Diversity of Genome Structure in Salmonella enterica Serovar Typhi Populations

Abstract

The genomes of most strains of Salmonella and Escherichia coli are highly conserved. In contrast, all 136 wild-type strains of Salmonella enterica serovar Typhi analyzed by partial digestion with I-CeuI (an endonuclease which cuts within the rrn operons) and pulsed-field gel electrophoresis and by PCR have rearrangements due to homologous recombination between the rrn operons leading to inversions and translocations. Recombination between rrn operons in culture is known to be equally frequent in S. enterica serovar Typhi and S. enterica serovar Typhimurium; thus, the recombinants in S. enterica serovar Typhi, but not those in S. enterica serovar Typhimurium, are able to survive in nature. However, even in S. enterica serovar Typhi the need for genome balance and the need for gene dosage impose limits on rearrangements. Of 100 strains of genome types 1 to 6, 72 were only 25.5 kb off genome balance (the relative lengths of the replichores during bidirectional replication from oriC to the termination of replication [Ter]), while 28 strains were less balanced (41 kb off balance), indicating that the survival of the best-balanced strains was greater. In addition, the need for appropriate gene dosage apparently selected against rearrangements which moved genes from their accustomed distance from oriC. Although rearrangements involving the seven rrn operons are very common in S. enterica serovar Typhi, other duplicated regions, such as the 25 IS200 elements, are very rarely involved in rearrangements. Large deletions and insertions in the genome are uncommon, except for deletions of Salmonella pathogenicity island 7 (usually 134 kb) from fragment I-CeuI-G and 40-kb insertions, possibly a prophage, in fragment I-CeuI-E. The phage types were determined, and the origins of the phage types appeared to be independent of the origins of the genome types.

Salmonella enterica serovar Typhi is host restricted, for it grows only in humans, where it causes typhoid enteric fever (13, 51). The annual global incidence of typhoid fever is estimated to be 21.6 million cases, with more than 220,000 deaths (10). The emergence of antibiotic-resistant strains (8) and the increased incidence of typhoid fever in human immunodeficiency virus type 1-infected persons are causes for concern. The genus Salmonella is separated into two species and more than 2,500 serovars (52) on the basis of the somatic and flagellar antigens. Many of the serovars, such as S. enterica serovar Typhimurium, are host generalists, growing in many different animal species and humans and causing gastroenteritis.

S. enterica serovar Typhi is more homogeneous than most serovars of Salmonella. Using multilocus enzyme electrophoresis, Reeves et al. (54) and Selander et al. (58) showed that S. enterica serovar Typhi strains constitute only one or two clones that are widely separated from the other serovars in subspecies I. Membrane protein profiles (15, 17) and plasmids (42) show homogeneity. Multilocus sequence typing of housekeeping genes has suggested that S. enterica serovar Typhi evolved only about 50,000 years ago from other Salmonella serovars (30).

The orders of orthologous genes in Escherichia coli K-12 and S. enterica serovar Typhimurium LT2 are almost identical, although the genera diverged about 100 to 160 million years ago (38, 48, 59, 62). Within the genus Salmonella, the gene order of S. enterica serovar Paratyphi B (33) and the gene order of S. enterica serovar Enteritidis (36) are very similar to the gene order of S. enterica serovar Typhimurium LT2, and chromosomes are also conserved in 17 independent strains of S. enterica serovar Typhimurium (39). During growth in laboratory culture, duplications of segments of the chromosome occur at high frequencies (10⁻² to 10⁻⁵) (4, 23), and some inversions, especially those with endpoints in the rrn operons, are common (22). Such remarkable conservation of the chromosome during evolution, in spite of the high frequency of rearrangements in culture, may have resulted from strong selective pressures that selectively removed rearranged genomes.

