doi: 10.1038/s41467-019-10330-w.
Role of network-mediated stochasticity in mammalian drug resistance
Affiliations
- PMID: 31235692
- PMCID: PMC6591227
- DOI: 10.1038/s41467-019-10330-w
Item in Clipboard
Role of network-mediated stochasticity in mammalian drug resistance
Nat Commun.
.
Abstract
A major challenge in biology is that genetically identical cells in the same environment can display gene expression stochasticity (noise), which contributes to bet-hedging, drug tolerance, and cell-fate switching. The magnitude and timescales of stochastic fluctuations can depend on the gene regulatory network. Currently, it is unclear how gene expression noise of specific networks impacts the evolution of drug resistance in mammalian cells. Answering this question requires adjusting network noise independently from mean expression. Here, we develop positive and negative feedback-based synthetic gene circuits to decouple noise from the mean for Puromycin resistance gene expression in Chinese Hamster Ovary cells. In low Puromycin concentrations, the high-noise, positive-feedback network delays long-term adaptation, whereas it facilitates adaptation under high Puromycin concentration. Accordingly, the low-noise, negative-feedback circuit can maintain resistance by acquiring mutations while the positive-feedback circuit remains mutation-free and regains drug sensitivity. These findings may have profound implications for chemotherapeutic inefficiency and cancer relapse.
Conflict of interest statement
The authors declare no competing interests.
Figures
Stress-dependent effect of network noise on drug resistance. a Tuning the induction (yellow gradient) of mammalian positive (mPF, left, red circle) or negative (mNF, left, blue circle) feedback synthetic gene circuits can confer high and low gene expression noise while the mean expression is identical. This enables decoupling gene expression noise amplitude (middle, standard deviation divided by the mean; σ/μ) from the mean within a decoupled noise regime (right, red and blue dashed lines) composed of decoupled noise points (right, red and blue arrows). b Schematic depictions to illustrate fractional viability under low or high levels of drug (stress, grey arrow) for cells with high (red distribution) or low (blue distribution) gene expression noise of a drug resistance gene. Relative survival of cells upon drug treatment will depend on network noise relative to the fitness function (dashed black line). If the fitness function is steep, noise hinders survival under low levels of drug while it is beneficial under high levels of drug (Supplementary Fig. 1)
Dose-response of the mPF-PuroR gene circuit. a Network schematic of the mPF-PuroR gene circuit induced by Doxycycline (Dox), which expresses the reverse tetracycline transactivator (rtTA) regulator, the Puromycin resistance gene (PuroR) and EGFP separated by the self-cleaving 2A elements. The rtTA regulator activates its own expression upon binding Dox (red dashed line). b Normalized mean expression under varying levels of Doxycycline induction. c Gene expression noise amplitude (normalized coefficient of variation, CV) in response to Doxycycline induction. Error bars denote the standard error of the mean. There is an x-axis break (//) between 50 and 500 ng/mL Doxycycline. All samples were measured in triplicate (n = 3). d Single-cell gene expression distributions of mPF-PuroR cells with broad peaks at intermediate levels of Doxycycline. The legend displays Doxycycline concentrations per distribution. Distributions are from representative replicates. Source data are provided as a Source Data file
Dose-response of the mNF-PuroR gene circuit. a The mNF-PuroR gene circuit controls the expression of a Puromycin resistance gene and the EGFP reporter gene through inducible negative auto-regulation (blue dashed line) of a humanized tetracycline repressor (hTetR) gene. The 2A peptides self-cleave after translation. b Normalized mean expression of mNF-PuroR cells under varying levels of Doxycycline (Dox). c Gene expression noise of mNF-PuroR cells in response to Doxycycline. Error bars denote the standard error of the mean. There is an x-axis break (//) between 50 and 500 ng/mL Dox. All samples were measured in triplicate (n = 3). d Single-cell gene expression distributions of the mNF-PuroR circuit. The legend indicates Doxycycline concentrations for each distribution. Distributions are from representative replicates. Source data are provided as a Source Data file
