Chromosome and plasmid-borne P(LacO3O1) promoters differ in sensitivity to critically low temperatures

1 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports Chromosome and plasmid-borne pLacO3O1 promoters differ in sensitivity to critically low temperatures Samuel M. D. Oliveira1, Nadia S. M. Goncalves1, Vinodh K. Kandavalli1, Leonardo Martins 1,2, Ramakanth Neeli-Venkata1, Jan Reyelt3, Jose M. Fonseca2, Jason Lloyd-Price4,5, Harald Kranz3 & Andre S. Ribeiro1,2 Temperature shifts trigger genome-wide changes in Escherichia coli’s gene expression. We studied if chromosome integration impacts on a gene’s sensitivity to these shifts, by comparing the singleRNA production kinetics of a PLacO3O1 promoter, when chromosomally-integrated and when singlecopy plasmid-borne. At suboptimal temperatures their induction range, fold change, and response to decreasing temperatures are similar. At critically low temperatures, the chromosome-integrated promoter becomes weaker and noisier. Dissection of its initiation kinetics reveals longer lasting states preceding open complex formation, suggesting enhanced supercoiling buildup. Measurements with Gyrase and Topoisomerase I inhibitors suggest hindrance to escape supercoiling buildup at low temperatures. Consistently, similar phenomena occur in energy-depleted cells by DNP at 30 °C. Transient, critically-low temperatures have no long-term consequences, as raising temperature quickly restores transcription rates. We conclude that the chromosomally-integrated PLacO3O1 has higher sensitivity to low temperatures, due to longer-lasting super-coiled states. A lesser active, chromosomeintegrated native lac is shown to be insensitive to Gyrase overexpression, even at critically low temperatures, indicating that the rate of escaping positive supercoiling buildup is temperature and transcription rate dependent. A genome-wide analysis supports this, since cold-shock genes exhibit atypical supercoiling-sensitivities. This phenomenon might partially explain the temperature-sensitivity of some transcriptional programs of E. coli. Escherichia coli has evolved sophisticated regulatory programs to adapt to fluctuating environments that allow tuning gene expression so as to trigger appropriate responses1,2. In general, gene expression regulation occurs during transcription initiation3 and it can be performed, e.g., by transcription factors4,5, which act locally, affecting specific genes, and by σ factors6–9, which have more genome-wide effects. Similarly, environmental changes can affect chromosomal DNA compaction, which is associated to supercoiling10,11 and is regulated by nucleoid associated proteins (NAPs)12,13. Interestingly, changes in DNA compaction has genome-wide effects13–15, causing the expression of some genes to increase while in others it decreases4,16–18. DNA compaction and supercoiling have distinct effects on plasmid-borne and chromosome integrated genes (see e.g.19). One reason for this is that the chromosome has topologically constrained segments that allow supercoiling buildup12,20–22, as transcription occurs, since this process generates positive supercoiling ahead of the RNA polymerase (RNAP) and negative supercoiling behind it23,24. Meanwhile, plasmids lack discrete constraints. Thus, when positive and negative supercoiling emerge, they freely diffuse in opposite directions and annihilate each other19. Thus, in general, the transcriptional activity in plasmids is only affected by transient constraints due to, e.g., transient protein binding19,25. Exceptions are, e.g., plasmids encoding membrane-associated proteins that, by anchoring to the membrane26–29, can form longer lasting constraints. Other exceptions are plasmids 1Laboratory of Biosystem Dynamics and Multi-Scaled Biodata Analysis and Modelling Research Community, Faculty of Medicine and Health Technology, Tampere University, Korkeakoulunkatu 7, 33720, Tampere, Finland. 2CA3 CTS/ UNINOVA. Faculdade de Ciências e Tecnologia, Universidade Nova de Lisboa, Quinta da Torre, 2829-516, Caparica, Portugal. 3Gene Bridges, Im Neuenheimer Feld 584, 69120, Heidelberg, Germany. 4Biostatistics Department, Harvard T.H. Chan School of Public Health, Boston, MA, 02115, USA. 5Infectious Disease and Microbiome Program, Broad Institute, Cambridge, MA, 02142, USA. Correspondence and requests for materials should be addressed to A.S.R. (email: [email protected]) Received: 2 October 2018 Accepted: 28 January 2019 Published: xx xx xxxx opeN 2 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ carrying tandem copies of one or two DNA-binding sites25,30 and plasmids carrying the T7 promoter, when expressed in topA mutant strains31. Nevertheless, it is worth noting that in vivo measurements suggest that, prior to annihilation, transient supercoiling changes can influence transcription rates of both plasmid-borne and chromosomally-integrated promoters31–33. Temperature shifts affect DNA supercoiling directly34,35 as well as indirectly, e.g., by affecting the interactivity between NAPs and chromosomal DNA36. This may explain why temperature down-shifts affect the activity of most chromosomal genes in E. coli37. Another temperature-dependent event in transcription is promoter escape38, the stage at which the RNAP is freed from the promoter and moves downstream towards the elongation region of the DNA template39. The stronger the binding between the RNAP and the promoter, the longer it usually takes for the RNAP to escape the promoter and begin elongation39. One reason for this is that, for escape to succeed, the RNAP needs to pull a sufficient