Pulsed-field gel electrophoresis (PFGE) permits rapid construction of genomic maps (16); partial digestion with the endonuclease I-CeuI shows the number and locations of the rrl genes for 23S rRNA and the order of adjacent fragments (34) (the rrn skeleton). I-CeuI, which is encoded by a class I mobile intron in the rrl gene for the large-subunit rRNA (23S-rRNA) in the chloroplast DNA of Chlamydomonas eugamatos (44), digests a 19-bp sequence in all seven rrl genes of enteric bacteria (43). The rrn skeleton is highly conserved in enteric bacteria, so related strains usually yield identical fingerprints (39).

Surprisingly, in view of the homogeneity in many properties, independent wild-type strains of S. enterica serovar Typhi show significant genomic rearrangements. The I-CeuI fragments of S. enterica serovar Typhi strains were shown by PFGE to be in many different orders, called genome types (40), that are mediated through recombination between the seven rrn genes that code for rRNA. Partial I-CeuI digestion can determine the order but not the orientation of the I-CeuI fragments.

Vi phage typing has high discriminatory power for subdivision of strains of S. enterica serovar Typhi, and the method of Craigie and Felix (9) has allowed over 100 different phage types to be recognized (49). Vi phage absorbs to the Vi (virulence) exopolysaccharide and adapts itself to the last strain in which it was propagated. In this study we determined the phage types of strains to determine if they are independent of or correlated with genome types.

In this study we used PFGE with I-CeuI to determine the genome types of a set of strains of S. enterica serovar Typhi which had been assembled from a variety of sources; we then determined the orientation of the I-CeuI fragments using PCR. In addition, we determined the Vi phage types and the flagellar antigens of the strains. The sizes of the seven fragments (the I-CeuI fingerprint) are indistinguishable in most strains, so they are characteristic of the species, but the sizes of a few fragments are increased or decreased due to insertions or deletions. We show here that the phage type is largely independent of the genome type. The need for chromosome balance between the two replichores and for maintaining appropriate gene dosage appears to restrict the range of genomic rearrangements.

MATERIALS AND METHODS

Bacterial strains and cultivation conditions.

The S. enterica serovar Typhi strains and their sources are shown in Table 1. All strains were maintained in 15% glycerol at −70°C in the collection of the Salmonella Genetic Stock Center (www.ucalgary.ca/∼kesander), and single-colony isolates were isolated prior to use. The strains were grown at 37°C in Luria-Bertani medium; solid media contained 1.5% agar.

TABLE 1.

Phenotypic and genotypic characteristics of S. enterica serovar Typhi strains used in this study

