Tuning gene expression variability and multi-gene regulation by dynamic transcription factor control

Many natural transcription factors are regulated in a pulsatile fashion, but it remains unknown whether synthetic gene expression systems can benefit from such dynamic regulation. Using a fast-acting, light-responsive transcription factor in Saccharomyces cerevisiae, we show that dynamic pulsatile signals reduce cell-to-cell variability in gene expression. We then show that by encoding such signals into a single input, expression mean and variability can be precisely and independently tuned. Further, we construct a light-responsive promoter library and demonstrate how pulsatile signaling also enables graded multi-gene regulation at fixed expression ratios, despite differences in promoter dose-response characteristics. Pulsatile regulation can thus lead to highly beneficial functional behaviors in synthetic biological systems, which previously required laborious optimization of genetic parts or complex construction of synthetic gene networks.


INTRODUCTION 25
The relationship between gene expression and cellular phenotype lies at the center of many 26 questions in different branches of biological research. While strong perturbations of gene 27 expression like knock-outs and overexpression led to a tremendous increase in our 28 understanding of protein function, graded gene expression regulation allows us to obtain a 29 quantitative understanding of the expression-phenotype mapping. Furthermore, conditional 30 and titratable gene expression is of major importance in biotechnology and synthetic biology. 31 Thus, a variety of tools for regulating cellular protein levels, such as gene expression 32 systems based on hormone or light-inducible transcription factors, were developed 1 . With a 33 few exceptions [2][3][4] , expression levels are regulated by adjusting the strength of an input, 34 leading to a graded and constant activation of a transcriptional regulator (Fig. 1a, from here 35 on referred to as amplitude modulation (AM)). In contrast, recent studies have shown that 36 many natural regulatory proteins, including transcription factors (TFs), exhibit pulsatile 37 activation patterns leading to a variety of phenotypic consequences 5 . 38 39 Motivated by the occurrence of pulsatile transcription factor regulation in natural systems, we 40 hypothesized that synthetic gene expression systems can benefit from such dynamic 41 regulation. To test this hypothesis, we constructed a fast-acting, and genomically integrated, 42 optogenetic gene expression system based on the bacterial light-oxygen-voltage protein 43 EL222 in Saccharomyces cerevisiae 4 . Fast kinetics of the optogenetic TF together with the 44 ability to control light intensity with high temporal precision allowed us to tune gene 45 expression using pulsatile TF inputs. In particular, we performed pulse-width modulation 46 (PWM) 3 , meaning that the duration of input pulses is varied to achieve different gene 47 expression levels, while keeping the period of the pulses constant (Fig. 1b). The ratio of 48 pulse duration to the period is referred to as duty cycle. PWM can be performed at different 49 input amplitudes and periods, providing further options for dynamic signal encoding to 50 regulate gene expression levels. We used a mathematical model to identify suitable PWM 51 periods and then showed experimentally that these can be exploited to tune gene 52 expression properties. By comparing this PWM approach to AM, we establish that dynamic 53 encoding of pulsatile signals can drastically increase the functionality of gene expression 54 systems. 55 56 57

59
Characterization and modeling of an EL222-based expression system in S. cerevisiae 60 61 In order to regulate gene expression using PWM, we implemented an optogenetic gene 62 expression system based on a previously described TF consisting of a nuclear localization 63 signal, the VP16 activation domain 6 , and the light-oxygen-voltage domain protein EL222 of 64 Erythrobacter litoralis (VP-EL222) 4 . Blue light illumination triggers structural changes in 65 EL222 leading to homodimerization and binding to its cognate binding site (Fig. 1c). An 66 EL222-responsive promoter was constructed by inserting five binding sites for EL222 4 67 upstream of a truncated CYC1 promoter (5xBS-CYC180pr) and was used to drive the 68 expression of the fluorescent protein (FP) mKate2 7 . For initial characterization, we 69 measured the expression levels of mKate2 in the dark and after 6h of blue light illumination 70 via flow cytometry (Fig. 1d). Illumination led to a VP-EL222 dependent increase in cellular 71 fluorescence of more than 250-fold. In the dark, the presence of VP-EL222 did not affect 72 gene expression. Neither the expression of VP-EL222 nor light-induction affected cell growth 73 or constitutive gene expression (Supplementary Fig. 1).