Decoupled noise regime and decoupled noise points before treatment. a Plotting the noise (coefficient of variation, CV) as a function of normalized mean gene expression for both gene circuits revealed two decoupled noise regimes. Black brackets indicate the expression range for the Doxycycline (Dox) concentrations used for noise-mean decoupling in (b). Error bars represent the standard error of the mean (n = 3). b Decoupled noise points (DNPs) at the beginning of the two drug treatment experiment sets. The noise was significantly different between gene circuits for both sets (**p value = 0.0022, n = 6, two-tailed Mann–Whitney U test). The mean expression was not significantly different for set 1 (~3%; p value = 0.0931, n = 6, two-tailed Mann–Whitney U test) while it had significance for set 2 (~8%; *p value = 0.0022, n = 6, two-tailed Mann–Whitney U test). The dashed lines display the range of statistically significant differences for gene expression noise and mean expression between the two gene circuits. c Gene expression distributions at the DNP from the first experiment set, filtered as described in the Methods. d Images of cells at the decoupled noise point from the first experiment set. The two-tailed Mann–Whitney U test inferred significance at p values < 0.05. Source data are provided as a Source Data file
Noise hinders resistance under low stress but aids it under high stress. a Experimental workflow of the Puromycin (Puro) treatment assay. Cells induced with Doxycycline (Dox) were treated with Puromycin in a parallel series of plates for imaging or for flow cytometry. Upon confluency, Puromycin was removed temporarily (see Fig. 7). b Illustration of a representative growth curve with three growth phases: (1) growth suppression, (2) regrowth (gray box), and (3) saturation (green box). c–g Growth curves for cells initially tuned to the DNPs under 0 (c), 10 (d), 22.5 (e), 35 (f), and 50 (g) μg/mL Puromycin. Dash-dot growth curves indicate data from the first experimental set while dash-dash growth curves are from the second experiment set. h Mean adaptation times corresponding to (c–g) (**p value = 0.0022, n = 6, two-sided Mann–Whitney U test; *p value = 0.0238, n = 3 for mPF-PuroR and n = 6 for mNF-PuroR, two-sided Mann–Whitney U test). The statistical test on the data from 35 μg/mL Puromycin included the non-growing mPF replicate 1 (infinite replicate adaptation time), which is not included in the mean (nmean = 2 for mPF-PuroR) in (h). The adaptation time error bars represent the standard error of the mean with replicates as described above. The two-tailed Mann–Whitney U test inferred significant differences at p values < 0.05. Source data are provided as a Source Data file
Modeling adaptation of mPF and mNF cells in various Puromycin concentrations. a Schematic depicting the effects of Puromycin concentration on CHO cell population composition and survival. Nongenetically Puromycin-resistant cells (green cells - brighter cells have higher PuroR expression level and are therefore more resistant) and nongrowing persister cells (magenta cells) can switch phenotypes (dashed bidirectional arrow). Persister cells and growing nongenetically resistant cells can also become stably Puromycin-resistant cells (black cells). When no Puromycin is present, a clonal population with heterogeneous gene expression exists (center). Under low Puromycin treatment conditions (left arrow), cells with low PuroR expression perish and a small fraction of the surviving clonal cells become persister cells. For high Puromycin treatment conditions (right arrow), only cells with high PuroR expression levels can survive drug treatment while the rest die (dark blue cells), and a higher fraction of the surviving cells become persisters. As persister and nongenetically resistant cells can become stably drug-resistant, the population on the right panel becomes increasingly heterogeneous over the course of treatment. b–f Representative growth curves for simulated mPF-PuroR and mNF-PuroR CHO cell populations under (b) 0, (c) 10, (d) 22.5, (e) 35, and (f) 50 μg/mL of Puromycin. Growth curves shown in panels in (b–f) correspond to: mPF subpopulations (left), mNF subpopulations (center), and full mPF and mNF populations (right). g Adaptation times corresponding to the mPF-PuroR and mNF-PuroR populations shown in panels (b–f). The model is described in the “Methods” section and parameter values are given in Supplementary Table 2
Temporary drug removal suggests PuroR-dependent resistance mechanisms. a Schematic for the drug removal and retreatment experiment. Doxycycline (Dox) was removed or maintained simultaneously with drug removal. b Mean expression for mNF-PuroR during temporary removal of 35 μg/mL Puromycin and after final retreatment (FT). c, d Mean expression for mPF-PuroR after removal of (c) 35 and (d) 50 μg/mL Puromycin and after final retreatment (FT). e–g Growth curves (top) and adaptation times (bottom) during re-treatment for (e) mNF-PuroR cells under 35 μg/mL, and mPF-PuroR cells under (f) 35 and (g) 50 μg/mL Puromycin. The solid and dashed black lines in panels b–d indicate uninduced and induced baseline mean expression levels, respectively, prior to treatment with Puromycin. The black growth curves are averaged untreated cell counts from Fig. 5c. Source data are provided as a Source Data file
References
Publication types
MeSH terms
Substances
Grants and funding
LinkOut - more resources
Full Text Sources
Other Literature Sources
Research Materials