Straina	SGSC no.b	Source (reference)c	Locality of isolation	Year of isolation	Flagellar antigend		Phage typee	Fragment order in genomef	PCR orderg	Genome typeh	Altered 1-Ceu-I fragment(s)i
					d	J
26.001	3124	NML	Manitoba	1994	+	−	E1	BCFEDG	A−C+	2
26.003	3126	NML	Alberta	1993	+	−	UT(Vi-ve)	BCDEFG	A−C+	1	G, −130 (SP17−)
26.004	3127	NML	British Columbia (India)	1993	+	−	I+IV	BCEFDG	A+C+	3
26.005	3128	NML	British Columbia (India)	1994	+	−	E1	BCFEDG	A−C+	2
26.006	3129	NML	British Columbia	1994	+	−	UT(Vi-ve)	GFCEDB	A−C+	16
26.007	3130	NML	British Columbia	1994	+	−	A	BCEFDG	A−C+	3
26.008	3131	NML	British Columbia	1994	+	−	D2	BCDFEG	A+C+	4
26.009	3132	NML	Manitoba	1994	+	−	B1	GFCEDB	A+C−	16
26.010	3133	NML	Manitoba	1994	+	−	E1	BCFEDG	A−C+	2
26.011	3134	NML	Alberta (Kenya)	1994	+	−	B2	BCEFDG	A−C+	3
26.012	3135	NML	Manitoba (India)	1994	+	−	O	BCFDEG	A+C+	6
26.015	3138	NML	British Columbia	1994	+	−	J1	BCEFDG	A+C+	3
26.016	3139	NML	British Columbia	1994	+	−	DVS	BCFDEG	A−C+	6
26.017	3140	NML	British Columbia	1994	+	−	B1	BCDFEG	A−C+	4
26.018	3141	NML	British Columbia	1994	+	−	B2	BCEFDG	A−C+	3
26.019	3142	NML	Alberta	1994	+	−	A	BCEDFG	A−C+	5
26.020	3143	NML	British Columbia	1994	+	−	Atypical	BCEDFG	A−C+	5
26.021	3144	NML	British Columbia	1994	+	−	O	BCFDEG	A−C+	6
26.022	3145	NML	Alberta	1994	+	−	I+IV	BCEFDG	A+C+	3
26.023	3146	NML	Alberta	1994	+	−	UT(Vi-ve)	BCEFDG	A−C+	3
26.024	3147	NML	Ontario	1994	+	−	E1	BCFEDG	A−C+	2
26.027	3150	NML	Quebec	1994	+	−	46	BCFEDG	A−C+	2
26.028	3151	NML	Quebec	1994	+	−	DVS	BCFDEG	A−C+	6	B, +88
26.029	3152	NML	Quebec	1994	+	−	UT(Vi-ve)	BCEFDG	A+C+	3
26.03	3153	NML	Quebec	1994	+	−	I+IV	BCEFDG	A−C+	3
26.031	3154	NML	Quebec	1994	+	−	Atypical	GDCEFB	A−C+	19
26.032	3155	NML	Quebec	1994	+	−	I+IV	GECFDB	A−C−	24
26.033	3156	NML	Quebec	1994	+	−	Atypical	BCEFDG	A+C+	3	B, +80
26.034	3157	NML	Quebec	1994	+	−	I+IV	BCEFDG	A−C+	3
26.035	3158	NML	Quebec	1994	+	−	I+IV	BCEFDG	A−C+	3
26.037	3160	NML	British Columbia	1994	+	−	I+IV	BCFEDG	A−C+	2
26.038	3161	NML	British Columbia	1994	+	−	E1	BFCEDG	A+C+	14
26.04	3163	NML	British Columbia	1994	+	−	M3	GDCEFB	A−C+	19	G, +50
26.041	3164	NML	Alberta (El Salvador)	1994	+	−	B3	BCEFDG	A+C+	3	B, +80
26.042	3165	NML	Alberta (California)	1994	+	−	B2	BCEFDG	A−C+	3	E, +40
26.043	3166	NML	Alberta	1994	+	−	DVS	BCEFDG	A+C+	3	B, +80
26.044	3167	NML	British Columbia	1994	+	−	F4	BCFEDG	A−C+	2	B, +90
26.045	3168	NML	British Columbia	1994	+	−	D8	BCEFDG	A+C+	3
26.047	3170	NML	British Columbia	1994	+	−	A variant	BCEFDG	A+C+	3
26.048	3171	NML	British Columbia	1994	+	−	B1	BCFEDG	A−C+	2
26.049	3172	NML	British Columbia	1994	+	−	B1	GCEDFB	A−C+	11
