Project Topics
Get Complete Project Material File(s) Now! »
Transcriptional regulatory network of E. coli consists of hierarchical strongly connected components
We have compared the transcription factor network from RegulonDB [45] and compared it to one we have derived from bioinformatics analysis of the global feedback structures from Ecocyc database [44] and found that (1) each global regulator is strongly connected to a cluster of regulators that mutually regulate each other, (2) these clusters have hierarchical relations associated with physiological tasks (oxidation, starvation…) and (3) all regulators are downstream combinations of these few clusters (Figure 9). This is contradictory to previous depictions of the transcriptional regulatory network, which showed it as a purely hierarchical structure [39]. We also depict the sigma factors to be at the bottom output layer of this network. Here, the two most dominate sigma factors in E. coli rpoD and rpoS show very different positions. The first strongly connected component is the only one which regulates rpoD, while rpoS is regulated by all of the strongly connected components except the MarRA/Rob cluster. This may reflect their roles in house-keeping and stress responses respectfully. It may be appropriate to put the sigma factors at the top of the hierarchy, however this would result in the entire network becoming one strongly connected component.
Interactions between Strongly Connected Components
Interactions between interacting components of the regulatory network may constrain the evolutionary trajectories of cells. This can occur when the fitness effects of mutations in one component are affected by the presence of another mutation (epistasis). If the mutation is beneficial in one instance and deleterious in another, or vice versa, this is referred to as sign epistasis. When this occurs in a component involved in adaption, it constrains the possible evolutionary trajectories, avoiding the deleterious mutation [26]. It is possible for two mutations to individually be deleterious but together are beneficial, referred to as reciprocal sign epistasis. Reciprocal sign epistasis is necessary for multiple peaks within a fitness landscape, and cells can become stuck on these peaks, unable to evolve towards a higher maximum fitness because any single mutation would be selected against. However these epistatic interactions are dependent on the environment, and changes in the environmental conditions provide an escape from fitness peaks within these rugged fitness landscapes [26]. If changes to environment provide additional evolutionary trajectories, it is possible that changes to selection pressures might as well. It is not clear if epistatic traits between genes would hold across multiple fitness measurements, such as growth or mobility, and if not could competing selection pressures overcome restrictions from reciprocal sign epistasis? Within the transcriptional regulatory network of E. coli, there are multiple strongly connected components. These are groups of transcription factors which are able to directly or indirectly regulate all other transcription factors within that group. The members of a strongly connected component appear to be functionally related. For example FNR, ArcA, Fur, SoxR, and SoxS are all in a single strongly connected component and all 5 transcription factors directly sense the oxidative state of the cell, which modulates their functionality. Sign epistasis has recently been shown within transcriptional cascades [25]. This motivated us to explore how perturbations in one strongly connected component would influence phenotypes that structurally we associate with another strongly connected component, and how the hierarchy between strongly connected components influences the epistasis of between genes in different strongly connected components. To this end, we chose to study the influence between the strongly connected component associated with the energy state of the cell and the strongly connected component associated with cellular mobility. In this context, we can quantify the proportion of upstream sign epistasis (that is within the energy cluster), downstream sign epistasis (within the mobility cluster), and reciprocal sign epistasis. We can also see if there is a tendency for sign epistasis to be coupled between fitness measurements, and if so, correspond to the regulatory network structure?
Growth Competition and Swimming Competition fitness
Initially we performed growth and swimming assays by inserting our fluorescent pCRRNA vectors into LC-E24 :: dcas9 2tetO HK022 attB. We performed Growth assays on LB media, and M63 media containing either Glucose or Lactose as a Carbon Source. In LB media, all the perturbations reduced the fitness of strain in competition assays. In our earlier tests with pCKDL vectors, the single crp, fis, and hns perturbations did not show significant differences in maximum growth rate to the reference strain. We saw much stronger negative fitness measurements when crp, fis, and hns were perturbed in competition assays. This is expected due to the much higher sensitivity of competition assays compared to growth curves. The biological replicates are separated on the Y axis, which indicates variations in the starting ratios of the perturbation strains and the reference strains on separate days. However this does not change the direction of the slope of the curves. We found that the competition fitness was media dependent. When we changed to a defined media M63, certain perturbations were advantageous. This further depended on the carbon source available. In M63 with Glucose as a carbon source, gadE perturbation became advantageous. We also observed sign epistasis between some pairs of genes, specifically crp-fliZ, ihfB-hns, and ihfB-rcsB, because these pairs of perturbations had a fitness advantage although all of the individual perturbations caused a fitness disadvantage. We also saw higher than expected fitness advantage for fis-gadE, and ihfB-gadE. In M63 with Glucose we do have cases when the slope was not reproduced in separated experiments, namely crp-gadE and fis-rcsB showed different fitness in different replicates. This could be due to individual mutants escaping the CRISPR-dCas9 system. When the carbon source was changed to lactose, we found even more perturbations which gave a fitness advantage. Here, 12 strains had a fitness advantage over the reference strain. Additionally, reproducibility was decreased, with 5 strains showing inconsistent phenotypes. Overall, the diversity of the growth competition phenotypes increases as we increased competitive stress by decreasing the richness of the growth media.
