Bacterial persisters are a stochastically formed subpopulation of low-energy cells.

Persisters represent a small subpopulation of non- or slow-growing bacterial cells that are tolerant to killing by antibiotics. Despite their prominent role in the recalcitrance of chronic infections to antibiotic therapy, the mechanism of their formation has remained elusive. We show that sorted cells of Escherichia coli with low levels of energy-generating enzymes are better able to survive antibiotic killing. Using microfluidics time-lapse microscopy and a fluorescent reporter for in vivo ATP measurements, we find that a subpopulation of cells with a low level of ATP survives killing by ampicillin. We propose that these low ATP cells are formed stochastically as a result of fluctuations in the abundance of energy-generating components. These findings point to a general "low energy" mechanism of persister formation.

Extensive impact of non-antibiotic drugs on human gut bacteria.

A few commonly used non-antibiotic drugs have recently been associated with changes in gut microbiome composition, but the extent of this phenomenon is unknown. Here, we screened more than 1,000 marketed drugs against 40 representative gut bacterial strains, and found that 24% of the drugs with human targets, including members of all therapeutic classes, inhibited the growth of at least one strain in vitro. Particular classes, such as the chemically diverse antipsychotics, were overrepresented in this group. The effects of human-targeted drugs on gut bacteria are reflected on their antibiotic-like side effects in humans and are concordant with existing human cohort studies. Susceptibility to antibiotics and human-targeted drugs correlates across bacterial species, suggesting common resistance mechanisms, which we verified for some drugs. The potential risk of non-antibiotics promoting antibiotic resistance warrants further exploration. Our results provide a resource for future research on drug-microbiome interactions, opening new paths for side effect control and drug repurposing, and broadening our view of antibiotic resistance.

Cell Lysis Directed by SulA in Response to DNA Damage in Escherichia coli.

The SOS response is induced upon DNA damage and the inhibition of Z ring formation by the product of the sulA gene, which is one of the LexA-regulated genes, allows time for repair of damaged DNA. On the other hand, severely DNA-damaged cells are eliminated from cell populations. Overexpression of sulA leads to cell lysis, suggesting SulA eliminates cells with unrepaired damaged DNA. Transcriptome analysis revealed that overexpression of sulA leads to up-regulation of numerous genes, including soxS. Deletion of soxS markedly reduced the extent of cell lysis by sulA overexpression and soxS overexpression alone led to cell lysis. Further experiments on the SoxS regulon suggested that LpxC is a main player downstream from SoxS. These findings suggested the SulA-dependent cell lysis (SDCL) cascade as follows: SulA→SoxS→LpxC. Other tests showed that the SDCL cascade pathway does not overlap with the apoptosis-like and mazEF cell death pathways.

GenoBase: comprehensive resource database of Escherichia coli K-12.

Comprehensive experimental resources, such as ORFeome clone libraries and deletion mutant collections, are fundamental tools for elucidation of gene function. Data sets by omics analysis using these resources provide key information for functional analysis, modeling and simulation both in individual and systematic approaches. With the long-term goal of complete understanding of a cell, we have over the past decade created a variety of clone and mutant sets for functional genomics studies of Escherichia coli K-12. We have made these experimental resources freely available to the academic community worldwide. Accordingly, these resources have now been used in numerous investigations of a multitude of cell processes. Quality control is extremely important for evaluating results generated by these resources. Because the annotation has been changed since 2005, which we originally used for the construction, we have updated these genomic resources accordingly. Here, we describe GenoBase (http://ecoli.naist.jp/GB/), which contains key information about comprehensive experimental resources of E. coli K-12, their quality control and several omics data sets generated using these resources.

Toward Network Biology in E. coli Cell.

E. coli has been a critically important model research organism for more than 50 years, particularly in molecular biology. In 1997, the E. coli draft genome sequence was published. Post-genomic techniques and resources were then developed that allowed E. coli to become a model organism for systems biology. Progress made since publication of the E. coli genome sequence will be summarized.

Effect of codon adaptation on codon-level and gene-level translation efficiency in vivo.

