated PACT1 promoter drove constitutive expression of TetR (Y2562, Y2573, Y2577). By this measure, TetR repressed P2tet by a factor of 12, P3tet by a factor of 1.5, and P5tet by a factor of 12 (Figure 1B and Figure 1—figure supplement 1). Absolute repressed signal from these promoters was 4.3, 33, and 3 times the autofluorescence background. Because our aim was to create a promoter with no expression when repressed, we viewed even small reductions in repressed expression as useful and therefore decided to use P5tet as a basis for further constructions.
Insertion of tetO1 sites in PTDH3 to create P5tet had reduced promoter maximum activity considerably. In order to regain the lost activity, we tested numerous constructs to find optimal placement for the tetO1 sites, optimized Rap1, Gcr1, and TATA sequences, and increased the number of Rap1 and Gcr1 sequences (see Appendix 1 and Appendix 1- Figure 1). This work resulted in P7tet.1, which carried two Rap1 and three Gcr1 sites, sequence optimized to generate higher promoter activity, and an alternative TATA sequence to that of PTDH3. By the assays described above, the new promoter P7tet.1 (Y2661) showed comparable maximum expression to PTDH3, 20-fold repression of Citrine signal, and absolute repressed activity (Y2663) of 4.3-fold over background (Figure 1C). We chose P7tet.1 as the promoter to develop our controller with.
We set out to optimize control of genes by P7tet.1. To do so, we tested the ability of different constructions that directed the synthesis of TetR to regulate P7tet.1-citrine directed fluorescence signal. Figure 2A shows the three different architectures. In Simple Repression (SR), the P7tet.1 controlled gene was repressed by TetR expressed from a constitutive promoter. In Autorepression (AR), the P7tet.1 controlled gene was repressed by TetR expressed from a second instance of P7tet.1, therefore creating a negative feedback loop. In Complex Autorepression (cAR), a second TetR gene expressed from a constitutive promoter was added to the AR architecture.
We compared the input-output relationship (i.e. dose response) for the three architectures. To do so, we constructed otherwise-isogenic strains with these architectures in which P7tet.1 directed Citrine expression (Y2663, Y2674, and Y2741). We used flow cytometry to quantify Citrine fluorescence signal from all strains 7 hr after addition of different concentrations of aTc and fitted a log logistic model to the median fluorescence (see Materials and methods) (Figure 2B&C).
Compared to the SR architecture, the AR architecture showed a more gradual dose response curve and a larger input dynamic range (the range of input doses for which the slope of the dose response curve was non-zero), from 3 to 400 ng/mL vs. 5–80 ng/mL aTc. This same flattening of the response curve and increased input dynamic range in autorepressing, synthetic TetR based eukaryotic systems has been described (Nevozhay et al., 2009), and we believe it operates in evolved prokaryotic systems including Tn10 and the E. coli SOS regulon, in which the TetR and LexA repressors repress their own synthesis (see Discussion). A broader input dynamic range allows more precise adjustment of protein levels, since small differences in inducer concentration (due for example to experimental errors, or differences in aTc uptake among cells) have smaller effects.
In these experiments, we also measured cell-to-cell variation (CCV) in the expression of the controlled gene. Many existing inducible gene expression systems show considerable variation in expression of the controlled gene, making it difficult to achieve homogenous phenotypes at the population level (Meurer et al., 2017; Elison et al., 2017; Ottoz et al., 2014). In S. cerevisiae and C. elegans, comparison of signals from strains with different constellations of reporter genes allows quantification of different sources of variation in protein dosage (Colman-Lerner et al., 2005; Pesce et al., 2018; Mendenhall et al., 2015). Here, we quantified overall variation in protein dosage by measuring the Coefficient of Variation (CoV) in fluorescent output from a single reporter (Figure 2—figure supplement 1), and we developed a second measure called Volume Independent Variation (VIV) (explained in Appendix 2) that normalized variation in dosage with respect to a key confounding variable, cell volume, to correct for its effect on protein concentration. In VIV, we estimated cell volume by a vector of forward and side scatter signals, and calculated the remaining (Residual) Standard Deviation of the single reporter output after normalization with this estimated volume (Figure 2D and Figure 2—figure supplement 2). By both measures, strains carrying the SR architecture showed high variation throughout the input dynamic range, with a peak around the mid-point (12 ng/mL aTc). Strains bearing the AR architecture showed low overall CCV, and no peak at intermediate aTc concentrations. This diminution of CCV in synthetic, autorepressing TetR based eukaryotic systems has previously been described (Becskei and Serrano, 2000; Nevozhay et al., 2009). In the SR architecture, variations in the amount of TetR in different cells cannot be buffered. In the AR architecture, such variations in repressor concentration are corrected for (see Discussion) and variation in expression of the controlled gene is at or around the same level as seen for constitutive expression driven by a number of native promoters (see Figure 2—figure supplement 4 for variation of commonly used promoters). This reduced cell-to-cell variation is useful for inferring single cell behaviors by observing population level responses (see Discussion).
Compared with cells bearing the SR architecture, otherwise-isogenic cells bearing the AR architecture showed increased basal expression (6.3 vs. 4.1-fold over autofluorescence background). The increased basal expression was a consequence of the fact that in the AR architecture P7tet.1 directs the synthesis of both the controlled Citrine gene and of TetR itself, so that, in uninduced cells, the steady state abundance of TetR was lower than in cells in which synthesis of TetR was driven by PACT1. More important, in the AR architecture, the fact that some amount of TetR expressed from P7tet.1 was needed to repress its own synthesis meant that it would not be possible to abolish P7tet.1-driven expression of the controlled gene completely. Since ability to abolish basal expression of the controlled gene was an important design goal, we constructed strains with a third architecture, cAR, in which a different constitutive promoter drove expression of a second TetR gene in order to drive basal expression lower. Compared to otherwise-isogenic AR strains, strains expressing Citrine controlled by the cAR architecture showed reduced basal expression (4.1-fold over autofluorescence), and, compared to the otherwise isogenic SR strain, showed reduced CCV and a more gradual dose response (Figure 2C&D and Figure 2—figure supplement 3). We therefore picked this cAR architecture for our controller.
To further decrease basal expression in the cAR architecture, we set out to create a more effective TetR derivative. Initially, we followed an approach that increased the size and nuclear concentration of TetR by fusing it to other inert bacterial proteins and nuclear localization sequences, but this approach was not enough to abolish all basal expression (see Appendix 3).
P3tet bears tetO1 sites only in its UAS. The fact that P3tet SR strains only showed weak repression (1.5-fold) suggested that TetR, and other inert derivatives described in the Appendix 3, exerted their effects on P7tet.1 mostly by their action at the tetO1 sites flanking the TATA sequence. We thus hypothesized that TetR derivatives that carried native, active yeast repressors might more effectively repress from sites in the UAS. The yeast repressor Tup1 complexes with Ssn6 (also called Cyc8) with a ratio of 4:1, forming a complex of 420 kDa (Varanasi et al., 1996), and this complex represses transcription through a number of mechanisms. These include repositioning and stabilizing nucleosomes to form an inacessible chromatin structure (Chen et al., 2013; Zhang and Reese, 2004; Ducker, 2000). Tup1 also blocks chromatin remodeling, masks activation domains, and excludes TBP (Wong and Struhl, 2011; Zhang and Reese, 2004; Mennella et al., 2003). LexA-Tup1 fusion proteins repress transcription when bound upstream of the Cyc1 promoter (Tzamarias and Struhl, 1994), and TetR-Tup1 fusions reduce uninduced expression in a dual TetR activator-repressor controller (Bellí et al., 1998). For P7tet.1, we imagined that as many as seven TetR-Tup1 dimers might bind to the promoter, potentially recruiting two additional Tup1 and one Ssn6 molecules per tetO1 site. The resulting ∼3mDa of protein complexes might block activation by one or more of the above mechanisms. We therefore measured the ability of a TetR-nls-Tup1 fusion to repress P7tet.1-driven Citrine signal in SR strains. When its expression was directed from PACT1 (Y2669), TetR-nls-Tup1 decreased uninduced fluorescence signal to background levels (Figure 3A). Because fusion of TetR to a mammalian repressor domain in mammalian cells had shown very slow induction kinetics (Deuschle et al., 1995), we checked whether the TetR-nls-Tup1 fusion showed increased induction time compared to TetR alone but found no such effect (Figure 3—figure supplement 1). Additionally, TetR-nsl-Tup1 abolished uninduced expression driven by P3tet (Figure 3—figure supplement 2) (77-fold repression), compared to repression in otherwise isogenic strains by TetR, which showed basal expression reduced by only 1.5-fold (Figure 1—figure supplement 1). By contrast, TetR-nls-Tup1 fusion repressed P2tet, where tetO1 flank only the TATA sequence, more strongly than TetR alone, but still showed basal expression. Our data thus suggested that the TetR-nls-Tup1 suppressed basal expression mainly by its effects in the UAS (see Discussion).
In the cAR architecture, the induction threshold, that is, the smallest concentration of inducer that can induce expression, is determined by the number of molecules of the repressors present before induction. We sought to lower the induction threshold in order to maximize the input dynamic range. Therefore, we constructed cAR controllers using TetR and TetR-nls-Tup1, to determine the lowest level of TetR-nls-Tup1 that could still abolish uninduced expression from P7tet.1. TetR-nls-Tup1 was driven by constitutive promoters of genes whose products were of decreasing abundance (Ho et al., 2018) (PACT1, PVPH1, PRNR2,PREV1) (Y2673, Y2684, Y2749, and Y2715). The PACT1, PVPH1 and PRNR2 strains showed no uninduced expression (Figure 3B), while the PREV1 strain did (Figure 3—figure supplement 3). Out of the three, Rrn2 protein is present at lower abundance, and the PRNR2-driven TetR-nls-Tup1 has the lowest induction threshold in a dose response experiment with strains bearing SR architectures (Y2669, 2676, 2717) (Figure 3C).
We therefore chose as our final controller the cAR architecture in which P7tet.1 directed the expression of both TetR and of the controlled gene, while PRNR2 directed the synthesis of TetR-nls-Tup1. We constructed plasmids such that the tetR and tetR-nls-tup1 components are encoded on a single integrative plasmid, and a separate plasmid can be used to generate PCR fragments bearing P7tet.1 for homologous recombination directed replacement of the promoter of any yeast gene. Due to its ability to give precisely regulated expression over a wide range of inducer concentrations, we called this construct a 'Well Tempered Controller' and gave it the number of Bach’s first Prelude and Fugue (Bach, Johann Sebastian, 1685-1750. The Well Tempered Clavier. Book I: 24 Preludes and Fugues, BWV 846, C Maj) (Figure 4A).
We measured the time-dependent dose response of fluorescent signal in Y2759, the WTC846::citrine strain during exponential growth using flow cytometry (Figure 4B&C). Without aTc, there was no signal above background. After induction, signal appeared within 30 min. Time to reach steady state, which will be shorter for proteins that degrade more quickly (see Appendix 4), was 7 hr for the stable protein Citrine. Steady state expression was adjustable over aTc concentrations from 0.5 ng/mL to 600 ng/mL, a 1200-fold input dynamic range. Maximum expression was similar to that for the PTDH3-citrine strain Y2683. Direct observation of Citrine and TetR expression by Western blotting showed no expression of Citrine in absence of aTc, adjustable Citrine levels over the same input dynamic range and TetR expression synchronized with Citrine (Figure 4—figure supplement 1). In all eight growth media tested, WTC846::citrine expression was precisely adjustable (Figure 4—figure supplement 2), and even very high induction of the WTC846 system in a strain where only the control plasmid bearing tetR and tetR-nls-tup1 was integrated (Y2761) had no significant effect on growth rates (Figure 4—figure supplement 3).
To better characterize the system, we also measured the shutoff speed of WTC846 driven expression. We reasoned that the time to observable phenotypic effect of WTC846 shutoff would depend on the speed of five processes: (i) aTc diffusion out of the cell, (ii) TetR binding to its operators, (iii) aTc sequestration by newly synthesized TetR, (iv) degradation and dilution of citrine mRNA, and (v) degradation and dilution of Citrine protein. Processes i-iii would lead to the cessation of transcript production by WTC846, and their speed would be the same for all WTC846 controlled genes, whereas the speed of processes iv and v determine perdurance of the gene product and will be different for different mRNAs and proteins.
To measure shutoff, we grew the WTC846::citrine strain (Y2759) to early exponential phase and then induced with a high concentration (600 ng/mL) of aTc and measured fluorescence signal every 30 min in flow cytometry. Additionally, after 30, 90, 150, and 210 min, we removed, washed, and resuspended a sample in (a) medium without aTc (to shut off expression from WTC846) and (b) medium without aTc but with cyclohexamide (to shut off both WTC846 and new protein synthesis). After shutoff, we expected to see an initial increase in signal, followed by decline from this peak. Increase in fluorescence after shutoff in (a) would depend on the time it took for WTC846 to stop producing new mRNA (processes i-iii), the time it took for the existing mRNA to be degraded (process iv), and on continued fluorophore formation by already synthesized but immature Citrine proteins, which has a maturation time of around 30 min (Nagai et al., 2002). Whatever increase in fluorescence in (a) observed above that baseline found after shutoff in cycloheximide (b) would be due to WTC846 shutoff speed and mRNA degradation speed.
As expected, we observed an initial increase of fluorescence in the shutoff samples (Figure 4—figure supplement 8A), which peaked for the samples in (a) (without cycloheximide) at around 60 min. A single-cell division takes 90 min and we therefore conclude that WTC846 shutoff (events i-iii) is rapid and occurs within one cell division, and likely within 30 min given that the time between Citrine production and observable fluorescence is around 30 min. Subsequent reduction in fluorescence, which fell to half after 120 min in all samples, is an estimation of process v, that is, Citrine degradation + dilution (Figure 4—figure supplement 8B), and is consistent with the idea that the continued reduction in Citrine signal is caused by dilution by cell division. Overall, we conclude that WTC846 shutoff is rapid, but the time required to see the phenotypic effects of the absence of the controlled gene product will primarily depend on the stability of the mRNA and expressed protein.