amount of downstream DNA into itself (so as to reach its active center), which involves breaking interactions between the RNAP and the promoter, and between the RNAP and initiation factors38, which are energy dependent processes. Given the above, we hypothesized that plasmid-borne and chromosomally-integrated genes can differ in sensitivity to temperature shifts and that these differences may be promoter strength-dependent. To test this, we compared quantitatively the effects of temperature shifts on the in vivo kinetics of transcription of the PLacO3O1 promoter, when on a plasmid and when chromosomally-integrated (Materials and Methods). Further, we assessed the effects on the native lac promoter, whose transcription rate is weaker than PLacO3O1, although we located it in the same position in the chromosome. For this, we used the MS2-GFP RNA tagging technique in E. coli, along with a recently proposed methodology to resolve the rate-limiting steps governing the in vivo dynamics of initiation of prokaryotic promoters (similar to established steady-state assays to resolve the in vitro dynamics)40. Further, we studied this process at critically low temperatures (below 23 °C), a regime in which most cellular processes exhibit significant differences due to, e.g., globally-altered transcription rates37 and increased cytoplasmic viscosity41. Using these techniques, we characterized, with single-RNA sensitivity, the RNA production dynamics of these constructs at various temperatures, as well in the presence of Gyrase and Topoisomerase I inhibitors and of DNP-based energy depletion. Also, we made use of stochastic modelling to show that the observed differences in transcription kinetics between chromosome and plasmid integrated promoters at low temperatures are consistent with current stochastic models of transcription initiation that account for supercoiling buildup, provided that such low temperatures result in the hindrance to escape from DNA super-coiling. Finally, we made use of information of what genes in E. coli have their activity induced following cold-shocks42 and of what genes are supercoiling sensitive14, to assess if these two features are strongly correlated, as our results would suggest. Results We studied at the single-RNA level if the kinetics of RNA production under the control of PLacO3O1 differs in response to temperature changes when the gene is single-copy F-plasmid-borne and when it is chromosome-integrated. For this, we made use of two identical constructs under the control of the PLacO3O1 promoter coding for multiple bindings sites for MS2-GFP followed by the coding region of mCherry. Both constructs, shown in FigsS1 and S2, are functional and responsive to the inducer (Fig.1). Also, control tests were performed to verify that spots detected in microscopy images correspond to MS2-GFP tagged RNA molecules (Fig.S3) and that, once appearing, their intensity does not change significantly during the measurement time (Supplementary section “Control tests of the RNA counting method”), as this would affect the counting of MS2-GFP tagged RNA molecules in each cell. For each strain (carrying the target gene in the plasmid or in the chromosome) and each temperature condition, we performed 3 or more biological repeats and counted RNA-spots (MS2-GFP tagged RNAs) in each cell from the microscopy images (Supplementary section “Image Analysis”). As we did not find statistically significant differences between repeats, the results shown here are from cells from all replicates. Since the two strains are not subject to the same antibiotics (only cells carrying the target gene in the single-copy plasmid are subject to chloramphenicol, see Methods), we tested whether their growth curves differ. Results in Fig.S5 show that these curves are not distinguishable. We also tested whether the two strains produce similar levels of MS2-GFP reporter proteins (as differences could result in different ability to count target RNAs). For this, inducing only the reporter gene, we measured the background fluorescence intensity of cells of both strains (almost exclusively due to MS2-GFP reporters). We then compared the two distributions of single-cell background fluorescence intensity by a Kolmogorov-Smirnov (KS) test of statistical significance and found no significant difference (p-value of 0.46). Induction of gene expression is similar, but not identical, in the chromosome and plasmid-integrated constructs. For each construct (chromosome integrated and single-copy plasmid borne) and IPTG concentration, we quantified the integer-valued RNA numbers in live cells at 30 °C by microscopy imaging, 1 hour after induction of the target promoter by IPTG (Materials and Methods). From Fig.1, we find that the mean integer-valued RNA numbers per cell, relative to maximum induction (1 mM IPTG) exhibits a similar fold change in both constructs (~4 for the plasmid and ~5 for the chromosome construct), as in previous studies43. Also, for both constructs, the transcription rate does not increase beyond 500 µM IPTG, i.e., 1 mM IPTG suffices for full induction, in agreement with previous studies40,43. However, the RNA production kinetics of the two constructs differs in some aspects. First, from Table1 and Fig.1, mean integer-valued RNA numbers per cell are higher in the plasmid construct for all induction conditions. Also, the increase in RNA numbers with IPTG concentration differs. Namely, from TableS3, in the plasmid 3 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ construct the RNA numbers increase gradually as IPTG is increased, allowing