Comparison of the Escherichia coli K-12 genome with sampled genomes of a Klebsiella pneumoniae and three Salmonella enterica serovars, Typhimurium, Typhi and Paratyphi

Abstract

The Escherichia coli K-12 genome (ECO) was compared with the sampled genomes of the sibling species Salmonella enterica serovars Typhimurium, Typhi and Paratyphi A (collectively referred to as SAL) and the genome of the close outgroup Klebsiella pneumoniae (KPN). There are at least 160 locations where sequences of >400 bp are absent from ECO but present in the genomes of all three SAL and 394 locations where sequences are present in ECO but close homologs are absent in all SAL genomes. The 394 sequences in ECO that do not occur in SAL contain 1350 (30.6%) of the 4405 ECO genes. Of these, 1165 are missing from both SAL and KPN. Most of the 1165 genes are concentrated within 28 regions of 10–40 kb, which consist almost exclusively of such genes. Among these regions were six that included previously identified cryptic phage. A hypothetical ancestral state of genomic regions that differ between ECO and SAL can be inferred in some cases by reference to the genome structure in KPN and the more distant relative Yersinia pestis. However, many changes between ECO and SAL are concentrated in regions where all four genera have a different structure. The rate of gene insertion and deletion is sufficiently high in these regions that the ancestral state of the ECO/SAL lineage cannot be inferred from the present data. The sequencing of other closely related genomes, such as S.bongori or Citrobacter, may help in this regard.

INTRODUCTION

Escherichia coli K-12 has been the primary model in bacterial genetics for decades. While E.coli K-12 is not pathogenic, a number of other E.coli strains are pathogens and other closely related species are major pathogens of humans and animals (1, and references therein). Many of the differences between these organisms have been located on ‘pathogenicity islands’. The E.coli K-12 genome sequence has been completed (2) and a series of related pathogen genomes are in the process of being sequenced to completion or have been extensively sample sequenced by our group. Those that are publicly available are listed in Table 1.

Table 1. Divergence of the sampled genomes, relative to E.coli K-12.

Organism	Abbreviation	Completion	Homologs of E.coli K-12 genesa	DNA homology of orthologous ORFsa	Reference
E.coli K-12	ECO	100%	4405 (100%)	100%	(2)
K.pneumoniae	KPN	∼95% sample	2423 (55%)	82%	WUSTL
Y.pestis	YPE	>99%	992 (23%)	73%	Sanger Centre
S.enterica serovars
Typhimurium LT2	STM	>99%	2986 (68%)	84%	WUSTL
Typhi CT18	STY	>99%	2955 (67%)	84%	Sanger Centre
Paratyphi A	SPA	∼95% sample	2872 (65%)	84%	WUSTL

^aSee Materials and Methods.

We have developed methods to visually portray DNA sequence information in a form that allows one completed genome, in this case E.coli K-12 (ECO), to be compared simultaneously with several sampled or completed genomes from related organisms (3). These tools are available for public use at http://globin.cse.psu.edu/ or via the Salmonella genome sequencing project at http://genome.wustl.edu/gsc/bacterial/newlistdisplay.pl.

Comparison of ECO with the other genomes allows the identification of all of the major differences between ECO and these organisms. We present a preliminary comparison of the E.coli K-12 genome with sample sequences of the genomes of Klebsiella pneumoniae (KPN) and the Salmonella enterica serovars Typhimurium, Typhi and Paratyphi A, collectively referred to as SAL. Genes that are present in ECO but absent in Salmonella or Klebsiella have been determined, clusters of ECO-specific genes have been located and the gross features of genetic reorganization have been examined.

MATERIALS AND METHODS

The E.coli K-12 sequence used is that reported by Blattner et al. (2). Sequence data for S.enterica serovars Typhimurium (STM) and Paratyphi A (SPA) and K.pneumoniae (KPN) were obtained from our sequencing project at ftp://genome.wustl.edu/pub/gsc1/sequence/st.louis/bacterial/Salmonella/, those for S.enterica serovar Typhi (STY) from ftp://ftp.sanger.ac.uk/pub/pathogens/st/ and Yersinia pestis (YPE) from ftp://ftp.sanger.ac.uk/pub/pathogens/yp/. Although they are not discussed further in this paper, the genomes for Vibrio cholerae (VCH) (www.tigr.org) and Pseudomonas aeruginosa (PAE) (www.pseudomonas.com) are also included in our comparative views. At the time of analysis two of the genomes, ECO and VCH, were complete, while the other genomes had been sequenced to varying extents. The numbers of melded contigs were: STM, 518; STY, 133; SPA, 887; YPE, 112; VCH, 2; PAE, 1.

Alignments of ECO with each of the other genomes were inspected with a web-based tool Enteric (3). For a user-supplied address in ECO the Enteric server scans pre-computed sets of pairwise alignments between ECO and each of the bacterial genomes and extracts those in the 20 kb region centered at the specified address. The alignments are computed by the program blastz (4). The Enteric server then produces a PDF document containing a hyperlinked, graphical representation of the selected alignments together with annotations that describe insertion, deletion and gene rearrangement events between the compared genomes. This consists of a set of percent identity plots (PIPs), as in Figure 1.

Figure 1.

Enteric alignments of E.coli K-12 with six other genomes. PIP alignments of E.coli with genomes of six other bacteria, in the 20 kb region at the beginning of the E.coli sequence. See Materials and Methods and Florea et al. (3) for a description of how to interpret PIP alignments.

