Real-time quantification of protein expression at the single-cell level via dynamic protein synthesis translocation reporters

Protein expression is a dynamic process, which can be rapidly induced by extracellular signals. It is widely appreciated that single cells can display large variations in the level of gene induction. However, the variability in the dynamics of this process in individual cells is difficult to quantify using standard fluorescent protein (FP) expression assays, due to the slow maturation of their fluorophore. Here we have developed expression reporters that accurately measure both the levels and dynamics of protein synthesis in live single cells with a temporal resolution under a minute. Our system relies on the quantification of the translocation of a constitutively expressed FP into the nucleus. As a proof of concept, we used these reporters to measure the transient protein synthesis arising from two promoters responding to the yeast hyper osmolarity glycerol mitogen-activated protein kinase pathway (pSTL1 and pGPD1). They display distinct expression dynamics giving rise to strikingly different instantaneous expression noise.

a. Stimulation with 0.2 M NaCl of two different clones carrying a pSTL1-dPSTR R2,3 containing the SynZips SZ2 and SZ3, which are not interacting. The non-functional pSTL1-dPSTR R2,3 does not display any nuclear enrichment. For comparison, the functional pSTL1-dPSTR R2,1 (yDA134) is plotted with a dashed blue line. b. Stimulation with 0.4 M NaCl of a strain carrying a pGAL1-dPSTR R , which is not expressing upon hyperosmotic shock, and does not display any nuclear enrichment. For comparison, the pSTL1-dPSTR R (yDA134) is plotted with a dashed red line. c. Comparison of pSTL1 expression induced by 0.2 M NaCl measured by flow cytometry (pSTL1-qVenus, cyan, N C~1 0'000) or microscopy for Venus expression or pSTL1-dPSTR R relocation (blue and red respectively, data from Fig. 1). Note the close overlap between the rise of transcription quantified with flow cytometry or the dPSTR. To allow for a direct comparison of the three methods, the fluorescent values measured are normalized. Percentage of expressing cells measured by the dPSTR or by the promoter-FP method, for each concentration. Cells having an expression output above the expression threshold plotted in a, c, d and e are defined as expressing cells. Note that the proportions are similar using the two methods. Each dot represents the mean of three experiments and the error bars, the standard deviations.  b. Difference between nuclear and cytoplasmic RFP concentrations expected for a fixed total RFP·SZ concentration (0.2 µM) and a range of NLS·SZ concentration (blue line). The dotted black line indicates the total NLS·SZ concentration. At low concentrations, there is a linear relationship between NLS·SZ level and nuclear minus cytoplasmic RFP concentration (red dash-dotted line). At higher concentration, the nuclear enrichment of the RFP saturates since an increasingly higher fraction of the RFP is already localized in the nucleus. If we tolerate 10% deviation from linearity (solid red curve), we obtain an upper limit for the concentration of NLS·SZ that can be measured (red arrow). The sensitivity of the detection allows to observe a 10% increase in RFP nuclear enrichment, thereby setting a lower limit to the concentration of NLS·SZ that can be measured (orange arrow). Therefore, the shaded area represents the sensitivity range of the assay for the selected RFP·SZ expression level. c. The lower and upper limits of sensitivity of the system are calculated for a range of RFP·SZ concentrations. An NLS·SZ sensitivity window is depicted with a shaded area bound by lower and higher limits (orange and red lines, respectively). The dotted black line indicates the case where NLS·SZ = RFP·SZ. d. Calibration of the protein number and concentration of RFP·SZ in the reporter strains (yDA134, blue squares and yDA93 orange, 6 biological replicates) relative to endogenously tagged proteins with different expression levels (blue circles, three biological replicates, see Supplementary Table 4). The mean and standard deviation of more than 500 single cells is plotted and fitted by linear regression (dashed black line). The expression level of RFP is found to be 4400±300 protein per cell or 0.18µM for yDA134 and 16'000±750 protein per cell or 0.66µM for yDA93 (dashed line in panel c and d). e. Normalized nuclear enrichment in course of time for yDA134 (blue) and yDA93 (orange) stimulated with 0.1 (resp. N C =558 and N C =341), 0.2 (resp. N C =655 and N C =297), or 0.4M NaCl (resp. N C =802 and N C =268). Note the good overlap between the responses of the two strains at 0.2 and 0.4M NaCl and the lower sensitivity at 0.1 M. c. The dynamics of total NLS·SZ production (dotted line) is compared to the RFP nuclear-cytoplasmic enrichment (solid line) in the case where the NLS·SZ is stable or if it has a half-life of 2 min. From this model, it can be observed that the dynamics of RFP nuclear enrichment can be limited by two reactions, the formation of the complex between NLS·SZ and the RFP·SZ and the import of the NLS-SZ in the nucleus. Both of these reactions happen with fast dynamics thus allowing a monitoring of the protein production in the sub-minute time scale. The deactivation of the system depends on three reactions, the repression of the active gene, the degradation of the mRNA and the degradation of the NLS-SZ. In our experimental data, the return of the dPSTR nuclear enrichment to pre-stimulus level takes place on the order of 10 minutes suggesting that the NLS·SZ degradation is not the limitting factor for this process. d. Characterization of destabilization sequence half-life. Cells bearing the pGPD1-dPSTR Y and pGPD1-dPSTR R were grown overnight in SD+1M sorbitol to increase the basal level of pGPD1 expression (note the differences in nuclear enrichment before time 0 compared to Fig. 5b). Cells were treated with cycloheximide (solid line, N C = 380) or not (dashed line, N C = 383). A half-life for the UbiY-NLS-SZ is estimated to 2.1±0.5 min (3 experiments, mean and standard deviation of the RFP and YFP dynamics) based on an exponential fit of the data (red line).   pDA137 S1a yDA112 ySP37 HTA2-CFP ura3: pGAL1-dPSTR R2-1 pDA169 S1b, S5d * The numbers in the superscript dPSTR R2-1 indicate the pair of SynZip used.