most conditions to differ significantly in a statistical sense, while in the chromosome construct the RNA numbers per cell only differ significantly between 0 µM IPTG and the other conditions (for 50 µM IPTG or higher, there is little increase in RNA numbers with additional increases in IPTG concentrations). Nevertheless, the same model of transcription (Supplementary Information, section “Model of transcription kinetics”, reactions 1–3) fits both constructs, as tuning kcc and/or kunlock suffices to account for differences between them in RNA numbers at 30 °C. Figure 1. Induction curves, measured by microscopy imaging and single RNA tagging by MS2-GFP, of the target promoter PLacO3O1 when integrated into the chromosome (light grey) and into a single-copy F-plasmid (dark gray) (E. coli strain BW25993). Shown are the mean integer-valued RNA numbers (relative to the reference case, 1 mM IPTG) in individual cells of the two constructs, 1 hour after induction at 30 °C. Data presented as relative mean to the reference case with 90% confidence intervals obtained from a two-tailed Student’s t-test. Sample size per condition, as IPTG is increased, is (chromosome) 665, 655, 675, 670, 660 and 645 cells, and (plasmid) 670, 670, 665, 655, 655, 675 cells. Also shown is the ratio between the mean integervalued RNA numbers per cell between cells with the target gene chromosome-integrated and on a single-copy plasmid. Absolute integer-valued RNA numbers per cell in each condition can be obtained from the absolute integer-valued RNA numbers per cell for 1 mM IPTG shown in Table1 along with the relative values shown here. Results are obtained from 3 biological repeats. Since these exhibited no statistically significant differences, the results presented here are composed of the data from the 3 biological replicates. Condition No. cells Mean integer-valued RNA no. per cell CV2 Chromosome construct 30 °C 645 2.08 2.60 27 °C 632 2.00 1.93 23 °C 668 1.74 2.30 20 °C 646 0.59 7.17 16 °C 668 0.22 15.81 10 °C 648 0.25 16.12 Plasmid construct 30 °C 675 2.86 1.06 27 °C 654 2.46 1.42 23 °C 665 1.63 2.73 20 °C 660 1.61 3.08 16 °C 663 1.50 2.99 10 °C 676 1.35 3.46 Table 1. Number of cells observed, mean, and squared coefficient of variation (CV2) of the absolute integervalued RNA numbers per cell for the chromosome-integrated and the plasmid-integrated constructs, when induced by 1 mM IPTG. Cells are induced and kept at 30 °C, 27 °C, 23 °C, 20 °C, 16 °C and 10 °C for 60 minutes prior to the acquisition of the results. Results are obtained from 3 biological repeats. Since these exhibited no statistically significant differences, the results presented here are composed of the data from the 3 biological replicates. 4 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ Transcription by the chromosome-integrated construct is noisier at lower temperatures. We next studied if temperature changes affect differently the chromosome and plasmid constructs. We measured integer-valued RNA numbers in cells under full induction (1 mM IPTG) by microscopy at various temperatures (30, 27, 23, 20, 16 and 10 °C). For each condition, from the absolute integer-valued RNA numbers in each cell, we calculated the mean and squared coefficient of variation (CV2) of the RNA numbers in single cells (Table1). To assess if the RNA production kinetics differs with temperature and between the two constructs, we performed KS tests. The P values comparing the single-cell distributions of RNA numbers between conditions for the chromosome and plasmid constructs and between the constructs at each temperature are shown in TablesS4 and S5, respectively. From Tables1 and S4, we find that PLacO3O1, when in the single-copy plasmid, is highly responsive to temperature decreases until 23 °C. Below this temperatures, changes in RNA numbers are only significant for temperature shifts wider than those considered in TableS4 (e.g. p < 0.01 for 23 °C and 10 °C, not shown in TableS4). This behavior is in line with previous reports for the PTetA and for the PLac-Ara-1 promoters, also on single-copy plasmids44. Meanwhile, when chromosome-integrated, PLacO3O1 activity decreases significantly for a wider range of temperatures. Namely, differences are detectable between all pairs of neighboring conditions, except between 16 °C and 10 °C. These results are supported by those in TableS5. The P values of the KS tests indicate that, below 23 °C, the plasmid and chromosome constructs differ from one another in all temperatures. For 23 °C and above, they only differ at 30 °C. From this and Table1, we conclude that the activity of the chromosome-integrated promoter is more heavily reduced as temperature is lowered, and that it remains sensitive to a wider range of temperature shifts. To validate these results, we used RT-qPCR (Supplementary Information) to obtain the mean RNA numbers relative to the control (30 °C) in cells under full induction (1 mM of IPTG) at 23, 16 and 10 °C (Fig.S6). From Table1, we calculated the same quantities from the microscopy measurements. Overall, both the chromosome and plasmid constructs exhibit the same qualitative behavior as temperature decreases when measured by microscopy and RT-qPCR. We next assessed if the weaker transcriptional activity of the chromosome-integrated promoter at the lowest temperatures could be explained by changes in the spatial distribution of RNAPs45 or of the nucleoids (Supplementary Information, section “Nucleoid staining with DAPI”). Measurements at 10 °C and 30 °C (Fig.S7) show no significant differences in these two features, allowing