In a PIP, a sequence match is shown as a line parallel to the horizontal axis at its relative position in the first (reference) sequence, positioned along the vertical axis at the coordinate corresponding to the percent identity of the match. Thus, a strongly conserved feature is represented as a horizontal line near the top of the PIP. Enteric produces a stacked array of PIPs, one for each of the comparisons between ECO and another bacterial genome, which show only alignments with 50–100% nucleotide identity.

Annotations of ECO genes are displayed above the PIPs, together with arrows indicating their orientations and with embedded hyperlinks to the corresponding entries in the WIT database (http://wit.IntegratedGenomics.com/IGwit/; 5). Other graphical features include colored rectangles or vertical bars, which denote discontinuities in the PIP alignments. When the cursor is placed over a particular region of the PIP, additional information is disclosed, such as the name of the aligning contig, the size of an inserted region or the ECO position of the adjacent gene in the other organism. More detailed instructions for how to read a PIP are presented in another paper (3). Users can generate their own PIPs using http://genome.wustl.edu/gsc/bacterial/newlistdisplay.pl or http://globin.cse.psu.edu/enterix/.

To measure the proportion of ECO genes present in each of the other genomes we proceed as follows. Each nucleotide in an ECO gene is assigned the highest percentage nucleotide identity found among all the alignments between ECO and the other genomes that include the nucleotide. This number is specified as 0% if no alignments contain that nucleotide. The average of these numbers over all nucleotides in an ECO gene provides a numerical value that assesses the strengths of those alignments for the whole gene. This value is adjusted where an unaligned ECO region is unsequenced in the other species. These gaps are identified as regions where the ends of sequenced contigs in the other species align with ECO (3). A 70% cut-off in this measurment of similarity was used to calculate the proportion of open reading frames (ORFs) shared between ECO and the other genomes in Table 1.

The percent sequence similarity among putative orthologous ORFs was calculated by first identifying ECO ORFs that appeared to have homologs (defined as at least 50% similar) in the SAL, KPN and YPE genomes. Then known transposons and repetitive elements were excluded from this list. Among the 1388 ORFs that remained the percent similarity was determined for the 1000 with the highest overall similiarity to ECO in each genome. This measurement of similarity between the orthologous parts of ECO and the other genomes is given in Table 1.

RESULTS AND DISCUSSION

The genomes will be referred to using the three letter abbreviations described in Table 1. In addition, the abbreviation SAL will be used to indicate features in common among all three Salmonella genomes, STM, STY and SPA. Note that ECO refers to the E.coli K-12 MG1655 genome only and does not include other E.coli genomes.

Over 1000 genes in E.coli K-12 are absent from Salmonella and Klebsiella

The ECO and SAL genetic maps are generally co-linear and quite similar (6). However, evidence of substantial differences in their genomes has mounted (see for example 7). With the completion of the ECO genome and the near completion of some of its relatives, we can now quantify these differences.

The alignment of ECO with the SAL genomes contains many rearrangements or insertion/deletion events. To partially quantify these phenomena, genes that appear to have no close counterpart in the SAL lineage, when compared to ECO, were determined. Only genes that are absent in all three members of subspecies 1 of S.enterica (STM, STY and SPA) were considered. A threshold of 70% DNA sequence identity was used. This cut-off was chosen because these genomes share >80% DNA identity in their closest 1000 putative orthologs (Table 1). Thus, most orthologs should have sequence similarities of >70%. To minimize the chance that an unsequenced portion of a gene would result in an artificially low percent identity, the unsequenced portions of genes were excluded from the calculations (see Materials and Methods). The extent of completion of the sampled genomes (>95%) results in a <100 bp average gap size. If one assumes that particular sequences are not being excluded from the libraries, then, on average, fewer than one gene >500 bp in length would fail to be sampled per genome. However, there are many documented cases where genes are excluded from shotgun libraries. Thus, although it is extremely unlikely that a gene present in all three SAL would be unsampled by chance in all three genomes, it is possible that some genes have been missed due to selection against them in the libraries.

The list of 1350 genes (30.6%) that are present in ECO but lack a homolog with >70% sequence similarity in SAL is available at our web site (http://globin.cse.psu.edu/ftp/dist/Enterix/sim_pct.p70/missing_SAL.p70). The proportion of genes in ECO that appear not to have a close homolog in the three SAL is very similar to our previous estimate of 32%, which was made using a sample of only 7% of the sequence of STY (8). Others predicted 755 genes (18%) as most likely to have been acquired from distant foreign hosts (9). This discrepancy will require further analysis.

1165 ORFs were found in ECO but not in any SAL or in KPN   (http://globin.cse.psu.edu/ftp/dist/Enterix/sim_pct.p70/ missing_SAL_KPN.p70). KPN is a very close outgroup, only slightly more divergent from ECO than the three SAL genomes (see the number of shared genes and the DNA similarity of the best conserved ORFs in Table 1). Only 138 (19%) of the 1165 ORFs have assigned gene names. In contrast, 2155 of 4405 genes (49%) are named in ECO. This difference may indicate that some of the genes found only in ECO have phenotypes in a narrow range of conditions that have not yet been found. For example, many genes in cryptic phage presumably contribute to fitness of ECO only indirectly, if at all. Table 2 lists the subset of ORFs found in ECO, but absent or divergent in SAL or KPN, for which a function has been attributed. Their locations in the ECO genome and brief explanations of their known or putative biological role are also given. Genes with attributed functions that are absent or divergent in all three SAL but are present in KPN are listed in Table 3. These tables are further annotated to show genes that have homologs with 50–70% DNA sequence identities. In some cases these genes with weaker homologies may be paralogs, but in other cases they probably represent genes that have experienced accelerated divergence or gene conversion events. Examples of the latter may include genes associated with proteins expressed on the surface (10), such as the flagella, pilin, fimbrial and lipopolysaccharide biosynthesis genes. These include the fli operon, flgA (2), ecpD (11), the rfa operon (13–15), rfbCX (16) and the fim operon (12). There is a fim operon in at least some Salmonella sp. that is not orthologous to the ECO fim operon. However, the Salmonella operon does occur, unannotated, in the ECO genome (12; J.Parkhill, unpublished results). This observation highlights the fact that genes that are unannotated in the reference genome, in this case ECO, will automatically be missed using the strategy we employed here.