Supplementary Table 2: List of plasmids used in this study
pMET-PP7-2xGFP -pRS304 ySP374 Supplementary   Figure 4). The model contains three species: the fluorescent protein linked to the SynZip (FP·SZ), the inducible peptide (NLS·SZ) and the complex formed between those two proteins via the SynZip (FP·SZ·NLS). The K d of the complex formed has been estimated to 10µM 2,3 . These three species are partitioned between two compartments: the nucleus and the cytoplasm. We estimate a passive diffusion of the FP·SZ into and out of the nucleus to be 0.005 s -1, 4 . Based on the level of nuclear enrichment of the Venus·SZ (Figure 1d), we estimate that the double NLS present on the NLS·SZ peptide allows at least a 10-fold enrichment of the proteins in the nucleus. Using these parameters, we have simulated the steady-state nuclear and cytoplasmic concentrations for each component for a range of total NLS·SZ and FP·SZ. For each condition, the nuclear to cytoplasmic difference is calculated as: Figure S4b displays the outcome of the model for a concentration of 0.2 µM of RFP·SZ. For low concentrations of the NLS·SZ, there is a linear correlation between nuclear enrichment of the RFP and the NLS·SZ concentration. However, there is a limit to the detection ability of the microscope and we estimate that only a 10% enrichment can be reliably detected at the single cell level (~10 -2 µM), thereby setting a lower limit to the NLS·SZ that can be measured with the dPSTR (1.5x10 -2 µM). If the NLS·SZ concentration reaches too high levels, the linear relationship with RFP nuclear enrichment is lost, as can be seen from the saturation of the blue curve in Supplementary Figure 4b. Allowing for 10% error in linearity between NLS·SZ level and nuclear enrichment measurement results in an upper limit in NLS·SZ that can be reached before saturation (1.6x10 -1 µM). These upper and lower limits were calculated for a range of RFP·SZ concentrations (Supplementary Figure 4c).
In parallel to this modeling effort, we have quantified the level of RFP·SZ expression in the dPSTR strain. We measured a calibration curve based on the fluorescence intensity of mCherry tagged proteins with known expression levels! 5, 6, ! 7 . After linear regression between protein numbers and fluorescent intensities, we can get a good estimate of the protein number of the constitutively expressed fluorescent protein in the dPSTR. At 4 400 proteins per cell or 0.18 µM (for a 40 fl volume 8,9 ), we can estimate the low and high limits of expression that can be quantified with the dPSTR between 1.2·10 -2 to 1.2·10 -1 µM of synthesized NLS·SZ. For comparison, we also quantified the fluorescence of another dPSTR strain with higher expression levels (16'000 prot. per cell, yDA93). The concentration range that can be quantified varies from 3.9·10 -2 to 6·10 -1 µM. These two strains have different sensitivity windows. As predicted by the model, we observe with this strain a lower sensitivity at detecting protein expression from the pSTL1 promoter at 0.1 M NaCl (Supplementary Figure 4e).

Model of NLS·SZ synthesis and degradation
We further developed this model to include a minimal transcriptional model in order to simulate the dynamics and level of nuclear import with a stable and unstable dPSTR (Supplementary Figure 5). The reactions and rate constants used in the model are presented in Supplementary Figure 5a. Briefly, the input to the model is the level of activated MAPK (MAPK P ), which increases after 10 min of simulation and declines in a linear fashion to reach zero after a time that is function of the level of activity upon stimulation. This mimics the different temporal windows of Hog1 activity upon various hyper-osmotic stresses. The active kinase turns the promoter into an active state. From this state, mRNA can be produced, which ultimately leads to the expression of the NSL·SZ. This peptide is produced in the cytoplasm, and it can bind to the FP·SZ and accumulate in the nucleus following the same reactions as described above. Once kinase activity returns to pre-stress levels, the promoting region of the DNA returns to its off-state, the mRNA molecules are degraded and the peptide NLS·SZ will be degraded as well.
To test the influence of the destabilizing sequence on the output of the model, we have selected two variants, the first one where the NLS·SZ is stable, and the second one where the NLS·SZ is actively degraded, which increases the rate of this reaction by 100 fold (Supplementary Figure 5c). To estimate the stability of the UbiY-NLS-SZ in cells, we have performed an experiment where protein transcription is blocked by cycloheximide and the degradation kinetics of the protein can thus be quantified. A strain bearing the pGPD1-dPSTR R and pGPD1-dPSTR Y was grown overnight in SD-full containing 1M Sorbitol, diluted in the morning with the same medium, placed under the microscope and treated with 0.1mg/ml cycloheximide (Supplementary Figure 5d). In this high osmolarity medium, the basal expression level of pGPD1 is increased due to the higher basal activity of Hog1. We measured a half-life of 2.1 ±0.5 min for the UbiY-NLS-SZ peptide which was used in the model. As expected and in agreement with our experimental measurements of the dPSTR, due to accumulation of all NLS·SZ expressed in the cell, the stable inducible peptide can lead to more saturation than the degraded one (Supplementary Figure 5c). However, the degraded peptide will be expressed at lower levels and thus nuclear enrichment of the fluorescent protein might be more difficult to detect at low concentrations. Therefore, the level of expression of the RFP·SZ might have to be adapted to accommodate the transcriptional output of the inducible promoter. We have achieved this by testing different combinations of ribosomal protein gene