rejecting these hypotheses. We also performed two additional tests for cells with the chromosome-integrated PLacO3O1. First, as the mean integer-valued RNA numbers per cell in induced cells at 10 °C (Table1) appears to be smaller than in non-induced cells at 30 °C (TableS1), we tested if this difference is statistically significant by performing a KS test between the distributions of single-cell RNA numbers in the two conditions. We obtained a p-value of 0.99 and, thus, we conclude that the RNA numbers in the two conditions do not differ, in a statistical sense (p-value larger than 0.01), implying that induced cells at 10 °C produce at least as much RNAs as non-induced cells at 30 °C. Second, we tested whether, at 10 °C, RNA numbers differ between induced and non-induced cells. A KS test between the distributions of RNA numbers in individual cells in the two conditions (Fig.2 and TableS1) shows that they can be distinguished in a statistical sense. Thus, we concluded that induction at 10 °C tangibly increases the RNA production rate of the chromosome-integrated PLacO3O1. Mean relative time prior to commitment to transcription increases in the chromosome-integrated construct at low temperature. To investigate why the two constructs responded differently to lowering temperatures, we assessed whether the changes in the kinetics of transcription with decreasing temperature occur prior to or following the commitment to open complex formation (Supplementary Information, section “Model of transcription kinetics”), by making use of Lineweaver–Burk plots46 of the inverse of the RNA production rate against the inverse of the RNAP concentration (Supplementary sections “Lineweaver-Burk Plots” and “Tuning intracellular RNAP concentrations”). For this, we measured RNAP levels in individual cells in each temperature condition, and verified that they differ statistically between conditions (KS-tests in TableS8). From these, we obtained the inverse of the RNAP concentrations relative to the 1X control condition (TableS7). Next, for the same condition, using the region of the target gene coding for mCherry (Fig.S2), we measured the RNA production rates of the two constructs at 10 and 30 °C by RT-qPCR, and obtained the inverse of these values (TableS9). Combining both measurements, we obtained Lineweaver–Burk plots for each construct and the two temperature conditions (example Fig.S10). Next, from these plots, using the same methodology as in40,43,47, we estimated the mean fraction of time between consecutive transcription events taken by the steps preceding Δ () t t prior and following Δ () t t after the commitment to open complex formation40,43 at the highest and lowest temperature, with Δt being the mean time length between consecutive transcription events in individual cells (Supplementary Information, section “Relative mean duration prior to and following commitment to transcription”). Results in Table2 show that, in all 4 conditions, the most rate-limiting events occur after commitment to open complex formation. However, Table2 also informs that the lowering temperatures do not cause the same effect in the two constructs. In particular, in the plasmid construct, in agreement with previous in vitro measurements for the synthetic PLac-UV5 promoter4, the reduction in RNA production rate with lowering temperature is mostly due to a reduction in the rate of the events after commitment to open complex formation (with tafter increasing from being 92% to 98% of the Δt as temperature is lowered). Meanwhile, in the chromosome construct, the opposite occurs (with tafter decreasing from 91% to 73% of the Δt as temperature is lowered) suggesting that, in this construct, the events 5 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ whose rates were most reduced occur prior to commitment to open complex formation, provided that the RNA production rate decreases with lowering temperature (as is the case, see Table1). Local DNA supercoiling in the chromosomally-integrated gene drives the differences between constructs. To explain the increased time-length of the events preceding the open complex formation in the chromosome-integrated construct at lower temperatures, we considered the model of transcription initiation (Supplementary Information, reactions 1–3). This model allows for this, provided that decreasing temperatures decrease the rate of unlocking (kunlock) from locked promoter states (reaction 3, Supplementary Information), or decrease the rate of unbinding of a repressor from a promoter (kON, reaction 2 in Supplementary Information), or both. Either of these possibilities is physically possible since lowering temperatures could affect the efficiency of repressors (see e.g.48), DNA packaging (known to differ between plasmid and chromosomes49, or DNA super-coiling35 (known to affect both packaging10 and transcription19,30,50,51). A third possibility would be that decreasing temperature modified the kinetics of closed complex formation, causing increased relative duration of this event, e.g. due to reduced k1 or k2, or instead increased k−1. However, this would result in reduced noise in RNA production40,52 and thus reduced CV2 in RNA numbers in individual Figure 2. Mean integer-valued RNA numbers in individual cells, relative to the last time moment, as a function of temperature, measured by microscopy with single RNA tagging by MS2-GFP, when PLacO3O1 is integrated into the chromosome (light grey) and in a single-copy F-plasmid (dark gray). (A) Cells are at 10 °C. (B) Cells are at 30 °C. Data presented as relative mean to the