Table 2. Escherichia coli K-12 genes with no close homologs in the K.pneumoniae and three Salmonella genomes^a.

Genesb	Startc	Endc	Putative functiond	References
agaRZeVeWASYeBCDI	3 275 497	3 284 666	N-acetylgalactosaimine (PTS)	(20)
aise	2 363 038	2 363 640	Induced by aluminium	(2)
alpA	2 756 665	2 756 877	Prophage CP4-57 regulatory protein	(32)
ampC	4 375 389	4 376 522	β-Lactamase	(39,40)
appA	1 039 840	1 041 138	Periplasmic phosphoanhydride phosphohydrolase precursor	(41)
appY	582 904	583 653	Transcriptional activator. DLP12 prophage	(42)
aqpZ	914 575	915 270	Water channel gene, aquaporin Z	(43)
aroLe	405 629	406 153	Shikimate kinase II	(2)
arp	4 217 880	4 220 066	Ankyrin-like regulatory protein; regulates acetyl-CoA synthetase	(2)
aslABe	3 980 571	3 983 620	Aryl-sulphate sulphohydrolase	(2)
asr	1 669 373	1 669 708	Acid-inducible RNA	(44)
atoSCeDAEBe	2 318 063	2 325 310	Acetyl coenzyme A:acetoacetyl coenzyme A transferase and thiolase II; required for growth on short chain fatty acids	(21)
b2854	2 992 021		Similar to Salmonella invasion protein iagB	(2)
bglJe	4 601 729	4 602 406	Activator transport operon: aromatic β-glucosides	(45)
bioDe	811 493	812 170	Dethiobiotin synthetase	(46)
bisZe	1 952 602	1 955 049	Biotin sulfoxide reduction	(47)
cadCe	4 357 974	4 359 512	Activator for lysine to cadaverine operon	(48)
cble	2 057 986	2 058 936	Regulatory protein	(49)
chpAR	2 908 778	2 909 361	Homologs of the pem locus; stable maintenance of plasmid R100	(50)
chpBS	4 446 018	4 446 619	Homologs of the pem locus; stable maintenance of plasmid R100	(50)
cmtAeB	3 075 490	3 077 349	PTS; mannitol (cryptic) permease iia components	(51)
cobTe	2 061 410	2 062 489	Nicotinate nucleotide dimethylbenzimidazole phosphoribosyltransferase	(52)
dicCAFeB	1 645 958	1 647 821	Cell division; Qin prophage	(53)
dive	2 435 970	2 436 965	Cell division	(2)
eaeH	313 581	314 468	Attaching and effacing protein homolog	(2)
ebgRAC	3 219 107	3 223 812	β-Galactosidase operon	(2)
ecpDe	155 461	156 201	Pilin chaperone	(11)
elaD	2 380 733	2 381 944	Putative sulfatase/phosphatase	(2)
emrE	567 538	567 870	DLP12 prophage; multidrug transporter	(25)
emrYeKe	2 478 658	2 481 359	Multidrug resistance emrAB homologs	(2)
entD	608 682	609 311	Enterochelin synthetase, component D (phoshpantetheinyltransferase)	(54)
envRe	3 410 440	3 411 102	Putative regulator, envCD	(2)
envYe	585 370	586 131	Porin thermoregulatory gene	(55)
erfKe	2 060 413	2 061 345	Cobalamin (coenzyme B12) biosynthetic gene?	(2)
farR	764 376	765 098	Fatty acid and fatty acyl-CoA-responsive DNA-binding protein	(56)
fecIRABCDeEe	4 508 261	4 515 803	Ferric citrate transport	(22)
fes	612 038	613 162	Enterochelin esterase	(57)
fhiAe	248 358	250 097	Putative transport protein, flagellar biosynthesis	(2)
fimZe	563 071	563 703	Type 1 pilin gene	(12)
flgAe	1 129 427	1 130 086	Flagellar basal body P-ring protein	(2)
fliCeDeSeTKeOe	2 000 133	2 019 891	Flagellar biosynthesis	(2)
flu	2 069 405	2 072 680	CP4-44 prophage; antigen 43; controls colony form variation and autoaggregation	(58)
flxA	1 644 429	1 644 761	Unknown	(59)
focBe	2 611 954	2 612 802	Putative formate transporter	(60)
frvRXBA	4 085 688	4 090 404	Sugar-specific PTS; fructose-like	(61)
gadA	3 663 810	3 665 210	Part of glutamate decarboxylasesystem; counteracts acidification during fermentative growth	(17,18)
gadB	1 568 669	1 570 069	Glutamate decarboxylasesystem	(17,18)
gatR_2eDeAeZeYe	2 169 417	2 175 230	Galactitol metabolism	(62)
glcCDFGB	3 119 650	3 127 051	Glycolate utilization	(19)
glf	2 105 248	2 106 351	Galactofuranose biosynthesis	(63)
