Development of a system for discovery of genetic interactions for essential genes in Escherichia coli K-12

ABSTRACT

Genetic interaction networks are especially useful for functional assignment of genes and gaining new insights into the systems-level organization of the cell. While studying interactions of nonessential genes can be relatively straight-forward via use of deletion mutants, different approaches must be used to reveal interactions of essential genes due to their indispensability. One method shown to be useful for revealing interactions of essential genes requires tagging the query protein. However, this approach can be complicated by mutational effects of potential hypomorphic alleles. Here, we describe a pilot study for a new scheme of systematically studying the interactions of essential genes. Our method uses a low-copy, F-based, complementing plasmid, pFE604T, from which the essential gene is conditionally expressed. The essential gene is expressed at lower levels, producing a moderate growth defect in a query host. Secondary mutations are introduced into the query host by conjugation and the resultant exconjugants are scored for growth by imaging them over time. We report results from studying five essential query genes: dnaN, ftsW, trmD, yrfF and yjgP, showing (on average) interactions with nearly 80 nonessential genes. This system should prove useful for genome-wide analyses of other essential genes in E. coli K-12.

INTRODUCTION

Genetic interactions occur when two mutations together result in an effect(s) different from either mutation alone. Interactions are positive (additive) when two mutations show a synergistic effect that is more extreme than when alone, for example, synthetic lethality; interactions are negative (diminished) when the phenotype of the double mutant is less severe than either mutation alone. Since genetic interactions often occur among genes in compensatory pathways or interlinked biological processes, studying them can provide useful information for elucidation of gene function (Mani et al., 2008). Models for genetic interaction networks have been constructed in both eukaryotic and prokaryotic model systems (Byrne et al., 2007; Bakal et al., 2008; Butland et al., 2008; Costanzo et al., 2010; Pan et al., 2007; Davierwala et al., 2005; Roguev et al., 2008; Schuldiner et al., 2005; Tong et al., 2001, 2004; Typas et al., 2008). These networks have revealed the global modular organization of gene products and functional interaction of bioprocesses in several model organisms (Dixon et al., 2009).

Analyses of protein-protein interaction networks have revealed that proteins acting as hubs are more likely to be essential than less connected proteins (Jeong et al., 2001). Therefore, genetic interactions of essential genes may unravel key new insights not revealed from studying genetic interactions of nonessential genes. Systematic efforts to map genetic interactions in E. coli K-12 have relied on the construction of double mutants by Hfr conjugation of donor and recipient cells with deletions and different antibiotic resistance markers (Butland et al., 2008; Typas et al., 2008). While these methods are useful for examining between nonessential genes, the same method cannot be used to study essential genes because such essential gene deletion mutants are non-viable. This experimental difficulty has been previously bypassed by adding a C-terminal SPA (Sequential Peptide Affinity) tag to the essential gene (Butland et al., 2008; Babu et al., 2011). An SPA tag integrated in the C terminal presumably alters the 3’-UTR and hence destabilizes certain transcripts (Babu et al., 2011). However, there are drawbacks to using SPA tags to affect a defect in an essential gene for genetic interaction studies. The tag often causes no measurable defect in the absence of secondary mutations. Furthermore, the mechanisms of defects are not clearly understood (Butland et al., 2008). In order to understand genetic interactions, one needs to understand the defect in the query genes.

Here we describe a pilot study for a new method for high-throughput genetic interaction analysis in E. coli between essential genes and nonessential genes. In this method, a chromosomal knockout mutant of the essential gene is crossed with a high-density array of nonessential deletions to construct by conjugation double mutants that express knock-down levels of the essential query protein (Fig. 1). The major advantage of the system over the SPA-tagging system is that the mutational defect of the chromosomal knockout mutant of an essential gene is known: the level of essential query gene protein is lowered, causing a moderate growth defect.

Fig. 1.

