56 4 Rational Efforts to Streamline the Escherichia coli Genome
(“genes that have not been shown to be nonessential”; http://ecoliwiki.net/
colipedia/index.php/Essential_genes 28 May 2013) (Figure 4.1) should obviously
be retained in the streamlining process.4.5.1.3 Comparative Genomics
Genome comparisons of related strains are highly informative and probably give
the best clues as to what to delete. Natural selection, within the genus, suppos-
edly conserved the basic set of genes, collectively called as the core genome,
which are needed for robust performance [33]. Interspersed, horizontally
acquired genomic islands carrying niche-specific and parasitic genes are obvious
choices for removal (Figure 4.1). Although non-orthologous gene displacement
might obscure shared functions [46], genome comparisons of more distantly
related species could also help finding deletion targets. For example, Buchnera
sp. is thought to be a naturally minimized version of E. coli, sharing a common
ancestor before switching to a symbiotic lifestyle. The 0.64 Mbp genome of
Buchnera could serve to identify genes common with E. coli and probably being
important for growth. Genes unique to E. coli could then be used as a smaller
pool to identify deletion candidates by other methods [47].4.5.1.4 In silico Models
Genome-scale metabolic network reconstructions coupled with constraint-
based modeling can contribute to rational strain design by predicting gene essen-
tiality and phenotypic consequences of gene deletions in microbes. Although
these large-scale computational models continue to be expanded and updated,
their predictive power to quantitatively assess cellular phenotypes in streamlin-
ing studies is still limited [48]. In particular, these models often fail to identify
groups of metabolic genes that are individually dispensable, but jointly essential.
The most widely used E. coli reconstruction, while covering 1366 metabolic
genes, still contains only a subset of the full gene complement of the cell [49].
Integration of other cellular systems (e.g., the machineries for replication, tran-
scription, translation, posttranslational modifications) and regulatory processes
is needed to more accurately compute complex cellular phenotypes [50, 51]. In
addition, there are still too many unknown gene functions to accurately build an
in silico interaction network that covers all key cellular processes.4.5.1.5 Architectural Studies
Genome streamlining does not equal simply minimizing the gene set. The mini-
mal set of genetic information necessary to sustain a functioning cell might con-
tain positional information as well: not only trans-acting genes but also cis-acting
chromosomal regions might be essential. In a comprehensive study [52], the
entire chromosome was scanned for cis-acting regions. Essential genes were
deleted from the chromosome in the presence of complementing plasmids car-
rying the particular gene. Surprisingly, the replication origin was found to be the
only essential cis-acting region. Other, reportedly cis-acting regions, like dif
(participating in resolution of replicated sister chromosomes) or migS (responsi-
ble for the polar movement of oriC) proved to be nonessential, and removal of
them caused only minor growth defects.