gltF	3 358 811	3 359 575	Putative periplasmic protein in glutamate biosynthesis	(64)
gntPe	4 547 522	4 548 865	Gluconate permease	(65)
goaG	1 363 574	1 364 839	4-Aminobutyrate transaminase	(2)
hcaR,A1,A2,C,B,D	2 666 026	2 671 266	Catabolism of 3-phenylpropionic acid and cinnamic acid	(23)
hdeABD	3 654 038	3 655 197	Stress response proteins	(24)
hdhA	1 695 297	1 696 064	α-Hydroxysteroid dehydrogenase gene	(66)
hofBeCe	114 522	117 099	Putative protein transport, pili biogenesis?	(2)
hofD	3 463 180	3 463 857	Leader peptidase, integral membrane protein	(2)
hofFeGeH	3 457 454	3 459 614	Putative general secretion pathway proteins	(2)
hrsAe	765 207	767 183	Thermoinduction of ompC	(67)
hsdSe	4 577 638	4 579 032	Type I restriction system	(68)
hslJe	1 439 345	1 439 767	Heat shock protein	(2)
htrC	4 187 364	4 187 903	Heat shock response	(69)
htrEe	152 829	155 426	Type II pilin porin?	(11)
htrL	3 790 453	3 791 325	Lipopolysaccharide biosynthesis	(2)
hyaEeFe	1 035 577	1 036 829	Hydrogenase-1 operon	(70)
hyfABCeDEFGeHeIeRe-focB	2 599 182	2 612 799	Putative nine subunit hydrogenase complex	(60)
inaA	2 346 842	2 347 492	Weak-acid-inducible	(71)
intAe	2 754 180	2 755 421	Prophage CP-457 integrase	(2)
intBe	4 494 318	4 495 508	Prophage P4 integrase	(2)
intE	1 198 902	1 200 029	Similar to bacteriophage P21 integrase	(2)
intF	294 920	296 320	Probable phage integrase	(2)
kch	1 307 040	1 308 293	Putative potassium channel protein	(2)
kdgTe	4 099 262	4 100 254	2-Keto-3-deoxygluconate permease	(2)
lacA	360 473	361 084	Galactoside acetyltransferase	(2)
lhr	1 727 111	1 731 727	Probable ATP-dependent helicase	(72)
lit	1 197 918	1 198 811	Similar to phage T4 late gene expression blocking protein (GPLIT); prophage e14	(2)
marB	1 618 013	1 618 231	Multiple antibiotic resistance	(73)
mbhAe	250 042	250 827	Putative chemotaxis protein	(2)
mcrA	1 209 569	1 210 402	Restriction of methylated DNA; prophage e14	(74)
mcrB	4 575 528	4 576 925	Restriction of methylated DNA	(75)
mcrC	4 574 482	4 575 528	Restriction of methylated DNA	(75)
mcrDe	4 573 163	4 574 425	Restriction of methylated DNA	(2)
mipBe	862 793	863 527	Transaldolase-like protein	(2)
molR_1,R_2,R_2	2 194 494	2 195 318	Molybdate metabolism regulator	(76)
mrre	4 584 519	4 585 433	Restriction of methylated DNA	(2)
nfrAB	587 205	592 401	Membrane proteins required for bacteriophage N4 adsorption	(77)
nikCeDEe	3 613 813	3 616 213	Nickel transport	(78)
ninE	572 144	572 314	Cryptic lambdoid prophage DLP12	(79)
nrfGe	4 291 122	4 291 718	Formate-dependent nitrite reduction	(80)
ogrKe	2 165 324	2 165 542	Phage related	(2)
ompFe	985 117	986 205	Outer membrane protein	(81)
ompG	1 379 971	1 380 876	Porin; large channels; cryptic	(28)
ompT	583 903	584 856	Porin; large channels; cryptic; DLP12 prophage	(28)
pdxKe	2 534 406	2 535 257	Pyridoxine (vitamin B6) kinase	(82)
perR	268 513	269 406	CP4-6 prophage; peroxide resistance protein	(2)
phnCeDENeQ	4 311 389	4 320 914	Phosphonate uptake and degradation	(83)
phpB	668 519	669 130	Cobalamin synthesis (putative)	(2)
pinO	3 451 145	3 451 564	Calcium-binding protein required for initiation of chromosome replication	(2)
pmrDe	2 371 292	2 371 588	polymyxin B resistance protein	(2)
ppdAeBe	2 961 175	2 962 199	Prepilin peptidase-dependent protein A	(2)
pphA	1 920 337	1 920 996	Protein phosphatase 1; signals protein misfolding	(2)
pphB	2 857 783	2 858 439	Ser/Thr protein phosphatase 2	(2)
pqqL	1 570 431	1 573 226	Redox cofactor for pyrroloquinoline quinone synthesis (cryptic in K-12)	(2)
pshM	3 462 695	3 463 180	Putative general secretion pathway protein M	(84)