We also quantified the cell-to-cell variation in Citrine expression using the single reporter VIV measure for the WTC846::citrine strain (Y2759) grown in YPD, and compared it to variation in a β-estradiol (LexA-hER-B112) activation based transcriptional control system we previously described, and the commonly used galactose activated PGAL1 (Figure 4D, Figure 4—figure supplement 4, Figure 4—figure supplement 6 and Figure 4—figure supplement 9). At increasing concentrations of aTc, VIV initially rose to 0.63 at 8 ng/mL, similar to the VIV measured for Citrine expression repressed by PRNR2-driven TetR-nls-Tup1 in an SR strain (Y2717, RSD of 0.67, Figure 4—figure supplement 5). At higher aTc inputs, VIV rapidly dropped below that seen in Y70, an otherwise-isogenic autofluorescence control strain, and reached the same low level (0.18) observed for Citrine whose expression was driven by PTDH3 (Y2683). Because the autofluorescence varied so greatly, absolute VIV for cells grown in different media could not be directly compared. However, under all growth conditions (Figure 4—figure supplement 7), VIV was highest at the similarly low concentrations of aTc and decreased at higher concentrations to the levels shown by the PTDH3-citrine strain (Figure 4—figure supplement 2). We interpret the peak of VIV in the input dynamic range as arising from the fact that the WTC846 architecture combines Simple Repression and Autorepression of the P7tet.1-controlled gene (here, Citrine). At low concentrations of inducer, in the SR regime, most repression of P7tet.1 was due to the constitutively expressed TetR-nls-Tup1, and the peak VIV was similar to that found for the strain where P7tet.1 was repressed by constitutively expressed TetR-nls-Tup1 (Figure 4—figure supplement 5 and see previous Results section). At higher concentrations of aTc, in the AR regime, P7tet.1 is derepressed, the concentration of TetR and the ratio of TetR to TetR-nls-Tup1 is large. At these inducer concentrations, TetR controls its own synthesis and variation is suppressed by this negative feedback, resulting in much lower cell-to-cell variation throughout the dynamic range compared to routinely used transcriptional controllers. Taken together, these results indicated that WTC846 fulfilled our initially stated criteria for an ideal conditional expression system.
We then assessed the ability of WTC846 to direct conditional expression of endogenous genes. We selected (i) genes that are essential for growth, but for which previously generated transcriptionally controlled alleles still formed colonies on solid medium (CDC42, TOR2, PBR1, CDC20) or continued to grow in liquid medium (PMA1) under uninduced conditions, (ii) essential genes for which existing transcriptionally controlled alleles did not show the expected overexpression phenotype (IPL1), or (iii) essential genes for which conditional expression alleles did not exist (CDC28) (Mnaimneh et al., 2004; Yu et al., 2006; Dechant et al., 2014). These genes encoded proteins with a variety of functions: stable (Cdc28) and unstable (Cdc20 and Cdc42) cell cycle regulators, a spindle assembly checkpoint kinase (Ipl1), a metabolic regulator (Tor2), a putative oxidoreducatase (Pbr1), and a high abundance membrane proton pump (Pma1). The encoded proteins spanned a range of abundance from ∼1000 (Tor2 and Ipl1) to >50,000 (Pma1) molecules per cell (Ho et al., 2018).
We constructed strains in which WTC846 controlled the expression of these genes. Before transformation the cells were grown in liquid medium containing aTc, and then plated on solid medium containing aTc (see Appendix 5 for a detailed protocol) (Figure 5A and Table 1, strains labeled WTC846-Kx::gene_name). To make these strains, we integrated a single plasmid-borne TetR-nls-Tup1 and autorepressing TetR construct into the LEU2 locus in a BY4741 background, and replaced sequences upstream of the ATG of the essential gene with a ∼1940 bp casette carrying an antibiotic selection marker and P7tet.1, without altering the sequence of the upstream gene or its terminator. In most cases we removed between 20 and 200 bp of the endogenous gene promoter. The cassette carried one of three different 15 bp translation initiation sequences (extended Kozak sequences; K1, K2, K3) as the last 15 bases before the ATG. These were designed to enable different levels of translation of the gene’s mRNA (Li et al., 2017). The predicted efficiency of the sequences was K1> K2> K3. If cells of a strain carrying a WTC846-controlled essential gene formed colonies on solid medium without aTc, we constructed an otherwise-isogenic strain with a lower efficiency Kozak sequence (data not shown).
We spotted serial dilutions of cultures of the final seven strains on YPD, YPE, SD, S Glycerol and SD Proline plates with and without inducer, and assessed the strains’ ability to grow into visible colonies at a single time point, at which cells of the parent strain formed colonies in all serially diluted spots (24 hr for YPD and SD, 42 hr for others.) (Figure 5B and Figure 5—figure supplement 1). On all these media, no strain formed colonies without aTc and at intermediate concentrations of aTc all strains did. This result showed that WTC846 alleles can produce null phenotypes.
At high aTc concentrations, the WTC846-K2::IPL1 strain formed colonies with lower plating efficiency than the parent strain. Ipl1 is a component of the kinetochore and is required for correct sister chromatid separation during mitosis. In mouse embryonic fibroblasts, overexpression of the orthologous Aurora B kinase causes aberrant chromosome segregation and increases duration of mitosis by activating the Spindle Assembly Checkpoint, which stops mitosis until correct spindle attachments to sister chromatids can be formed (González-Loyola et al., 2015). In a previous study in S. cerevisiae, however, PGAL1-driven overexpression of Ipl1 did not decrease plating efficiency, did not cause accumulation of cells with 2 n DNA content unable to complete mitosis, and did not cause aberrant chromosome segregation as assessed by microscopy, unless Ipl1 was overexpressed simultaneously with another kinetochore component (Sli15) (Muñoz-Barrera and Monje-Casas, 2014). We asked whether WTC846-driven Ipl1 overexpression alone could cause missegregation phenotypes in S. cerevisiae. We cultured WTC846-K2::IPL1 cells for 18 hr in YPD with a high concentration of aTc (400 ng/mL), and measured total DNA content in flow cytometry to assess cell cycle state. In these cultures compared to the parent with WT Ipl1, many cells were in the G2/M phase with 2 n DNA content, indicative of an inability to complete mitosis, and a significant portion of the population showed aberrant chromosome numbers above 2 n (Figure 5—figure supplement 2). That is, WTC846-driven Ipl1 overexpression in S. cerevisiae caused a previously undescribed phenotype, which resembled that caused by Aurora B overexpression in mammalian cells. To determine why WTC846-driven Ipl1 overexpression caused this phenotype while PGAL1-driven overexpression did not, we compared WTC846-driven Citrine expression with Citrine driven by PGAL1 carried on a centromeric plasmid. Compared with WTC846-driven expression, centromeric PGAL1 plasmid expression was twofold lower, and cell-to-cell variation was ∼4.5-fold higher (Figure 5—figure supplement 7). Either the lower expression or the higher variation, or both, might account for the fact that PGAL1 driven Ipl1 overexpression does not result in the mammalian Aurora B phenotype in S. cerevisiae.
We tested whether adjustable expression of metabolic and essential genes could be used to titrate growth rates. We constructed strains with WTC846 alleles of Tor2, a low abundance, stable, essential protein necessary for nutrient signalling and actin polarization (Bartlett and Kim, 2014), Pma1, an abundant, essential proton pump that regulates the internal pH of the cell (Ambesi et al., 2000), and Tpi1, a highly abundant, non-essential glycolytic enzyme (Fraenkel, 2003) (Y2773, 2828, 2849). We cultured WTC846::TOR2, WTC846::PMA1, and WTC846::TPI1 cells in different liquid media over a large input dynamic range of aTc, and measured growth by scattered light intensity in a growth reader as a proxy for culture density (Biolector or GrowthProfiler) (Figure 5C for Tor2 and Figure 5—figure supplement 3 for all three proteins). All strains showed distinct growth rates at different aTc concentrations. For all strains, we identified an aTc concentration that resulted in the same growth rate as the otherwise-isogenic strain bearing the native gene promoter. In order to assess whether the WTC846::PMA1 strain showed the expected hypomorphic phenotype of defective daughter cell separation (Cid et al., 1987), we used flow cytometry and Sytox Green staining to quantify DNA content. At low aTc concentrations, cells showed an apparent increase in ploidy and cell size and microscopic observation showed that each mother had multiple daughters attached to it (Figure 5—figure supplement 4). Observation of WTC846::TOR2 strains revealed a novel overexpression phenotype: at high aTc concentrations, cells bearing the higher translational efficiency TOR2 allele (WTC846-K1::TOR2) grew more slowly than the otherwise-isogenic control parent strain with WT TOR2 (Figure 5—figure supplement 3D, compare 600 ng/mL line to blue dashed line). The strain with the less efficient WTC846-K3::TOR2 allele did not show this overexpression phenotype. These results demonstrate that researchers can adjust input to WTC846 alleles to tune protein levels and different growth rates with a level of precision not achievable until now, and that the dynamic range of phenotypic outputs can be further expanded by the ability to construct WTC846 alleles with alternative Kozak sequences to observe phenotypes at the two dosage extremes.
We then tested whether adjustable gene expression could precisely regulate cell size. In S. cerevisiae, Whi5 regulates the volume at which unbudded cells commit to a round of division and start forming buds. whi5Δ cells are smaller, and cells expressing Whi5 under PGAL1 control are larger than otherwise-isogenic cells (de Bruin et al., 2004). Whi5 controls cell volume by a complex mechanism and unlike most other proteins, its abundance does not scale with cell volume (Schmoller et al., 2015). Whi5 mRNA and protein are expressed during S/G2/M (in haploids, at about 2500 molecules), and Whi5 is imported into the nucleus in late M phase (Taberner et al., 2009), where it suppresses transcription of the G1 cyclins needed to commence a new round of cell division (de Bruin et al., 2004; Taberner et al., 2009). During G1, as cells increase in volume, the nuclear concentration of Whi5 falls due to dilution (Schmoller et al., 2015) and slow nuclear export (Qu et al., 2019) until a threshold is reached, after which Whi5 is rapidly exported from the nucleus, and cells enter S phase. To test whether we could control cell volume by controlling Whi5, we constructed haploid and diploid WTC846::WHI5 strains (Y2791, Y2929). In these strains, we expected Whi5 to be expressed throughout the cell cycle, but that import of the protein into the nucleus during late M phase, and diminution of nuclear concentration to below the threshold needed to START as cell volume increased in G1, should remain unaffected. We expected that the volume of these cells should scale with the concentration of the aTc inducer. We grew these strains along with otherwise isogenic control strains in S Ethanol to exponential phase at different aTc concentrations, and measured cell volume using a Coulter counter. Increasing Whi5 expression resulted in increasingly larger cells (Figure 5D). Without aTc, diploid WTC846::WHI5 cells were about the same volume as haploid controls (median 27fL vs 25fL), whereas haploid WTC846::WHI5 cells were only slightly smaller at 24fL. At around 10 and 12 ng/mL aTc, both haploid and diploid strains had about the same volume as controls. At full induction, both WTC846::WHI5 strains had a median volume of around 60fL, almost twice as large as the diploid control, yielding a more than twofold range of possible cell volumes attainable using WTC846 for both haploid and diploid cells. We also calculated the CoV of cell volume to assess cell-to-cell variation of this WTC846 directed phenotype. For most of the volume range, the CoV was around the same level as for the control strains with WT Whi5 (Figure 5—figure supplement 5A&B). Both diploid and haploid cells (especially haploids) expressing high levels of Whi5 showed increased variation in volume. We quantified DNA content of the haploid strain in a high aTc concentration using Sytox staining and found an increase in the number of aneuploid cells (>2n) (Figure 5—figure supplement 5C). We therefore believe that overexpression of Whi5 leads to endoreplication, and the increased variation in volume at high aTc concentrations in the haploid strain originates from these endoreplicated cells.
Finally, we tested the ability of WTC846 to exert dynamic control of gene expression by constructing a WTC846-K3::CDC20 strain (Y2837) and using this allele to synchronize cells in batch culture by setting Cdc20 expression to zero and then restoring it (Juanes, 2017). Cdc20 is an essential activator of Anaphase Promoting Complex C, which once bound to Cdc20, initiates the mitotic metaphase to anaphase transition (Pesin and Orr-Weaver, 2008), and is then degraded during anaphase. Upon depletion of Cdc20, for example by shift of ts strains to the restrictive temperature, or transcriptionally controlled alleles to non-inducing medium, cells arrest in metaphase with large buds and 2 n DNA content. When Cdc20 is restored by switching to the permissive condition, cells enter the next cell cycle simultaneously (Cosma et al., 1999; Shirayama et al., 1998). For an investigator to be able to use WTC846-K3::CDC20 to synchronize the cells in a culture, the investigator would need to shut off Cdc20 expression completely, and then re-express it in all the cells in a population. To test the feasibility of this, we diluted exponentially growing WTC846-K3::CDC20 cells into YPD medium without aTc (0.5 million cells/mL) and took samples for Sytox staining and flow cytometry analysis for DNA content at fixed intervals. Within 480 min, the entire culture had arrested at the G2/M phase with 2 n DNA content (Figure 5E). Microscopic inspection confirmed that cells had arrested with large buds, as is expected upon a G2/M arrest. We next added 600 ng/mL of aTc. As assayed by Sytox staining and flow cytometry and confirmed by microscopy, cells then re-entered the cell cycle within 35 min and went through one cell cycle completely synchronously. Induction of WTC846 is thus rapid, indicating that diffusion of aTc into the cell, and TetR unbinding of tetO are also rapid.
We also determined the arrest time of cells pre-cultured with a lower concentration (3 ng/mL) of aTc (Figure 5—figure supplement 6). These cells had a lower concentration of Cdc20 before aTc was removed, and therefore required less time to reach complete arrest (∼210 min as opposed to ∼480 min). This suggests that the predominant contribution to the time to reach complete arrest is the concentration and stability of Cdc20. Given this, and the rapid shutoff kinetics of WTC846 presented in Figure 4—figure supplement 8, we conclude that the shutoff dynamics of WTC846 controlled phenotypes depend mostly on the speed of degradation of the controlled protein. Additionally, when compared to published data (Tavormina and Burke, 1998; Cosma et al., 1999; Ewald et al., 2016), arrest at G2/M using the WTC846-K3::CDC20 strain is more penetrant than that obtained using temperature-sensitive (∼25% unbudded cells) and transcriptionally controlled (∼10% unbudded) alleles of CDC20. Release is at least just as fast as that observed for the temperature-sensitive (∼35 min) and the transcriptionally controlled allele (∼40 min).
Conditional expression of genes and observation of phenotype remain central to biological discovery. Many methods used historically, such as suppression of nonsense mutations, or conditional inactivation of temperature sensitive mutations, do not facilitate titration of graded or intermediate doses of protein. More current methods for graded expression do not allow experimenters to adjust and set protein levels and show high cell-to-cell variation of protein expression in cell populations, limiting their utility for elucidating protein-dosage-dependent phenotypes. Moreover, most such methods also have secondary consequences including slowing of cell growth. In order to overcome these limitations, we developed for use in S. cerevisiae a 'Well-tempered Controller'. This controller, WTC846, is an autorepression-based transcriptional controller of gene expression. It can set protein levels across a large input and output dynamic range. As assessed by Citrine fluorescence readout, WTC846 alleles display no uninduced basal expression, and uninduced WTC846 alleles of poorly expressed proteins display complete null phenotypes. WTC846 alleles also exhibit high maximum expression, low cell-to-cell variation, and operation in different media conditions without adverse effects on cell physiology.
