Effects of resource sharing on gene expression dynamics Patrick M. Caveney1,2 S. Elizabeth Norred1,2 Charles W. Chin1,2 Jonathan B. Boreyko1,2,3 Brandon S. Razooky2,4 Scott T. Retterer1,2,5 C. Patrick Collier2 Michael L. Simpson1,2,6 1. 2. 3. 4. 5. 6. Bredesen Center, University of Tennessee, Knoxville Center for Nanophase Materials Sciences, Oak Ridge National Laboratory Department of Biomedical Engineering and Mechanics, Virginia Tech Laboratory of Immune Cell Epigenetics and Signaling, The Rockefeller University Biosciences Division, Oak Ridge National Laboratory Joint Institute for Biological Sciences, University of Tennessee, Knoxville and Oak Ridge National Laboratory Introduction Method Results Episodic gene expression, bursting, is defined by periods of high expression, ON, separated by periods of no expression, OFF, and has been observed in all kingdoms of life[1]. When a gene is ON it draws heavily from a shared pool of reusable resources (e.g. polymerase, ribosomes) for a short time while using no resources during OFF periods. Bursting is characterized by burst size (i.e. duration of a burst) and burst frequency (i.e. number of bursts per time). Resource sharing leads to interactions between genes with no regulatory relationships[2]. We use a synthetic, microfabricated platform to investigate the effect of sharing limited, reusable resources on burst size and frequency. Lithographically defined, actuatable, PDMS chambers (height of ~5 µm and diameters of 10, 5, and 2 µm with volumes 393, 99, and 16 fL) contained an EGFP expressing cell-free protein synthesis system[3]. Reactions were also confined in POPC water-in-water vesicles prepared by oil emulsion[4] with volumes ranging from ~20 fL to ~2000 fL. The amount of protein in each chamber was measured for one hour. The deterministic trend of EGFP growth was removed to isolate the noise of protein expression in each chamber. Full details in[3]. The key finding is that larger shared resource pools prefered to produce protein by extending existing bursts (i.e. increasing burst size) instead of initiating more bursts (i.e. increasing burst frequency). This result was confirmed in two different systems of cell-sized reactions, PDMS chambers and POPC vesicles. Simulations of many genes pulling resources from a common pool indicated increased burst size in larger reactions was responsible for increased protein abundance. This happened because the first few mRNA created gathered the largest portion of resources, regardless of the reaction size, and in larger reactions there were more resources available to gather. Thus, in agreement with in vivo experiments[5], protein abundance is preferentially modulated by burst size not burst frequency. How is bursting affected by resource sharing? Figure 1: A) Bursty gene expression is caused by genes transitioning between ON and OFF states. While in the ON state the genes draw heavily from a common pool of resources. B) We experimentally vary the way resources are shared, either by increasing the number of discrete pools or increasing the size of one pool. C) We observe how protein abundance may increases either by increased burst frequency or increased burst size. D) Changes in burst size and frequency can be distinguished on a noise plot of CV2 vs protein abundance[3]. A 15 min 30 min 60 min Figure 4: (left) CV2 vs. protein abundance for 2, 5, and 10 μm diameter chambers. The small filled data points represent individual chambers while the large filled data points show the mean behaviors for all chambers of a given size. The dotted line is of the form a/Abundance and where a is a constant that is calibrated so the line passes through the mean of the 2 μm diameter chambers (large filled orange triangle). The open orange triangles show combinations of 2 μm chambers which share resources in the same manner as single 2 μm chambers. The leftmost open triangle shows the average behavior of sums of 2 individual 2 μm chambers, while the rightmost open triangle shows the average behavior of sums of 6 individual 2 μm chambers. CV2 for these combinations of 2 μm chambers closely follow the a/Abundance trend showing individual 2 μm chambers were statistically independent and burst frequency increased with increasing abundance while burst size was constant. In contrast, the individual 5 μm chambers deviate strongly (solid line) from the a/Abundance trend even though their volume and protein abundance were about equal to six 2 μm chambers (red box). The 5 μm chambers had proportionally larger resource pools, and shared these resources