pspEe	1 367 713	1 368 027	Phage shock protein	(85)
recEe	1 412 813	1 415 410	Exonuclease VIII; Rac prophage; part is present in S.typhi	(86)
recT	1 412 008	1 412 817	DNA-pairing protein; RecA-independent recombination	(86)
relEeBe	1 643 370	1 643 896	Toxin-antitoxin	(87)
rem	1 642 675	1 642 926	Hypothetical rem protein	(2)
rfaBeIeJeKLSYeZe	3 794 575	3 802 743	Lipopolysaccharide core biosynthesis	(13–15)
rfbCeX	2 106 359	2 107 606	O-antigen synthesis	(16)
rhsA	3 759 810	3 763 943	Rhs multigene family	(26,27)
rhsB	3 616 823	3 621 058	Rhs multigene family	(26,27)
rhsC	728 806	732 999	Rhs multigene family	(26,27)
rhsD	522 485	526 765	Rhs multigene family	(26,27)
rhsE	1 525 914	1 527 962	Rhs multigene family	(26,27)
rimLe	1 496 962	1 497 501	Acetylation of ribosomal protein L12, converting it into L7	(88)
rnae	643 420	644 226	Ribonuclease I	(89)
rpiRB	4 309 680	4 311 378	Ribose catabolism	(90)
rtne	2 268 746	2 270 302	Resistance to phage N4 and lambda	(2)
ruse	572 594	572 956	Crossover junction endodeoxyribonuclease; cryptic lambdoid prophage DLP12	(91)
sbcCe	411 831	414 977	ATPase involved in DNA repair	(92)
sbm	3 058 870	3 061 014	Methylmalonly-coA mutase α-subunit	(2)
sbmC	2 078 811	2 079 284	Stationary phase induced SOS gene; gyrase inhibitor	(93)
sfa	1 051 070	1 051 300	Suppressor of the unsaturated fatty acid locus fabA6(Ts)	(94)
sfmAeCeDeHeFe	557 402	563 068	SFM fimbrial proteins	(2)
sgcEe	4 524 473	4 525 105	Putative epimerase	(2)
sieB	1 416 572	1 417 183	Superinfection exclusion protein B	(95)
slp	3 651 558	3 652 157	Outer membrane lipoprotein; carbon starvation-inducible and stationary phase inducible	(96)
smf_1e	3 430 436	3 431 197	Unknown	(2)
sohA	3 274 643	3 274 978	HtrA suppressor protein (protein prlf)	(97)
sugEe	4 374 303	4 374 770	Chaparone; small multidrug resistance family	(2)
t150e	3 718 827	3 719 678	IS150 putative transposase	(2)
tap	1 967 407	1 969 008	Taxis toward peptides	(98)
tdcRe	3 264 895	3 265 239	Threonine dehydratase regulatory protein	(99)
thiMe	2 182 533	2 183 321	Hydoxyethylthiazole kinase	(100)
tnaLABe	3 886 064	3 887 774	Tryptophanase operon	(101)
tolAe	775 565	776 830	Tolerance to phage and colicins	(102)
torDe	1 061 022	1 061 621	Trimethylamine N-oxide (TMAO) operon	(103)
torSe	1 052 657	1 055 371	TorS-TorR two component system	(103)
trge	1 490 494	1 492 134	Methyl-accepting chemotaxis protein III; ribose and galactose sensor receptor	(104)
trkG	1 421 806	1 423 263	Uptake of K+; Rac prophage	(105)
ttdAeBe	3 204 104	3 205 617	Tartrate dehydratase	(106)
uidRABC	1 689 610	1 694 486	Glucuronide operon (gus)	(2)
umuDe	1 229 990	1 230 409	Translesion DNA synthesis	(2)
wbbHIJK	2 101 413	2 105 248	O-antigen production	(107,108)
wzzBe	2 095 343	2 096 359	Regulator of length of O-antigen component of lipopolysaccharide chains	(2)
xasA(gadC)B	1 566 978	1 570 069	Glutamate decarboxylase and permease; protects from cytoplasmic acidification	(17,18)
xylEe	4 238 358	4 239 833	Xylose transport protein	(109)
ygeG	2 989 290	2 989 781	Similar to Shigella and Yersinia plasmids IPPI and LCRH	(2)
ygeH	2 990 116	2 991 492	Similar to Salmonella invasion protein iagA	(2)
ygeK	2 992 482	2 992 928	Similar to luxR/uhpA family of transcriptional regulators	(2)
ygeV	3 002 030	3 003 808	Hypothetical σ54-dependent transcriptional regulator	(2)
ybiL/b0805	838 472	840 754	Outer membrane protein; putative ferrisiderophore receptor	(2)
ygeW	3 004 356	3 005 447	Similar to ornithine carbamoyltransferase	(2)
ygeY	3 006 785	3 007 996	Putative dacetylase	(2)
yqeI	2 986 524	2 987 333	Similar to V.cholerae toxR	(2)