High-throughput method to find essential-nonessential gene interactions. Hfr donor strains are chromosomal knockout mutants with an essential query gene deletion marked with cat (chloramphenicol resistance cassette; blue box) but complemented by the IPTG inducible knock-down level of essential query protein from pFE604T-ORF (green triangle). Recipient strains are nonessential gene deletions marked with kan (kanamycin resistance cassette; pink box). The marked essential query gene deletion and pFE604T-ORF are transferred to recipient strains en masse by the Hfr conjugation gene transfer system. The DNA carrying the essential query gene deletion recombines with the recipient’s chromosome via homologous recombination. The oriT site present on pFE604T-ORF allows the transfer of the plasmid to the recipient cells. E represents essential gene and N represents nonessential gene.

RESULTS

Construction of low-copy, F-based complementing plasmid pFE604T

pFE604T is a single-copy mini-F plasmid (Fig. 2) which was purposefully designed for systematic construction of chromosomal knockout mutants of essential genes for their high-throughput genetic interaction analyses. The essential gene is cloned into pFE604T using SfiI restriction sites downstream of a T5 promoter and lacI-O repressor-operator region, where is inducible by IPTG. Since E. coli genes in the ASKA library are flanked by SfiI sites (Kitagawa et al., 2005), any essential E. coli gene can readily be cloned into pFE604T. We refer to the pFE604T that can express an essential query gene as pFE604T-ORF. The viability of the chromosomal knockout mutant is sustained by expression of the essential query gene from the single-copy pFE604T-ORF. Construction of pFE604T-ORF is shown in Supplementary Fig. S1 and described in detail in the Supplementary Methods.

Fig. 2.

Diagram of pFE604T. The complementing plasmid, pFE604T, is used in both the construction of chromosomal knockout mutant of essential gene and in high throughput genetic interaction analysis. It is a single-copy mini-F derivative that contains features for plasmid stability maintenance (green colors), selection markers (blue colors), expression of the essential gene (red colors), and conditional replication (orange color). A query gene can be cloned into the ORF region and its expression is under the regulation of the lacI^q repressor system (yellow colors) and a T5 promoter.

Control of DNA replication, copy number, incompatibility, and partition from pKV713 (Kawasaki et al., 1990) are stringently controlled by the ori2, incC, repE, and sopABC genes present on pFE604T (Ogura and Hiraga, 1983; Mori et al., 1986; Uga et al., 1999). Antibiotic resistance markers (gentamycin and tetracycline) are incorporated into pFE604T for selection. pFE604T has two origins of replication: ori2 and oriRγ. ori2 is the default origin of replication, resulting in a single copy of the plasmid during cell division, while oriRγ is a conditional origin of replication that requires the trans-acting Π protein (encoded on the chromosome by pir) for replication and results in multiple plasmid copies each division. Due to having these two origins of replication, pFE604T can replicate at a single or medium plasmid copy number in non-pir or pir⁺ E. coli hosts, respectively. A non-pir host is used for genetic interaction experiments, while a pir⁺ host is used for producing additional plasmid. An oriT site is present on pFE604T, enabling the one-step transfer of pFE604T-ORF from an Hfr host to an array of nonessential gene deletion mutants en masse via conjugation. Transfer of pFE604T-ORF to the nonessential gene mutant is critical before or while transferring the second, essential gene mutation to the nonessential gene mutant in order for complementation of the essential gene to occur in the double mutant.

We began the construction of each chromosomal knockout mutant of an essential query gene by transforming E. coli K-12 BW25113 with the pFE604T-ORF containing the essential gene. Next, the plasmid-borne essential query gene is expressed from pFE604T-ORF under the induction of IPTG, allowing disruption of the query gene on the chromosome by one-step homologous recombination (Datsenko and Wanner, 2000).

The behavior of the doubling time of a chromosomal knockout mutant could in principle be modeled as a parabolic curve. Because bacteria tend to be programmed to express proteins at optimal levels (Dekel and Alon, 2005), the minimum of that curve would correspond to an IPTG concentration resulting in expression of the essential gene at optimal, physiological levels. IPTG concentrations less than “optimum” result in a subphysiological level of expression of the essential gene. Under this model, we can prove a concentration of IPTG produces subphysiological levels of expression of the essential gene by showing that a higher concentration of IPTG results in a faster doubling time than the original concentration. In fact, we observed that for five out of five pFE604T-ORF constructs tested, doubling times were faster when using 1 mM IPTG than when using 0.1 mM IPTG (Fig. 3). We conclude that a concentration of 0.1 mM IPTG produces subphysiolical expression levels of the essential gene from pFE604T-ORF. Full complementation from pFE604T-ORF was not observed even at 1 mM IPTG, likely due to the plasmid being present as only a single copy. Because there are observable growth defects in the chromosomal knockout mutants, they are amenable for genetic interaction analysis.