reference case with 90% confidence intervals obtained from a twotailed Student’s t-test. Sample size per condition, as time progresses is: (A) Chromosome at 10 °C (610, 611, 615, 610, 609, 602 and 605 cells), and Plasmid at 10 °C (615, 610, 610, 612, 608, 606 and 606 cells). (B) Chromosome at 30 °C (604, 615, 610, 610, 605, 608 and 605 cells), and Plasmid at 30 °C (610, 615, 606, 615, 610, 613 and 609 cells). For each time point, new cells were taken from the original culture. Results are obtained from 3 biological repeats. Since these exhibited no statistically significant differences, the results presented here are composed of the data from the 3 biological replicates. Finally, at t = 0 min, the mean absolute number of RNA molecules per cell was (A) 0.1 for chromosome and 0.9 for plasmid, and (B) 0.2 for the chromosome and 0.9 for the plasmid. Condition ∆ tprior t ∆ tafter t 30 °C Chromosome construct 0.09 0.91 Plasmid construct 0.08 0.92 10 °C Chromosome construct 0.27 0.73 Plasmid construct 0.02 0.98 Table 2. Relative mean duration of the rate limiting steps in transcription initiation at 30 °C and 10 °C. Shown are the mean durations, relative to the mean time-length of the intervals between transcription events (∆t), of the rate-limiting steps prior Δ () t t prior and after Δ () t t after commitment to open complex formation for the chromosome-integrated and the plasmid-integrated constructs. 6 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ cells (since, at 30 °C, most time between transcription events is spent in open complex formation, Table2). The data on CV2 in RNA numbers in Table1 disproves this possibility. Meanwhile, in the first possibility, where kunlock or kON are decreased with decreasing temperature, this would result in increased noise in RNA production53 and, thus, increased CV2 in RNA numbers in individual cells, which was observed (Table1). To determine whether it is kunlock or kON that is decreased, consider that a change in repressors’ efficiency with temperature (i.e. a change in kON) should affect both the chromosome and plasmid constructs similarly since both constructs are affected by this mechanism. However, we observed divergent responses between these two constructs to the lowest temperatures, with the plasmid-borne construct being unable to turn off its RNA production as efficiently as the chromosome-integrated construct (Table1). Thus, we conclude that the stronger decrease in the chromosome construct in RNA production rate with lowering temperature (at the lowest temperature conditions tested) is likely due to an increased amount of time required to remove the promoter from the locked state, which does not occur in the plasmid construct (i.e. changes in kunlock with lowering temperature are the most likely explanation for the observed behaviors). It is further possible to assess if the changes in kunlock, causing different behaviors of the two constructs in response to lowering temperatures, are associated to DNA packaging and/or super-coiling. For that, we measured the nucleoid size in cells with one nucleoid (Supplementary Information) in the various temperature conditions. If decreasing temperature (in the ranges shown in Table1) affects DNA packaging significantly, we expect differences in the mean and/or variability of the nucleoid size. However, we found no significant differences between 10 °C and 30 °C (TableS10). Similar results were reported in41. Thus, we discard DNA packaging as the main cause for the differences between chromosome and plasmid response to temperature shifts. Given all of the above, we hypothesize that the difference in response of the chromosome and plasmid-integrated genes with lowering temperature is due to an increased rate of accumulation of local DNA supercoiling in the chromosome-integrated gene, which increases the escape times from locked states (reactions 1 and 3, Supplementary Information). To validate this hypothesis, we performed several experiments. First, we compared the numbers of RNAs produced over time by cells of each strain. We expect this number to increase near-constantly in the cells carrying the plasmid-borne gene, but not in the cells of the other strain. For this, from the moment of activation of the target gene (t = 0 minutes), we measured integer-valued RNA numbers in individual cells at 10 °C every 15 minutes for 90 minutes (for each time point, new cells were taken from the original culture). If the weaker activity of the chromosome-integrated promoter is due to increased propensity to be in the locked state due to the accumulation of DNA super-coiling, we expect its transcription activity to be blocked after a few events. At a population level, this would result in a sharp decrease in the rate of increase of RNA numbers in the cells, some time after the start of the measurements. Meanwhile, in the plasmid construct, we expect a constant RNA production rate over time, due to the lack of accumulation of local DNA super-coiling19. Results in Fig.2A confirm these predictions. Cells at 10 °C with the chromosome construct only exhibit production in the first 30 minutes, while the plasmid construct shows approximately constant RNA production rate throughout the measurement. We also performed measurements at 30 °C. Given the similar dynamics of transcription of the two constructs in this condition (Table2), we expect the RNA production rate to be constant in time in both constructs. Results in Fig.2B confirm this. To further test the hypothesis, we next compared the activity of the two constructs at 30 °C when subjecting