Table 3. Escherichia coli K-12 genes with close homologs in K.pneumoniae but absent in the three Salmonella genomes^d.

Geneb	Startc	Endc	Putative functiond	References
afuC	276 980	278 038	Putative ferric transport ATP-binding protein	(2)
aldA	1 486 256	1 487 695	Aldehyde dehydrogenase	(2)
aldH	1 360 767	1 362 254	Putative aldehyde dehydrogenase	(2)
araFGeH	1 980 581	1 984 151	l-Arabinose transport system	(110)
arsRBCe	3 646 158	3 648 289	Arsenical resistance operon	(111)
ascFBG	2 837 547	2 839 004	PTS; arbutin, salicin and cellobiose	(112)
b1145	1 201 482	1 202 156	Putative phage repressor	(2)
b2442	2 556 791	2 558 086	Putative prophage integrase	(2)
betTIBeA	324 801	330 720	Osmoprotectant glycine betaine from choline	(113)
bglBeFeGe	3 899 920	3 904 195	Transport and phosphorylation of β-glucosides and antiterminator; cryptic operon	(114)
chaC	1 271 709	1 272 425	Cation transport protein	(2)
cynReTeSXe	357 015	360 370	Cyanase and carbonic anhydrase	(115)
ddpX/b1488	1 560 519	1 561 100	d-Ala-d-Ala dipeptidase, Zn-dependent	(2)
dinDe	3 815 375	3 816 211	DNA damage inducible, regulated by the LexA-RecA	(116)
evgAeS	2 481 775	2 485 987	Two component signal transduction system	(117)
exuReT	3 242 744	3 245 065	Negatively controls uxuAB operon	(118)
feaReB	1 444 402	1 447 042	Regulates expression of tynA and padA; phenylacetaldehyde dehydrogenase	(2)
fimBEAeIeCeDeFeGH	4 538 525	4 547 279	Type 1 pilin; mannose-sensitive hemagglutination; role in pathogenesis	(12)
folX	2 419 345	2 419 707	Dihydroneopterin-triphosphate epimerase	(119)
fruLe	87 860	87 946	fruR leader peptide	(2)
GapC_1e,C_2	1 487 737	1 488 389	Glyceraldehyde 3-phosphate dehydrogenase	(2)
glvGeBC	3 858 976	3 861 491	Probable 6-phospho-β-glucosidase and PTS system; arbutin-like components	(2)
hipAeBe	1 588 878	1 590 466	Replication or cell division?	(120)
hypAe	2 848 670	2 849 020	Hydrogenase 3 subunit	(121)
idi/b2889e	3 031 085	3 031 633	Isopentenyl diphosphate isomerase	(2)
intC	2 464 565	2 465 722	Putative prophage sf6-like integrase	(2)
lacIZY	361 150	366 734	Lactose operon	(2)
malIeXYe	1 696 176	1 700 153	Maltose regulon	(122)
maoC	1 449 621	1 451 666	NADH-dependent dehydrogenases	(123)
mhpRABCDEFTe	366 811	375 894	3,3-(Hydroxyphenyl) propionate catabolism	(124)
nac	2 059 038	2 059 955	Nitrogen assimilation regulatory protein	(125)
nikAB	3 611 298	3 613 816	Nickel transport	(78)
nlpAe	3 836 802	3 837 620	Lipoprotein-28	(2)
ordL	1 362 256	1 363 536	Probable oxidoreductase	(2)
phnFGHIJKLMP	4 311 389	4 320 914	Phosphonate uptake and degradation	(83)
phoA	400 902	402 386	Alkaline phosphatase precursor	(2)
pppA/b2972	3 111 560	3 112 492	Bifunctional prepilin peptidase	(2)
shiAe	2 051 665	2 052 981	Shikimate transport system	(126)
tam/b1519	1 605 370	1 606 128	Trans-aconitate methyltransferase	(2)
tauABeCD	384 399	387 870	Taurine operon	(127)
tesAe	518 363	518 989	Thioesterase I	(128)
tynA	1 447 100	1 449 373	Copper amine oxidase precursor (tyramine oxidase) (maoA)	(2)
uxaACe	3 239 467	3 242 381	Hexuronate degradation	(118,129)
uxaB	1 607 253	1 608 704	Hexuronate degradation	(118,129)
Wzy/rfc	2 104 085	2 105 248	O-antigen polymerase	(130)
xylFGeH	3 728 760	3 732 527	Xylose transport operon	(131,132)
yaiP	381 963	383 159	Polysaccharide metabolism
ybcS	576 836	577 333	Bacteriophage λ lysozyme homolog	(2)
ybcU	577 823	578 116	Bacteriophage λ Bor protein homolog	(2)
ydeP/b1501	1 582 231	1 584 510	Efflux of cysteine biosynthetic pathway	(2)
ygfG	3 061 971	3 062 798	Methylmalonyl-CoA decarboxylase, biotin-independent	(2)
ygfH	3 062 822	3 064 300	Propionyl-CoA:succinate-CoA transferase	(2)
ygfP	3 023 787	3 025 106	Guanine deaminase	(2)

^aNote that genes may have been excluded from the sample by negative selection in the library, so this data does not guarantee their absence. Furthermore, genes that appear to be present by our criteria may be recently inactivated pseudogenes or close paralogs with a different function.

^bAn E.coli gene is identified as ‘absent’ in a given genome if the best DNA sequence identity for the portion of a gene that appeared to be sequenced in the other genome is <70% (see Materials and Methods). 1350 E.coli genes are absent from all three Salmonella species and 1165 genes are absent from both SAL and KPN. Only named genes and some unnamed genes with assigned putative functions are listed here. A complete list can be found at http://globin.cse.psu.edu/ftp/dist/Enterix/sim_pct.p70/missing_SAL.p70 and http://globin.cse.psu.edu/ftp/dist/Enterix/sim_pct. p7 0/missing_SAL_KPN.p70, respectively.

^cStart and stop positions for ECO. When the sequence consists of a set of genes, the span includes all of these genes.

^dPutative functions are from the reference paper, GenBank (http://www.ncbi.nlm.nih.gov:80/entrez/query.fcgi?db=Protein), WIT (http://wit.mcs.anl.gov/WIT2/) or KEGG (http://www.genome.ad.jp/kegg/).

^eGenes with weak homologs in at least one SAL or KPN, defined as 50–70% DNA sequence identity. These are possible paralogous comparisons, gene conversion or accelerated evolution.

Among the genes and operons confirmed to be absent in SAL and KPN are those involved in glutamate metabolism (gadABC/xasAB) (17,18), glycolate utilization (glcCDFGB) (19), the glucuronide operon (uidRABC, gus) (2), the aga operon (20), the ato operon (21), the fec (ferric citrate transport) operon (22), the hca operon (23), the hde operon (24), emrE (a multidrug transporter) (25) and the entire dispersed family of ORFs of unknown function (rhsABCDE) (26,27). However, a set of distant family members does occur (J.Parkhill, unpubished results). The absence of certain other genes was less expected. OmpG is a cryptic phage gene in ECO that is expressed only after a mutation in the gene. Gene expression was reported at low levels in SAL (28), but we found that this gene and its close homologs are absent from SAL, indicating that the reported expression may be of a different gene.