Fig. 3.

Doubling time of chromosomal knockout mutants as a function of IPTG concentration. Chromosomal knockout mutants have longer doubling times compared to the wild type in both IPTG concentrations. Error bars represent standard deviations in the measurements of the doubling times.

Development of a system for genetic interaction analysis involving essential genes

The rationale of our high-throughput system for the determination of genetic interaction is adopted from the previous studies utilizing the Hfr conjugation gene transfer system: eSGA and GIANT-coli (Babu et al., 2011; Typas et al., 2008). Hfr donors were made by integrating a conjugative F plasmid (CIP) into the chromosome of chromosomal knockout mutants (Fig. 3 of Typas et al. (2008)). The recipient strains were arrayed onto three 1536-density plates, covering 3845 Keio deletions (Baba et al., 2006) and 61 small RNA deletions marked with a kanamycin resistance cassette, for a total of 3906 nonessential gene deletions (see Supplementary Information (Yamamoto et al., 2009, Nomura et al, unpublished results).

A schematic flowchart illustrating the procedure for high throughput generation of double mutants by conjugation is shown in Fig. 4. In step 1, an Hfr donor strain carrying pFE604T-ORF was mated with recipient strains harboring a single nonessential gene deletion (marked with kan) resulting in the transfer of the complementing plasmid (marked with tet) to the recipients. This was accomplished in a high-throughput manner by transferring the nonessential gene deletion library onto a bacterial lawn of the donor strain using replicator pins. In step 2, successful conjugants (single nonessential gene mutants harboring pFE604T-ORF) were selected by transferring colonies onto plates containing both kanamycin (Km) and tetracycline (Tc) using replicator pins and allowing colonies to grow 24 hours. In step 3, the successful conjugants from step 2 were transferred to an IPTG containing agar plate that was previously covered with the chromosomal knockout mutant of Hfr donor. The second conjugation ensued, and chromosomal DNA containing the essential gene deletion (marked with cat) was transferred to recipient strains, in which recombination created double mutants. Mating occurred over 12 hours. In step 4, colonies were transferred onto plates containing chloramphenicol (Cm), tetracycline and IPTG for an intermediate selection that reduced further mating and avoided partial duplication artifacts. Colonies grew for 24 hours. In step 5, colonies from intermediate selection were pinned onto plates containing chloramphenicol, tetracycline, kanamycin and 0.1 mM IPTG.

Fig. 4.

Schematic of method for double mutant construction. E represents essential gene and N represents nonessential gene.

Computational processing

After 24 hours, images of the plates were obtained by scanning. An image-processing program computed pixels for the density of colonies on the plate. These raw data were normalized for the plate edge effect (Baryshnikova et al., 2010) and for plate-by-plate variation in average density. The genetic interaction scores (GI scores) we next calculated controlled for differences in density due to the nonessential gene deletion and the presence of the plasmid by normalizing average double mutant (DKO) colony densities with colony densities of the single nonessential gene deletion plus the query-specific plasmid (SKOp). Normal distribution parameters were estimated for modeling the distribution of non-interacting gene pairs. Values deviating from this distribution, as determined by cut-offs of the genetic interaction score (using FDR = 0.1), were designated as either negative interactions (low GI score) or positive interactions (high GI score). Genes that have negative interactions or positive interactions with three or more query genes were filtered out of the lists of specific genetic interactions (Supplementary Table S1). The mutants of these common interacting genes are usually either slow growers, which are sensitive to random perturbations, or have mutations that affect conjugation and recombination efficiency. Unfortunately, reported negative interactions of genes located nearby the query gene on the chromosome are not reliable since they could result from linkage effects during recombination. When linkage effects occur during recombination, the intact sequence of the secondary deletion gene from the donor’s chromosome replace the kanamycin resistance gene of the nonessential gene deletion strain (Babu et al., 2011). This creates a strain that is incapable of growing in kanamycin and creates the appearance of a genetic interaction during the screens. Therefore, genes located within 35 kb of the query gene were not included in lists of significantly interacting genes even if their GI score qualifies as significant, and they were excluded from further analysis.