75
In order to achieve a quantitative understanding of the system and investigate potential 76 PWM regimes, we derived a simple mathematical model of VP-EL222 mediated gene 77 expression (Fig. 1E, for details see Supplementary Note 1). The model was fitted to the 78 data of three characterization experiments, namely a gene expression time-course as well 79 as dose response curves to AM and PWM with a period of 7.5 min (Supplementary Fig. 2). 80 Analyzing the model showed the importance of fast VP-EL222 deactivation kinetics for 81 successful PWM (Supplementary Note 1.3). For a fixed pulse width, slow deactivation rates 82 require long PWM periods to achieve purely pulsatile TF regulation (Fig. 1f). However, such 83 periods may result in significant temporal variation of downstream gene expression / input 84 tracking (Fig. 1g). Here, the half-life of the active VP-EL222 state was inferred to be lower 85 than 2 minutes (Fig. 1f). Measurements of transcription upon a blue light pulse using 86 smFISH lead to results consistent with the fast VP-EL222 kinetics (Supplementary Fig. 3, 87 Supplementary Note 1.4). For the inferred deactivation rate, the model predicts strongly 88 pulsatile TF activity for a 30 min PWM period and a 50% duty-cycle, whereas for the same 89 duty-cycle TF activity does not return to baseline when a 7.5 min PWM period is used (Fig.  90   1f). Importantly, even for a 30 min period, temporal changes in protein expression at steady 91 state are expected to be minor for a wide range of protein half-lives (Fig. 1g,  92 Supplementary Note 1.3). We confirmed experimentally that there is no measurable input 93 tracking for a stable fluorescent protein (Fig. 1g). Thus, the fast kinetics of the VP-EL222 94 based system together with its tight regulation, and apparent lack of toxicity, makes it an 95 ideal gene expression tool for a variety of applications and enables the regulation of protein 96 levels by PWM. 97 Pulsatile input signals achieve coordinated multi-gene expression 98 99 Given that most cellular phenotypes are a result of the coordinated regulation of many genes 100 whose protein expression ratios can be of high importance for achieving these phenotypes 101 8 , we explored the use of AM and PWM for achieving graded expression of multiple proteins, 102 each at a different level, with a single gene expression system. Eukaryotic genes are usually 103 monocistronic and thus, promoter libraries are typically used to adjust relative expression 104 levels 9 . Hence, we built a set of light-responsive promoters differing in the promoter 105 backbone and EL222 binding site number. The resulting promoters showed a wide range of 106 maximal expression levels with promoters based on both the GAL1 and the SPO13 107 backbone exhibiting very low basal expression (Fig. 2a, Supplementary Fig. 4). However, 108 when we analyzed the response of two promoters differing in the number of EL222 binding 109 sites to AM, we found that they show different dose-response behaviors (Fig. 2b). In 110 contrast, PWM with a period of 30 min resulted in coordinated expression with an almost 111 linear relationship between the duty-cycle and the protein output (Fig. 2c). Thus, only PWM 112 is compatible with the use of a simple promoter library for graded multi-gene expression at 113 constant ratios (Fig. 2d). We observed the same behavior when both reporters were located 114 in a single cell (Supplementary Fig. 6a). The use of shorter PWM periods resulted in 115 intermediate levels of coordinated promoter regulation, allowing for input-mediated tuning of 116 expression ratios (Fig. 2d, Supplementary Fig. 6b,c for modeling results). We note, that 117 Elowitz and colleagues have shown that frequency modulation of TF activity can coordinate 118 multi-gene expression in S. cerevisiae 10 . Thus, our work demonstrates how we can learn 119 from natural systems to better regulate gene expression in synthetic systems using simple 120 strategies. 121 122 Reducing and tuning expression variability with pulsatile signals 123 124 While we have so far only analyzed the average response of cells to input signals, gene 125 expression can exhibit a substantial amount of heterogeneity 11 . For many applications, 126 precise single cell regulation of gene expression is desirable 12 . However, the ability to tune 127 variability may allow for the analysis of its phenotypic consequences 11 . To date, variability 128 regulation was achieved by the construction of synthetic gene networks 13-16 -namely 129 feedback loops and cascades-as well as the tuning of promoter features, such as TATA 130 boxes 17 . While variability reduction via frequency modulation of TF activity was proposed 131 theoretically, such behavior has not yet been shown experimentally 18,19 . 132 133 For the synthetic gene expression system, PWM led to reduced cell-to-cell variability in 134 protein levels compared to AM for the same mean expression (Fig. 3a). Furthermore, 135 changing the PWM period enabled tuning of expression heterogeneity with a single input 136 and no change in network architecture (Fig. 3a). In order to investigate the mechanism 137 behind this noise reduction, we performed a dual reporter experiment (see Methods for 138 details). This assay allows for the decomposition of expression variability stemming from 139 stochastic events at the promoter level (intrinsic) and global differences between cells 140 (extrinsic) 20,21 . We found that PWM reduces both extrinsic and intrinsic variability (Fig. 3b,  141 c). However, for most expression levels, extrinsic variability is the dominant source of 142 heterogeneity in the synthetic expression system. Given that TF variability is thought to be a 143 major determinant of extrinsic variability 22 , we hypothesized that PWM leads to lower gene 144 expression heterogeneity by operating in a promoter-saturating regime, where transmission 145 of TF variability to gene expression output is minimal (Fig. 3d). 146  147  To approximate this phenomenon with our simple mathematical model, we performed  148  simulations in which we drew TF concentration from a log-normal distribution describing the  149 single-cell distribution of a mCitrine-tagged 23 version of VP-EL222 (Supplementary Note 150 1.6, Supplementary Fig. 10a). This model can qualitatively recapitulate the experimental 151 data (Fig. 3e, Supplementary Fig. 10b,c). We further showed experimentally that PWM 152 reduced the slope of the correlation between VP-EL222 expression levels and mKate2 153 output (Fig. 3f, Supplementary Fig. 10d). Next, we expressed VP-EL222 from a 154 centromeric plasmid to increase TF variability by introducing plasmid copy number variation 155 (Supplementary Fig. 11) 24 . Under these conditions, AM led to a wide-spread multi-modal 156 protein distribution at intermediate expression levels (Fig. 3g). In contrast, PWM resulted in 157 the merging of these distributions. Thus, the use of PWM does not only reduce 158 heterogeneity as measured by the CV but may also lead to qualitatively different distributions 159 by attenuating the effects of TF variability on downstream gene expression. 160 161 Finally, we sought to map the expression level of the metabolic enzyme URA3 on cell growth 162 in the absence of uracil. We found that the dose response of mean expression to growth 163 depends on expression heterogeneity with tight regulation enabling maximal growth at lower 164 expression levels (Fig. 3h). This result exemplifies the importance of precise regulation for 165 the analysis of expression-phenotype relationships and for the adjustment of optimal protein 166 levels for synthetic biology applications in which metabolic burden is non-negligible 25 . 167 168