BACKGROUND: There is a significant difference between synonymous codon usage in many organisms, and it is known that codons used more frequently generally showed efficient decoding rate. At the gene level, however, there are conflicting reports on the existence of a correlation between codon adaptation and translation efficiency, even in the same organism. RESULTS: To resolve this issue, we cultured Escherichia coli under conditions designed to maintain constant levels of mRNA and protein and subjected the cells to ribosome profiling (RP) and mRNA-seq analyses. We showed that the RP results correlated more closely with protein levels generated under similar culture conditions than with the mRNA abundance from the mRNA-seq. Our result indicated that RP/mRNA ratio could be used as a measure of translation efficiency at gene level. On the other hand, the RP data showed that codon-specific ribosome density at the decoding site negatively correlated with codon usage, consistent with the hypothesis that preferred codons display lower ribosome densities due to their faster decoding rate. However, highly codon-adapted genes showed higher ribosome densities at the gene level, indicating that the efficiency of translation initiation, rather than higher elongation efficiency of preferred codons, exerted a greater effect on ribosome density and thus translation efficiency. CONCLUSIONS: These findings indicate that evolutionary pressure on highly expressed genes influenced both codon bias and translation initiation efficiency and therefore explains contradictory findings that codon usage bias correlates with translation efficiency of native genes, but not with the artificially created gene pool, which was not subjected to evolution pressure.

Colony-live--a high-throughput method for measuring microbial colony growth kinetics--reveals diverse growth effects of gene knockouts in Escherichia coli.

BACKGROUND: Precise quantitative growth measurements and detection of small growth changes in high-throughput manner is essential for fundamental studies of bacterial cell. However, an inherent tradeoff for measurement quality in high-throughput methods sacrifices some measurement quality. A key challenge has been how to enhance measurement quality without sacrificing throughput. RESULTS: We developed a new high-throughput measurement system, termed Colony-live. Here we show that Colony-live provides accurate measurement of three growth values (lag time of growth (LTG), maximum growth rate (MGR), and saturation point growth (SPG)) by visualizing colony growth over time. By using a new normalization method for colony growth, Colony-live gives more precise and accurate growth values than the conventional method. We demonstrated the utility of Colony-live by measuring growth values for the entire Keio collection of Escherichia coli single-gene knockout mutants. By using Colony-live, we were able to identify subtle growth defects of single-gene knockout mutants that were undetectable by the conventional method quantified by fixed time-point camera imaging. Further, Colony-live can reveal genes that influence the length of the lag-phase and the saturation point of growth. CONCLUSIONS: Measurement quality is critical to achieving the resolution required to identify unique phenotypes among a diverse range of phenotypes. Sharing high-quality genome-wide datasets should benefit many researchers who are interested in specific gene functions or the architecture of cellular systems. Our Colony-live system provides a new powerful tool to accelerate accumulation of knowledge of microbial growth phenotypes.

GlgS, described previously as a glycogen synthesis control protein, negatively regulates motility and biofilm formation in Escherichia coli.

Escherichia coli glycogen metabolism involves the regulation of glgBXCAP operon expression and allosteric control of the GlgC [ADPG (ADP-glucose) pyrophosphorylase]-mediated catalysis of ATP and G1P (glucose-1-phosphate) to ADPG linked to glycogen biosynthesis. E. coli glycogen metabolism is also affected by glgS. Though the precise function of the protein it encodes is unknown, its deficiency causes both reduced glycogen content and enhanced levels of the GlgC-negative allosteric regulator AMP. The transcriptomic analyses carried out in the present study revealed that, compared with their isogenic BW25113 wild-type strain, glgS-null (ΔglgS) mutants have increased expression of the operons involved in the synthesis of type 1 fimbriae adhesins, flagella and nucleotides. In agreement, ΔglgS cells were hyperflagellated and hyperfimbriated, and displayed elevated swarming motility; these phenotypes all reverted to the wild-type by ectopic glgS expression. Also, ΔglgS cells accumulated high colanic acid content and displayed increased ability to form biofilms on polystyrene surfaces. F-driven conjugation based on large-scale interaction studies of glgS with all the non-essential genes of E. coli showed that deletion of purine biosynthesis genes complement the glycogen-deficient, high motility and high biofilm content phenotypes of ΔglgS cells. Overall the results of the present study indicate that glycogen deficiency in ΔglgS cells can be ascribed to high flagellar propulsion and high exopolysaccharide and purine nucleotides biosynthetic activities competing with GlgC for the same ATP and G1P pools. Supporting this proposal, glycogen-less ΔglgC cells displayed an elevated swarming motility, and accumulated high levels of colanic acid and biofilm. Furthermore, glgC overexpression reverted the glycogen-deficient, high swarming motility, high colanic acid and high biofilm content phenotypes of ΔglgS cells to the wild-type. As on the basis of the present study GlgS has emerged as a major determinant of E. coli surface composition and because its effect on glycogen metabolism appears to be only indirect, we propose to rename it as ScoR (surface composition regulator).