Pilot study of five genes

In our pilot study of the genetic interaction method using pFE604T, we studied five essential query genes: dnaN, ftsW, trmD, yjgP, and yrfF. dnaN is the beta subunit of DNA polymerase III, ftsW is involved in cell division, trmD is required for tRNA methylation, yjgP is a LPS transporter and yrfF has unknown function. There were a large number of genes found genetically interacting with the five essential query genes in the pilot study (Supplementary Tables S2 and S3). Each screen resulted in an average of 77 interactions. Thus, these essential genes are required for buffering many cellular processes.

Functional enrichment analysis (Table 1) showed some expected and unprecedented regulation mechanisms for functions of the query genes. dnaN, showed negative interactions with genes that are enriched in DNA-related functions. yjgP, a gene that encodes a LPS transporter, shows negative interactions with genes enriched in lipopolysaccharide core region biosynthetic process. Other observations of functional enrichment are not readily apparent from the primary function of the essential query gene. Some interactions may be due to secondary effects caused by low levels of the essential protein. We found that the recC and dnaT genes, which are important for the repair of double strand breaks in DNA by recombination, show negative interaction with dnaN. This result suggests that knock-down of dnaN expression might result in double strand breaks, leading to induction of the SOS response. To investigate this hypothesis, a plasmid expressing a GFP-sulA fusion protein (pTN175) was used for monitoring the SOS response (Nakayashiki et al., 2013). We found that the fluorescence signals of sulA-GFP in the dnaN chromosomal knockout mutant are about 2020 ± 20 compared to 670 ± 15 in wild type, suggesting an approximately 3-fold increase of SOS response. The induction of SOS response was further evidenced by the observation of filamentous morphology under the microscope of the cells by microscopy.

Table 1. Functional enrichment analysis of gene interactions

Query genes	Functional enrichment of interacting genes	Type of interaction
dnaN	Peptidoglycan based cell wall	Positive
	Organelle envelope
dnaN	Homologous recombination	Negative
	Pyrimidine metabolism
	DNA replication
ftsW	Oxidative phosphorylation	Negative
	Organelle envelope
	Quinone
	TCA cycle
trmD	Oxidative phosphorylation	Positive
	Organelle envelope
	Purine nucleotide biosynthetic process
yjgP	Iron sulfur protein	Positive
	ATP-biosynthesis
yjgP	Lipopolysaccharide core region process	Negative
yrfF	Organelle inner membrane	Positive
	Cell membrane
yrfF	Organelle envelope	Negative
	RNA degradation

DISCUSSION

Genetic interaction analysis of essential genes is vital for the systems-level understanding of E. coli. The genetic interactions of essential genes in E. coli have not been satisfactorily examined, owing, at least in part, to the difficulty of the essential gene perturbation. While the reported genetic interactions involving essential genes in E. coli determined using SPA-tagged mutants has been an important step in this direction (Babu et al., 2011), previously mentioned shortcomings in the method of SPA-tagging may limit the applicability of results obtained from these studies. Here we described a system for the systematic determination of nonessential-essential gene pairs in E. coli. The interpretation of the interactions is more straightforward as compared to studies relied on SPA-tagging, because the knock-down effect is obvious: the shortage of the essential query protein.

Our system was demonstrated in five essential query genes encoding proteins involved in different cellular functions. We found that these essential genes interact with high number of nonessential genes with similar or dissimilar functions. In the future, it is readily possible to obtain lists of genetic interactions for all 328 essential genes of E. coli using the pFE604T plasmid. These lists can serve as an interaction catalogue for biologists to look for potential candidates to study further. Also, the construction of a library of chromosomal knockout mutants with down regulated of essential query gene will be a valuable experimental resource for the E. coli community.

