Leaner and meaner genomes in Escherichia coli
Genome Biology volume 7, Article number: 237 (2006)
A 'better' Escherichia coli K-12 genome has recently been engineered in which about 15% of the genome has been removed by planned deletions. Comparison with related bacterial genomes that have undergone a natural reduction in size suggests that there is plenty of scope for yet more deletions.
Why should one want to design a better bacterium? One answer is that this is one way of really testing our understanding of how a living cell works - by making predictions, manipulating the object, and seeing what we get. This is the province of synthetic biology, whose ultimate goal is to understand life by constructing it from scratch; it is hoped that along the way will emerge an understanding of the properties of living cells and organisms that is difficult to arrive at by conventional investigation of the organisms themselves [1, 2]. Much progress has been made recently towards designing and synthesizing novel biological organisms from a set of standardized parts , such as protein-coding genes, regulators, and terminators, as listed on the BioBrick website .
In contrast, in work recently published in Science, Posfai et al.  have taken a 'deconstructionist' approach to redesigning life. Specific regions of the Escherichia coli K-12 genome were targeted for deletion with the intention of improving the production potential of this model organism. As an unanticipated side effect, they have come up with a bacterium that is even better than the parental strain for some purposes, in that it is more efficiently electroporated and accurately propagates unstable recombinant genes and plasmids. It is interesting to compare these engineered reduced genomes  with the genomes of other bacteria within the Enterobacteriaceae, some of which are endosymbionts whose genomes have become dramatically reduced during evolution.
Smaller is indeed often better, as people who travel frequently or who worry about buying fuel for their cars know. Posfai et al.  chose which genes and genomic regions to remove on the basis of several criteria, including "troublesome sequences" such as insertion sequence (IS) sites and transposable elements that appear to code only for their own replication, and repeat regions that can cause homologous recombination. They also removed some regions that are not present in all E. coli genomes, and so are unlikely to be essential for basic properties such as growth. There are many large regions throughout the E. coli K-12 genome that are not conserved among other E. coli genomes, but given the variation in genome size between different strains, with differences of more than 1 million base pairs (20% of the genome) being common, this is perhaps not surprising.
To make the deletions, synthetic oligomers containing regions homologous to target sites flanking the desired region were used. Regions were deleted by recombination mediated by the phage lambda Red system, and done in a way that gave 'scarless' deletions where no marker sequence was left. The strains with deletions were then tested for growth in minimal media. Finally, as one last step to check for quality, the reduced strains were hybridized to tiling microarrays of the E. coli K-12 genome. The first reduced strain yielded surprising results. In the words of the authors: "Alarmingly, we found five unexpected copies of IS that had transposed to new locations since the project began in 2002." Thus new strains were made, which were shown to be free of IS elements. The engineered strains had similar growth rates to their parent strain, and the electroporation efficiency of engineered strain MDS42 was 100 times greater than for the original E. coli MG1655 K-12. Furthermore, plasmid genes that were unstable in MG1655 were found to be completely stable in the engineered strains. IS mutagenesis is a natural defense against deleterious genes, and is normally helpful to bacteria in the wild, but is detrimental when one wishes to grow these genes in laboratory strains of E. coli.
Natural genome reduction
If synthetic biology can be used to design a reduced E. coli genome with some desirable new functions, what about 'non-synthetic' biology - that is, evolution? Is there anything that we can learn from evolutionary biology about how to make a reduced E. coli genome? Soon after Posfai et al.  published their paper, a study by Wu et al.  appeared on the reduced genomes of two very different bacteria living inside an insect called the glassy-winged sharpshooter (Homalodisca coagulata) . One of these bacteria (Baumannia cicadellinicola) belongs phylogenetically to the same group as E. coli, which inspired me to make a comparison of all of the engineered E. coli genomes of Posfai et al.  with other enteric bacterial genomes sequenced so far (Table 1).
The B. cicadellinicola genome is towards the bottom of the table, but there are four known genomes in this family that encode an even smaller number of proteins. The genome at the bottom of the list (Buchnera aphidicola strain BBp) codes for only 504 proteins, or less than 10% of the number of proteins encoded by the larger E. coli genomes (5,379 proteins in E. coli CFT073). The smallest 'normal', free-living enterobacter (apart from the newly engineered E. coli genomes) is a Yersinia pestis strain (Mediaevalis), with 3,895 genes, or only 72% of the number of genes found in the largest enterobacterial genome. Furthermore, only slightly more than half of the Y. pestis Mediaevalis genes have homologs in the CFT073 genome (52% - that is, 2,938 Y. pestis genes/5,379 E. coli CFT073 genes). Thus, just on the basis of gene diversity within the enteric bacteria, it seems that perhaps half or more of the genes in the larger enterobacterial genomes might be expendable - at least under laboratory growth conditions. Indeed, only 620 E. coli K-12 genes have been found experimentally to be essential for growth in rich media, while 3,126 genes were found to be dispensable for growth under this condition . This indicates that there could well be more room for engineered deletions in the E. coli genomes.
Figure 1 compares the proteins encoded by some of the genomes in Table 1. E. coli K-12 isolate MG1655 was the genome used as a starting point by Posfai et al.  to make the reduced genomes, and it is used as the reference here. The locations of all the deletions engineered by Posfai et al.  are indicated outside the circles. The location of each gene from the E. coli MG1655 genome is used to visualize the best hit based on a BLAST search of all the proteins encoded by the different genomes indicated in the legend. A strong hit is represented by a solid bar so the extensive solid regions in the circles represent regions of homology with the other genomes. Each circle represents a different genome, with the outer five circles representing nonreduced genomes, and the inner three circles reduced genomes. Thus, the outermost circle is for the Y. pestis genome, followed by a Salmonella genome, and then three E. coli genomes. Note that many of the planned deletions made by Posfai et al.  correspond to gaps in the otherwise mostly solid dark circles; these gaps represent large chromosomal regions lacking homology in terms of protein sequences in the other genomes.