Some genes are absent from only one SAL or KPN. These genes are not reported in Tables 2 and 3 but some are of particular interest, such as rffH, which is explored further elsewhere (3). It is likely that at least a few of the genes that are absent in our SAL and KPN genomes will prove to be present in the genomes of other strains of S.enterica or K.pneumoniae. Nevertheless, we can extrapolate from the very few genes that appear to be shared by ECO and only one out of three SAL (less than 30 such genes) that the lack of a homolog in our data analysis will usually mean lack of a homolog in virtually all the genomes in a species.

There are 394 events where one or more adjacent genes are absent in all three SAL relative to ECO (Table 4). 199 of these events involve more than one gene. Similarly, there were 216 cases where SPA has a sequence of >400 bases that is not found in ECO, 261 in STM and 257 in STY. There are at least 160 cases where the three SAL share sequences of >400 bases that are not found in ECO (Table 4). This number is in line with our previous estimate of a few hundred insertion/deletion events that distinguish these genomes, which was extrapolated from small samples of the genomes (8,29) and estimates based on the codon adaptation index (9).

Table 4. Genome differences between E.coli K-12 and the three Salmonella spp.

No. of adjacent genes present in E.coli K-12 and absent from Salmonellaa	Frequency
1	195
2	67
3	27
4	25
5	13
6	8
7	7
8	15
9	6
11	3
12	6
13	5
14	3
15	2
17	2
18	2
19	2
20	1
22	1
23	1
28	1
30	2
Total	394
Sequences present in Salmonella and absent in E.coli K-12b	160

^aAt a threshold of 70% DNA sequence similarity.

^bInsertions in the three Salmonella genomes of >400 bp in length within 200 bp of each other in the E.coli K-12 sequence.

If one discounts phage and transposable elements, both of which are probably not found in all ECO strains, and accepts that the SAL and ECO genomes diverged ∼150 million years ago (30), then the rate at which ECO has acquired a sequence not found in SAL (through an insertion in ECO or deletion in SAL) approaches 2 per million years. On average, each event resulted in the insertion or deletion of about three ORFs. These are conservative estimates, as they do not take into account sequences gained and then lost during the past 150 million years.

Most of the genes unique to E.coli K-12 are in large clusters

It seemed likely that some of the 199 clusters of ECO-specific genes might be associated with known cryptic phage (see for example 31). These genes would presumably be absent from KPN as well as from the three SAL. The list of 1165 such genes (at a 70% DNA identity threshold) was examined for regions in which genes unique to ECO were separated by no more than 3 kb, to account for occasional close paralogs and recent insertions that might disrupt contiguous regions of sequence unique to ECO. This examination revealed 28 regions over 10 kb that consisted primarily of genes unique to ECO (Table 5). Four known cryptic lambdoid phage were identified within these largest clusters: Qin (within a 20 kb region), Rac (21 kb), e14 (27 kb) and DLP12/qsr′ (20 kb). A P4-like phage (32) may reside within a 32 kb cluster. Another cluster of 26 kb may include the CP4-6 prophage (33). The largest cluster, 46 kb, encodes genes with amino acid homology to Salmonella invasion genes and Vibrio toxR, though below the 70% DNA sequence similarity cut-off we have used to distinguish probable orthologs from other homologs. Many other very large clusters have no known association with phage. Perhaps some of these are plasmid integrations (34).

Table 5. Escherichia coli K-12 gene clusters not found in the three Salmonella or the K.pneumoniae genomes^a.

First gene in cluster	Start	End	Size	Named genes in cluster	References
b0245	262 552	273 216	10 664	?
afuC	276 980	303 406	26 426	Phage CP4-6, intF	(33)
ykgL	311 336	324 588	13 252	eaeH
yahA	331 595	344 873	13 278	?
sfmA	565 195	573 562	16 160	sfmACDHF, fimZ, emrE, ninE, rus	(31)
ybcR	576 621	596 686	20 065	Phage DLP12, qsr′, appY, ompT, envY, nfrA, nfrB	(79)
ycbE	992 500	1 003 880	11 380	?
appA	1 039 840	1 055 371	15 531	appA, sfa, torS
ymfD	1 196 090	1 223 130	27 040	Phage e14	(31)
b1345	1 410 024	1 431 698	21 674	Phage Rac	(31)
ydbA_1	1 463 416	1 480 225	16 809	?
b1483	1 555 136	1 596 110	40 974	xasA, gadB, pqqL
ydfK	1 631 063	1 650 732	19 669	Phage Qin	(31)
molR_1	2 194 494	2 208 964	14 470
atoS	2 318 063	2 334 712	16 649	atoSCDAEB, inaA
emrY	2 478 658	2 493 312	14 654	emrYK, evgAS
hyfA	2 599 182	2 612 802	13 620	hyfABCDEFGHIR, focB
hcaR	2 666 026	2 682 076	16 050	hcaR, A1, A2, CBD
intA	2 754 180	2 786 670	32 490	P4-like phage	(32)
ygcN	2 890 650	2 901 396	10 746	?
yqeH	2 985 498	3 031 633	46 135	Similar to iagAB, toxR
yghK	3 117 613	3 132 838	15 225	glcBGFDC
ygjH	3 218 556	3 228 880	10 324	ebgRAC
sohA	3 274 643	3 290 073	15 430	agaRWASBCDI
pinO	3 451 145	3 467 490	16 345	hofFGD, pshM
slp	3 651 558	3 665 210	13 652	hdeBAD, gadA
yjcP	4 297 143	4 311 796	14 653	rpiRB, phnQ
intB	4 494 318	4 525 105	30 787	fecIRABC

^aAt a threshold of 70% DNA sequence similarity. Clusters may include segments of up to 2.7 kb shared with any of the other genomes. Twenty-eight clusters >10 kb are shown.