The central component of WTC846 is an engineered TATA containing promoter, P7tet.1. We and others had shown that prokaryotic repressors including LexA (Brent and Ptashne, 1984), TetR (Murphy et al., 2007), λcI (Wedler and Wambutt, 1995) and LacI (Hu and Davidson, 1987; Figge et al., 1988) can block transcription from engineered TATA-containing eukaryotic promoters, when those promoters contain binding sites between the UAS (or, for vertebrate cells, the enhancer) and the TATA (Brent and Ptashne, 1984) or downstream of or flanking the TATA. To develop P7tet.1, we placed seven tetO1 TetR-binding sites in the promoter of the strongly expressed yeast gene TDH3. Two of the sites flank the TATA sequence, the other five abut binding sites for an engineered UAS that binds the transcription activators Rap1 and Gcr1. In WTC846, one instance of P7tet.1 drives expression of the controlled gene, while a second instance of P7tet.1 drives expression of the TetR repressor, which thus represses its own synthesis.
We believe that repression of P7tet.1 by TetR is due mainly to its action at the two tetO1 sites flanking the TATA sequence, because TetR represses a precursor promoter that only carries such sites to the same extent. The mechanism(s) by which binding of repressors near the TATA might interfere with PIC formation, transcription initiation, or early elongation remain unknown, as well as why binding of larger presumably transcriptionally inert TetR fusion proteins results in stronger repression. However, examination of the Cryo-EM structure of TBP and TFIID bound to mammalian TATA promoters (Nogales et al., 2017) suggests that binding of TetR and larger derivatives of it to these sites might simply block PIC assembly. Studies of repression of native Drosophila melanogaster promoters by the en and eve homeobox proteins show that a similar, steric occlusion based mechanism can block eukaryotic transcription by binding of the repressors to sites close to the TATA sequence (Ohkuma et al., 1990; Austin and Biggin, 1995).
In WTC846, when inducer is absent, measured basal expression of the controlled gene is abolished by a second TetR derivative, a fusion bearing an active repressor protein native to yeast. Because the same TetR-nls-Tup1 fusion protein fully represses a precursor promoter that only carries TetR operators in the UAS, we believe that the main zeroing effect of TetR-nls-Tup1 is manifested through binding the tetO1 sites in the UAS. Native Tup1 repressor complexes with Ssn6 (also called Cyc8) to form a 420 kDa protein complex (Varanasi et al., 1996), and TetR binds DNA as a dimer. In gcr1Δ cells, in which transcription from PTDH3 is severely diminished, the native PTDH3 promoter has two nucleosomes positioned between the UAS and the transcription start site (Pavlović and Hörz, 1988). It is thus possible that in the UAS as many as five very large dimeric TetR-nls-Tup1 complexes block binding of Gcr1 and Rap1, mask their activating domains, or some combination of these, resulting in similar placement of two nucleosomes in P7tet.1. One of these nucleosomes could then be positioned at the 294nt stretch between the UAS and the TATA sequence. It also seems possible that binding of the TetR-nls-Tup1 repressor might shift the position of the second nucleosome further downstream, so that it obscures the transcription start site.
Both the increased input dynamic range and the lower cell-to-cell variation in expression from WTC846 arise from the fact that the TetR protein that represses the controlled gene also represses its own synthesis. This autorepression architecture is common in prokaryotic regulons (Smith and Magasanik, 1971; Brent and Ptashne, 1980) including Tn10, the source of the TetR gene used here, and it has been engineered into eukaryotic systems (Becskei and Serrano, 2000; Nevozhay et al., 2009). In self-repressing TetR systems, the input (here, aTc) and TetR output together function as a comparator-adjustor (Andrews et al., 2016). In such systems, aTc diffuses into the cell. Intracellular aTc concentration is limited by entry. Inside the cell, aTc and TetRfree concentrations are continuously compared by their binding interaction. If TetRfree is in excess, it represses TetR expression, and total intracellular TetR concentration is reduced by dilution, cell division, and active degradation of DNA-bound TetR until an equilibrium determined by the intracellular aTc concentration is again reached. The consequence of this autorepression is that the WTC846 requires more aTc to reach a given level of controlled gene expression than strains in which TetR is expressed only constitutively. Autorepression flattens the dose response curve, increases the range of aTc doses where a change in promoter activity can be observed, and buffers the effects of stochastic cell-to-cell variations in TetR concentration, thereby reducing cell-to-cell variation in expression of the controlled gene throughout the input dynamic range.
We further extended the output dynamic range of WTC846-controlled genes by developing three different Kozak sequences, K1, K2, and K3 (Li et al., 2017), to allow controlled genes to be translated at different levels. We used these sequences to construct strains bearing conditional alleles of the essential genes CDC28, IPL1, TOR2, CDC20, CDC42, PMA1, and PBR1. These strains all showed graded expression of growth and other phenotypes, from lethality at zero expression to penetrant expression of previously reported phenotypes at higher protein dosage (Mnaimneh et al., 2004; Yu et al., 2006; Dechant et al., 2014; Muñoz-Barrera and Monje-Casas, 2014). We used controlled expression in the WTC846::CDC20 strain to bring about G2/M arrest followed by synchronous release, with low cell-to-cell variation in induction timing, demonstrating that WTC846 can be used in experimental approaches that require dynamic control of gene expression.
Both induction and shutoff with WTC846 are rapid, as both Citrine and Cdc20 expression occur within 30 min of induction, and shutoff of Citrine expression is observed within 60 min. However, time to steady state expression after induction, reached when degradation and dilution through cell division balance new synthesis, takes longer. Time to steady state will depend on the stability of the controlled protein. This is 6–7 hr for the stable protein Citrine, and the majority of yeast proteins have similar stability (Wiechecki et al., 2018). Those proteins with shorter half-lives will reach steady state faster. Furthermore, we showed in WTC846::IPL1 strains that high level expression of this spindle assembly checkpoint kinase arrests cells at G2/M with 2 n or higher DNA content. This phenotype, thought to be due to disruption of kinetichore microtubule attachments, is displayed in mammalian cells when the homologous Aurora B is overexpressed (González-Loyola et al., 2015), but had not been observed previously in S. cerevisiae when Ipl1 was overexpressed from PGAL1 (Muñoz-Barrera and Monje-Casas, 2014). We also showed that in WTC846::WHI5 strains, different levels of controlled expression of Whi5 can constrain cell sizes within different limits.
Cell-cell variation in WTC846-driven expression is highest at low aTc levels, because control in this regime depends mostly on the higher variation Simple Repression by TetR-Tup1 expressed from the PRNR2. This variation at low input doses in the WTC846represented a trade-off between the design goals of abolition of basal expression and suppression of cell-to-cell variation. The Autorepression (AR) architecture better suppressed cell-to-cell variation in controlled gene expression at low inducer inputs, but, because of the fact that TetR and the controlled gene were both under the control of the same repressible promoter, the controlled gene still showed considerable basal expression when uninduced. Given that suppression of basal expression of the controlled gene was critical to generating 'reversible null' phenotypes, we developed the AR architecture further. The resulting cAR configuration of WTC846, had low cell-to-cell variation, equivalent to the variation at the lowest expression levels that AR could achieve. Importantly, because transcription of WTC846-controlled genes is synchronized to that of the autorepressing TetR gene, transcription and mRNA abundance of WTC846-controlled genes should be steady throughout the cell cycle. This autorepressing circuitry operationally defines WTC846 as an 'expression clamp', a device for adjusting and setting gene expression at desired levels, and maintaining it with low cell-to-cell variation, and so allowing expressed protein dosage in individual cells to closely track the population average.
Taken together, our results show that WTC846 controlled genes define a new type of conditional allele, one that allows precise control of gene dosage. We anticipate that WTC846 alleles will find use in cell biological experimentation, for example in assessment of phenotypes now incompletely penetrant due to variable dosage of the causative gene products (Casanueva et al., 2012), and for sharpening the thresholds at which dosage dependent phenotypes manifest. We also hope that genome wide collections of WTC846 alleles might enable genome wide gene-by-gene and gene-by-chemical epistasis for interactions that depend on gene dosage. In S. cerevisiae, recent development of strains and methods (such as the SWAP-Tag [Weill et al., 2018]) that facilitate installation of defined N and C terminal genetic elements after cycles of mating, sporulation, and selection of desired haploids should allow generation of whole genome WTC846 strains for this purpose. Epistasis screens rely on measurement of colony size on plates or culture density in liquid media. For two proteins whose effect on growth rate was identical, a one-generation difference in achievement of steady state expression could result in a twofold difference in number of cells in a colony or well, and thus in a 1.26-fold difference in colony diameter. We therefore suggest that growth rate-based assays using WTC846 or any other inducible system pre-induce cells several generations before plating or pinning. WTC846 alleles may find use in engineering applications such affinity maturation of antibodies expressed by yeast surface display, where precise ability to lower surface concentration should aid selection for progressively higher affinity binders. WTC846 can also be a useful complement to boost the efficiency of methods that act at the protein level such as induced degradation or AnchorAway techniques. Such techniques could be used in conjunction with WTC846 to achieve rapid and sustained shutoff from a well-maintained steady state level. This would also allow fast step function decreases in abundance. For example, an experimenter might simultaneously induce depletion of the product of a controlled gene by such a method while adjusting aTc downward to rapidly reset the level of an expressed protein to a new, lower level. Implementation of the WTC846 control logic in mammalian cells and in engineered multicellular organisms should allow similar experimentation now impossible due to cell-to-cell variation and imprecise control.
Information on plasmids, and promoter and protein sequences used in this study can be found in Supplementary file 1 - Tables S2 and S4. Plasmids with auxotrophic markers were constructed based on the pRG shuttle vector series (Gnügge et al., 2016) using either restriction enzyme cloning or isothermal assembly (Gibson et al., 2009). Inserts were generated either by PCR on existing plasmids or custom DNA synthesis (GeneArt, UK). Oligos for cloning and for strain construction were synthesized by Thermofisher, UK. Plasmids used to generate linear PCR products for tagging transformations were based on the pFA6 backbone (Janke et al., 2004). Plasmids necessary to construct WTC846 strains are available through Addgene. Plasmid structures and a detailed protocol for strain construction can be found in Appendix 5.
pRG shuttle vector series backbones used for integrative transformations have T7 and T3 promoters flanking the insert (Gnügge et al., 2016). During cloning, the insert of plasmids bearing TetR were cloned such that the insert promoter was closer to the T7 promoter and the terminator was near the T3 promoter of the backbone. In plasmids bearing Citrine, the insert was flipped onto the opposite strand, such that the insert promoter was near the T3 promoter, and the terminator near the T7 promoter. This inversion was done to avoid homologous recombination during subsequent integration of these plasmids into the same strain, since in many strains TetR and Citrine were flanked by the same promoter and the same terminator.
Strains used in this study can be found in Supplementary file 1 - Table S1. Strains used for fluorescent measurements and the WTC846-K3::TPI1 strain are based on a BY4743 derivative haploid background (MATa his3Δ leu2Δ met15Δ ura3Δ lys2Δ). Strains where P7tet.1 replaced endogenous promoters were based on the haploid BY4741 background with the modifications whi5Δ::WHI5-mKOkappa-HIS3, myo1Δ::MYO1-mKate(3x)-KanMX and so were resistant to G418. The oligos used to replace the promoters of the different endogenous genes with WTC846-controlled P7tet.1 can be found in Supplementary file 1 - Table S3. Correct replacement of the endogenous promoter with P7tet.1 was checked using colony PCR with the protocol from the Blackburn lab (also detailed in Gnügge et al., 2016), and subsequent sequencing (Microsynth, Switzerland). For colony PCR, we used a standard forward oligo annealing to P7tet.1, and gene specific reverse oligos annealing within the tagged gene. Oligo sequences for colony PCR can be found in Supplementary file 1 - Table S3. A comprehensive protocol on how to generate strains where WTC846 controls endogenous genes can be found in Appendix 5.
YPD/YPE was prepared with 1% yeast extract (Thermofisher, 212720), 2% bacto-peptone (Thermofisher, 211820), and 2% glucose (Sigma, G8270) / ethanol (Honeywell, 02860). Synthetic (S) media except SD Proline contained 0.17% yeast nitrogen base (without amino acids and ammonium sulfate) (BD Difco, 233520) with 0.5% ammonium sulfate (Sigma, 31119) as nitrogen source, complete complement of amino acids and adenine and uracil, except for SD min which contained only the necessary amino acid complements to cover auxotrophies. SD Proline media contained 0.17% yeast nitrogen (without amino acids and ammonium sulfate), only the amino acids necessary to cover auxotrophies and 1 mg/mL proline as the sole nitrogen source. The carbon source was 2% glucose for SD and SD Proline, 2% ethanol for S Ethanol, 3% glycerol for S Glycerol (Applichem, A2957), 2% fructose for S Fructose, 2% Raffinose for S Raffinose and 2% Galactose together with 2% Raffinose for S GalRaf. Experiments were performed in YPD media unless otherwise specified. Solid medium plates were poured by adding 2% agar (BD Sciences, 214040) to the media described above.
aTc was purchased from Cayman Chemicals (10009542) and prepared as a 4628.8 ng/mL (10 mM) stock in ethanol for long term storage at −20°C and diluted in water for experiments as necessary.
When constructing strains where P7tet.1 replaces endogenous promoters, a PCR fragment containing P7tet.1 and an antibiotic marker (either Nourseothricin (Werner BioAgents, clonNAT) or Hygromycin (ThermoFisher,10687010)) was transformed for homologous recombination directed replacement of the endogenous promoter. Cells were plated on YPD + antibiotic plates for selection. Whenever the promoter of an essential gene was being replaced, transformations were plated on multiple plates with YPD + antibiotic and 10/50/100/500 ng/mL aTc.
For spotting assays of cell growth and viability, cells were precultured in YPD media with 20 ng/mL aTc (except for WTC846-K2::IPL1 strain which was precultured in 10 ng/mL aTc) and the necessary antibiotic to stationary phase, and diluted into YPD + antibiotic without aTc at a concentration of 0.8x106 cells/mL. Six hr later, cells were spun down and resuspended in YPD. Cells were spotted onto plates containing different media and aTc concentrations prepared as described above such that the most concentrated spot has 2.25x106 cells, and each column is a 1:10 dilution. Pictures were taken after 24 hr for the YPD and SD plates, and 42 hr for SD Proline, S Glycerol and YPE plates.
Cells were diluted 1:200 from dense precultures and cultured to early exponential phase (2–5 x 106 cells/mL) in 96 deep-well plates at 30°C before induction with aTc if necessary. For aTc dose responses, samples were taken at times indicated. For experiments where no dose response was necessary, cells were measured at least 4 hr after dilution of precultures, but always before stationary phase. Samples were diluted in PBS and measured using a LSRFortessa LSRII equipped with a high-throughput sampler. PMT voltages for the forward and side scatter measurements were set up such that the height of the signal was not saturated. Citrine fluorescence was quantified using a 488 nm excitation laser and a 530/30 nm emission filter. PMT voltage for this channel was set up such that the signal from PTDH3 expressed Citrine did not saturate the measurement device, except for basal level measurements in Figure 3B and Figure 3—figure supplement 3, where PMT voltage for the Citrine channel was increased to maximum. Side scatter was measured using the 488 nm excitation laser and 488/10 nm emission filter.