The reduced genomes are shown in the inner circles. B. aphidicola strain BBp is the smallest genome, and is depicted as the orange inner circle, which has few hits, as expected, as this genome encodes so few proteins. The B. cicadellinicola genome is the next circle (red), and the third is Sodalis glossinidius, which is a genome that is undergoing reduction, but still contains about 2,500 genes, as well as about 1,000 pseudogenes . This circle contains more hits, although it is still a bit sparse compared to the inner three circles, which have large regions where nearly all of the proteins are conserved.
These reduced genomes contain only about 10% of the genes in the larger E. coli genomes from which they originated long ago. This raises many questions. What about the remaining 90%? Does E. coli really not need most of these genes? Some are certainly redundant - a necessary condition for robust systems  - and the definition of 'essential genes' might include some genes that do not give a lethal phenotype when deleted .
Is it possible to model which genes would remain, and which 90% or so could be removed, under the right conditions? A model of E. coli metabolism was recently used to generate reduced genomes in silico , and to compare these genomes with the endosymbiotic genomes shown at the bottom of Table 1. The idea was to use a known metabolic environment and then to model random gene loss, and evaluate relative viability. If the gene loss had no apparent effect, then another gene would be removed, and this process was repeated until a minimal genome was obtained.
Two different endosymbiotic bacterial environments were modeled - those of Buchnera and Wiggelsworthia - and the model predicted the gene content of the two genomes to about 80% accuracy .
There are, of course, several different ways to arrive at the same reduced genome, but by looking at which genes are necessary to perform core metabolic activities (for a given endosymbiotic environment, it should be stressed), it is possible in general to predict the genes that are likely to remain in a reduced genome. This information can then be used in future experiments to design better genomes, tailor-made for specific applications. Posfai et al.  were not intending to manufacture a 'minimal genome' such as the highly reduced ones discussed here, but rather they simply wanted to engineer an E. coli genome that would be a better 'workhorse' - that is, it would be easier to get DNA into the cells, and the DNA and its gene products would be stable once it was there. There are others, however, who do have the aim of using synthetic biology to design and manufacture a minimal genome . Perhaps the time is near when mircobiology will join the engineering sciences.
Benner SA, Sismour AM: Synthetic biology. Nat Rev Genet. 2005, 6: 533-543. 10.1038/nrg1637.
de Lorenzo V, Serrano L, Valencia A: Synthetic biology: challenges ahead. Bioinformatics. 2006, 22: 127-128. 10.1093/bioinformatics/btk018.
Fu P: A perspective of synthetic biology: assembling building blocks for novel functions. Biotechnol J. 2006, 1: 690-699. 10.1002/biot.200600019.
The BioBricks Foundation. [http://openwetware.org/wiki/The_BioBricks_Foundation]
Posfai G, Plunkett G, Feher T, Frisch D, Keil GM, Umenhoffer K, Kolisnychenko V, Stahl B, Sharma SS, de Arruda M, et al: Emergent properties of reduced-genome Escherichia coli. Science. 2006, 312: 1044-1046. 10.1126/science.1126439.
Wu D, Daugherty SC, Aken SE, Pai GH, Watkins KL, Khouri H, Tallon LJ, Zaborsky JM, Dunbar HE, Tran PL, et al: Metabolic complementarity and genomics of the dual bacterial symbiosis of sharpshooters. PLoS Biol. 2006, 6: e188-. 10.1371/journal.pbio.0040188.
Gerdes SY, Scholle MD, Campbell JW, Balazsi G, Ravasz E, Daugherty MD, Somera AL, Kyrpides NC, Anderson I, Gelfand MS, et al: Experimental determination and system level analysis of essential genes in Escherichia coli MG1655. J Bacteriol. 2003, 185: 5673-5684. 10.1128/JB.185.19.5673-5684.2003.
Toh H, Weiss BL, Perkin SAH, Yamashita A, Oshima K, Hattori M, Aksoy S: Massive genome erosion and functional adaptations provide insights into the symbiotic lifestyle of Sodalis glossinidius in the tsetse host. Genome Res. 2006, 16: 149-156. 10.1101/gr.4106106.
Kafri R, Levy M, Pilpel Y: The regulatory utilization of genetic redundancy through responsive backup circuits. Proc Natl Acad Sci USA. 2006, 103: 11653-11658. 10.1073/pnas.0604883103.
Fang G, Rocha E, Danchin A: How essential are nonessential genes?. Mol Biol Evol. 2005, 22: 2147-2156. 10.1093/molbev/msi211.
Pal C, Papp B, Lercher MJ, Csermely P, Oliver SG, Hurst LD: Chance and necessity in the evolution of minimal metabolic networks. Nature. 2006, 440: 667-670. 10.1038/nature04568.
Smith HO, Hutchison CA, Pfannkoch C, Venter JC: Generating a synthetic genome by whole genome assembly: phiX174 bacteriophage from synthetic oligonucleotides. Proc Natl Acad Sci USA. 2003, 100: 15440-15445. 10.1073/pnas.2237126100.
CBS Genome Atlas Database. [http://www.cbs.dtu.dk/services/GenomeAtlas]
I would like to thank Michael Sismour for useful comments about synthetic biology and the Danish Center for Scientific Computing for funding.
Authors’ original submitted files for images
Below are the links to the authors’ original submitted files for images.