Tuning of mRNA stability through altering 3′-UTR sequences generates distinct output expression in a synthetic circuit driven by p53 oscillations

Synthetic biological circuits that can generate outputs with distinct expression dynamics are useful for a variety of biomedical and industrial applications. We present a method to control output dynamics by altering output mRNA decay rates. Using oscillatory expression of the transcription factor p53 as the circuit regulator, we use two approaches for controlling target gene transcript degradation rates based on the output gene’s 3′-untranslated region (3′-UTR): introduction of copies of destabilizing AU-rich elements into the 3′-UTR or swapping in naturally occurring 3′-UTRs conferring different transcript stabilities. As a proof of principle, we apply both methods to control the expression dynamics of a fluorescent protein and visualize the circuit output dynamics in single living cells. We then use the naturally occurring 3′-UTR approach to restore apoptosis in a tunable manner in a cancer cell line deficient for caspase-3 expression. Our method can be readily adapted to regulate multiple outputs each with different expression dynamics under the control of a single naturally occurring or synthetically constructed biological oscillator.

to their placement in noncoding regions as well as their relatively short sequence lengths, 3′-UTR AREs provide a potentially useful strategy for altering mRNA expression dynamics in a controlled manner.
In this study, we show that synthetic target expression dynamics in response to p53 oscillations can be regulated through directed perturbation of mRNA stability. We perturb target mRNA stability through two methods: using naturally occurring 3′-UTRs of different stabilities engineered to novel cDNAs; and destabilizing an mRNA by introducing varying copies of AREs into an engineered 3′-UTR. As a proof of principle, we first applied both strategies to alter the expression dynamics of a fluorescent protein reporter under the control of a p53-responsive synthetic promoter induced by DNA double strand breaks. We then used CASP3 (encoding the pro-apoptotic factor caspase-3) as the target gene with a variety of 3′-UTRs to restore apoptosis in a regulated manner in the caspase-3-deficient breast carcinoma cell line MCF-7. Our study highlights the potential of using an oscillatory biological circuit to generate desired target gene expression dynamics for bioengineering and translational medicine applications.

Results
Altering output transcript stability with AU-rich elements generates distinct expression dynamics from a biological oscillator. For genes induced by a transcription factor with oscillatory expression dynamics, mRNA expression can exhibit a range of dynamics from oscillatory to continuously rising ( Fig. 1; ref. 2 ). The decay rate of the target mRNA relative to the frequency of the transcription factor oscillations is a key factor in determining the dynamics (Fig. 1). 3′-UTRs can confer different decay rates to mRNAs, and therefore we predicted that the expression dynamics of a long-lived protein encoded by mRNAs with distinct 3′-UTRs would show a range of absolute levels of expression and rates of accumulation (Fig. 1).
We tested this prediction using the p53 oscillatory response to DNA double strand breaks as the input for several synthetic gene circuits. As a proof of principle, we first used a gene encoding the red fluorescent protein mCherry as the circuit output, as its dynamics can be readily visualized in single, live cells over time enabling precise quantification of expression dynamics ( Fig. 2A). We constructed synthetic promoter reporter output plasmids for the p53 oscillator by expressing the cDNA encoding mCherry under the control of a p53-responsive reduced MDM2 promoter. Downstream of the mCherry cDNA, we inserted the sequence for the TP53I3 3′-UTR. The TP53I3 transcript is relatively long-lived compared with other p53 target gene transcripts, and its 3′-UTR has a relatively short sequence, making it a suitable candidate as a synthetic circuit component. We made analogous circuit output reporter constructs by altering the 3′-UTR through the addition of one or two copies of the ARE sequence 5′-TTATTTATTTATTATTTATTTATT-3′ ( Fig. 2A) from the mouse Tnf 3′-UTR. Transfecting the plasmids individually into an MCF-7 p53-Venus clonal cell line previously characterized as expressing p53 oscillations in individual cells 7 , we developed three distinct stable MCF-7 cell lines expressing the mCherry output reporters.
To directly determine the efficacy of AREs in destabilizing the mCherry transcript, we induced expression of mCherry mRNA with the radiomimetic drug neocarzinostatin (NCS; ref. 15 ) to generate DNA double strand breaks (DSBs). Simultaneously, we supplemented cell culture medium with 5-ethynyl uridine (EU) to pulse label newly synthesized transcripts for 3 h and then harvested cells upon chasing with fresh growth medium without EU 16 . We observed that the addition of ARE sequences increased the decay rate of the synthetic output transcripts, as predicted ( Fig. 2B-D).