Genetically stable kill-switch using "demon and angel" expression construct of essential genes.

Genetic instability of synthetic genetic devices is a key obstacle for practical use. This problem is particularly critical in kill-switches for conditional host killing. Here, we propose a genetically stable kill-switch based on a "demon and angel" expression construct of a toxic essential gene. The kill-switch conditionally overexpresses the toxic essential gene. Additionally, the identical essential gene is deleted in the genome. The essential gene is expressed at a low level to maintain host survival in the OFF state and kills the host by the overexpression in the ON state. The single expression construct is responsible for both killing the hosts and maintaining viability, reducing the emergence of loss-of-function mutants. We constructed the kill-switch using the toxic essential gene encoding tyrosyl-tRNA synthetase, tyrS, in Escherichia coli. The bacteria harboring the kill-switch were conditionally suicidal over 300 generations. Toxic overexpression of essential genes has also been found in other organisms, suggesting that the "demon and angel" kill switch is scalable to various organisms.

Genetic mutation in Escherichia coli genome during adaptation to the murine intestine is optimized for the host diet.

Mammalian gut microbes colonize the intestinal tract of their host and adapt to establish a microbial ecosystem. The host diet changes the nutrient profile of the intestine and has a high impact on microbiota composition. Genetic mutations in Escherichia coli, a prevalent species in the human gut, allow for adaptation to the mammalian intestine, as reported in previous studies. However, the extent of colonization fitness in the intestine elevated by genetic mutation and the effects of diet change on these mutations in E. coli are still poorly known. Here, we show that notable mutations in sugar metabolism-related genes (gatC, araC, and malI) were detected in the E. coli K-12 genome just 2 weeks after colonization in the germ-free mouse intestine. In addition to elevated fitness by deletion of gatC, as previously reported, deletion of araC and malI also elevated E. coli fitness in the murine intestine in a host diet-dependent manner. In vitro cultures of medium containing nutrients abundant in the intestine (e.g., galactose, N-acetylglucosamine, and asparagine) also showed increased E. coli fitness after deletion of the genes-of-interest associated with their metabolism. Furthermore, the host diet was found to influence the developmental trajectory of gene mutations in E. coli. Taken together, we suggest that genetic mutations in E. coli are selected in response to the intestinal environment, which facilitates efficient utilization of nutrients abundant in the intestine under laboratory conditions. Our study offers some insight into the possible adaptation mechanisms of gut microbes.IMPORTANCEThe gut microbiota is closely associated with human health and is greatly impacted by the host diet. Bacteria such as Escherichia coli live in the gut all throughout the life of a human host and adapt to the intestinal environment. Adaptive mutations in E. coli are reported to enhance fitness in the mammalian intestine, but to what extent is still poorly known. It is also unknown whether the host diet affects what genes are mutated and to what extent fitness is affected. This study suggests that genetic mutations in the E. coli K-12 strain are selected in response to the intestinal environment and facilitate efficient utilization of abundant nutrients in the germ-free mouse intestine. Our study provides a better understanding of these intestinal adaptation mechanisms of gut microbes.

Genome-wide screening with hydroxyurea reveals a link between nonessential ribosomal proteins and reactive oxygen species production.

We performed a screening of hydroxyurea (HU)-sensitive mutants using a single-gene-deletion mutant collection in Escherichia coli. HU inhibits ribonucleotide reductase (RNR), which leads to arrest of the replication fork. Surprisingly, the wild-type was less resistant to HU than the average for the Keio Collection. Respiration-defective mutants were significantly more resistant to HU, suggesting that the generation of reactive oxygen species (ROS) contributes to cell death. High-throughput screening revealed that 15 mutants were completely sensitive on plates containing 7.5 mM HU. Unexpectedly, translation-related mutants based on COG categorization were the most enriched, and three of them were deletion mutants of nonessential ribosomal proteins (L1, L32, and L36). We found that, in these mutants, an increased membrane stress response was provoked, resulting in increased ROS generation. The addition of OH radical scavenger thiourea rescued the HU sensitivity of these mutants, suggesting that ROS generation is the direct cause of cell death. Conversely, both the deletion of rpsF and the deletion of rimK, which encode S6 and S6 modification enzymes, respectively, showed an HU-resistant phenotype. These mutants increased the copy number of the p15A-based plasmid and exhibited reduced basal levels of SOS response. The data suggest that nonessential proteins indirectly affect the DNA-damaging process.