Many differences involve insertion/deletion of single genes

Studies of the differences between E.coli and Salmonella have often focused on large groups of genes found in one species or the other (see for example 35–37). However, even at a relaxed threshold of 50% sequence identity there are almost 100 sites where a single gene is absent in SAL but present in ECO and flanked by genes found in both genomes. None of these are known transposable elements nor do they have significant homologies to such elements. These observations add to the growing indications that it will not be sufficient to study a few large ‘pathogenicity islands’ in order to understand the myriad of genetic differences between Salmonella and E.coli. Deletions can also be associated with increased pathogenesis (38). It will be interesting to explore whether any of the 394 sequences that are missing in all three SAL have this effect.

Most of the genome is stable while the rest has sustained multiple changes

The ECO, KPN and SAL genomes are approximately equally divergent, with KPN appearing as a very close outgroup and Yersinia as a distant outgroup to these three genera, as indicated by the numbers of shared genes (Table 1) and the divergence from ECO exhibited by presumptive orthologs (84.5, 84.1, 84.2, 82.1 and 72.9% for STM, STY, SPA, KPN and YPE, respectively) (Table 1). Thus, it should be possible to predict the most parsimonious ancestral state when an ECO sequence is not present in at least one of the other lineages. This is important because the structure of the common ancestors of ECO and SAL and their common ancestor with KPN could help determine what functions were acquired on each lineage as it became specialized to its particular niche over a period that may be in excess of 150 million years (30). We inspected the PIPs for deletions shared by all three Salmonella genomes that were unchanged in the ECO–KPN comparison. Such deletions would be suspected to have occurred in the SAL lineage. Sequences missing from both SAL and KPN would most likely indicate insertions in the ECO lineage. However, with the exception of some of the very large insertions attributable to prophage in ECO, there were relatively few simple insertion/deletion events that occurred in only one of the lineages. In many cases (20 of the first 50 we inspected) all three genomes appeared to have undergone insertions, deletions and rearrangements in the same or overlapping regions.

Two examples can be seen in Figure 1, located at 5 and 16 kb on the ECO genome. A simple absence of a gene occurs in the SAL genomes at 5 kb in the ECO genome, symbolized by the red box. [See Florea et al. (3) for details about PIPs.] However, the absence of this gene in KPN is accompanied by the insertion of an additional sequence of >400 bp (symbolized by the two blue stripes). An insertion also occurs at the same position in the distant outgroup YPE. However, analysis of the insertion in KPN and YPE shows that they are different from each other and not similar to any sequences yet found in any genome (not shown). Thus, there have been different events at this location in all of the genomes, making the ancestor impossible to determine without further comparisons with other related genomes, and perhaps not even then.

At 16 kb on the ECO map is another example where the SAL, KPN and ECO genomes all differ. The ECO sequence at this position includes a suspected insertion sequence and hence the ancestral state may be suspected to be absence of the genes. This hypothesis is supported by referral to the structure of YPE, a very distant enterobacteria, where this segment is missing. However, the SAL have a different sequence inserted at this site, leading to no prediction about the state of this region in the ECO–SAL ancestor.

The ability to determine the ancestral state of the progenitor of these genera will presumably be aided by sample sequencing of other Enterobacteria, especially those other putative Escherichia and Salmonella, such as Escherichia blattae, Escherichia fergusonii, Escherichia hermannii, Escherichia vulneris and Salmonella bongori, as well as other closely related genera, particularly Citrobacter. However, it is likely that the state of the most recent common ancestor of ECO and SAL will remain unresolved at many positions even then.

It is biologically illuminating that so many insertion/deletion events cannot be explained by a single event. If these insertion/deletion events were randomly distributed over the genome one would expect two different insertions/deletions to occur only rarely in the same place in multiple lineages. The observation that the rearrangement events are concentrated in the same place in multiple lineages strongly implies that they are, at a minimum, excluded from large parts of the genome or perhaps in some cases targeted to a subset of the genome. It is these ‘hot-spots’ that appear to have sustained rearrangement events in separate lineages since the divergence of SAL, ECO and KPN.

In summary, an initial comparison of the ECO genome with the SAL and KPN genomes reveals the set of ECO genes that have no close homolog in SAL or KPN. Many genes found in ECO, but not in SAL or KPN, occur in very large clusters, some of which contain cryptic phage but some of which consist primarily of genes of unknown function. Nevertheless, a surprising number of differences involve single genes. Rearrangements appear to be concentrated in areas of the genome where the rate of rearrangement is rapid, relative to the divergence times of these organisms, so that many sites of rearrangement have sustained changes in more than one lineage. The data presented here comprise only the first of a number of possible genome comparisons. It will be interesting to use each of the other genomes in turn as the reference genome, in order to determine the sequences unique to each of the other lineages.

An additional feature of the data is that the genomes of K.pneumoniae (KPN) and S.enterica serovar Paratyphi A are samples with about 4× coverage and no immediate plans for completion. The data we were able to extract from these and the other incomplete genomes helps to demonstrate the utility of the relatively inexpensive sampled genome sequences for comparative genomics. However, one must bear in mind the caveats that some genes may be under-represented due to negative selection in the shotgun libraries and sequencing errors in samples means that one often cannot distinguish genes from pseudogenes. Nevertheless, with these caveats, the utility of sampled genomes, such as the KPN and SPA genomes that we have generated, and of the analysis of partial sequences of genomes undergoing completion, such as STM, STY and YPE, becomes increasingly evident as tools that can exploit these resources are developed.