We next determined if the addition of destabilizing sequences to the 3′-UTR of the output gene could alter the protein expression dynamics for the synthetic circuit output. mCherry expression was monitored in single live cells via long-term time-lapse fluorescence microscopy for 24 h following activation of the p53 oscillator with NCS. Cells expressing the mCherry output mRNA without AREs, and therefore with the most stable transcript, showed overall higher basal expression, greater dynamic range of activation, and more step-like or rising mCherry protein expression (Fig. 2E,F). These expression dynamics were comparable to those for endogenous p53 target genes such as TP53I3 previously characterized as having "rising" expression dynamics 2 . The addition of one or two AREs to the mCherry-TP53I3 3′-UTR resulted in decreases in the basal expression, dynamic range, and rate of accumulation of mCherry ( Fig. 2G-J), as predicted (Fig. 1).
Using 3′-UTRs of endogenous p53 target genes with differing stability generates distinct expression dynamics for a synthetic p53 target gene. As a second approach to tune the expression dynamics of our synthetic output gene, we made use of naturally occurring 3′-UTRs of genes with differing mRNA decay rates. Previously, target genes of p53 had been categorized as having rising, weakly pulsing, or strongly pulsing expression dynamics, correlating with decreasing mRNA stability, respectively 2 . We selected p53 target genes from each of these three categories that had relatively short 3′-UTRs -the rising gene DDB2, the weakly pulsing gene TRIAP1, and the strongly pulsing gene GADD45A. Insertion of these 3′-UTR sequences downstream of the mCherry cDNA (Fig. 3A) resulted in mCherry transcripts with different decay rates ( Fig. 3B-D).
We next determined the output mCherry expression dynamics for each of the p53 target 3′-UTR synthetic constructs when expression was driven by p53 oscillations. We generated three cell lines expressing the synthetic output circuits consisting of the MDM2 reduced promoter, mCherry cDNA, and a p53 target gene 3′-UTR (Fig. 3A). We monitored mCherry protein expression in live cells for 24 h following induction of DNA DSBs with NCS ( Fig. 3E-J). We observed that mCherry expression had the highest basal expression level, greatest dynamic range of expression, and rising dynamics for cells with the mCherry output component having the DDB2 3′-UTR, corresponding to the most stable transcript (Fig. 3E,F). A decrease in basal expression, dynamic range, and less characteristic rising dynamics were observed for 3′-UTRs that conferred lower stability on the mCherry output transcript (Fig. 3G-J). Comparison of integrated mCherry levels as a function of integrated p53-Venus levels in individual cells over 24 hours indicated a weak positive correlation for the TRIAP1 (Pearson's linear correlation ρ = 0.22) and GADD45A (Pearson's linear correlation ρ = 0.25) 3′ UTR circuits, although the DDB2 3′UTR circuit www.nature.com/scientificreports www.nature.com/scientificreports/ was uncorrelated ( Supplementary Fig. S1). Use of clonal cell lines genetically identical for the mCherry gene, rather than a stably transfected but genetically variable population of cells, could potentially be used to increase the correlation, if desired.
Taken together, our results provide two methods, addition of AREs and use of naturally occurring 3′-UTRs conferring distinct mRNA decay rates, for tuning the basal expression levels, dynamic range, and qualitative shape of the dynamic response for synthetic biological circuits.
Altering transcript 3′-UtRs can restore apoptotic activity in a tunable manner. Having used 3′-UTRs to alter transcript stability and circuit output dynamics in a tunable manner for a fluorescent reporter, we next developed analogous synthetic circuits with a more translationally relevant goal, restoring apoptotic activity in caspase-deficient cells. MCF-7 cells have a CASP3 mutation, preventing effective activation of apoptosis 17 . We therefore constructed plasmids using CASP3, coding for the uncleaved pro-caspase 3 protein, as the circuit output gene (Fig. 4A). To the CASP3 cDNA, we cloned 3′-UTRs corresponding to those for the p53 targets DDB2, TRIAP1, and GADD45A (Fig. 4A), and established stable MCF-7 cell lines expressing the plasmids. Quantification of mRNA decay rates for the CASP3 constructs in response to DNA double strand breaks showed that addition of the different 3′-UTRs altered mRNA stability ( Fig. 4B-D Figure 1. An oscillating transcription factor generates distinct output expression dynamics as a function of mRNA decay rates. A model predicts that distinct mRNA and protein expression dynamics can be generated by altering the transcript stability of target genes induced by an oscillating transcription factor. 3′-UTRs can confer different transcript stabilities. Based on the relationship between the transcript decay rate and the oscillator frequency, output gene transcripts will have rising, weakly pulsing, or strongly pulsing mRNA expression dynamics, leading to alterations in the rate of accumulation of an output protein product. www.nature.com/scientificreports www.nature.com/scientificreports/ We next determined whether apoptosis could be restored in a tunable manner using the caspase-3 circuits when driven by p53 oscillations. We generated three different cell lines each with stable expression of a different CASP3 -p53 target gene 3′-UTR output plasmid. We treated cells with NCS to induce DNA damage and initiate p53 oscillations, and we then stained cells with propidium iodide and annexin-V 24 h after NCS treatment. We observed the greatest percentage of apoptotic cells for cells expressing CASP3 with the DDB2 3′-UTR (Fig. 4E), corresponding to the most stable transcripts (Fig. 4B). Fewer apoptotic cells were detected for cells expressing CASP3 with the TRIAP1 or GADD45A 3′-UTRs (Fig. 4E), corresponding to less stable transcripts (Fig. 4C,D). These results suggest that 3′-UTRs can be used to alter the stability of output mRNA transcripts, providing a tunable method to alter a variety of biological circuit dynamics.