Cells were grown to stationary phase with the indicated aTc concentration. 5 mL of cell culture was centrifuged and resuspended in 1 mL 70% ethanol. Fixed cells were again centrifuged, and resuspended in 200 uL Trupage LDS loading buffer (Merck, PCG3009) supplemented with 8M urea. Cells were broken using glass beads and a bead beater, and boiled at 95°C for 30 min. Proteins were separated using SDS-Page with Trupage precast 10% gels (Merck, PCG2009-10EA) and the associated commercial buffer, and transferred onto a nitrocellulose membrane (GE Healthcare Life Sciences, 10600008).
We used mouse monoclonal primary antibodies for detecting TetR (Takara, Clone 9G9), and Citrine (Merck, G6539), both diluted 1:2000 in Odyssey Blocking buffer (PBS) (LI-COR Biosciences) + 0.2% Tween 20. The secondary antibody was the near-infrared fluorescent IRDye 800CW Goat anti-Mouse IgG Secondary Antibody from Li-Cor (926–32210), diluted 1:5000 in the same manner. We used Chameleon Duo pre-stained Protein Ladder as our molecular weight marker (928-60000). We used the SNAP i.d. 2.0 system which uses vacuum to drive reagents through the membrane, and the Odyssey CLx (LI-COR) detector for imaging. Images were processed using the Fiji software to obtain black and white images with high contrast (Schindelin et al., 2012).
Cells were precultured in YPD (with aTc in the case of strains where WTC846 controlled essential genes) to stationary phase, then diluted into fresh media at a concentration of 50.000 cells per mL and induced with the necessary aTc concentrations, except for YP Ethanol and S Ethanol media where the concentration was 500,000 cells per mL. The Growth Profiler 960 (EnzyScreen) with 96-well plates and 250 µL volume per well, or Biolector (m2p-labs) with 48 well plates and 1 mL volume per well was used to measure growth curves. These are commercial devices that quantify culture density by detecting the light that is reflected back by the liquid culture.
WTC846-K3::CDC20 and the appropriate control strains were precultured in YPD (pH 4) with with the indicated aTc concentration to a concentration of 2x106 cells/mL, then centrifuged and diluted 1:3 into YPD (pH 4) without aTc. We found that low pH (pH4) of the media was necessary for efficient mother-daughter separation upon completion of cytokinesis, potentially due to the low pH optimum of the chitinase CTS1 (Hurtado-Guerrero and van Aalten, 2007), which plays a role in separation. For the experiment presented in Figure 5E, to prevent the culture from becoming too dense, 25% of the media was filtered and returned to the culture after 4 hr of growth without aTc, which removed 1/4th of the cells. If release was performed, this was done after 8 hr of arrest by adding 600 ng/mL aTc to the culture. Samples were taken at indicated time points before, and every 5 min after aTc was added to the culture, and fixed with 70% ethanol. For the experiment presented in Figure 5E, to aid mother-daughter separation, the samples were sonicated for 1 min in a water bath before fixation.
Samples for DNA staining were digested with 5 mg/mL proteinase K for 50 min at 50°C, followed by 2 hr of RNase A (Applichem, A2760,0500) treatment at 30°C. Samples were stained for DNA content using SYTOX Green (Thermofisher, S7020) diluted 1:5000 in PBS, and were sonicated in a water bath for 25 s before flow cytometry. Fluorescence was detected using a 488 nm excitation laser and a 525/15 nm emission filter. The PMT voltage was set up such that the sample with the highest expected ploidy did not saturate the signal.
Cells were grown to early exponential phase(∼3 million cells/mL) in YPD at 30°C with shaking and induced with 600 ng/mL aTc. Two mL samples were taken at indicated time points. To remove excess aTc, cells were spun down for 20 s, supernatant was removed and cells were resuspended in YPD. This process was repeated three times. After the 3rd resuspension, the 2 mL sample was divided between two wells of a 96 deep-well plate. Cycloheximide was added to one of the wells at a final concentration of 70 μg/mL. The plate was continuously shaken at 30°C. Citrine fluorescence was measured every 30 min using flow cytometry as explained above.
All analysis was performed using R (R Development Core Team, 2013), and the packages Bioconductor (Ellis et al., 2009), dplyr (Wickham et al., 2018), drc (Ritz et al., 2015), MASS (Kafadar et al., 1999), mixtools (Benaglia et al., 2009), and ggplot2 (Ginestet, 2011). All raw data that is not provided as source data here is available publicly at doi.org/10.3929/ethz-b-000488967.
Flow cytometry data was not gated except when necessary to remove debris. For aTc dose response experiments, median fluorescence of the entire population was used to fit a five-parameter dose response curve with the drm() command and the fplogistic formula from the drc package. Parameters p_1 and p_2 were fixed individually for each curve, the rest of the parameters were estimated by the drm command. Parameter values can be found in Supplementary file 1 - Table S5. The cytometry cell volume proxy was always calculated as the magnitude of the vector of the FSC-W and SSC-H signals (), since forward and side scatter signals provide information about cell volume and budding state. The forward scatter width and side scatter height were chosen because this combination (as opposed to other combinations involving FSC-H/SSC-W or area of the signals) showed the most separation between measured signal peaks corresponding to spherical calibration beads of known diameter.
For single-reporter quantification of VIV, we calculated the residual standard deviation (RSD) of a linear model describing the relationship between the cytometry cell volume proxy and fluorescence of the population. To do this, the rlm() command from the MASS package was used to generate the linear model, and the residual standard deviation given by the same rlm() command was used as our measure of VIV. See Appendix 2 for a detailed explanation of the method.
Where shown, error bars for median fluorescence and the RSD were calculated using bootstrapping. The original set of data points was sampled with replacement and median fluorescence or RSD was calculated. 95% confidence intervals were calculated based on 1000 repetitions of this sampling process and plotted as error bars.
To generate a linear model describing the relationship between the volume proxy measurements done by flow cytometry and volume measurements by Coulter Counter, first the two data sets were sampled with replacement 5000 times. Then these samples were ordered by increasing volume proxy or volume and merged. The lm() command in the R package stats was used to fit the linear model. Sub-populations from Gates 4, 6, and 9 were not included in the fitting, as these medians were deemed suboptimal representations of the bimodal distributions. The resulting linear fit had a slope of 471 and an intercept of 62032.
When creating P7tet.1, the final promoter used in WTC846, we optimized the placement of the tetO1 sequences, the sequences of the endogenous Gcr1 and Rap1-binding sites, the number of these sites present in the promoter, and the TATA sequence. Our goal was to increase maximum expression from the promoter, because PTDH3 derivatives we had constructed with tetO1 sites had shown reduced expression. For these optimizations we used as a starting promoter P4tet. This promoter is a variant of P5tet (Figure 1B) from which the tetO1 sequence immediately upstream of the TATA was removed. P4tet showed a higher basal expression level than P5tet, allowing us to better observe subtle differences in basal expression (Appendix 1—figure 1B). We tested repression of P4tet and derivatives in the SR architecture. We constructed the derivatives as follows. We extended the Rap1 site and the upstream Gcr1 site by one base pair (P4tet.1 and P4tet.2), to account for the possibility that we initially truncated the endogenous binding sites Metzger et al., 2015, replaced the downstream Gcr1 binding site with the same extended Gcr1 site (P4tet.3), since the upstream Gcr1 binding site was closer to the reported consensus sequence Huie et al., 1992, and tested alternative TATA sequences (P4tet.4 and P4tet.5) Mogno et al., 2010. Four of these optimizations resulted in increased maximum activity (Appendix 1—figure 1B): the Rap1 site extension (P4tet.1), replacement of the downstream Gcr1 site (P4tet.3), and the TATA sequence optimizations (with P4tet.5 driving expression more strongly than P4tet.4). P4tet.2 with the extended Gcr1 site showed reduced maximum expression.
We therefore chose the following modifications: the single base pair extension of the upstream Rap1 site, the replacement of the downstream Gcr1 site with the original upstream Gcr1 sequence, and the TATA sequence TATAAATA. We implemented these modifications to P5tet to generate P5tet.1. By the assays described in the main text for testing the other promoter derivatives, compared to P5tet, the new promoter P5tet.1 (Y2659) showed 95% of maximum expression driven by PTDH3, and increased repression (15 fold vs. 12 fold) with only a slight increase in basal activity when fully repressed (Y2656, Figure 1B and Appendix 1—figure 1C). In order to increase maximum expression even further, we took advantage of previous work showing that increasing the number of transcription factor binding sites in a promoter could increase its strength Ottoz et al., 2014. We investigated whether adding additional Rap1 and Gcr1 sites to P5tet.1 would increase promoter activity. We created P7tet by duplicating the Rap1 site, and P7tet.1 by duplicating both the Rap1 and one of the two Gcr1 sites in P5tet.1, while keeping the same tetO1 placements at these duplicated sites (Appendix 1—figure 1C). P7tet.1 had a higher maximum activity (116% vs 99% of PTDH3 activity) and fold repression than P7tet (20-fold vs 18-fold), with only minimal increase in absolute repressed activity (4.3-fold vs fourfold above autofluorescence). We therefore chose P7tet.1 as the promoter for further use.
Appendix 1—figure 1—source data 1
In yeast and C. elegans, comparison of signals from strains with different combinations of different reporter genes allows the different contributions to variation to be independently quantified (Colman-Lerner et al., 2005; Mendenhall et al., 2015). One of these contributions, individual differences in general ability to express genes into proteins, contributes to phenotypic variation in genetic penetrance and expressivity (Burnaevskiy et al., 2019). Quantification of this and other sources of variation benefits from ability to measure output of single cells over time (Colman-Lerner et al., 2005) and, in flow cytometry, requires measurement of outputs of multiple reporters (Pesce et al., 2018). Here, however, we were interested in the overall variability rather than specific sources of variability. Therefore, we only had a single reporter protein (Citrine). However single reporter studies have a major, confounding contribution to measured variation in gene expression that multi-reporter studies don’t: Fluorescent proteins in yeast are degraded very slowly unless they have degradation tags attached (Gordon et al., 2007) and therefore, if constitutively expressed, their abundance increases over time (Cookson et al., 2010). Thus, in cycling populations of budding yeast that continually express fluorescent proteins, a major source of cell-to-cell variation in fluorescent signal is that small, new-born cells have not had time to accumulate much fluorescent protein, while larger cells have. This source of variability normally affects all reporter proteins in the cell in a similar fashion, and therefore does not require correcting in multi-reporter studies. On the other hand in single reporter studies with flow cytometry in yeast, as for higher cells, this volume related variation in fluorescent protein expression is generally corrected for by gating; that is filtering the data to select only a narrow subset of cells with similar forward and side scatter, and thus volume, which increases with cell cycle progression. Such gating disregards data from the majority of the cells whose values fall outside the gated range. Here, in order to avoid discarding data, we established a single-reporter measure of cell-to-cell variation that corrects for variation due to fluorescent protein accumulation without gating.
We first established that forward and side scatter signals can be used to distinguish smaller cells from larger ones. We sorted cells on a BD FACS Aria III flow cytometer. We set different gates on the FSC and SSC signals (shown in Appendix 2—figure 1) to collect 10 sorted sub-populations, each containing about 100,000 exponentially growing Y2683 cells. We then immediately measured (a) FSC and SSC from the collected subpopulations on a different instrument, the LSRII Fortessa LSR used for the flow cytometric measurements in this work, and (b) volume in fL with a Coulter Counter (Appendix 2—figure 2). The raw data acquired by the two methods can be seen in Appendix 2—figure 2A,B. For the flow cytometry data, we used the width of the FSC and height of the SSC to calculate a volume proxy using the formula as explained in Materials and methods. Figure 2C shows a linear relationship between the medians of the sub-populations as measured by the two methods, that is, that the flow cytometric measurement is a proxy for volume, and that two volume measurements qualitatively agree. The three sub-populations (4, 6, and 9) where a slight deviation from the linear relationship is observed are all bimodally distributed, meaning the population is a mixture of large and small cells and the median is not a good representation of this sub-population. Overall, for the cell-to-cell variation calculations outlined below, this relative relationship is enough to distinguish new-born, smaller cells from larger cells that have had time to accumulate fluorescent protein.
Appendix 2—figure 2—source data 1
We then used this information to calculate the CCV in fluorescent protein expression that could not be attributed to differences in cell volume/cell cycle progression, and called this method Volume Independent Variation (VIV). We began with Y2683 cells, which express Citrine from wild type PTDH3. We then plotted the volume proxy vs. the fluorescence signal observed in the entire population measured by flow cytometry (Figure 2—figure supplement 2). Then we performed a robust linear fit on the cell volume proxy vs. the Citrine fluorescence signal. This linear model allowed us to correct for the differences in cell volume and calculate the RSD of the fit as explained in the Materials and methods. This RSD value quantifies the variation in the population that is not due to differences in cell volume between the cells. While use of this measure is in principle akin to measuring variation in expression of fluorescent proteins using a very narrow gate on the measured FSC vs SSC signals of the population, it avoids the need to discard data. Moreover, it could also be used when comparing populations with different cell volume distributions.
We reasoned that basal expression from P7tet.1 in the WTC846 architecture might arise because (a) the nuclear concentration of TetR might be too low for all of the tetO1 TetR binding sites in P7tet.1 to be occupied at all times, and/or (b), that TetR derivatives might fully occupy all of the operators and yet not repress completely. We tested the first idea by increasing the nuclear concentration of TetR proteins by expressing derivatives that contained a second SV40 Nuclear Localization Sequence. We tested the second idea by fusing TetR to other protein moieties that might aid repression. Specifically, we added to TetR portions of prokaryotic proteins that we could presume to be inert, hoping that these bulkier TetR derivatives might repress more strongly, for example by better sterically interfering with the binding of transcription factors, or with contacts between Gcr1 and Rap1 at the UAS and the transcription apparatus at the core promoter. We tested the efficacy of these new molecules by expressing them from PACT1 in the SR architecture (Y2681, Y2664, Y2665, Y2666, and Y2667, Appendix 3—figure 1). For smaller repressors, addition of a second NLS decreased uninduced expression whereas for larger repressors it did not. The strain carrying the TetR-nls-MBP (Maltose Binding Protein, the E. coli malE gene product), showed the most repression, but still exhibited uninduced expression signal of 2.2-fold above autofluorescence background.