169
We present a highly inducible, fast-acting optogenetic expression system for S. cerevisiae 170 which enables the regulation of protein levels by PWM. Learning from the use of pulsatile 171 regulation in a natural system 10

219
The model consists of three ordinary differential equations describing VP-EL222 / TF activity, TF-mediated mRNA production with a transcription rate modeled by Hill kinetics,    were fit to the data for guidance (see Supplementary Table 5 for parameters). Effect of AM and PWM on the CV is shown in Supplementary Fig. 12 All experiments except growth and smFISH were performed in the following way. 326 Cultures were grown overnight starting from single yeast colonies, subcultured in fresh 327 medium and grown for at least 16 h in the dark while maintaining an optical density at 700 328 nm (OD 700 ) lower than 0.4. At the start of the experiment, cells were diluted to an OD 700 of 329 0.005 in 4 ml of medium. Before measurement, cell samples were incubated in SD with 0.1 330 mg/ml cycloheximide for 3.5 h at 30 °C to ensure full fluorescent protein maturation. 331 Samples were analyzed using a LSRFortessaTM LSRII cell analyzer (BD Biosciences, 332 Germany). To measure mKate2 fluorescence, a 561 nm excitation laser and a 610/20 nm 333 emission filter and for mCitrine, a 488 nm excitation laser and a 530/11 nm emission filter 334 were used. Data was analyzed using R with the flowCore package 32 . Cells were gated 335 based on forward and side scatter to remove debris and cell aggregates. For strains 336 containing centromeric plasmids, a budded cell population was selected by gating based on 337 the forward and side scatter width 33 . We found that this population shows a higher 338 percentage of responsive cells, which likely results from a higher degree of plasmid 339 retention. Strong outliers were removed from the data as follows: First, the fluorescence 340 values were log-transformed. Outliers were defined as data-points with an absolute deviation 341 from the fluorescence distribution median of greater than 3-fold of the median absolute 342 deviation. 343 For the analysis of gene expression heterogeneity, fluorescent levels were normalized by 344 side scatter area to reduce the effect of cell size (see Supplementary Fig. 8  The image analysis procedure was performed using custom Matlab scripts and consists of 439 three steps: segmenting individual nuclei (based on DAPI images), locating fluorescent spots 440 in the nuclear regions, and quantifying the intensity of these spots. 441 Nuclei were first enhanced by using the difference of Gaussians algorithm. Nuclear regions 442 were then segmented by manually optimized thresholding. Detected regions that were too 443 big or small to represent nuclei were removed. For each nuclear region, a Difference of 444 Gaussian algorithm was used to enhance spots in the CY3 images and spots were identified 445 using thresholding. In order to quantify the intensity of the nuclear spots, the sum of a two-446 dimensional Gaussian function and a 2D-plane was fitted in a square area around the 447 identified spot with an edge length of 19 pixels. If no spot was detected, the same procedure 448 was performed at the center of the nuclear region. Spot intensity was then defined as the 449 integral of the Gaussian function. For each nucleus / cell, the spot with the highest intensity 450 was defined as the transcription site.