Discussion
Synthetic biological circuits are useful for many applications, from probing the regulation and function of naturally occurring processes to engineering completely novel metabolic pathways 18 . Since the likelihood of failure for a synthetic circuit increases as more components are added, methods for controlling multiple outputs with distinct thresholds and temporal characteristics is desirable. In this study, we present relatively straight-forward methods for generating different outputs from biological oscillators. Due to manipulating the decay rates of mRNAs through noncoding regions, our methods should be readily adaptable to a wide range of systems since they do not alter protein coding sequences.
Numerous examples of biological oscillators have been identified in natural signaling and transcriptional circuits 1,[5][6][7][8][19][20][21][22][23][24][25][26][27][28][29] and robust synthetic oscillators are some of the earliest examples of constructed circuits 30,31 . We previously showed that natural oscillators can drive a wide spectrum of qualitatively distinct output expression dynamics 2 . Adapting this strategy from the p53 system to generate diversity in transcriptional responses for other oscillators should be relatively simple, whether for naturally occurring oscillators such as NF-κB and circadian circuits or for synthetic oscillators such as a bacterial repressilator 30 .
The steady state levels of a transcript depend on a balance of mRNA production and degradation rates. Regulation of output transcript decay rates through alteration of 3′-UTR sequences also provides a method for tuning steady state expression levels for a wide range of synthetic transcriptional circuits. This method is effective regardless of the characteristics of the promoter driving synthesis of the transcript of interest, whether, for example, from a promoter regulated by an oscillating transcription factor, a constitutive promoter, or a promoter with a high level of burstiness [32][33][34][35][36] . A benefit gained from controlling the degradation rate of transcripts regulated by       www.nature.com/scientificreports www.nature.com/scientificreports/ an oscillating transcription factor is that it provides a potential mechanism for generating a wider range of qualitatively distinct expression dynamics prior to the steady state.
There are many ways in which our initial designs might be improved to yield more optimal outputs or more complex regulatory control. For example, alternative AREs of varying numbers might enable more precise regulation of mRNA decay rates. Additionally, a larger survey of naturally occurring 3′-UTRs could potentially yield a variety of useful sequences based on size and degradation rate. Optimal choices for noncoding sequences may depend on characteristics of the specific output mRNAs for any given synthetic circuit. Finally, alteration of the protein decay rates would provide an additional level of regulation for controlling the ultimate protein expression output for many synthetic circuits. Future studies refining and building upon our methods will likely provide additional tools for the synthetic biology toolbox.
Human cell lines and culture. MCF-7 breast carcinoma cells were cultured in base growth medium of RPMI, 10% fetal bovine serum (FBS), 100 U/mL penicillin G, 100 mg/mL streptomycin sulfate, and 250 ng/mL amphotericin B. MCF-7 p53-Venus cells 7 that express fluorescently tagged p53 were cultured in base growth medium containing 400 mg/mL neomycin.