Genetic interaction analysis of three out of our five query genes have been studied previously using SPA-tagging (Babu et al., 2011). Surprisingly, there is no overlap between the genetic interaction lists of these studies. An explanation for the lack of consistency might be that the perturbation of the essential gene used in this study is vastly different than the previously used perturbation of SPA-tagging. Furthermore, the laboratory environment external to experiment might affect the outcome of the genetic interaction analysis (Michaut and Bader, 2012). Although the two methods do not produce reproducible lists of interactions, the lists may be thought of as complementary and providing insight into different dimensions of genetic interaction.

We observed that genes of disparate function interact with the same essential gene. Such interactions might not be uncommon, as they were also evident in yeast (Davierwala et al., 2005). In fact, essential genes tend to have more functional links than nonessential genes (Babu et al., 2011; Davierwala et al., 2005), and the tendency of having proximate functional relationship between interacting genes is not as strong as it is for nonessential genes (Davierwala et al., 2005). The average genetic interactions per query gene for essential genes in this study are approximately 77, as compared to 20 in the nonessential genes (Babu et al., 2011). Thus, essential genes might be much more functionally involved in the cellular processes than what we have thought.

MATERIALS AND METHODS

Strains and growth conditions

The library of 3906 strains carrying a single nonlethal mutation consisted of mutants from the Keio collection (Baba et al., 2006) and a small RNA deletion mutant library (Nomura et al., unpublished data). Non-pir BW25113 was used as the background to construct the chromosomal knockout mutants (referred to as “wild type” in the text). The pir⁺ strain BW25141 (Datsenko and Wanner, 2000) was used to increase plasmid copy number. All the strains used were routinely grown in LB medium containing 1% Bacto Tryptone (Difco), 0.5% yeast extract (Difco), and 1% NaCl (Wako, Osaka, Japan) with or without antibiotics at 50 μg/ml for ampicillin, 30 μg/ml for kanamycin, 5 μg/ml for gentamycin, 12.5 μg/ml for tetracycline, 25 μg/ml for chloramphenicol, 15 μg/ml for streptomycin at 30℃ or 37℃. All antibiotics and IPTG were from Wako (Osaka, Japan). Chromosomal knockout mutants of essential genes were grown in LB medium containing tetracycline, chloramphenicol and 0.1 mM IPTG at 37℃. Hfr donor strains were grown in media that included streptomycin.

Plasmids

pAH143 (Haldimann and Wanner, 2001), pLZ2210-CAS8 (Zhou and Wanner, unpublished data), pKD46 and pKD3 (Datsenko and Wanner, 2000), pTN175 (Nakayashiki et al., 2013), and ASKA ORF clones (Kitagawa et al., 2005) have been described. Development of CIP plasmids (Typas et al., 2008, Takeuchi et al., unpublished results) will be described elsewhere.

Construction of double mutant involving a chromosomal knockout mutant of essential gene using automated strain arraying

We arrayed colonies using a replica-pinning ROTOR HDA bench-top robot (Singer instrument) to automate the process of construction of double mutants. An outline of the technique is shown in Fig. 4.

Data analysis

Images of each plate were scanned using an EPSON GT-X970 scanner. Raw colony densities were quantified from plate images using an in-house developed image analysis program (Takeuchi et al, unpublished results), producing a 32 × 48 matrix of pixel values. Data from plates was categorized as either a single deletion plus plasmid (SKOp) or as a double mutant (DKO). Because colony sizes tend to be larger in outer layers due to a position effect (Baryshnikova et al., 2010), we normalized entries the outermost three columns/rows of the matrix by multiplying the colony density by the median colony across all plates in the category (SKOp or DKO) divided by the mean colony central density within only the particular layer the data point is located in, averaged across all plates in the category. We next normalized all colony sizes within each data matrix by dividing by the mean value of the values in the matrix.