Supplementary Note 1: Mathematical Modeling of the VP-EL222 mediated expression
In the following section, the specifics about the ordinary differential equation (ODE) model, parameter inference and model analysis are explained.
Suppl. Note 1.1: A simple ODE model describing VP-EL222 mediated gene expression In order to obtain a quantitative understanding of the VP-EL222 based gene expression system in S. cerevisiae , we constructed a mathematical model of this process. The main purpose of the model is to describe the dynamics of the system in order to understand how dynamic inputs can be used to shape the gene expression output. We thus decided to employ a simplistic gene expression model consisting of three ODEs. These ODEs describe: (1) Light dependent activation and of the TF VP-EL222.
For simplicity, we assume that TF multimerization and promoter binding occur on fast timescales compared to the transcription process and we thus use Hill-type kinetics to model the effect of activated VP-EL222 (TF on ) on the transcription rate.
The model is described by the following ODEs (I denotes the blue light input and TF tot denotes the total amount of cellular VP-EL222): Suppl. Note 1.2: Fitting model parameters to experimental data The model possesses 10 parameters (9 rate parameters and the total concentration of the TF / VP-EL222 (TF tot )). The value TF tot acts as a scaling factor that can be by completely compensated for by changes in other parameters (k on and K d ) and does not affect the dynamics of the system. We thus fixed this value to 2000 molecules/cell. All characterization experiments were performed using the VP-EL222 mediated expression of the stable fluorescent protein (FP) mKate2. We thus equate the protein degradation rate (k degP ) to the cellular growth rate of 0.007 min -1 (results of growth rate measurements are shown in Supplementary Fig. 1a ). Thus, we end up with 8 free parameters that need to be estimated.
For this purpose, we performed three characterization experiments using the strain DBY43, expressing mKate2 from the 5xBS-CYC180 promoter.
1. We performed time-course measurements of mKate2 expression under constant illumination conditions. This experiment was performed to elucidate the kinetics of VP-EL222 activation and (to a larger extent) that of mRNA accumulation / degradation. The kinetics of protein accumulation are given by having fixed k degP . 2. We analyzed the dependence of mKate2 expression on light intensity, i.e. AM. This experiment gives us information about the mapping of light intensity to active transcription factor and finally transcription / protein expression. 3. We analyzed the dependence of mKate2 expression on the duty cycle in a PWM experiment with a short, 7.5 min period. The rationale behind this experiment is that it provides us with information about the kinetics of VP-EL222 activation and deactivation.
The results of these experiments are shown in Supplementary Fig. 2 . We note again, that we are mainly interested in the dynamics of the gene expression system and not the absolute values of cellular mRNA or protein contents. For the model fitting, we thus assume a direct relation between fluorescence measured by flow cytometry and protein expression in the model. Parameters were estimated by fitting the model to the mean of three independent experiments of each class of characterization experiments. To do so, we used a simplexbased search (Nelder-Mead algorithm, "fminsearch" function in Matlab) to minimize the sum of squared residuals (SSR) between the model and the data. This procedure was performed for different initial parameter values. The parameters resulting in the minimal SSR between all runs were used in this study and are reported in Supplementary Table 1 . The model fits are shown in Supplementary Fig. 2 .
Suppl. Note 1.3: Using the model to analyze functional regimes for PWM The goal of PWM in this study is to regulate TF activity in a pulsatile fashion, while leading to close to constant protein levels over time at steady state. We used the mathematical model to analyze how these properties are affected by different parameters, mainly k off , k degP , and the PWM period. In order to ensure that we are not analyzing transient model behavior, all metrics described below are calculated after running the model for a simulated time of 720 min.
Effects of pulsatile TF regulation via PWM can be expected to be most pronounced when the concentration of active TF directly follows the light input, meaning that cellular TF activity itself shows either the maximal desired value or its basal level at any given time. However, in every realistic scenario, the temporal TF activity will deviate from this behavior to an extent that depends on the kinetics of TF activation / deactivation as well as the PWM period. We hence analyzed how the PWM period and the TF deactivation rate (k off ) affect TF pulsing. In order to quantify this behavior, we use a tracking score defined by the ratio between the integrated TF activity during the light pulse and the whole period ( Fig. 1f ). This metric is 1 if the TF activity perfectly tracks the input and equals the duty-cycle if TF activity does not change over the PWM period. We calculated values of this metric for a duty cycle of 50% ( Fig. 1f ) As expected, the model shows that longer PWM periods are required with decreasing k off to achieve a similar tracking score. The model further shows that the inferred rate of k off for VP-EL222 (0.34 min -1 , equivalent to a on-state half-life of about 2 min) is sufficiently large for performing PWM with reasonable periods. For a period of 30 min cellular TF activity is predicted to be at its maximal level during much of the light pulse and to return to basal levels in the dark before the next pulse ( Fig. 1f ). In contrast, when the period is reduced to 7.5 min, TF activity is predicted to reach its maximal activity before the end of the light pulse and to not return to the basal level in the dark ( Fig. 1f ). This leads effectively to gene regulation via mixed contributions of constant TF activity and weak pulsing. We found experimentally that this difference has strong functional consequences for the ability to use PWM for gene co-regulation ( Fig. 2d, Supplementary Fig. 7 ) and noise reduction ( Fig.  3a,e ).
The model shows that pulsatile TF regulation can be more easily achieved with long PWM periods. However, using long PWM periods to regulate gene expression can potentially result in significant temporal fluctuation on the protein level, which is often not desirable. We thus sought to quantify the temporal response of protein expression to PWM. We use a score defined by the ratio of the maximal expression difference during the period divided by the mean expression level ( Fig. 1g ). Using this score, we found that even for a 30 min period, temporal changes in protein expression at steady state are expected to be minor for a wide range of protein half-lives ( Fig 1g ). We confirmed experimentally that there is no measurable input tracking for a stable fluorescent protein ( Fig 1g ). For the median protein half-life of ≈40 min in S. cerevisiae 1 , PWM is predicted to lead to a maximal temporal fluctuation of about 6% for a 10% duty cycle. We note that this value is also affected by the mRNA degradation rate. Parameter estimation resulted in a value of 16.5 min for the mRNA half-life, which is close to the median half-life in S. cerevisiae (10 -20 min) 2,3 . Thus, modeling suggests that VP-EL222 should enable PWM-based regulation of a large percentage of yeast proteins. For short-lived proteins, the system would need to be optimized by introducing mutations that increase the dark-reversion rate. Such mutations were identified previously 4,5 . We further note that VP-EL222 was previously employed in higher eukaryotes 6 , where both mRNA and protein degradation rates were measured to be significantly lower than in S. cerevisiae 7,8 . It is thus likely that PWM can be very successfully applied in these organisms and that PWM should also be possible with other tools for gene expression regulation that work on longer time-scales.