The tRNA thiolation pathway modulates the intracellular redox state in Escherichia coli.

We have performed a screening of hydroxyurea (HU)-sensitive mutants using a single-gene-deletion mutant collection in Escherichia coli K-12. HU inhibits ribonucleotide reductase (RNR), an enzyme that catalyzes the formation of deoxyribonucleotides. Unexpectedly, seven of the mutants lacked genes that are required for the incorporation of sulfur into a specific tRNA modification base, 5-methylaminomethyl-2-thiouridine (mnm(5)s(2)U), via persulfide relay. We found that the expression of RNR in the mutants was reduced to about one-third both in the absence and presence of HU, while sufficient deoxynucleoside triphosphate (dNTP) was maintained in the mutants in the absence of HU but a shortage occurred in the presence of HU. Trans-supply of an RNR R2 subunit rescued the HU sensitivity of these mutants. The mutants showed high intracellular ATP/ADP ratios, and overexpression of Hda, which catalyzes the conversion of DnaA-ATP to DnaA-ADP, rescued the HU sensitivity of the mutants, suggesting that DnaA-ATP represses RNR expression. The high intracellular ATP/ADP ratios were due to high respiration activity in the mutants. Our data suggested that intracellular redox was inclined toward the reduced state in these mutants, which may explain a change in RNR activity by reduction of the catalytically formed disulfide bond and high respiration activity by the NADH reducing potential. The relation between persulfide relay and intracellular redox is discussed.

Glycogen is the primary source of glucose during the lag phase of E. coli proliferation.

In the studies of Escherichia coli (E. coli), metabolomics analyses have mainly been performed using steady state culture. However, to analyze the dynamic changes in cellular metabolism, we performed a profiling of concentration of metabolites by using batch culture. As a first step, we focused on glucose uptake and the behavior of the first metabolite, G6P (glucose-6-phosphate). A computational formula was derived to express the glucose uptake rate by a single cell from two kinds of experimental data, extracellular glucose concentration and cell growth, being simulated by Cell Illustrator. In addition, average concentration of G6P has been measured by CE-MS. The existence of another carbon source was suggested from the computational result. After careful comparison between cell growth, G6P concentration, and the computationally obtained curve of glucose uptake rate, we predicted the consumption of glycogen in lag phase and its accumulation as an energy source in an E. coli cell for the next proliferation. We confirmed our prediction experimentally. This behavior indicates the importance of glycogen participation in the lag phase for the growth of E. coli. This article is part of a Special Issue entitled: Computational Methods for Protein Interaction and Structural Prediction.

Functional analysis of Rickettsia monacensis strain humboldt folA dihydrofolate reductase gene via complementation assay.

Nutritive symbiosis between bacteria and ticks is observed across a range of ecological contexts; however, little characterization on the molecular components responsible for this symbiosis has been done. Previous studies in our lab demonstrated that Rickettsia monacensis str. Humboldt (strain Humboldt) can synthesize folate de novo via the folate biosynthesis pathway involving folA, folC, folE, folKP, and ptpS genes. In this study, expression of the strain Humboldt folA gene within a folA mutant Escherichia coli construct was used to functionally characterize the strain Humboldt folA folate gene in vivo. The strain Humboldt folA folate gene was subcloned into a TransBac vector and transformed into a folA mutant E. coli construct. The mutant containing strain Humboldt folA subclone and a pFE604 clone of the knocked-out folA gene was cured of pFE604. Curing of the folA mutant E. coli construct was successful using acridine orange and 43.5 °C incubation temperature. The plasmid curing assay showed curing efficiency of the folA mutant at 100%. Functional complementation was assessed by growth phenotype on minimal media with and without IPTG between strain Humboldt folA and E. coli folA. Large and homogenous wild-type colony growth was observed for both strain Humboldt and E. coli folA on minimal media with 0.1 mM IPTG, wild-type growth for strain Humboldt folA and pin-point growth for E. coli folA on 0.01 mM IPTG, and pin-point growth without IPTG for both strain Humboldt and E. coli folA. This study provides evidence substantiating the in vivo functionality of strain Humboldt folA in producing functional gene products for folate biosynthesis.