A single genetic interaction score was calculated for each DKO combination by dividing the mean normalized colony density of the DKO by the mean normalized colony central density of the corresponding SKOp. We assume the genetic interaction scores of non-interacting gene pairs will form a normal distribution. We estimate the parameters of this distribution using a least-squares fit of the graph of a normal probability distribution function with the density plot of the genetic interaction scores of each query gene. Density plots are obtained using the default function in R (http://www.r-project.org/). Fitting only considers the central 50% density of the genetic interaction scores, with the density plot scaled so that the area under curve is the same for density plot and the fitted curve. We then separately consider for negative and positive interactions the false discovery rates calculated by this model when using various cutoffs for significance. We chose cut-offs for positive and negative interactions that result in a predicted 10% false discovery rate.

We filtered out “genetic interactions” that are non-specific to individual query genes by ignoring genes that have significant genetic interaction scores for three or more of the query genes in downstream analysis. Additionally, we filtered out interactions with genes located within 35 kb of the query, since such “interactions” are often due to overwriting of the nonessential gene deletion and antibiotic resistance gene during recombination with DNA carrying the deletion of the query gene (Babu et al., 2011). Functional enrichment was tested for sets of negative interactions and sets of positive interactions using the Database for Annotation, Visualization, and Integrated Discovery (DAVID) v6.7 (Huang et al., 2009a, b).

Growth curve analysis of chromosomal knockout mutants of essential query genes

Overnight cultures of chromosomal knockout mutants were inoculated into 96-well microtitre plates containing 200 μl of liquid medium supplemented with 1 mM or 0.1 mM of IPTG, incubated at 37℃ for 24 hours with optical density (600 nm) measured every 30 min using an automated SpectraMax^® GEMINI EM (Molecular Devices Inc). Doubling time of each strain cultivated in different IPTG concentrations was calculated using an online calculator: http://www.doubling-time.com/compute.php.

Flow cytometry analysis

The dnaN chromosomal knockout mutant harboring a sulA-GFP plasmid, pTN175, was grown overnight to stationary phase and diluted 1000× with 1× phosphate-buffered saline. Data were collected using Accuri^TM C6 flow cytometer (Becton Dickinson) with a 488-nm argon laser and a 515- to 545-nm emission filter (FL1) at high flow rate.

Author’s contributions

HTY designed and carried out the experiment and wrote a draft of the manuscript. NY and TN participated in the design of the study. KAD and YH participated in plasmid design, especially work carried out by HTY, NY, and RK at Purdue University. RT developed the colony quantification program. TC participated in computational analysis. TN provided direct supervision, helpful comments, and discussion throughout this study. BLW and HM jointly oversaw all aspects of this research and rewrote the draft manuscript for publication.

ACKNOWLEDGMENTS

This study was supported in part by a Grant-in-Aid for Scientific Research (A), (C) and a Grant-in-Aid for Scientific Research (Kakenhi) on Priority Areas System Genomics from the Ministry of Education, Culture, Sports, Science, and Technology of Japan to the Nara Institute of Science and Technology to HM, NIH GM077905 from U. S. Public Health Service and Award 106394 from the US National Science Foundation to BLW.