Appendix 3—figure 1—source data 1
In the experiments in Figure 4, WTC846-controlled Citrine takes around 7 hr to reach steady state concentration. Here we present a simple ODE model to demonstrate that the time to steady state will change based on the stability of the controlled protein. In the model (Equation 1) the protein of interest is produced at a constant rate a and lost with a lumped linear rate (dilution + degradation) d. Analytical solution (Equation 2) of this model shows that the only constant that affects the time variable is the degradation + dilution rate. Dependence on this variable is also evident in simulations based on this model (Appendix 4—figure 1). The smaller d is, (i.e. the more stable the protein is), the longer it takes to reach steady state. On the other hand, changes in the production rate a have no effect on time to steady state, although both rates affect the maximum level. Citrine is a remarkably stable protein (see Figure 4—figure supplement 8), but recent data suggests that most (somewhere between 50–85% depending on the data set) of the yeast proteome is just as stable Wiechecki et al., 2018. Therefore most other WTC846-controlled endogenous proteins will likely have a time to steady state around 7 hr, except those with a shorter half-life which will exhibit a shorter time to steady state.
The WTC846 is a two unit transcriptional control system for S. cerevisiae (Appendix 5—figure 1A). An inducible promoter (P7tet.1) is placed in front of the Gene of Interest (GOI). The promoter is based on an engineered version of the strong constitutive promoter of TDH3. It was made repressible by placing TetR-binding sites next to the binding sites for the transcriptional machinery. As a result, binding of the TetR protein can prevent binding of the endogenous proteins which normally drive transcription. The repressors TetR and TetR-Tup1 are found on one integrative repressor plasmid (Appendix 5—figure 2A). TetR is expressed under the above described promoter (P7tet.1) creating an autorepression loop. TetR-nls-Tup1 abolishes the basal activity of P7tet.1 and is expressed under the control of the weak, constitutive RNR2 promoter.
To create a functional system, we advise to first integrate the repressor plasmid. The P7tet.1 can then be placed in front of any gene in the genome using PCR tagging Janke et al., 2004. The tagging plasmid (based on Janke et al., 2004) is used as a template (Appendix 5—figure 3). We provide two versions of the P7tet.1 followed by a flag tag followed by a linker composed of eight glycine residues; either cloned in a plasmid providing a HygR marker (P2350), or a NAT marker (P2375). For PCR-based tagging, the 5’ and 3’ ends of the PCR fragment need to be complementary to a sequence upstream of the GOI and to the beginning of the GOI, respectively. This is ensured by using primers with tails complementary to these regions. We tested the plasmid for use with and without the flag tag.
Induction of the tagged gene can then be controlled by aTc, a small molecule that causes TetR to dissociate from its binding sites on P7tet.1. An example is seen in Appendix 5—figure 1B, where Citrine expression was controlled across a large expression range using aTc. Tetracycline or Doxycycline can also be used, although they will likely require different concentrations compared to aTc.
The basal activity of P7tet.1 can be controlled by the Kozak sequence (last 15 bp before the start codon of the gene of interest). The provided sequence in the P2350 and P2375 plasmids shows no detectable basal Citrine expression. However, even a small basal expression level can become an issue if the GOI encodes a protein that is required in very small numbers. We encountered this problem with Tor2, Cdc28 and similarly low abundance, stable proteins. In this case, changing the translation efficiency by modifying the Kozak sequence allowed us to abolish all basal expression. The protocol below also explains how to achieve this.
Transform the repressor plasmid in a strain that has the correct auxotrophic marker deletion. The plasmid should be linearized using AscI digestion for integrative transformation Gnügge et al., 2016. The repressor plasmids are given in Appendix 5—table 1:
Design primers to create the tagging fragment from the tagging plasmid (P2350 or P2375).
Forward primer: Use the sequence agcttgccttgtccccgcc as the annealing part of the forward primer. Select 40 base pairs anywhere upstream of the GOI, and use this sequence as the 5’ tail of your forward primer. Remember that the region between these 40 base pairs and the start codon of the gene will be deleted during the transformation. You can thus remove the entire natural promoter of the gene, but this is not mandatory.
Reverse primer option 1 - without flag tag: Take tttattcgaaactaagttcttggtg as the annealing portion of your reverse primer, which will anneal to the sequence caccaagaacttagtttcgaataaa on the plasmid. Then use the reverse complement of the first 40 base pairs (including ATG) of the GOI, followed by the reverse complement of the desired Kozak sequence as your 5’ tail, such that the primer reads: 5’-reverse GOI sequence-CAT-reverse Kozak sequence-annealing portion-3’.
Kozak sequence to modulate expression: The Kozak sequences that we have tested, in decreasing order of translation efficiency are (reverse complement is given in parentheses):
Reverse primer option 2 - including flag tag: If you would like to include the flag tag at the start of the gene, use the sequence catcgatgaattctctgtcgg as the annealing portion of your reverse primer, which will anneal to the standard S4 primer binding site on the plasmid (ccgacagagaattcatcgatg) . In this case the Kozak sequence cannot be altered, and the one already on the tagging plasmid has to be used (this is the first one in the list above). Use the reverse complement of the first 40 bases (after ATG) of the gene as the 5’ tail of your reverse primer such that the primer reads: 5’-reverse GOI sequence-annealing portion-3’.
Perform the tagging PCR to generate the tagging fragment using the primers designed in the previous step. If your standard PCR protocol fails (occasionally happens due to the long tails on the oligos), use the PCR protocol detailed below (adapted from Janke et al., 2004):Reaction Setup (200μL): 20μL Taq/Vent Buffer*, 35μL 2mM dNTPs, 2μL tagging plasmid, 0.5μL 100μM forward primer, 0.5μL 100μM reverse primer ,0.8μL Taq polymerase, 0.4μL Vent polymerase, 134.4μL ddH2O. (*Buffer composition: 500mM Tris/HCl(pH=9.0) 22.5mM MgCl2 160mM NH4SO4). PCR program: (a) 95°C 5min (b) 95°C 1min (c) Ta 30sec (d) 68°C 1min per kb (e) 95°C 1min (f) Ta 30sec (g) 68°C 1min/kb + 20sec per cycle (h) 9°C hold. Repeat steps (b-d) 10x, (e-g) 20x.
Gel isolate and transform the tagging fragment into the strain created in step 1. Select on solid medium with the appropriate antibiotic and aTc. If the GOI is an essential gene, the transformation efficiency will be low. In order to increase transformation efficiency, pre-culture, recovery media for the cells and selection plate should all contain aTc. Note that the required aTc concentrations are around five times higher in solid media than in liquid media to achieve the same expression level.
Correct integration can be confirmed using colony PCR. Use sequence cagttcgagtttatcattatcaatactg as the forward primer (binds at the start of P7tet.1), and a reverse primer that anneals within the GOI. The fragment length will depend on where in the GOI the reverse primer anneals. (This forward primer will work for all cases except when the TDH3 promoter is being replaced. Since P7tet.1 is based on the TDH3 promoter, this primer will anneal to the promoter whether or not the replacement was successful.) Integration efficiency is low when tagging essential genes (about 10% of colonies screened), but a positive PCR result generally is enough to indicate correct integration. However it is best to isolate the PCR fragment and sequence the entire promoter to confirm correct integration.
All relevant sequences are included in the supporting files for reproducibility. All raw flow cytometry data is publicly available at https://doi.org/10.3929/ethz-b-000488967. All other source data is included in the manuscript and supporting files.
Biogenesis and function of the yeast plasma-membrane H(+)-ATPaseJournal of Experimental Biology 203:155–160.https://doi.org/10.1242/jeb.203.1.155
Insight into Tor2, a budding yeast microdomain proteinEuropean Journal of Cell Biology 93:87–97.https://doi.org/10.1016/j.ejcb.2014.01.004
Mixtools: an R package for analyzing finite mixture modelsJournal of Statistical Software 32:1–29.https://doi.org/10.18637/jss.v032.i06
Cell cycle-dependent variations in protein concentrationNucleic Acids Research 38:2676–2681.https://doi.org/10.1093/nar/gkp1069
Tetracycline-reversible silencing of eukaryotic promotersMolecular and Cellular Biology 15:1907–1914.https://doi.org/10.1128/MCB.15.4.1907
SoftwareflowCore: Basic structures for flow cytometry data, version 2.4.0flowCore.
ConferencePhysiological studies of conditional lethal mutants of bacteriophage T4DCold Spring Harbor Symposia on Quantitative Biology. pp. 375–394.https://doi.org/10.1101/SQB.1963.028.01.053
The top genes: on the distance from transcript to function in yeast glycolysisCurrent Opinion in Microbiology 6:198–201.https://doi.org/10.1016/S1369-5274(03)00023-7
Transcriptional repression in eukaryotes: repressors and repression mechanismsCellular and Molecular Life Sciences 60:721–741.https://doi.org/10.1007/s00018-003-2260-3
ggplot2: elegant graphics for data analysisJournal of the Royal Statistical Society: Series A 174:245–246.https://doi.org/10.1111/j.1467-985X.2010.00676_9.x
Aurora B overexpression causes aneuploidy and p21Cip1 repression during tumor developmentMolecular and Cellular Biology 35:3566–3578.https://doi.org/10.1128/MCB.01286-14
Reducing the environmental sensitivity of yellow fluorescent protein. mechanism and applicationsThe Journal of Biological Chemistry 276:29188–29194.https://doi.org/10.1074/jbc.M102815200
Sequential gene function in the initiation of Saccharomyces cerevisiae DNA synthesisJournal of Molecular Biology 84:445–461.https://doi.org/10.1016/0022-2836(74)90451-3
Methods of synchronization of yeast cells for the analysis of cell cycle progressionMethods in Molecular Biology 1505:19–34.https://doi.org/10.1007/978-1-4939-6502-1_2
Nucleotides upstream of the Kozak sequence strongly influence gene expression in the yeast S. cerevisiaeJournal of Biological Engineering 11:25.https://doi.org/10.1186/s13036-017-0068-1
Gene regulation at the right operator (OR) bacteriophage lambda. I. OR3 and autogenous negative control by repressorJournal of Molecular Biology 139:147–161.https://doi.org/10.1016/0022-2836(80)90302-2
Systems for applied gene control in Saccharomyces cerevisiaeBiotechnology Letters 30:979–987.https://doi.org/10.1007/s10529-008-9647-z
Gene regulation at the right operator (OR) of bacteriophage λJournal of Molecular Biology 139:195–205.https://doi.org/10.1016/0022-2836(80)90304-6
TATA is a modular component of synthetic promotersGenome Research 20:1391–1397.https://doi.org/10.1101/gr.106732.110
Structural insights into the eukaryotic transcription initiation machineryAnnual Review of Biophysics 46:59–83.https://doi.org/10.1146/annurev-biophys-070816-033751
Single-cell profiling screen identifies microtubule-dependent reduction of variability in signalingMolecular Systems Biology 14:e7390.https://doi.org/10.15252/msb.20167390
Regulation of APC/C activators in mitosis and meiosisAnnual Review of Cell and Developmental Biology 24:475–499.https://doi.org/10.1146/annurev.cellbio.041408.115949
SoftwareR: A language and environment for statistical computingR Foundation for Statistical Computing, Vienna, Austria.
Functional expression of the cre-lox site-specific recombination system in the yeast Saccharomyces cerevisiaeMolecular and Cellular Biology 7:2087–2096.https://doi.org/10.1128/MCB.7.6.2087
Fiji: an open-source platform for biological-image analysisNature Methods 9:676–682.https://doi.org/10.1038/nmeth.2019
The Cyc8 (Ssn6)-Tup1 corepressor complex is composed of one Cyc8 and four Tup1 subunitsMolecular and Cellular Biology 16:6707–6714.https://doi.org/10.1128/MCB.16.12.6707
A temperature-sensitive lambda cI repressor functions on a modified operator in yeast cells by masking the TATA elementMolecular and General Genetics MGG 248:499–505.https://doi.org/10.1007/BF02191651
Softwaredplyr: A Grammar of Data Manipulation, version 0.8.0.1Dplyr.
Integrative Meta-Analysis reveals that most yeast proteins are very stableSSRN Electronic Journal 234:36.https://doi.org/10.2139/ssrn.3155916
The UAS of the yeast GAPDH promoter consists of multiple general functional elements including RAP1 and GRF2 binding sitesJournal of Veterinary Medical Science 56:235–244.https://doi.org/10.1292/jvms.56.235
A survey of essential gene function in the yeast cell division cycleMolecular Biology of the Cell 17:4736–4747.https://doi.org/10.1091/mbc.e06-04-0368
Redundant mechanisms are used by Ssn6-Tup1 in repressing chromosomal gene transcription in Saccharomyces cerevisiaeJournal of Biological Chemistry 279:39240–39250.https://doi.org/10.1074/jbc.M407159200
Naama BarkaiSenior and Reviewing Editor; Weizmann Institute of Science, Israel
In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.
[Editors' note: this paper was reviewed by Review Commons.]https://doi.org/10.7554/eLife.69549.sa1
Summary review comments and responses
Both reviewers were obviously expert and their comments positive and constructive. Taking them together, the major comments we received on our manuscript fall into three different categories.
1. Need for clarification, either in terms of wording (Reviewer 1, #1) or more detailed explanation of the design process (R1, #2)
2. Requests for us to further elaborate on our results and their implications (R1, #3,4,6,7,9,10,11)
3. Requests for additional characterization of the system (R1, #5,8). In this category there was also one minor comment from Reviewer 2 that requested additional controls.
Actions we have taken.
We have addressed all major and minor comments and performed additional experiments where necessary.
1. We adjusted the wording as suggested and added text to clarify the design process.
2. We responded to each comment in this category in detail.
a. R1#3, we added text and a simple computational model for clarification.
b. R1#4,7,9,10,11 we added text to further elaborate on the points raised.
c. R1#6, we added text and performed an additional experiment to further elaborate on our results.
3. We performed all the requested additional characterization and control experiments.
a. R1#5, we performed two additional experiments to further characterize the shut-off speed of the system.
b. R1#8, as requested, we characterized two other transcriptional control systems and compared them to our system in terms of cell-to-cell variation.
c. R2#m1, we performed the control experiment requested.
In addition, we addressed all minor comments as suggested by the reviewers.
Detailed comments and responses
Reviewer #1 (Evidence, reproducibility and clarity (Required)):
Azizoglu et al. describe a novel system for highly regulated, conditional expression of any gene in budding yeast, with the goal of facilitating quantitative functional genetic analysis. The authors aimed to meet 6 criteria in their system: (1) operation in all growth media; (2) regulation by an exogenous small molecule with minimal effect on cell function; (3) zero basal expression to obtain null phenotypes; (4) large range of precisely adjustable expression; (5) very high maximal expression; and (6) low cell-to-cell variability at any output level. The data shown indicate that they have more or less achieved these aims and that the system they describe is an advance on current methods. The logic behind the development of the system is very clearly presented, as well as its relationship to previous work. In addition, the authors place their work in a very broad historical perspective, which is unusual but much appreciated by this reviewer. Finally, they complement their description of the system (which employs a fluorescence read-out) with several case studies on a variety of different yeast genes that demonstrate its utility and highlights its advantages relative to the current state-of-the-art methods. This study thus provides an important new tool for quantitative analysis of the effect of expression level on phenotype that should be of wide interest in the yeast genetics community as a discovery tool.