Fitness Landscape of interactions between ‘Energy’ and ‘Mobility’ Regulatory Clusters
For each pair of perturbations we plotted the growth and swimming fitness measurements for the reference strain, the individual perturbations and the combination of perturbations for LB media, M63 with Glucose and M6 with Lactose. Since we have normalized the fitness of the reference strain to 1, the predicted fitness of the double perturbation strain (AB) with no epistasis is fAB = fref * fA * fB. This prevents negative fitness from being possible, as it would be impossible to have a negative swimming diameter or a negative growth rate. With this we calculated all interactions that would have an expected fitness with no epistasis. We plotted the fitness measurements for each pair of genes against each other in each media (Figure 14). The direction and the shape of the trajectories indicates the type of epistasis between the two genes. If the trajectories for a given gene are in the opposite direction in relation to one or two axis, they indicate sign epistasis for that gene in the corresponding fitness. If the arrows have a different length but the same direction, they indicate magnitude epistasis. We identified 6 basic shapes within our data. Parallelogram, Obtuse, Acute, Triangles, Concave, and Hourglasses. Parallelograms always indicate no epistasis. Obtuse quadrilaterals indicate magnitude epistasis. Acute quadrilaterals will indicate sign epistasis on the two side edges with acute angles with the long edge as long as neither edge is parallel with one of the axis, these sometimes occur as Acute Trapezoids (for example CRP + RcsB in LB). Triangles have two strains with the same fitness measurements resulting in one edge length near zero, when it is a single and double mutant it indicates that one perturbation is completely masking the influence of the other perturbation. When it is the reference strain and a single mutant, it indicates that one perturbation has no influence without the second perturbation. Concave shapes indicate reciprocal sign epistasis for at least one fitness measurement, and depending on the orientation, possibly both. Finally, Hourglass shapes have fitness that cross over. The base of the hourglass will have sign epistasis in both fitness metrics unless the parallel with one axis in which case it will only be sign epistasis in one fitness. Depending on the rotation of the hourglass, it may also have sign epistasis for the other gene, resulting in reciprocal sign epistasis, however it can only have reciprocal sign epistasis for one fitness. It is important to note that the shape alone does not determine the epistasis, but the rotation of the shape must also be taken into account.
Table of contents :
Table of Contents
1 Introduction
1.1 E. coli Transcriptional Regulatory Network
1.2 Global Transcriptional Regulators
1.3 Mathematical and Computation Methods for modeling Regulatory Networks
1.4 Large scale control and programming of gene expression using CRISPR
2 Interaction between Transcription Factors in Different Regulatory Clusters
2.1 Transcriptional regulatory network of E. coli consists of hierarchical strongly connected components
2.2 Interactions between Strongly Connected Components
2.3 Growth Competition and Swimming Competition fitness
2.4 Fitness Landscape of interactions between ‘Energy’ and ‘Mobility’ Regulatory Clusters
2.5 Methodology
2.6 Supplementary Figures
3 Multiplexed Knockdowns of Global Transcription Regulators
3.1 CRISPR-Cas9 Knockdowns
3.2 CKDL Vector for Multiplexed Knockdowns
3.3 Fitness and Epistasis of pCKDL Perturbations
3.4 Transcriptional Profiles of pCKDL Perturbations
3.5 Logical Modelling of E. coli Regulatory Network
3.6 Methodology
3.7 Supplementary Figures
4 Molecular Barcoding for Screening of Antibiotic Combinations
4.1 Tracking Droplet Combinations with DNA barcodes
4.2 Quantification of barcode efficiency
4.3 Methodology
5 Single Cell Transcriptional Analysis of E. coli
5.1 Microfluidic Design for Single Cell Transcriptional Analysis
5.2 Cell lysis and Isolation of cDNA
5.3 Single Cell RNA sequencing of MG1655
5.4 Single Cell analysis of Antibiotic Persister Cells
5.5 Methodology
5.6 Supplementary Figures
6 Perspectives
7 Bibliography