REFERENCES

•  Baba,  T.,  Ara,  T.,  Hasegawa,  M.,  Takai,  Y.,  Okumura,  Y.,  Baba,  M.,  Datsenko,  K. A.,  Tomita,  M.,  Wanner,  B. L., and  Mori,  H. (2006) Construction of Escherichia coli K-12 in-frame, single-gene knockout mutants: the Keio collection. Mol. Syst. Biol. 2, 2006.
•  Babu,  M.,  Diaz-Mejia,  J. J.,  Vlasblom,  J.,  Gagarinova,  A.,  Phanse,  S.,  Graham,  C.,  Yousif,  F.,  Ding,  H.,  Xiong,  X.,  Nazarians-Armavil,  A., et al. (2011) Genetic interaction maps in Escherichia coli reveal functional crosstalk among cell envelope biogenesis pathways. PLoS Genet. 7, e1002377.
•  Bakal,  C.,  Linding,  R.,  Llense,  F.,  Heffern,  E.,  Martin-Blanco,  E.,  Pawson,  T., and  Perrimon,  N. (2008) Phosphorylation networks regulating JNK activity in diverse genetic backgrounds. Science 322, 453–456.
•  Baryshnikova,  A.,  Costanzo,  M.,  Kim,  Y.,  Ding,  H.,  Koh,  J.,  Toufighi,  K.,  Youn,  J. Y.,  Ou,  J.,  San Luis,  B. J.,  Bandyopadhyay,  S., et al. (2010) Quantitative analysis of fitness and genetic interactions in yeast on a genome scale. Nat. Methods 7, 1017–1024.
•  Butland,  G.,  Babu,  M.,  Diaz-Mejia,  J. J.,  Bohdana,  F.,  Phanse,  S.,  Gold,  B.,  Yang,  W.,  Li,  J.,  Gagarinova,  A. G.,  Pogoutse,  O., et al. (2008) eSGA: E. coli synthetic genetic array analysis. Nat. Methods 5, 789–795.
•  Byrne,  A. B.,  Weirauch,  M. T.,  Wong,  V.,  Koeva,  M.,  Dixon,  S. J.,  Stuart,  J. M., and  Roy,  P. J. (2007) A global analysis of genetic interactions in Caenorhabditis elegans. J. Biol. 6, 8. PM:17897480.
•  Costanzo,  M.,  Baryshnikova,  A.,  Bellay,  J.,  Kim,  Y.,  Spear,  E. D.,  Sevier,  C. S.,  Ding,  H.,  Koh,  J. L.,  Toufighi,  K.,  Mostafavi,  S., et al. (2010) The genetic landscape of a cell. Science 327, 425–431.
•  Datsenko,  K. A., and  Wanner,  B. L. (2000) One-step inactivation of chromosomal genes in Escherichia coli K-12 using PCR products. Proc. Natl. Acad. Sci. USA 97, 6640–6645.
•  Davierwala,  A. P.,  Haynes,  J.,  Li,  Z.,  Brost,  R. L.,  Robinson,  M. D.,  Yu,  L.,  Mnaimneh,  S.,  Ding,  H.,  Zhu,  H.,  Chen,  Y., et al. (2005) The synthetic genetic interaction spectrum of essential genes. Nat. Genet. 37, 1147–1152.
•  Dekel,  E., and  Alon,  U. (2005) Optimality and evolutionary tuning of the expression level of a protein. Nature 436, 588–592.
•  Dixon,  S. J.,  Costanzo,  M.,  Baryshnikova,  A.,  Andrews,  B., and  Boone,  C. (2009) Systematic mapping of genetic interaction networks. Annu. Rev. Genet. 43, 601–625.
•  Haldimann,  A., and  Wanner,  B. L. (2001) Conditional-replication, integration, excision, and retrieval plasmid-host systems for gene structure-function studies of bacteria. J. Bacteriol. 183, 6384–6393.
•  Huang,  D. W.,  Sherman,  B. T., and  Lempicki,  R. A. (2009a) Systematic and integrative analysis of large gene lists using DAVID Bioinformatics Resources. Nature Protoc. 4, 44–57.
•  Huang,  D. W.,  Sherman,  B. T., and  Lempicki,  R. A. (2009b) Bioinformatics enrichment tools: paths toward the comprehensive functional analysis of large gene lists. Nucleic Acids Res. 37, 1–13.
•  Jeong,  H.,  Mason,  S. P.,  Barabasi,  A. L., and  Oltvai,  Z. N. (2001) Lethality and centrality in protein networks. Nature 411, 41–42.
•  Kawasaki,  Y.,  Wada,  C., and  Yura,  T. (1990) Roles of Escherichia coli heat shock proteins DnaK, DnaJ and GrpE in mini-F plasmid replication. Mol. Gen. Genet. 220, 277–282.