26.05	3173	NML	Alberta	1994	+	−	I+IV	GCFEDG	A−C+	2	B, +80
26.051	3174	NML	British Columbia	1994	+	−	DVS	BCDEFG	A−C+	1
26.054	3177	NML	Quebec	1994	+	−	E1	BCFEDG	A−C+	2	B, +15
26.055	3178	NML	Quebec	1994	+	−	Atypical	BCFEDG	A−C+	2
26.056	3179	NML	Quebec	1994	+	−	F1	GECDFB	A−C+	23
25.035	2667	NML	Manitoba	1994	+	−	F1	BDCFEG	A−C−	18	G, +80
25.036	2668	NML	Alberta	1993	+	−	E1	BCEFDG	A−C+	3
25.037	2669	NML	British Columbia	1993	+	−	D5	BCEFDG	A−C−	3
25.039	2671	NML	British Columbia	1993	+	−	O	BCFDEG	A+C+	2
25.04	2672	NML	British Columbia (Guatamala)	1993	+	−	E1	BCDFEG	A+C+	4
25.041	2673	NML	Alberta	1993	+	−	B2	BCEFDG	A−C+	3	E, +40; B, +20
25.042	2674	NML	Ontario	1993	+	−	E1	BCFEDG	A−C+	2
ST60	2770	Pang	Malaysia	1986	+	−	C4	BCFEDG	A−C+	2	G, +20
ST24A	2771	Pang	Malaysia	1986	+	−	DVS	BCEFDG	A+C+	3
ST143	2773	Pang	Malaysia	1994	+	−	D2	BCDFEG	A+C+	4
ST145	2774	Pang	Malaysia	1994	−	−	I+IV	BCEFDG	A+C+	3	E, +30
ST308	2775	Pang	Malaysia	1987	+	−	E1	BCEFDG	A+C+	3
ST1002	2776	Pang	Malaysia	1987	+	−	E1	GCEFDB	A−C+	9
ST168	2777	Pang	Malaysia	1987	+	−	UT(Vi-ve)	BCEFDG	A−C+	3	E, +40; G, −70
ST309	2779	Pang	Malaysia	1987	+	−	E1	BCEFDG	A+C+	3	B, +15
ST1106	2780	Pang	Malaysia	1987			D1	BCDFEG	A−C+	4	E, +40
ST495	2781	Pang	Malaysia	1987	+	−	B1	BCEFDG	A−C+	3
In14	2782	Pang	Indonesia	1994	+	−	Atypical	BCEFDG	A−C+	3
In15	2783	Pang	Indonesia	1994	+	−	D2	BCEFDG	A−C+	3
3123	3184	Pang	Chile	1983			rough	BCEFDG	A+C+
3125	3185	Pang	Chile	1983	+	−	46	BCEFDG	A+C+	3
T189	3187	Pang	Thailand	1990	+	−	N	GCEDFB	A−C+	11
T202	3189	Pang	Thailand	1990	+	−	UT(Vi-ve)	BCEFDG	A+C+	3	G, −130 (SP17−)
T104	3188	Pang	Thailand	1990	+	−	UT(Vi-ve)	BCDEFG	A−C+	1	G, −130 (SP17−)
In4	3190	Pang	Indonesia	1992	+	−	53	BDCEFG	A+C+	17	B, +80
In20	3191	Pang	Indonesia	1992	+	−	A	GCEFDB	A−C+	9	E, +40
In24	3192	Pang	Indonesia	1992	+	−	C3	BCEFDG	A−C+	3
PNG30	3193	Pang	Papua New Guinea	1994	+	−	D2	BCFEDG	A−C+	2
PNG31	3194	Pang	Papua New Guinea	1994	+	−	D2	BCFEDG	A−C+	2
PNG32	3195	Pang	Papua New Guinea	1994	+	−	D2	BCEFDG	A+C+	2
ST1	2728	Pang			+	−	I+IV	BDCFEG	A−C+	18	G, −130 (SP17−)
CC6	3198	Ho	Thailand		+	−	A	GCDEFB	A−C+	7
CC7	3199	Ho	Thailand		+	−	A	GCDEFB	A−C+	7
382-82	2664	CDC	Marshall Islands		+	−	M1	BFCDEG	A+C+	13
9032-85	2663	CDC	Taiwan		+	−	UT(Vi-ve)	BFDCEG	A+C−	23	G, +15
1707-81	2661	CDC	Liberia		+	−	UT(Vi-ve)	BCEFDG	A+C+	3
1196-74	2662	CDC	Mexico		+	−	A	BCFDEG	A−C+	6	B, +80
3434-73	2658	CDC	Peru		+	−	G1	BECFDG	A−C+	22
3137-73	2660	CDC	India		+	−	K1	BCFDEG	A+C+	6
3815-73	2659	CDC	Unknown		+	−	T	BCEFDG	A−C+	3
SARB63	2520	Selander	Dakar, Senegal	1988	+	−	A	CBEFDG	A−C+	25
SARB64	2521	Selander	Dakar, Senegal	1988	+	−	UT(Vi-ve)	BDCEFG	A−C+	19	G, −130 (SP17−)