All cell lines were incubated and grown at 37 °C and 5% CO 2 .
mRNA decay rate measurements. For MCF7 p53-Venus cell lines containing p53RE-mCherry-DDB2 3′-UTR, p53RE-mCherry-TRIAP1 3′-UTR or p53RE-mCherry-GADD45A, and for MCF7 cell lines with constructs of p53RE-CASP3-DDB2 3′-UTR, p53RE-CASP3-TRIAP1 3′-UTR or p53RE-CASP3-GADD45A 3′-UTR, 4 × 10 5 cells were plated on two 6-cm dishes for each cell line and each time point of the experiment, two days prior to treatment. For p53-Venus cell lines containing p53RE-mCherry-TP53i3 3′-UTR, p53RE-mCherry-TP53i3 3′-UTR::1ARE or p53RE-mCherry-TP53i3 3′-UTR::2ARE, 4 × 10 5 cells were plated on a 6-cm dish for each (2019) 9:5976 | https://doi.org/10.1038/s41598-019-42509-y www.nature.com/scientificreports www.nature.com/scientificreports/ cell line and each time point of the experiment, two days prior to treatment. On the day of the treatment, cells were treated with 0.2 mM 5-ethynyl Uridine (EU) from the Click-iT Nascent RNA Capture Kit (ThermoFisher, C10365) and 400 ng/mL NCS (Sigma, N9162) for 3 hours, those of which were used for estimating newly synthesized RNA fraction and inducing double-strand breaks (DSB), respectively. Following 3 hours of treatment, cells were harvested by scraping at time points of 0, 1, 2 and 4 hours and frozen in a dry ice-ethanol bath. RNA of the samples was extracted using QIAshredder and RNeasy Kits (Qiagen), and RNA concentrations were measured by a UV spectrophometer to ensure that equal amount of 2.5 μg RNA was used to perform the Click-iT reaction of each sample, following the manufacturer's protocol. RNA concentrations were measured again, so that 150 ng of each biotinylated RNA sample was used for the RNA binding reaction using the same kit and for the subsequent 20 μl cDNA synthesis reactions using an Invitrogen High Capacity cDNA Reverse Transcription Kit (ThermoFisher, 4374966). Samples and assays were put through thermal mixing (70 °C for 40 min, 60 °C for 30 s), hot start (95 °C for 1 min) and 35 cycles of PCR (96 °C for 5 s, 64 °C for 20 s), then melt curve acquisition at 64-95 °C with 0.5 °C resolution. The mRNA expression at the indicated time points for all samples was quantified in triplicate via qPCR with a CFX96 Real-Time PCR machine (Bio-Rad). Each qPCR reaction mixture contained 1ul of a sample, Maxima SYBR Green/Fluorescein Master Mix (Thermo Scientific, K0241), forward and reverse primers specific for the genes expressed, and water up to a 10 ul total reaction volume. For the constructs expressing mCherry, 5′-CACGAGTTCGAGATCGAGGG-3′ forward and 5′-CCCTTGGTCACCTTCAGCTT-3′ reverse primers were used; for those expressing CASP3, 5′-AAATACCAGTGGAGGCCGAC-3′ forward and 5′-ATGGCACAAAGCGACTGGAT-3′ reverse primers were used; and for GAPDH measurements, 5′-ACATCGCTCAGACACCATG-3′ forward and 5′-TGTAGTTGAGGTCAATGAAGGG-3′ reverse primers were used.
Flow cytometry. Two days prior to treatment, 2 × 10 5 MCF7 cells containing the constructs of p53RE-CASP3-DDB2 3′-UTR, p53RE-CASP3-TRIAP1 3′-UTR or p53RE-CASP3-GADD45A 3′-UTR were plated on 2 6-cm dishes each, and one set of plates was incubated with 400 ng/mL NCS 24 hours before the experiment. Cells were harvested by trypsinization and stained with 5 μL of the FITC annexin V dye and 10 μL of the 100 μg/mL PI dye following the manufacturer's protocol (APC Annexin V Apoptosis Detection Kit with PI; BioLegend). Stained cells were analyzed by flow cytometry on a FacsCanto (BD Biosciences).
Fluorescence expression measurements at single cell level. Time-lapse microscopy was used for the expression measurements of mCherry-tagged MCF7 p53-Venus cell lines. Two days prior to the microscopy experiment, cells were grown on poly-D-lysine-coated glass-bottom plates (MatTek Corporation) in selection medium. Approximately 45 mins before the experiment, medium of all samples were changed to transparent RPMI medium supplemented with 2% fetal bovine serum (FBS), 100 U/mL penicillin G, 100 mg/mL streptomycin sulfate, and 250 ng/mL amphotericin B. Cells were imaged on a Nikon Eclipse TiE inverted fluorescence microscope with a 20x plan apo objective (NA 0.75) using an iXon Ultra-888 camera (Andor) with constant temperature, CO 2 concentration, and humidity maintained. Images were acquired every 20 min over a 24-hour period. The mCherry filter set contained filters of 540-580 nm for the excitation light, 585 nm for the dichroic beam splitter, and 593-668 nm for the emission light (Chroma). We analyzed images using NIS-Elements software (Nikon) and custom written ImageJ (NIH) and MATLAB software (Mathworks), which is available upon request.