Suppl. Note 1.4: Estimating nascent RNA accumulation
Single-molecule FISH (smFISH) allows for the quantification of nascent transcripts, which is a fast readout of VP-EL222's transcriptional activity. We performed an smFISH experiment in which we measure the transcriptional response of 5xBS-CYC180pr to a 20 min light pulse ( Supplementary Fig. 3 ). To evaluate whether the identified model parameters describing VP-EL222 activity and the transcription process are consistent with this data, we introduce an ODE describing nascent RNA accumulation: Here, TF-dependent nascent RNA production is modeled using Hill-type kinetics with the same parametrization as for mRNA production ( Equation 2 , Supplementary Table 4 ). The rate at which a nascent RNA escapes from the transcription site (k esc ) is given by the RNA dwell-time which includes elongation and termination, leading to k esc = (elongation time) -1 + (termination time) -1 . The termination time was set to the literature value of 70 seconds 9 and the elongation time was set to 100 seconds based on the transcript length of 2000 bases and an average elongation rate of 20 bases per second 9 . We found that the predicted dynamics of nascent RNA accumulation closely resemble the experimental data ( Supplementary Fig. 3c ). We note that this model is very simplistic -it does for example assume that nascent RNAs are observable (via smFISH) directly after transcription initiation.
Suppl. Note 1.5: Refitting of promoter-specific model parameters In order to describe protein expression from the 2xBS-CYC180 promoter with the mathematical model identified above, we need to re-fit the promoter-specific parameters, k basal , k max , K d , and n. To do so, we performed two characterization experiment for this promoter, namely we measured the expression response to AM and PWM with a 7.5 min period. Model fitting was performed as described in Suppl. Note 1.2 . Experimental results and fits are shown in Supplementary Fig. 5 .