We appreciate this generous summary.
We note that the approach to setpoint gene regulation is general beyond yeast. We expect it to be applied to vertebrate systems and we expect it to be applied to plants. We also note that, although its primary application is genetic discovery, it is finding use in some biotechnology applications, for example for tuning the expression of genes in multistep biochemical pathways to find a level that results in maximum synthesis of an end product.
Primary comments and suggestions
1. The phrase "Expression of genes conditionally into phenotype…", used in the Abstract, and in other guises elsewhere in the text, is awkward and confusing. A more comprehensible alternative could be: "The relationship between gene expression level and phenotype remains central to biological research. Current methods enable either on/off or imprecisely controlled graded expression…"
We thank the reviewer. In response, we changed the wording.
New abstract now reads "Conditional expression of genes and observation of phenotype remain central to biological discovery".
The variant of the awkward phrase occurs in one other place, in the first sentence of the introduction. In response to the reviewer, we changed the revised text to read
“… means to express genes conditionally to permit observation of the phenotype have remained.…”
Old text read:
“… means to express genes conditionally into phenotype have remained.…”
2. It seems to me that the AR strain (Figure 2A, and Supplementary Figure 4) displayed less cellto-cell variation than the cAR variant. Why was this strain not pursued further as an alternative for cases in which this feature might be more desirable?
We thank the reviewer for this question. We should have been clearer.
Yes, the AR strain displays lower cell-to-cell variation compared to the cAR strain in Figure 2A, especially at lower doses of aTc. But this was not true for the final cAR constructions that comprise WTC846 (compare Figure 4 and the AR strain in Figure 2). When comparing the AR strain in Figure 2 and the final WTC846 strain in Figure 4, at lowest expression levels, cell-to-cell variation is almost the same.
There are two differences between the cAR strain presented in Figure
2 and the final WTC846 strain: the former has TetR instead of TetRTup1 as the constitutively expressed element of the architecture, and more importantly, the promoter for this constitutively expressed repressor in the original cAR strain (PACT1) is much stronger than in the final WTC846 strain (PRNR2). As a consequence, the starting cAR strain constitutively expressed a much larger amount of repressor, which was not subject to the cell-to-cell variation reducing effects of the negative feedback loop. The lower constitutive expression of repressor in the WTC846 strain means that a greater proportion of the cellular complement of repressor is autorepressed, and variation is suppressed.
The answer to why we pursued the cAR architecture after the experiments in Figure 2 lies in the initial criteria we had set for our ideal transcriptional controller. These are enumerated in the Introduction section. The two relevant criteria read: "(3) manifest no basal expression of the controlled gene in absence of inducer, allowing generation of null phenotypes” and "(6) exhibit low cell-to-cell variability at any set output…” In the case of our AR strain, we reasoned we could never fulfil criterion (3), as long as both the repressor and the controlled gene were expressed from the same TetR repressible promoter. If we wished the gene of interest to have no basal expression, then by definition TetR should have no basal expression, which would then mean there is no TetR to repress the promoters.
The cAR strain therefore embodies a trade-off between being able to supress basal expression and allowing higher cell-to-cell variation within a small portion of the dynamic range of the controller. We went with this trade-off because we judged that the ability to zero out basal expression was important for generating “reversible knockout” phenotypes, and perhaps in some synthetic biology applications where complete absence of a protein might be required for proper function of an engineered synthetic genetic circuit or to keep a kill switch off. Additionally, we went with the trade off because the means by which we would achieve it, the "zeroing repressor" concept, would be applicable to other WTC846-like controlled systems. As a backup, we made parts available via Addgene that enable construction of strains with the other architecture, AR, by one step cloning.
In response to the reviewer comments, we made the following changes.
Last paragraph of the Results section titled “Complex autorepressing (cAR) controller architecture expands the input dynamic range and reduces cell-to-cell variation”:
Previously the text read:
“Compared with cells bearing the SR architecture, otherwise-isogenic cells bearing the AR architecture showed increased basal expression (6.3 vs. 4.1fold over autouorescence background). […] We therefore picked this cAR architecture for our controller.”
The text in the Results section now reads:
“Compared with cells bearing the SR architecture, otherwise-isogenic cells bearing the AR architecture showed increased basal expression (6.3 vs. 4.1fold over autofluorescence background). […] We therefore picked this cAR architecture for our controller.”
We also made the following addition to the Discussion:
“Cell-cell variation in WTC846-driven expression is highest at low aTc levels, because control in this regime depends mostly on the higher variability Simple Repression by TetR-Tup1 expressed from PRNR2. […] This autorepressing circuitry operationally defines WTC846 as an ”expression clamp”, a device for adjusting and setting gene expression at desired levels and maintaining it with low cell-to-cell variation, and so allowing expressed protein dosage in individual cells to closely track the population average.”
We also made the following addition to the legend in the Plasmids table (Supplementary Table 2), to make it clear that we have provided information and materials sufficient to make strains with the other architectures used in this study.
“Plasmids used in this study. * indicates plasmids available through Addgene. These plasmids are sufficient to allow construction of WTC846 strains carrying the cAR architecture without any further construction. […] Construction of SR strains would require deletion of the negative feedback-controlled TetR from (P2365/2370/2371/2372/2374), and construction of AR strains would require deletion of the constitutively expressed TetR-Tup1 from the same plasmids.”
3. Induction with the WTC846 system using aTc is a relatively slow process, at least judged by the Citrine read-out (Figure 4B), taking place over >360 minutes (>3 cell divisions in rich medium). The authors should discuss why this is so, with reference to known transcriptional and/or translational kinetics in this complex feedback system.
We thank the reviewer for directing us to address induction or induction kinetics. We had not done so previously, partly because the story is complicated, and partly because what we knew was probative, not dispositive.
We need to distinguish between speed of induction and time to steady state. Reviewer notes that it takes >3 cell divisions to reach steady state after induction. In WTC846 strains, induction, judged by Citrine fluorescence signal (Figure 4B) is detected within 30 minutes after aTc addition, and the release after Cdc20 depletion (Figure 5B) occurs within 35 minutes after aTc addition. Given that induction is rather fast, why do WTC846 strains require 6 hours to reach steady state? Expression of the controlled gene is synchronized to expression of the autorepressing TetR. And yet, we know that the long time to steady state is not a consequence of the autorepressing feedback, as we observe that Simple Repression (SR) configurations also require a similar length of time (6 hours) to reach steady state (See Supplementary Figure 6). This time to steady-state derepressed expression is also comparable to that for a previously described TetR based system, which depends on galactose for activation, which required around 400 minutes to reach steady state (doi: 10.1038/nature01546). This time to steady state is still short compared to activation-based systems like the β-estradiol dependent system previously constructed in our lab (doi: 10.1093/nar/gku616).
Our explanation for the long time to equilibrium is that steady state concentration of the controlled protein (Citrine) is reached when degradation of existing protein and its dilution by partition into daughter cells, balances new synthesis. Therefore, the time to steady state will depend on the stability of the controlled protein. If the half-life is longer than doubling time, steady state will be reached in 6-7 hours. If the protein has a half-life shorter than doubling time, steady state will be reached faster. Citrine is long lived and is lost mainly due to dilution by cell division (the longevity of Citrine is shown in our Supplementary Figure S25). Estimates vary as to the stability of the yeast proteome. A recent meta-analysis of 4 different datasets measuring half-lives suggests most proteins (between 50% and 85%) are just as stable and are lost mainly due to dilution by cell division (http://dx.doi.org/10.2139/ssrn.3155916).
Equation 1 is a simple ODE model where a protein is produced at constant rate a and lost with a lumped linear rate (dilution + degradation) d. The analytical solution (Equation 2) of this model shows that the only constant that affects the time variable is the degradation + dilution rate. Dependence on this variable is also evident in the simulations we provide in Appendix 4—figure 1, based on this model. The smaller d is, (i.e., the more stable the protein is), the longer it takes to reach steady state. On the other hand, changes in the production rate a have no effect on time to steady state, although both rates affect the maximum level as one would expect.
In order to make the distinction between induction and steady state kinetics clearer, we made the following changes in the first paragraph of the 4th subsection of results:
Old text read:
“After induction, signal appeared within 30 minutes and reached steady state within 7 hours.”
New text reads:
“After induction, signal appeared within 30 minutes. Time to reach steady state, which will be shorter for proteins that degrade more quickly (see Appendix 4), was 7 hours for the stable protein Citrine.”
We also added the simulations mentioned above as Appendix 4—figure 1, and we refer to it as seen below:
“Time to steady state depends on the stability of the controlled protein. In the experiments in Figure 4, WTC846-controlled Citrine takes around 7 hours to reach steady state concentration. […] Therefore most WTC846-controlled endogenous proteins will likely have a time to steady state around 7 hours, except those with a shorter half-life which will exhibit a shorter time to steady state.
We also made the following addition to the 7th paragraph of the Discussion:
“Both induction and shutoff with WTC846 are rapid, as both Citrine and
Cdc20 expression occur within 30 minutes of induction, and shutoff of Citrine expression is observed within 60 minutes. […] Those proteins with shorter halflives will reach steady state faster.”
4. They should also discuss the potential implications of this for screens based upon plate measurements of single-colony growth rates or growth of robotically "pinned" cell cultures. They show a limited number of growth ("spot") assays on solid media that would appear to indicate abrupt cut-offs for colony formation, though the data show only one time point in the growth assay and might obscure growth rate differences.
First, viability experiments. Our plate-based assays using WTC846, were not meant to measure growth rates but only the ability of single cells to divide and form colonies (a measure of cell viability) at a terminal time point.
The revised results text in the 3rd paragraph of the last Results section now reads:
“We spotted serial dilutions of cultures of the final seven strains on YPD, YPE, SD, S Glycerol and SD Proline plates, with and without inducer, and assessed the strains' ability to grow into visible colonies at a single time point, at which cells of the parent strain formed colonies in all serially diluted spots(24h for YPD and SD, 42h for others.).”
But this is not what the reviewer asks. We would like to see WTC846 used in plate based and robotically pinned growth dependent screens. We know there are differences in time to approach steady state expression for different strains expressing different proteins from WTC846 – or by any other inducible system. These differences will lead to different effects on growth rate for the first doublings on solid medium or liquid culture. This fact admits to an obvious workaround, which is to suggest that experimenters pre-induce the strains before plating cells or pinning them robotically into wells on a dish.
In response, we added wording to the last paragraph of the Discussion section about the suitability of WTC846 for pinned growth rate experiments. The text now reads:
“Taken together, our results show that WTC846 controlled genes define a new type of conditional allele, one that allows precise control of gene dosage.[…] WTC846- alleles may find use in engineering applications, for example in directed evolution of higher affinity antibodies expressed in yeast surface display, where precise ability to lower surface concentration should aid selection of….”
5. The authors should examine shut-off kinetics, ideally by as direct a measure of RNA Pol II initiation as possible. This could be of interest in exploring the function of essential genes, for example.
Yes. The reason we were reluctant to explore shutoff kinetics is that it is another complex story. In WTC846, elimination of protein from the cell (which is what we care about for assessment of phenotype) depends both on shutoff of new expression and on perdurance.
We expected (and expect) that the speed of shutoff of WTC846 controlled genes would be affected primarily by the rates of 3 reversible processes: aTc diffusion out of the cell, TetR binding to its operators, and sequestration of free aTc by newly synthesized TetR. These processes are fast: TetR binding and unbinding to DNA is fast (all binding sites were occupied within ~5 minutes of addition of doxycycline to a reverse TetR system in mammalian cells) (doi: 10.1038/ncomms8357 (2015)). Moreover, in the same experiment, reverse TetR started associating with specific sites seconds after addition of doxycycline, indicating that, at least in mammalian cells inducer diffusion and inducer-TetR binding are also rapid. For this reason (although it is possible that diffusion out of yeast cells might take slightly longer due to the cell wall), we expect that shutoff of WTC846-controlled phenotypes will be limited by perdurance, governed by the degradation kinetics of the mRNA and the encoded protein. Given the rapidity of induction (again, fluorescence signal after 30 minutes, Figure 4B, release from CDC20 arrest within 35 minutes, Figure 5E), we reasoned that the same reversible phenomena that govern WTC846 induction (namely aTc diffusion, TetR-DNA interactions, and sequestration by newly synthesized TetR) will govern shutoff, so that shutoff of the WTC846 expressed phenotype will be dominated by perdurance of the controlled gene products. The consequence would be that for controlled genes (as in Figure 5E with CDC20), the predominant factor in achieving zero protein and observing the associated phenotype in the cell upon aTc removal will be the stability of the mRNA and protein of interest.
Given the relative complexity of this explanation, in response to this reviewer we performed additional direct experiments to (a) determine how quickly WTC846 stopped producing the protein of interest upon aTc removal and (b) confirm that perdurance is the predominant factor in how quickly the desired loss of function is seen.
For (a) we grew a strain in which Citrine expression was under WTC846 control to exponential phase and then induced with a high concentration (600ng/mL) of aTc. We then measured fluorescence signal every 30 minutes in flow cytometry. Additionally, after 30, 90, 150, and 210 minutes, we removed, washed, and resuspended a sample from the culture in (a) medium without aTc (to shut off WTC846) and (b) without aTc but with cyclohexamide (to shut off both WTC846 and new protein synthesis). After shutoff, we expected to see an initial increase in signal, followed by decline from this peak. The initial increase in fluorescence after shutoff would be due to 3 factors: the time it took for WTC846 to stop producing new transcripts, the time it took for the existing mRNA to be degraded, and continued fluorophore formation by already-synthesized but immature Citrine proteins. Whatever increase in fluorescence we observed above that found in the cycloheximide sample would be due to the other 2 factors outlined above, namely WTC846 shutoff speed and mRNA degradation speed.
Our results are presented as a supplementary figure in the revised submission. Shutoff of WTC846 directed expression is rapid, as we see stabilization of fluorescence within 60 minutes. And assuming a ~30 minute maturation time for Citrine (since its parent EYFP matures in ~30 minutes after denaturation (https://doi.org/10.1038/nbt0102-87)), WTC846 shutoff likely occurs within the first 30 minutes. These results also show that a stable protein like Citrine takes a long time to be eliminated from the cells, given that its level is halved every 90-120 minutes, consistent with dilution through cell division.
In (b), we revisited cell cycle arrest and release by Cdc20. In the original experiment in Figure 5E, arrest of the WTC846::CDC20 strain took 8 hours (See Materials and Methods). Given that CDC20 is a very unstable protein, we attributed this long time to arrest to the high aTc concentration in which we precultured the cells (20ng/mL), which likely brought about high pre-arrest concentrations of Cdc20 that took a long time to become depleted. To confirm this hypothesis, we precultured the WTC846::CDC20 strain at a much lower aTc concentration (3ng/mL) and then placed cells in medium without aTc. At this lower concentration, arrest began at 1.5 hours after removal of aTc, consistent with the shutoff time of WTC846 established above, and between 3-4 hours all cells had arrested. The large difference between the arrest times of cells grown with 3 and 20ng/mL aTc, as well as the short shutoff time of WTC846 [WTC846 – Citrine expression] presented above, support the notion that perdurance of protein of interest will in most cases dominate the time to complete “phenotypic” shutoff, i.e. the time needed to see the zero expression "null" phenotype.