cells to Novobiocin, an inhibitor of Gyrase activity (Methods)19,54. Gyrase releases positive supercoiling55 but not negative supercoiling56. According to the twin-supercoiled-domain model24, which predicts that negative/positive supercoils should accumulate in the absence of supercoil-relaxing enzymes, we expect cells with the chromosome construct to exhibit a similar behavior as when at 10 °C. Meanwhile, cells with the plasmid construct should again exhibit a constant rate of transcription over time19. Figure3A confirms these predictions. In this regard, in both strains, the gene acrA is present, and thus, Novobiocin is not expected to affect cell division rates57. To test this, we measured cell growth rates by OD600 for varying Novobiocin concentrations (0, 50, 75, 100 and 150 ng/μl). We found the growth rates to not differ significantly between conditions (data not shown). These results also show that 100 ng/μl Novobiocin concentration suffices to affect (but not halt) the transcription rate of the chromosome-integrated gene (compare the results for this construct in Figs2B and 3A, at 30 °C). Subsequently, we subject cells with the chromosome construct to Novobiocin when at 10 °C. Results in Fig.3B, when compared to Figs2A and 3A, show that transcription in cells carrying the chromosome is more strongly blocked when combining Novobiocin and low temperatures. I.e. while at 10 °C alone and subject to Novobiocin alone, RNA numbers increase by a factor of 4 (from 0.25 to 1) in a period of 90 min., when subjecting cells to both 10 °C and Novobiocin, the RNA numbers increase only by a factor of 2 (from 0.5 to 1) in the same period of time. Meanwhile, in cells with the plasmid construct, we observe the same RNA production as in Fig.2A, meaning that, in these cells, Novobiocin has no effect at either temperature. Given this, we suggest that the transcription activity of the chromosome-integrated promoter at 10 °C is hampered by an increased difficulty in unblocking the DNA from supercoiled states that is not due to a loss of functionality of Gyrases. Then, we observed cells with the chromosome integrated gene at 30 °C, when subject to Topotecan, an inhibitor of Topoisomerase I activity58,59 (Methods). Topoisomerase I releases negative, but not positive supercoiling60. We expect these cells to exhibit a similar behavior as when at 10 °C, which Fig.3C confirms. Also, we observed the same cells subject to Topotecan when at 10 °C. From Fig.3D, transcription is again blocked more strongly than when at 10 °C but not subject to Topotecan and when at 30 °C subject to Topotecan. Namely, while in the latter two conditions RNA numbers increased by a factor of 4 (from 0.25 to 1) in 90 min., when subjecting cells to both 10 °C and Topotecan the RNA numbers increase only by a factor of 2 (from 0.5 to 1) in the same period of time. These results suggest that the activity of the chromosome-integrated gene at 10 °C is hampered by an increased difficulty in unblocking the DNA from supercoiled states, rather than due to a loss of functionality 7 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ of Topoisomerases I (or Gyrases). TableS15, with the results of the KS tests between the distributions of RNA numbers in individual cells at 10 °C and 30 °C, when subject to Novobiocin or Topotecan, confirm that the distributions differ with temperature, in a statistical sense. Finally, in comparison, subjecting cells with the single-copy F-plasmid to Topotecan causes, qualitatively, the same behavior as adding Novobiocin (at 30 °C and 10 °C) (Fig.3A–D). Promoter escape from supercoiling buildup is similarly hampered if cellular energy is depleted. If the escape from DNA supercoiling buildup in the chromosomally-integrated construct at low temperatures is due to energy deficiency at low temperatures (the energy required for the necessary endothermic reactions to occur should be higher in such conditions), it should be possible to mimic the phenomena by, instead of lowering temperature, depleting cells of energy via DNP treatment45 (Methods). In particular, we expect cells subject to this treatment to, even at 30 °C, be less able to maintain the chromosome integrated promoter active over time when compared to the control, similar to when at 10 °C. To test this, we subjected cells to DNP for 90 minutes (at 30 °C) prior to imaging (Methods). As expected, we observed similar RNA production dynamics (Fig.4), as in untreated cells at 10 °C with a chromosome integrated PLacO3O1 (Fig.2A). I.e., beyond 30 minutes, there is little to no transcription, suggesting that, in this condition, the activity is also being hampered by increased difficulty in escaping from supercoiled states. Low temperatures have no long-term consequences on transcription blocking by DNA supercoiling. We performed an additional test to support the hypothesis that the escape from DNA supercoiling buildup in the chromosome construct at low temperatures is due to energy deficiency. Namely, we hypothesized that changing temperature to near-optimal conditions (e.g. 30 °C) should restore the cells’ ability to relax DNA supercoiling (as at higher temperature this process is expected to require less energy). To test this, we subjected cells with a Figure 3. Mean integer-valued RNA numbers in individual cells, relative to the last time moment, as a function of temperature and gyrases and topoisomerases I inhibitors, measured by microscopy with single RNA tagging by MS2-GFP, when PLacO3O1 is integrated into the chromosome (light grey) and in a single-copy F-plasmid (dark gray). (A) Cells