SA4825	2655	ProvLab	Calgary, Canada	1993	+	−	B2	BCEFDG	A−C+	3	E, +40
PL27566	2990	ProvLab			+	−	M1	ECBFDG	A−C+	26	B, +20
PL45838	2991	ProvLab		1994	+	−	DVS	BCEFDG	A+C+	3	B, +80
PL73203	2992	ProvLab		1995	+	−	A	BCFEDG	A−C+	2
57639-199	2682			1990	+	−	O	BCFDEG	A−C+	6
Lysin+SA4	2683				+	−	I+IV	BCFDEG	A−C+	6
ISP1820 (ATCC 55047)	2406	Hone (25)	Chile	1983	+	−	46	BFECDG	A−C+	19
Ty2	2408	Hone (25)	USSR	1918	+	−	E1	GCEFDB	A−C+	9
200Ty	2272	Stocker			+	−	E1	GCEFDB	A−C+	9
403Ty	2273	Stocker						BCEFDG	A−C+	3	G, −130
541Ty	2758	Stocker						GCDEFB	A−C+	7
R1637	2692	ProvLab			+	−	E2	BFCEDG	A−C+	14
R1962	2693	ProvLab					UT(Vi-ve)	BCDEFG	A−C+	1	G, −130 (SP17−)
R1167	2694	ProvLab			+	−	A	GDCEFB	A−C+	19
R136	2695	ProvLab			+	−	D9	BCEFDG	A+C+	3	E, +40
R2101	2696	ProvLab			+	−	M1	BCEFDG	A−C+	3
R70	2697	ProvLab			+	−	46	BCFEDG	A+C+	2
R196	2698	ProvLab			+	−	E1	BCFEDG	A−C+	2
414Ty	3212	Stocker	Australia	1981	−	+	I-IV	BCEFDG	A−C+	3	E, +40
415Ty	3213	Stocker	The Netherlands	1982	−	+	UT(Vi-ve)	BCEFDG	A−C+	3	E, +40; G, −130 (SP17−)
416Ty	3214	Stocker	Japan	1982	−	+	UT(Vi-ve)	BCEFDG	A−C+	3	E, +40
417Ty	3215	Stocker	New Caledonia	1982	−	+	HIV	BECFDG	A−C−	22	E, +40; G, −130
418Ty	3216	Stocker	The Netherlands	1988	−	+	I+IV	BCEFDG	A−C+	3	E, +40
419Ty	3217	Stocker	The Netherlands	1988	+	−	I+IV	BCEFDG	A−C+	3	E, +40
420Ty	3218	Stocker	Japan	1982	+	−	UT(Vi-ve)	BCEFDG	A−C+	3	E, +40
421Ty	3219	Stocker	France	1984	+	−	UT(Vi-ve)	BCEFDG	A−C+	3	E, +40; G, −130 (SP17−)
422Ty	3220	Stocker	The Netherlands	1988	+	−	I+IV	BCEFDG	A−C+	3	E, +40
423Ty	3221	Stocker	Australia	1981	−	+	I+IV	BCEFDG	A−C+	3	E, +40
424Ty	3222	Stocker	The Netherlands	1988	+	−	I+IV	BCEFDG	A−C+	3	E, +40
425Ty	3223	Stocker			−	+	I+IV	BCEFDG	A−C+	3	E, +40
444Ty	3224	Stocker			+	−	I+IV	BCEFDG	A−C+	3	E, +40
445Ty	3225	Stocker						BCEFDG	A−C+	3	E, +40
446Ty	3226	Stocker			−	+	I+IV	BCEFDG	A−C+	3	E, +40
447Ty	3227	Stocker			+	−	I+IV	BCEFDG	A−C+	3	E, +40
701Ty	3485	Stocker						CEFBDG	A−C+	27	E, +40
702Ty	3486	Stocker						BCEFDG	A−C+	3	E, +40
TYT1668	3487	Mora	Chile		+	−	M1	BECDFG	A+C+	21
TYT1669	3488	Mora	Chile		+	−	UT(Vi-ve)	BCFDEG	A+C+	6	G, −170 (part of SP17 deleted)
TYT1670	3489	Mora	Chile		+	−	46	BCEFDG	A−C+	3
TYT1671	3490	Mora	Chile					BCEFDG	A−C+	3
TYT1672	3491	Mora	Chile		+	−	E1	GECDFB	A−C+	23
TYT1673	3492	Mora	Chile		+	−	F8	BCEFDG	A−C+	3
TYT1674	3493	Mora	Chile		+	−	E1	GECDFB	A−C+	23
TYT1675	3494	Mora	Chile		+	−	E1	GECDFB	A−C+	23
TYT1676	3495	Mora	Chile		+	−	E1	GECDFB	A−C+	23
TYT1677	3496	Mora	Chile		+	−	F8	BCEDFG	A−C+	5	G, +80
ST318	2778	Pang	Malaysia		+	−	Atypical	BCEFDG	A−C+	3
CT18	4072	Sanger Centre (50)	Vietnam						A+C+	9