Investigating the effects of perturbations to pgi and eno gene expression on central carbon metabolism in Escherichia coli using (13)C metabolic flux analysis.

BACKGROUND: It has long been recognized that analyzing the behaviour of the complex intracellular biological networks is important for breeding industrially useful microorganisms. However, because of the complexity of these biological networks, it is currently not possible to obtain all the desired microorganisms. In this study, we constructed a system for analyzing the effect of gene expression perturbations on the behavior of biological networks in Escherichia coli. Specifically, we utilized (13)C metabolic flux analysis ((13)C-MFA) to analyze the effect of perturbations to the expression levels of pgi and eno genes encoding phosphoglucose isomerase and enolase, respectively on metabolic fluxes. RESULTS: We constructed gene expression-controllable E. coli strains using a single-copy mini F plasmid. Using the pgi expression-controllable strain, we found that the specific growth rate correlated with the pgi expression level. (13)C-MFA of this strain revealed that the fluxes for the pentose phosphate pathway and Entner-Doudoroff pathway decreased, as the pgi expression lelvel increased. In addition, the glyoxylate shunt became active when the pgi expression level was almost zero. Moreover, the flux for the glyoxylate shunt increased when the pgi expression level decreased, but was significantly reduced in the pgi-knockout cells. Comparatively, eno expression could not be decreased compared to the parent strain, but we found that increased eno expression resulted in a decreased specific growth rate. (13)C-MFA revealed that the metabolic flux distribution was not altered by an increased eno expression level, but the overall metabolic activity of the central metabolism decreased. Furthermore, to evaluate the impact of perturbed expression of pgi and eno genes on changes in metabolic fluxes in E. coli quantitatively, metabolic sensitivity analysis was performed. As a result, the perturbed expression of pgi gene had a great impact to the metabolic flux changes in the branch point between the glycolysis and pentose phosphate pathway, isocitrate dehydrogenase reaction, anaplerotic pathways and Entner-Doudoroff pathway. In contrast, the impact of perturbed eno expression to the flux changes in E. coli metabolic network was small. CONCLUSIONS: Our results indicate that the response of metabolic fluxes to perturbation to pgi expression was different from that to eno expression; perturbations to pgi expression affect the reaction related to the Pgi protein function, the isocitrate dehydrogenase reaction, anaplerotic reactions and Entner-Doudoroff pathway. Meanwhile, eno expression seems to affect the overall metabolic activity, and the impact of perturbed eno expression on metabolic flux change is small. Using the gene expression control system reported here, it is expected that we can analyze the response and adaptation process of complex biological networks to gene expression perturbations.

Past, Present, and Future of Genome Modification in Escherichia coli.

Escherichia coli K-12 is one of the most well-studied species of bacteria. This species, however, is much more difficult to modify by homologous recombination (HR) than other model microorganisms. Research on HR in E. coli has led to a better understanding of the molecular mechanisms of HR, resulting in technical improvements and rapid progress in genome research, and allowing whole-genome mutagenesis and large-scale genome modifications. Developments using λ Red (exo, bet, and gam) and CRISPR-Cas have made E. coli as amenable to genome modification as other model microorganisms, such as Saccharomyces cerevisiae and Bacillus subtilis. This review describes the history of recombination research in E. coli, as well as improvements in techniques for genome modification by HR. This review also describes the results of large-scale genome modification of E. coli using these technologies, including DNA synthesis and assembly. In addition, this article reviews recent advances in genome modification, considers future directions, and describes problems associated with the creation of cells by design.

Deep sequencing reveals as-yet-undiscovered small RNAs in Escherichia coli.