Suppl. Note 1.6: Modeling the effects of VP-EL222 variability on gene expression
Previous studies suggest that two major sources of extrinsic gene expression variability in S. cerevisiae are heterogeneity in TF expression and the cell cycle 10,11 . Due to the fact that we directly affect TF dynamics, we thought to introduce cell-to-cell variability in TF concentration to our model and analyze the resulting CV-mean relationship for AM and PWM. As performed in other studies 12 , we modeled protein / TF variability by running 10,000 ODE simulations differing only in the value for TF tot for each input condition (consequences of this modeling choice are described below). Here, each simulation represents a single cell.
To do so, we first measured the fluorescence distribution of mCitrine tagged VP-EL222 to estimate heterogeneity of TF expression. We find that this distribution can be well described by a log-normal distribution with a CV of roughly 0.2 ( Supplementary Fig. 10a ). We then drew values for TF tot from a log-normal distribution with a CV of 0.2 and a mean value of 2000 (the TF tot value used for parameter estimation) and ran ODE simulations for a simulated time of 360 min. We ran simulations for different types of inputs (AM, and PWM with a 7.5, 15, and 30 min period) and different promoters (5xBS-CYC180pr and 2xBS-CYC180pr). Results of these simulations are shown in Fig. 3e and Supplementary  Fig. 10b.

VP-EL222 -dependent promoter sequences / reporter constructs
For all following promoter sequences, EL222 binding sites (BS; called C120 in the original publication) are underlined and promoter backbones are green. A sequence containing 5 binding sites for EL222 was amplified from pcDNA-C120-mCherry 6 . All other binding site combinations were constructed by oligonucleotide annealing to obtain a single plasmid containing a single EL222 binding site, followed by duplications of this sequence using restriction enzyme cloning.

pDB60 / 5xBS-CYC180pr-Kozak-mKate2-ADH1t
pDB60 is used to express mKate2 15 under control of the 5xBS-CYC180 promoter. The promoter consists of a sequence containing 5 EL222 binding sites as well as a 180 bp sequence derived from the CYC1 promoter ( CYC180 ). CYC180 was amplified from BY4741 genomic DNA 16 . A consensus Kozak sequence was inserted upstream of the start codon to enhance translation. The mKate2 reporter gene is inserted into pFA6a-His3MX6 17 using PacI and AscI sites (upstream of the ADH1 terminator ). 5xBS-CYC180pr is inserted using HindIII and PacI. All other VP-EL222 dependent promoters (see below) were characterized using the same plasmid backbone and were integrated into HindIII/PacI digested plasmid.

aagcttTTTAATTATATCAGTTATTACCCGGTACCCCCCTCGAG GAATTTTCAAAAATTCT
In order to achieve light-dependant gene expression with very low basal expression, we inserted EL222 binding sites upstream of the basal SPO13 promoter 19  Further sequences expressed from VP-EL222 dependent promoters As shown above for mKate2 in pDB60, all sequences were integrated into PacI, AscI digested pFA6a-His3MX6-derived plasmids. Furthermore, all sequences possess a Kozak consensus sequence directly upstream of the start codon.

pDB78 / mKate2 -24xPP7SL
A sequence containing 24 tandem repeats of the PP7 stem loop was amplified from pDZ416 20 and was inserted after the mKate2 stop codon.