by increasing burst size, not burst frequency. (right) CV2 vs. protein abundance for vesicles ranging in diameter from 4 μm to 19 μm. Each data point (gray or colored) represents an individual vesicle. The orange points are vesicles with diameters 9-10 μm, and the blue points have diameters 18-19 μm. The solid line is a power law fit to all points. While abundance varies by 3 orders of magnitude, CV 2 only decrease by about 1 order of magnitude. Dashed lines show fits to individual volumes (orange and blue) where CV2 goes as 1/Abundance2. Gillespie simulation of resource sharing 120 min 135 min mRNA made early gather disproportionate amounts of resources B EGFP production in cell-free reactions confined to cell-sized PDMS chambers and POPC vesicles was observed 0 min Figure 3. Protein abundance after 1 hour scaled linearly with volume in both the chambers (left) and vesicles (right). The volume range of the vesicles was larger than the chambers (orange band). Increasing protein abundance is driven by increased burst size in PDMS chambers and POPC vesicles C D Protein abundance scales linearly with volume 150 min 180 min Figure 2: Cell-free reactions were confined in both PDMS microfluidic chambers and POPC vesicles of varying volumes. Reactions in both constitutively expressed EGFP, and fluorescence increased over one hour. For both methods fluorescence was measured every 3 minutes. Noise was extracted by subtracting a general trend (gray line) scaled by a gain factor from each individual trace[3]. The resulting noise traces were quantified with the coefficient of variation squared (CV2 = variance / mean2) and plotted against fluorescence. Figure 5: Schematic of Gillespie simulation of resource sharing. Genes bursted ON and OFF with rates kON and kOFF, respectively. While ON, genes produced mRNA at rate . Ribosomes in a common pool (i.e. reusable resource molecules) associated with mRNA with rate constant kb. The number of genes varied between 5 and 50, and the number of resources varied proportionally with 300 resources per gene. Figure 6: (left) CV2 vs protein abundance from the model described in Figure 5. Colors represent the size of the reaction from 5 to 50 genes. Large points are means of multiple runs of the same reaction size. The solid line is a power law fit to all data points while the dashed line is of the form a/Abundance2 where a is selected so that the line passes through the mean of a 35 gene reaction. As in the experimental chambers and vesicle data, CV2 from this model was relatively insensitive to increases in abundance driven by changes in reaction sizes. This shows that between reaction sizes abundance increases were driven by increased burst size. In contrast, CV2 is highly sensitive to increases in abundance that occur within a single reaction size, predicting that within a single reaction size increased abundance is the result of increased burst frequency and decreased burst size. This behavior was also seen experimentally in the chambers and vesicles (Figure 3). (right) mRNA molecules are ranked in the order of the time they were created. The yaxis shows the fraction of the total protein translated from each mRNA molecule. Points are colored by the reaction size (small reaction sizes are more blue, larger ones are more red). mRNA molecules made early, regardless of the reaction size, collected a disproportionate amount of resources and made a disproportionate amount of the total protein. Acknowledgements This research was conducted at the Center for Nanophase Materials Sciences, which is sponsored at Oak Ridge National Laboratory by the Scientific User Facilities Division, Office of Basic Energy Sciences, U.S. Department of Energy. The authors would also like to thank the following individuals for their fruitful discussions Jennifer Morrell-Falvey, Carmen Foster, and Roy Dar. References 1. A. Sanchez, I. Golding, Genetic determinants and cellular constraints in noisy gene expression. Science. 342, 1188–1193 (2013). 2. Gyorgy, Andras, et al. "Isocost lines describe the cellular economy of genetic circuits." Biophysical Journal 109.3 (2015): 639-646. 3. Norred, Sarah Elizabeth, et al. "Sealable Femtoliter Chamber Arrays for Cellfree Biology." Journal of visualized experiments: JoVE 97 (2015). 4. K. Nishimura, S. Tsuru, H. Suzuki, T. Yomo, Stochasticity in gene expression in a cell-sized compartment. ACS Synth. Biol. (2014) 5. R. D. Dar, B. S. Razooky, L. S. Weinberger, C. D. Cox, M. L. Simpson, The Low Noise Limit in Gene Expression. PloS One. 10, e0140969 (2015).
© Copyright 2026 Paperzz