We’ve included both experiments as supplementary figures (also presented here) and added the following text to the manuscript.
As the second paragraph of the Results section titled “WTC846 fulfils the criteria of an ideal transcriptional controller” we added:
“To better characterize the system, we also measured the shutoff speed of WTC846 driven Citrine expression. […] Overall, we conclude that WTC846 shutoff is rapid, but the time required to see the phenotypic effects of the absence of the controlled gene product will primarily depend on the stability of the mRNA and expressed protein.”
And we made the following changes in the last paragraph of the Results section titled “WTC846 allows precise control over protein dosage and cellular physiology”.
“Finally, we tested the ability of WTC846 to exert dynamic control of gene expression by constructing a WTC846-K3::CDC20 strain (Y2837) and using this allele to synchronize cells in batch culture by setting Cdc20 expression to zero and then restoring it. […] Given this, and the rapid shutoff kinetics of WTC846 presented in Figure 4—figure supplement 8, we conclude that the shutoff dynamics of WTC846 controlled phenotypes depend mostly on the speed of degradation of the controlled protein.”
We also added a short comment on the speed of shutoff within the third to last paragraph of the Discussion section:
“Both induction and shutoff with WTC846 are rapid, as both Citrine and Cdc20 expression occur within 30 minutes of induction, and shutoff of Citrine expression is observed within 60 minutes.”
We also adjusted our Materials and methods to explain these additional experiments.
6. The expected results for the Whi5 experiment and how they might differ from the actual findings (Figure 5D, pg. 10) are not clear and should be developed further, perhaps in the Results section itself. As the authors point out: "Whi5 controls cell volume by a complex mechanism", but they don't explain what I believe is a key feature, namely that its translation does not scale with cell volume (according to the Skotheim lab), which is an unusual feature. It would be interesting to examine WTC846 regulation of other genes implicated in cell size control.
We thank the reviewer for the suggested clarification. We agree that the decoupling of the amount of translation from cell volume is a key feature of Whi5 control. We changed the results (6th paragraph of the last Results section). Revised text now reads.
“Whi5 controls cell volume by a complex mechanism and unlike most other proteins its abundance does not scale with cell volume. […] To test whether we could control cell volume by controlling Whi5, we constructed haploid and diploid WTC846::whi5 strains (Y2791, Y2929)…”
In terms of our results, we believe that while we alter the cell cycle stage at which Whi5 mRNA is expressed (by making it expressed continually), the nuclear import/export regulation of Whi5 remains intact. The continued import/ export regulation allows cell size regulation by Whi5 to remain functional, and so that cell volume simply scales with the amount of Whi5 expression we induce via WTC846. We clarified this point by adding the text below to the Results:
“Here, we constructed haploid and diploid WTC846::WHI5 strains (Y2791, Y2929). […] We expected that the volume of these cells should scale with the concentration of the aTc inducer.”
While we observe that the cell size scaled with aTc concentration, we also found that overexpression of Whi5 in the haploid strain led to increased cell-to-cell variation in cell volume. In order to develop this aspect of the results further, we analysed DNA content of haploid cells grown in presence of high aTc and noticed an increase in the number of cells that had >2n DNA content. We believe this indicates a misregulation of the normal progression of the cell cycle, leading to endoreplication, which could then lead to the observed increase in cell-to-cell variation. We show this experiment in Figure 5—figure supplement 5. We also made the following addition to the Whi5 paragraph of the Results (6th paragraph of the subsection titled “WTC846 alleles allow precise control over protein dosage and cellular physiology”):
“Both diploid and haploid cells (especially haploids) expressing high levels of Whi5 showed increased variation in volume. […] We therefore believe that overexpression of Whi5 leads to endoreplication, and the increased variation in volume at high aTc concentrations in the haploid strain originates from these endoreplicated cells.”
It will be interesting to examine WTC846 regulation of other cell size control genes, and we are attempting to recruit a student to address this work.
New figure legend reads:
“(C) DNA content of the haploid WTC846-K1::WHI5 strain cells (Y2791, red) at mid-exponential phase in S Ethanol media, grown with 4 different aTc concentrations as marked on top of each plot. […] Without Whi5 there was a reduction of G1 cells with 1n DNA content in the Y2791 strain, and upon overexpression of Whi5 aneuploid cells with >2n DNA content are observed.”
7. In Figure 5E, it would be great to show, or at least explain clearly in the text, how the WTC846::CDC20 strain compares to the standard cdc20-ts allele used in the field.
We did so. We added the following text together with the relevant references to our Results section to compare our Cdc20 allele to already existing ones:
“When compared to published data, arrest at G2/M using the WTC846::CDC20 strain is more penetrant than that obtained using temperature sensitive (~25% unbudded cells) and transcriptionally controlled (~10% unbudded) alleles of cdc20. Release is at least just as fast as that observed for the temperature sensitive (~35 minutes) and the transcriptionally controlled allele (~40 minutes).”
8. In Discussion (pg. 10, bottom right) the authors need to show clearly that WTC846 actually gives much lower cell-to-cell variation than all other current methods (at least where this has been measured).
We've done the best we can. When evaluating gene expression tools, cell-to- cell variation has not been routinely measured. This is possibly because all systems thus far have relied on a simple repression or activation configuration, and cell-to-cell variation has been quite high. Moreover, the best means to quantify the different contributions to cell-to-cell variation in expression require microscopic quantification of expression of two reporters in sets of hundreds of single cells tracked over time (Colman-Lerner et al. 2005) and even the best flow cytometric methods for achieving variation (Pesce 2018) involve dual reporter strains rather than the single reporter strains used here.
To address this point, we have included (in references 13, 19, 20), relevant instances in which variation in single reporter strains can be assessed. In these, cell-to-cell variation was not explicitly quantified, but induction resulted in highly variable (even bimodal) distributions of expression. One of those references is to a transcriptional controller induced by β-estradiol (LexA-hER-B112) we published previously. In response to the reviewer's direction, we quantified cell-to-cell variation in gene expression by this controller compared to expression by WTC846. The β-estradiol induced system had high cell to cell variation (>0.4) throughout its entire dynamic range, whereas WTC846 shows high variation only at a small portion of its dynamic range at low expression levels. Additionally, we constructed a strain where the commonly used, galactose inducible PGAL1 promoter directed Citrine expression. We quantified cell-tocell variation in a dose response to galactose, where upon induction variability was always very high (>0.6). We included these comparisons as a supplementary figure.
We also included the following text in our Results section (last paragraph of the section titled “WTC846 fulfills the criteria of an ideal transcriptional controller”):
“We quantified the cell-to-cell variability in Citrine expression using the single reporter (ViV) measure for the WTC846::citrine strain (Y2759) grown in YPD, and compared it to variation in a β-estradiol (LexA-hER-B112) activation based transcriptional control system we previously described, and the commonly used galactose activated PGAL1 promoter(Figure 4D, Figure 4—figure supplements 4, 6 and 9).”
9. In the Discussion (pg. 11 top right) the authors state that "WTC846 alleles display no uninduced basal expression" (and earlier in Results they repeated speak of "abolished" basal expression). I think that it would be more correct to point out, at least in the
Discussion, that WTC846 alleles display no uninduced basal expression as measured here, by the Citrine fluorescent read-out. Producing null phenotypes in cases where this was not possible with previous systems is a valuable demonstration of the utility of the new method, but not a demonstration of complete repression. The authors might try to measure mRNA steady-state levels by RT-qPCR to estimate copy number per cell, and together with any information on mRNA stability, to estimate transcription rate, if they really want to pin a number on the basal expression level.
The reviewer's qualification is of course correct. Since WTC846 is intended as a tool for controlling gene expression, and our functional tests with low expressed genes did not show any problematic basal expression, we opted not to further quantify mRNA, but to change the wording so that it says what we did observe. Discussion now reads.
“It can set protein levels across a large input and output dynamic range. […] WTC846 alleles also exhibit high maximum expression, low cell-to-cell variation, and operation in different media conditions without adverse effects on cell physiology.”
Instead of the old text which read:
“It can set protein levels across a large input and output dynamic range. WTC846 alleles display no uninduced basal expression, high maximum expression, low cell-to-cell variation, and operate in different media conditions without adverse effects on cell physiology.”
10. Since one of the major potential applications of the WTC846 system described here, as pointed out by the authors, is genome-wide screening, that authors should at least discuss how this might be efficiently implemented. The generation of a library of WTC846 fusions to each protein-coding gene in yeast by standard methods would be extremely laborious and expensive. An alternative worth mentioning is the SWAp-Tag strategy described by Schuldiner and colleagues (Yofe et al.  Nature Methods). There may very well be others.
We thank the reviewer for the suggested addition. As the reviewer suggested, one good way to generate a library of WTC846 controlled endogenous genes would be to use existing whole genome collections such the SWAP-Tag as a starting point. SWAP-Tag allows insertion of N or C-terminal sequences to most ORFs in the yeast genome by creating standard sites for homologous recombination. Generating a whole genome collection of WTC846 alleles would be accomplished crossing the existing SWAP-Tag N terminal collection with a WTC846 donor strain, followed by sporulation on the appropriate selective medium. We’ve therefore added a short section and the necessary references in our discussion on this subject using the text below.
Last paragraph of the Discussion previously read:
“Taken together, our results show that WTC846- controlled genes define a new type of conditional allele that allows precise control of gene dosage. […] We also expect them to find use in genome-wide gene-by-gene and gene-bychemical epistasis screens to detect protein dosage independent interactions,…”
New version reads:
“Taken together, our results show that WTC846- controlled genes define a new type of conditional allele, one that allows precise control of gene dosage. S. cerevisiae […] We therefore suggest that growth rate-based assays using WTC846 or any other inducible system pre-induce cells several generations before plating or pinning. WTC846- alleles may find use in engineering applications, for example in directed evolution of higher affinity antibodies expressed in yeast surface display, where precise ability to lower surface concentration should aid selection of….”
11. Although the authors emphasize the utility of WTC846 regulation for generating precisely graded expression levels of a given protein with minimal secondary effects growth or physiology, they devoted considerable effort towards achieving maximal shut-off without much discussion of the utility of this feature. In light of anchoring and induced degradation methods that operate at the protein level to generate loss-of-function effects, the authors should consider how WTC846 regulation might be a useful complement to these methods, particularly when their efficiency or rapidity is in question.
We thank the reviewer for the suggestion. WTC846 could certainly a useful complement to such methods, especially in cases where rapid shutoff of expression is required together with precise dosage control. We have added this to the last part of our discussion using the text below:
“WTC846 can also be a useful complement boost the efficiency of methods that act at the protein level such as auxin induced degradation or AnchorAway techniques. […] For example, an experimenter might simultaneously induce depletion of the product of a controlled gene by such a method while adjusting aTc downward to rapidly reset the level of an expressed protein to a new, lower level.”
Additional comments and suggestions
1. Pg. 5 right column: "…and we developed a second measure (explained in SI) that normalized variation in dosage with respect to a key confounding variable, cell volume, to correct for its effect on protein concentration. In this, we measured the Residual Standard Deviation (RSD) in signal from cells after normalization of output for volume estimated by a vector of forward and side scatter signals…"
I think it would be a good idea to give the second measure a name here and make it clearer that Figure 2D reports the RSD of this value if I understand correctly. The sentence "In this, we measured the RSD in signal…" is awkward and unclear. Please re-word.
Yes. We named the method "Volume Independent Variation". We reworded the unclear sentence as shown below, and clarified in the legend that Figure 2D reports this measure. We modified all other figure legends in the manuscript and the supplementary material where RSD was reported to reflect the new name for the measure.
Old text read:
“In this, we measured RSD in signal from cells after normalization of output for volume estimated by a vector of forward and side scatter signals.”
New text reads:
“In ViV, we estimated cell volume by a vector of forward and side scatter signals, and measured the remaining (Residual) Standard Deviation of the single reporter output after normalization with this estimated volume.”
The new Figure 2 legend now reads:
“(D) Cell-to-cell variation of expression by these three architectures. We calculated single-reporter cell to cell variation (VIV) as described. […] Dot-dash line indicates VIV of the strain where Citrine is constitutively expressed from PTDH3 and dashed line indicates VIV of autofluorescence in the parent strain without Citrine.”
2. In Figure 3A the fluorescence curves should be better aligned with the strain designations.
Yes, with thanks. We have made the necessary change to the alignment.
3. Why is there no curve shown for repressed P3tet in Figure 1B? (numbers should be aligned correctly also).
We thank the reviewer for pointing out the alignment issue and have fixed it in the figure.
There is no curve for P3tet because the curve is centred around fluorescence value ~20000 and thus lies outside the range of the xaxis we used in main Figure 1. If we extend the x-axis to include this curve, the difference between all the other curves becomes difficult to perceive. Therefore, we show the data for P3tet separately again in Supplementary Figure 1 (and we direct the reader to the Supplementary Figure at the relevant point in the text).
We have now added a note also in the figure legend (see below) directing the reader to Figure S1.
“Repressed activity of P3tet is above the x axis depicted in this figure, but can be seen in Figure S1.”
4. Yeast gene names should be in italics throughout (including the figure legends). Deletions that a recessive should also by italics, and with the "Δ", essentially an allele designation, placed after the gene name.
Necessary changes made in text and the supplementary material.
5. Reference no. 34 refers to Rap1 and Grf2 binding sites (not Gcr1). Please clarify.
We thank the reviewer for pointing this out. Yes, the paper refers to Rap1 and Grf2 binding sites and finds them in the PTDH3 UAS, but also includes in the last portion of its Results section the identification of the CATCC consensus binding sites. The authors of the paper suggest in their discussion that these are likely to be binding sites for the recently discovered Gcr1. This paper was the earliest reference we could find to the CATCC binding sites in the PTDH3 UAS, and therefore we cited it in this context.
We added a more recent reference corroborating the GCR1 binding sites in the P TDH3 UAS in the main text, next to the original reference 34. We also added a short note “Gcr1 binding sites were found in reference 13 and confirmed in reference 14“ in the legend of Figure S21.
6. Bottom pg. 11, right, I would suggest "…when inducer is absent, the measured basal expression of the controlled gene is abolished…"
We thank the reviewer and have made the suggested change.
7. References no. 28-30 and 35 are incomplete (no journal, volume, or page number).
We thank the reviewer for pointing this out. We fixed the references.
8. Bottom pg. 5, left should be >50,000.
We thank the reviewer for pointing this out. We fixed the symbol.
9. Page 8 right: "…according to,53 to enable…" could be replaced by "…to enable…" with the reference placed at the end of the sentence.
We thank the reviewer for the suggested change which we have implemented in the text.