at 30 °C and subject to Novobiocin. (B) Cells at 10 °C and subject to Novobiocin. (C) Cells at 30 °C and subject to Topotecan. (D) Cells at 10 °C and subject to Topotecan. Data presented as relative mean to the reference case with 90% confidence intervals obtained from a two-tailed Student’s t-test. Sample size per condition, as time progresses is: (A) Chromosome, 30 °C, Novobiocin (608, 606, 605, 610, 613, 610 and 615 cells) and Plasmid, 30 °C, Novobiocin (615, 605, 613, 610, 610, 601 and 602 cells); (B) Chromosome, 10 °C, Novobiocin (615, 610, 615, 625, 605, 620 and 620), and Plasmid, 10 °C, Novobiocin (615, 620, 615, 610, 620, 620 and 615 cells); (C) Chromosome, 30 °C, Topotecan (615, 610, 612, 610, 605, 603 and 610 cells) and Plasmid, 30 °C, Topotecan (662, 623, 626, 606, 643, 659 and 647 cells) and, finally, (D) Chromosome, 10 °C, Topotecan (620, 610, 620, 610, 610, 610 and 615 cells) and Plasmid, 10 °C, Topotecan (679, 629, 649, 645, 642, 601 and 632 cells). For each time point, new cells were taken from the original culture. Results are obtained from 3 biological repeats. Since these exhibited no statistically significant differences, the results presented here are composed of the data from the 3 biological replicates. In all cases, Novobiocin or Topotecan was added to the culture at the same time as the inducer of the target gene, IPTG. Finally, at t = 0 min, the mean absolute number of RNA molecules per cell was (A) 0.3 for chromosome and 0.9 for plasmid, (B) 0.1 for chromosome and 0.9 for plasmid, (C) 0.1 for chromosome and 0.9 for plasmid, and (D) 0.1 for chromosome and 0.9 for plasmid. 8 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ chromosome integrated PLacO3O1 to two temperature shifts, first from high (30 °C) to low (10 °C) and then from low (10 °C) to high (30 °C), and measured the mean integer-valued RNA numbers in the cells over time. Results in Fig.5 show that both temperature shifts caused smooth transitions in the RNA production rates that are consistent with changes in the kinetics of locking/unlocking of promoters from positive supercoiling buildup. In detail, cells at 30 °C have constant RNA production, as shown previously. Once temperature is shifted to 10 °C, after 15–30 minutes, little to no RNA production is observed (as in Fig.2A). More importantly, once high temperatures are restored (to 30 °C), RNA production is quickly restored to nearly the original rate. The fast transition between behaviors and the ability to quickly restore the original dynamics reinforce the conclusion that the activity of the chromosome integrated promoter at 10 °C is hampered by an increased difficulty in unblocking the promoter from supercoiled states (due to an increase in the energy required). Note that in this particular experiment, following the shift from 30 °C to 10 °C, it does not follow a transient of ~15–30 minutes of reduced transcription activity that is visible in Figs2, 3 and 4. This is because, in this case, when the shift occurs, the cells already contain sufficient IPTG to achieve full transcription rates, while in the previous experiments the inducer was added immediately before the microscopy measurements began, and thus, a transient time to reach quasi-equilibrium RNA production rates is expected, due to the non-negligible time that cells need to intake inducers from the media38, particularly at low temperatures. In the case of IPTG, previous measurements suggest that this transient is ~15–30 minutes long61, in agreement with the results in Figs2A, 3A–D and 4. Stochastic modelling also suggests increased long-lasting super-coiled states at critically low temperatures to be the cause for enhanced sensitivity to shifts to critically low temperatures. We tested whether the increase in the expected time for promoters to escape from a supercoiling state across the cell population is, in accordance with current stochastic models of transcription in E. coli40,43 a plausible explanation for the change with decreasing temperature in the average RNA numbers over time in cells with the chromosome integrated promoter (Fig.2). For this, we use the stochastic model of transcription initiation (Supplementary Information, reactions 1–3), derived from multiple studies, including genome-wide studies of variability in transcript counts62,63 and studies of the transcription dynamics of individual genes40,43. All parameter values (TableS11) are from single-cell, single-RNA empirical data on the activity of lac derivative promoters19,40. Mean RNAP numbers are set to correspond to the RNAp concentration reported in40. Finally, from the results above, we assume that the increase in Δ () t t prior as temperature decreases (Table2) is mostly due to a decrease in kunlock. Thus, the remaining rate constants are, for simplicity, unchanged. For each value of kunlock tested, we performed 500 independent simulations, each 75 minutes long. Data was collected every 15 minutes, as in the experiments (Fig.2). The values of kunlock were selected as follows: the highest value, corresponding to high temperatures (30 °C), is reported in19. This value was then gradually lowered until the mean number of RNAs per cell at the end of the measurement period was similar to that observed in cells at 10 °C. We assessed if the model was able to reproduce the observed RNA numbers over time at both high and low temperatures, and if there is a gradual behavioral change between these extreme conditions. For this, the initial Figure 4. Mean integer-valued RNA numbers in individual cells at 30 °C subject to DNP treatment relative to the last time moment, measured by microscopy with single RNA tagging by MS2-GFP, when PLacO3O1 is integrated into the chromosome (light grey). Data presented as relative mean to the reference case with 90% confidence intervals obtained from a two-tailed Student’s t-test. Sample size per condition, as time progresses is 601, 610, 601, 605, 610, 605 and 608 cells. For each time point, new cells were taken from the original culture. Results are obtained from 3 biological repeats. Since these exhibited no statistically significant differences, the results presented here are composed of the data from the 3 biological replicates. Finally, at t = 0 min, the mean absolute number of RNA molecules per cell was 0.2. 