•  Kitagawa,  M.,  Ara,  T.,  Arifuzzaman,  M.,  Ioka-Nakamichi,  T.,  Inamoto,  E.,  Toyonaga,  H., and  Mori,  H. (2005) Complete set of ORF clones of Escherichia coli ASKA library (a complete set of E. coli K-12 ORF archive): unique resources for biological research. DNA Res. 12, 291–299.
•  Mani,  R.,  St. Onge,  R. P.,  Hartman,  J. L.,  Giaever,  G., and  Roth,  F. P. (2008) Defining genetic interaction. Proc. Natl. Acad. Sci. USA 105, 3461–3466.
•  Michaut,  M., and  Bader,  G. D. (2012) Multiple genetic interaction experiments provide complementary information useful for gene function prediction. PLoS Comput. Biol. 8, e1002559.
•  Mori,  H.,  Kondo,  A.,  Ohshima,  A.,  Ogura,  T., and  Hiraga,  S. (1986) Structure and function of the F plasmid genes essential for partitioning. J. Mol. Biol. 192, 1–15.
•  Nakayashiki,  T.,  Wanner,  B. L., and  Mori,  H. (2013) Genome-wide screening with hydroxyurea reveals a link between nonessential ribosomal proteins and reactive oxygen species production. J. Bacteriol. 195, 1226–1235.
•  Ogura,  T., and  Hiraga,  S. (1983) Partition mechanism of F plasmid: two plasmid gene-encoded products and a cis-acting region are involved in partition. Cell 32, 351–360.
•  Pan,  X.,  Yuan,  D. S.,  Ooi,  S. L.,  Wang,  X.,  Sookhai-Mahadeo,  S.,  Meluh,  P., and  Boeke,  J. D. (2007) dSLAM analysis of genome-wide genetic interactions in Saccharomyces cerevisiae. Methods 41, 206–221.
•  Roguev,  A.,  Bandyopadhyay,  S.,  Zofall,  M.,  Zhang,  K.,  Fischer,  T.,  Collins,  S. R.,  Qu,  H.,  Shales,  M.,  Park,  H.-O.,  Hayles,  J., et al. (2008) Conservation and rewiring of functional modules revealed by an epistasis map in fission yeast. Science 322, 405–410.
•  Schuldiner,  M.,  Collins,  S. R.,  Thompson,  N. J.,  Denic,  V.,  Bhamidipati,  A.,  Punna,  T.,  Ihmels,  J.,  Andrews,  B.,  Boone,  C.,  Greenblatt,  J. F.,  Weissman,  J. S., and  Krogan,  N. J. (2005) Exploration of the function and organization of the yeast early secretory pathway through an epistatic miniarray profile. Cell 123, 507–519.
•  Tong,  A. H.,  Evangelista,  M.,  Parsons,  A. B.,  Xu,  H.,  Bader,  G. D.,  Page,  N.,  Robinson,  M.,  Raghibizadeh,  S.,  Hogue,  C. W.,  Bussey,  H.,  Andrews,  B.,  Tyers,  M., and  Boone,  C. (2001) Systematic genetic analysis with ordered arrays of yeast deletion mutants. Science 294, 2364–2368.
•  Tong,  A. H.,  Lesage,  G.,  Bader,  G. D.,  Ding,  H.,  Xu,  H.,  Xin,  X.,  Young,  J.,  Berriz,  G. F.,  Brost,  R. L.,  Chang,  M., et al. (2004) Global mapping of the yeast genetic interaction network. Science 303, 808–813.
•  Typas,  A.,  Nichols,  R. J.,  Siegele,  D. A.,  Shales,  M.,  Collins,  S. R.,  Lim,  B.,  Braberg,  H.,  Yamamoto,  N.,  Takeuchi,  R.,  Wanner,  B. L., et al. (2008) High-throughput, quantitative analyses of genetic interactions in E. coli. Nat. Methods 5, 781–787.
•  Uga,  H.,  Matsunaga,  F., and  Wada,  C. (1999) Regulation of DNA replication by iterons: an interaction between the ori2 and incC regions mediated by RepE-bound iterons inhibits DNA replication of mini-F plasmid in Escherichia coli. EMBO J. 18, 3856–3867.
•  Yamamoto,  N.,  Nakahigashi,  K.,  Nakamichi,  T.,  Yoshino,  M.,  Takai,  Y.,  Touda,  Y.,  Furubayashi,  A.,  Kinjyo,  S.,  Dose,  H.,  Hasegawa,  M., et al. (2009) Update on the Keio collection of Escherichia coli single-gene deletion mutants. Mol. Syst. Biol. 5, 335.