10. Page 8, right: again, replace inverted "?" with ">" (?)
We thank the reviewer for catching these formatting errors. We have replaced the symbol.
Reviewer #1 (Significance (Required)):
The increasing interest in quantitative analysis of cell function emphasizes the need for systems that allow the graded and precise control of gene expression to address dosage effects on phenotype. Although systems of this sort have been developed in yeast, they all have specific drawbacks, as clearly described here. The WTC846 system described here represents a genuine improvement on current methods and is likely to be an important addition to the yeast toolbox for gene function analysis. It may also motivate similar developments in mammalian cell culture systems, and eventually in whole animals.
We thank the reviewer for this assessment. Just as qualitative means to bring about conditional gene expression enabled experiments that produced important qualitative insights into cell function, so development of precise quantitative means to bring about conditional gene expression will enable better quantitative analysis of cell function. It will also enable new technical approaches such as the threshold phenotypes, epistasis approaches, and evolutionary methods mentioned above. We suspect that development of analogous methods in in vertebrate cells will likely prove even more important in making vertebrate models of human diseases.
Reviewer #2 (Evidence, reproducibility and clarity (Required)):
Provide a short summary of the findings and key conclusions (including methodology and model system(s) where appropriate).
In this manuscript, Azizoglu and co-authors developed an autorepression-based transcriptional controller of gene expression for a given gene of interest in Saccharomyces cerevisiae. This system is very precise and allows for dynamic gene expression by adjusting the amount of inducer (aTc) and/or by changing the translation initiation sequences. In addition, it is equally good at being turned on and off in all the media tested and has low cellto-cell variation. The authors tested their transcriptional controller using different genes (including CDC20, IPL1, WHI5, TOR2, and others) and recapitulated various known phenotypes. They also identified a novel phenotype when overexpressing IPL1 in yeast which mimics a previously identified phenotype in mammalian cells when overexpressing Aurora B. This system will be very useful for the yeast community.
We also want to stress that now that we have proof of concept, we can do this in higher cells and multicellular organisms including vertebrates, which need it more.
Are the key conclusions convincing?
Should the authors qualify some of their claims as preliminary or speculative, or remove them altogether?
No. They do a good job backing up their claims in their supplementary files.
Would additional experiments be essential to support the claims of the paper? Request additional experiments only where necessary for the paper as it is, and do not ask authors to open new lines of experimentation.
No. See above.
Are the suggested experiments realistic in terms of time and resources? It would help if you could add an estimated cost and time investment for substantial experiments.
No suggested experiments.
Are the data and the methods presented in such a way that they can be reproduced?
Are the experiments adequately replicated and statistical analysis adequate?
Specific experimental issues that are easily addressable. Could the authors add the pGAL1-10-IPL1 control to their WTC846-driven IPL1 overexpression experiment (Figure S17)? It might be easier for the reader to understand why the exciting phenotype the authors found was not reported before.
We thank the reviewer for the suggested control. Since we believe that the WTC846::IPL1 allele manifests an overexpression phenotype because it produces more protein than PGAL1 would have, replacing the promoter of Ipl1 with PGAL1 would be the best control. However, since IPL1 is an essential gene, and its misregulation affects chromosomal integrity, we weren't confident we could replace its promoter with PGAL1 and still achieve the right expression level to keep the cells healthy. In fact, in previously published work on IPL1, overexpression of IPL1 was achieved by introducing a PGAL1-IPL1 producing plasmid into an IPL1+ background.
To address the reviewer's concern we therefore took the same approach. We placed the PGAL1 promoter upstream of Citrine in a centromeric plasmid, then grew the strain that carried the (PGAL1Citrine) plasmid in the same (Gal Raf) media used by M. MunozBarrera and F. Monje-Casas 2014, to observe effects of Ipl1 overexpression on growth rate and colony formation. The authors did not observe detrimental effects of Ipl1 overexpression in either case.
Here, we used flow cytometry to measure single cell fluorescence from the strain bearing the PGAL1-Citrine plasmid, as well as the WTC846::citrine strain grown in YPD with 400ng/mL aTc (the same concentration we used in the Ipl1 overexpression experiment).
Our results showed that the median expression level of the centromeric PGAL1-Citrine strain was approximately two fold lower than that in the integrated WTC846::citrine strain, and that it displayed much higher cell to cell variation. As measured by the single reporter cell-to-cell variation measure (ViV), PGAL1-Citrine variation was 0.9, whereas WTC846::citrine variation was only 0.19. Ipl1 overexpression phenotype does not lead to a complete arrest of growth in the entire population- some cells are affected while others are not. With PGal1 driven overexpression, overall Ipl1 level is lower compared to WTC846 driven overexpression. Additionally, due to high cell-to-cell variability only a handful of cells are actually expressing Ipl1 at this level, which would have made it difficult to observe the growth arrest/delay phenotype that is already not a uniform phenotype in the population. Therefore, we believe the combination of lower expression and higher cell-to-cell variation in the PGAL1-Citrine strain likely accounts for the failure of Munoz-Barrera and Monje-Casas to observe a growth arrest phenotype.
We present the results in Figure 5—figure supplement 7. We explain the results in additional sentences at the end of the 4th paragraph of the Results section titled “WTC846 alleles allow precise control over protein dosage and cellular physiology“.
The paragraph in the main text now reads:
“At high aTc concentrations, the WTC846-K2:IPL1 strain formed colonies with lower plating efficiency than the parent strain. […] Either the lower expression or the higher variation, or both, might account for the fact that PGAL1 driven Ipl1 overexpression does not result in the mammalian Aurora B phenotype in S. cerevisiae."
Are prior studies referenced appropriately?
Are the text and figures clear and accurate? The authors used the strain names in the text and it is hard to figure out what each strain contains. They could transmit this information by using the construction "strain Y### (contains [specify relevant details]).
We thank the reviewer for the suggestion. During the revision process, we took care to mention the relevant details as much as possible together with the yeast strain number within the text.
In Figure 3, specify that the sentence that explains the errors bars refers to section A and B. or move the sentence before C.
Right. There are actually error bars also around the data points in C, although they are very narrow and difficult to see. This sentence therefore refers to Figure 3C as well. In response to this comment, we adapted the figure to make it easier to see the error bars in Figure 3C.
In page 3, the authors mention ts and cs mutations. The authors should define what ts and cs are and italicize them.
We thank the reviewer for pointing this out. We have included the long format of these abbreviations where they were first mentioned, and have italicized the abbreviations.
Old text read:
“During the 20th century, workhorse methods to ensure the presence or absence of gene products have included use of ts and cs mutations within genes, for example to give insight into ordinality of cell biological events.”
New text reads:
“During the 20th century, workhorse methods to ensure the presence or absence of gene products have included use of temperature sensitive (ts) and cold sensitive (cs) mutations within genes, for example to give insight into ordinality of cell biological events.”
On page 4, the authors defined PIC and aTc, but both terms were previously defined.
On page 4, in the following sentence "We also combined the operators in these
constructs to generate P5tet (Figure 2B)", the figure is wrongly referenced, the reference should be to Figure 1B.
We thank the reviewer for catching this, and have fixed the reference.
On page 7, in the following sentence "The PACT1, PVPH1, PRNR2 strains showed no uninduced expression,.…" please add a reference to Figure 3B.
We thank the reviewer for the suggestion and have added the reference.
On page 10, in the following sentence "…at high aTc concentrations, cells bearing the higher translational efficiency TOR2 allele (WTC846-K1::TOR2) grew more slowly than the otherwise-isogenic control parent strain with WT TOR2 (Figure S18C)" the figure is wrongly referenced, the reference should be to Figure S18D. I also suggest the authors add the following text: (Figure S18D, compare the 600ng/mL line to blue dashed line).
We thank the reviewer for pointing this out and have corrected the reference. We also added the suggested text for clarification.
On page 10, in the following sentence "…Whi5 is rapidly exported from the nucleus, and the cells enter START", I think they mean S-phase.
We thank the author for the comment and have made the correction to the text which now says S-phase.
On page 10, the Kozak sequence is missing in the following sentence: "Use of WTC846::CDC20 to synchronize the cells…"
We thank the reviewer. We have added the correct Kozak sequence information to the strain name.
Throughout the paper, gene deletions should be written as follow: whi5Δ.
Throughout the paper, genes should be italicized (e.g. HIS3).
We thank the reviewer for the correction of the naming nomenclature. We have fixed the gene names throughout the paper and the supplementary material.
Throughout the paper, the authors have random question marks (which may be the result of converting a file in another format to a PDF) that may be removed.
We thank the reviewer for catching these formatting errors. We replaced the question marks with the correct symbols.
On Supplementary Text page 1, the following sentence: "However single reporter studies have a major, confounding contribution to measured variation in gene expression that multi-reporter studies don't:" is missing a comma after However and change don't to do not.
We thank the reviewer for both corrections which we have implemented as suggested.
In Supplemental Figure 7, the authors may want to add P2tet and P3tet for better main text that the promoter for essential genes was replaced with another promoter in the presence of the inducer (aTc) before referencing Figure 5, or in the legend of figure 5.
We believe the reviewer wanted to raise two separate points here. One about Figure S7, the other about clarifying the essential endogenous gene promoter replacement protocol we used. We address these two points separately below.
First. We thank the reviewer for the suggestion about Figure S7. In Figure S7, inclusion of the unrepressed promoter fluorescence curves would widen the x-axis range we need to show, and therefore would make it more difficult for the reader to perceive the small differences between the repressed expression of P2tet and P3tet compared to the autofluorescence. Therefore, given that the focus of this supplementary figure is narrowly on the repressed expression, we believe it is best to omit the unrepressed expression curves in this figure.
It is important to note that, although P3tet is weaker than P2tet, this fact is not the only reason that P3tet shows a lower level of repressed expression. Unrepressed P2tet is 1.14 times stronger than P3tet (data from main figure 1). However, when repressed, P2tet expression is 1.38 higher than that for P3tet. This suggests that promoter strength is not the explanation for the lower repressed expression from P3tet. To emphasize this point, in response to the reviewer direction, we added fold comparison numbers to the figure legend.
The new text in the figure legend of Supplementary Figure 7 reads as follows:
“For comparison, repressed expression in the P2tet strain (Y2662) was 1.38 fold higher than in the P3tet strain (Y2702), even though the expression in the unrepresssed P2tet strain was only 1.14 fold more than P3tet. This fact shows that, even given the difference in promoter strength, P3tet is repressed more effectively than P2tet.”
Second, we thank the reviewer for the suggested clarification to the strain generation protocol. We changed the main text to indicate that the strains were generated by replacing the promoter of the endogenous gene in cells growing in medium containing aTc., using the text presented below:
“We constructed strains in which WTC846 controlled the expression of these genes by replacing the promoter of the endogenous gene. Before transformation the cells were grown in liquid medium containing aTc, and then plated on solid medium containing aTc (see Supplementary Materials for a detailed protocol).”
Reviewer #2 (Significance (Required)):
Describe the nature and significance of the advance (e.g. conceptual, technical, clinical) for the field. The authors provide a new tool to dynamically control the expression of genes. The previous systems that have been developed are hard to modulate or are "leaky". The system developed in this work allows for precise control of gene dosage, has low cell-to-cell variation, and has a true OFF state.
Static steady state expression around an adjustable setpoint, i.e. clamped expression is good too.
Place the work in the context of the existing literature (provide references, where appropriate).
Regulating gene expression in a targeted way has been a common method in yeast research for decades. However, the current gene expression systems have deficiencies (e.g. leaky expression, not tunable expression, high cell-to-cell variation). The authors in this paper developed a new system that shore up these deficiencies by enabling tight on and off controlled gene expression.
State what audience might be interested in and influenced by the reported findings. The yeast community in general, as it is a transcriptional controller developed for S. cerevisiae.
We also believe the ability to make alleles of this type will be extended higher cells and multicellular organisms, particularly vertebrates.
Define your field of expertise with a few keywords to help the authors contextualize your point of view. Indicate if there are any parts of the paper that you do not have sufficient expertise to evaluate. Genetics, yeast, evolution, chromosome segregation.
- Asli Azizoglu
- Fabian Rudolf
- Roger Brent
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
We thank Jörg Stelling for conceptual discussions, Hans Michael-Kaltenbach for discussions on the cell-to-cell variation measure, Kristina Elfström for initial construction of P2tet constructs, Gnügge for initial characterization of repressibility of operator placements around the TATA sequence by LacI, and Mattia Gollub, Justin Nodwell, Alan Davidson, Kohtaro Tanaka, and Alejandro Colman-Lerner for valuable discussions over the course of the work. This work was supported by the Swiss National Science Foundation as part of the Molecular Systems Engineering NCCR and by grant R21CA223901 from the NCI to RB.
- Naama Barkai, Weizmann Institute of Science, Israel
© 2021, Azizoglu et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Auxin-inducible degrons are a chemical genetic tool for targeted protein degradation and are widely used to study protein function in cultured mammalian cells. Here we develop CRISPR-engineered mouse lines that enable rapid and highly specific degradation of tagged endogenous proteins in vivo. Most but not all cell types are competent for degradation. By combining ligand titrations with genetic crosses to generate animals with different allelic combinations, we show that degradation kinetics depend upon the dose of the tagged protein, ligand, and the E3 ligase substrate receptor TIR1. Rapid degradation of condensin I and condensin II - two essential regulators of mitotic chromosome structure - revealed that both complexes are individually required for cell division in precursor lymphocytes, but not in their differentiated peripheral lymphocyte derivatives. This generalisable approach provides unprecedented temporal control over the dose of endogenous proteins in mouse models, with implications for studying essential biological pathways and modelling drug activity in mammalian tissues.
CRISPR technology has made generation of gene knock-outs widely achievable in cells. However, once inactivated, their re-activation remains difficult, especially in diploid cells. Here, we present DExCon (Doxycycline-mediated endogenous gene Expression Control), DExogron (DExCon combined with auxin-mediated targeted protein degradation), and LUXon (light responsive DExCon) approaches which combine one-step CRISPR-Cas9-mediated targeted knockin of fluorescent proteins with an advanced Tet-inducible TRE3GS promoter. These approaches combine blockade of active gene expression with the ability to re-activate expression on demand, including activation of silenced genes. Systematic control can be exerted using doxycycline or spatiotemporally by light, and we demonstrate functional knock-out/rescue in the closely related Rab11 family of vesicle trafficking regulators. Fluorescent protein knock-in results in bright signals compatible with low-light live microscopy from monoallelic modification, the potential to simultaneously image different alleles of the same gene, and bypasses the need to work with clones. Protein levels are easily tunable to correspond with endogenous expression through cell sorting (DExCon), timing of light illumination (LUXon), or by exposing cells to different levels of auxin (DExogron). Furthermore, our approach allowed us to quantify previously unforeseen differences in vesicle dynamics, transferrin receptor recycling, expression kinetics, and protein stability among highly similar endogenous Rab11 family members and their colocalization in triple knock-in ovarian cancer cell lines.