9 Scientific RepoRts | (2019) 9:4486 | https://doi.org/10.1038/s41598-019-39618-z www.nature.com/scientificreports www.nature.com/scientificreports/ numbers of all molecular species were set to zero, with the exception of PON (set to 1, corresponding to one active promoter per cell), RNAp (as noted above), and RNA. Initial RNA numbers were drawn randomly from a Poisson distribution (0.7 RNA/cell) to match the here observed outcome of spurious RNA production events. We observed also (empirically, Supplementary TableS1) that this number did not differ with temperature, as expected, since, prior to moment 0, cells were at the same temperature (30 °C) in both measurements. In Fig.6, we compared the results of the model with those in Fig.2A (10 °C) and Fig.2B (30 °C) for the chromosome-integrated promoter. For simplicity, as noted, we ignored the first time moment of the empirical data (0 minutes following induction) since, at this stage, the cells did not yet have fully active transcription61. This removed the need to model the intake process for the inducers61. Results in Fig.6 support the earlier conclusions. The accuracy with which the model reproduces the measurements suggests that the difference in mean RNA production rates between cells with the chromosome-integrated promoter at critically low (10 °C) and at high (30 °C) temperatures can be explained by a reduced ability to release chromosome-integrated promoters from the effects of DNA supercoiling at critically low temperatures. Finally, note that setting kunlock to infinite in reaction 3 (equivalent to having a model that does not allow promoter locking) results in a similar behavior to that of the plasmid-borne construct, and thus to the chromosome integrated promoter at 30 °C (data not shown). Transcription by a lesser active chromosome-integrated promoter construct is less sensitive to gyrase overexpression and temperature shifts. The influence of positive supercoiling buildup on the dynamics of a chromosome-integrated gene differs with its location on the chromosome due to, among other, differences in the expression rates of operons in different DNA loops64, which will cause the effects of positive supercoiling buildup to differ. Meanwhile, we observed that temperature downshifts have weaker effects on the plasmid-borne PLacO3O1 than on the chromosome-integrated PLacO3O1, since the latter is affected by positive supercoiling buildup. This implies that temperature down-shifts affect the various steps in transcription initiation by different degrees (as suggested in44), with escape from positive supercoiling buildup being one of the most affected. Further, from above, it is reasonable to hypothesize that the effects of temperature downshifts on RNA production rates may be positively correlated with the expression rate of the chromosome-integrated promoter of interest. Namely, consider that the expected time for a Gyrase to intervene is determined, among other, by its intracellular concentration. As such, when increasing the transcription rate of a gene, it should become less likely for Gyrase to remove positive supercoiling buildup between transcription events. Similarly, one can also hypothesize that a chromosome-integrated promoter similar in functioning to, but with weaker activity than PLacO3O1, should be relatively less sensitive to temperature downshifts. I.e., its RNA production rate should be less reduced by a temperature downshift, relative to the control condition (i.e. more similar in behavior to the plasmid-integrated PLacO3O1). Figure 5. Mean RNA numbers in individual cells, relative to the last time moment, as a function of temperature shifts, measured by microscopy with single RNA tagging by MS2-GFP, when PLacO3O1 is integrated into the chromosome (light grey). In these measurements of integer-valued RNA numbers of PLacO3O1’s activity when integrated in the chromosome, first, the cells are kept at 30 °C for 30 minutes. Next, they are kept at 30 °C and measurements are conducted (starting point of the measurements is defined as moment t = 0). 30 minutes after starting the measurements, the temperature is changed to 10 °C and then kept constant until reaching moment 120 min. Then it is altered again to 30 °C and kept constant until the end of the measurements. Data presented as relative mean to the reference case with 90% confidence intervals obtained from a two-tailed Student’s t-test. Sample size per condition, as time progresses is 600, 601, 603, 615, 613, 610, 603, 614, 611, 608, 607 cells. For each time point, new cells were taken from the original culture. Results are obtained from 3 biological repeats. Since these exhibited no statistically significant differences, the results presented here are composed of the data from the 3 biological replicates. Finally, at t = 0 min, the mean absolute number of RNA molecules per cell was 1.3.