BACKGROUND: In Escherichia coli, approximately 100 regulatory small RNAs (sRNAs) have been identified experimentally and many more have been predicted by various methods. To provide a comprehensive overview of sRNAs, we analysed the low-molecular-weight RNAs (< 200 nt) of E. coli with deep sequencing, because the regulatory RNAs in bacteria are usually 50-200 nt in length. RESULTS: We discovered 229 novel candidate sRNAs (≥ 50 nt) with computational or experimental evidence of transcription initiation. Among them, the expression of seven intergenic sRNAs and three cis-antisense sRNAs was detected by northern blot analysis. Interestingly, five novel sRNAs are expressed from prophage regions and we note that these sRNAs have several specific characteristics. Furthermore, we conducted an evolutionary conservation analysis of the candidate sRNAs and summarised the data among closely related bacterial strains. CONCLUSIONS: This comprehensive screen for E. coli sRNAs using a deep sequencing approach has shown that many as-yet-undiscovered sRNAs are potentially encoded in the E. coli genome. We constructed the Escherichia coli Small RNA Browser (ECSBrowser; http://rna.iab.keio.ac.jp/), which integrates the data for previously identified sRNAs and the novel sRNAs found in this study.

The Escherichia coli K-12 ORFeome: a resource for comparative molecular microbiology.

BACKGROUND: Systems biology and functional genomics require genome-wide datasets and resources. Complete sets of cloned open reading frames (ORFs) have been made for about a dozen bacterial species and allow researchers to express and study complete proteomes in a high-throughput fashion. RESULTS: We have constructed an open reading frame (ORFeome) collection of 3974 or 94% of the known Escherichia coli K-12 ORFs in Gateway entry vector pENTR/Zeo. The collection has been used for protein expression and protein interaction studies. For example, we have compared interactions among YgjD, YjeE and YeaZ proteins in E. coli, Streptococcus pneumoniae, and Staphylococcus aureus. We also compare this ORFeome with other Gateway-compatible bacterial ORFeomes and show its utility for comparative functional genomics. CONCLUSIONS: The E. coli ORFeome provides a useful resource for functional genomics and other areas of protein research in a highly flexible format. Our comparison with other ORFeomes makes comparative analyses straighforward and facilitates direct comparisons of many proteins across many genomes.

Systematic phenome analysis of Escherichia coli multiple-knockout mutants reveals hidden reactions in central carbon metabolism.

Central carbon metabolism is a basic and exhaustively analyzed pathway. However, the intrinsic robustness of the pathway might still conceal uncharacterized reactions. To test this hypothesis, we constructed systematic multiple-knockout mutants involved in central carbon catabolism in Escherichia coli and tested their growth under 12 different nutrient conditions. Differences between in silico predictions and experimental growth indicated that unreported reactions existed within this extensively analyzed metabolic network. These putative reactions were then confirmed by metabolome analysis and in vitro enzymatic assays. Novel reactions regarding the breakdown of sedoheptulose-7-phosphate to erythrose-4-phosphate and dihydroxyacetone phosphate were observed in transaldolase-deficient mutants, without any noticeable changes in gene expression. These reactions, triggered by an accumulation of sedoheptulose-7-phosphate, were catalyzed by the universally conserved glycolytic enzymes ATP-dependent phosphofructokinase and aldolase. The emergence of an alternative pathway not requiring any changes in gene expression, but rather relying on the accumulation of an intermediate metabolite may be a novel mechanism mediating the robustness of these metabolic networks.

Comprehensive analysis of the effects of Escherichia coli ORFs on protein translation reaction.

Protein synthesis is one of the most important reactions in the cell. Recent experimental studies indicated that this complex reaction can be achieved with a minimum complement of 36 proteins and ribosomes by reconstituting an Escherichia coli-based in vitro translation system with these protein components highly purified on an individual basis. From the protein-protein interaction (PPI) network of E. coli proteins, these minimal protein components are known to interact physically with large numbers of proteins. However, it is unclear what fraction of E. coli proteins are linked functionally with the minimal protein synthesis system. We investigated the effects of each of the 4194 E. coli ORF products on the minimal protein synthesis system; at least 12% of the entire ORF products, a significant fraction of the gene product of E. coli, affect the activity of this system. Furthermore 34% of these functional modifiers present in the PPI network were shown by mapping to be directly linked (i.e. to interact physically) with the minimal components of the PPI network. Topological analysis of the relationships between modifiers and the minimal components in the PPI network indicated clustering of the minimal components. The modifiers showed no such clustering, indicating that the location of functional modifiers is spread across the PPI network rather than clustering close to the minimal protein components. These observations may reflect the evolutionary process of the protein synthesis system.