^aDesignation used in the laboratory which provided the strain.

^bStrain number used at the Salmonella Genetic Stock Centre.

^cNML, National Microbiology Laboratory, Winnipeg, Canada (formerly Laboratory Centre for Disease Control) (R. Khakhria and David Woodward); Pang, T. Pang, Research Policy and Cooperation, World Health Organization, Geneva, Switzerland; Ho, M. Ho, Department of Microbiology and Infectious Diseases, University of Calgary; CDC, Centers for Disease Control (J. J. Farmer) (54); Selander, Robert K. Selander, Pennsylvania State University (part of SARB set, Salmonella Reference B) (6); ProvLab, Southern Alberta Provincial Lab (C. Anand); Hone, D. Hone, Center for Vaccine Development, Baltimore, Md. (25) (H238.2 is ISP1820 with aroD1013, and H251.1 is Ty2 with aroC101); Stocker, B. A. D. Stocker, Stanford University; Mora, G. Mora, University of Chile, Santiago, Chile.

^dFlagellar antigens d and j are two alternative states of the phase 1 antigen.

^ePhage typing with Vi typing phage II of Craigie and Felix (9) was done at the Laboratory Centre for Disease Control in Ottawa, Canada.

^fFragment order was determined by partial digestion of DNA with endonuclease I-CeuI and separation of the fragments by PFGE; the data are the order of I-CeuI fragments.

^gOrder and orientation of the A and C fragments as determined by PCR. A plus sign indicates that the fragment is in the normal (uninverted) orientation, and a minus sign indicates that the fragment is in the inverted orientation.

^hSpecific order of I-CeuI fragments, as shown in Fig. 4.

ⁱAltered I-CeuI fragments and their sizes (in kilobases) are indicated. All of the strains of S. enterica serovar Typhi yielded seven fragments following complete digestion with I-CeuI, and in most strains the sizes of these fragments were indistinguishable from the sizes observed in strain Ty2 (12) and CT18 (50). In certain strains the sizes of the fragments differed by the amounts indicated; larger sizes are indicated by a plus sign, and smaller sizes are indicated by a minus sign. Deletion of the I-Ceul-G fragment resulted in loss of all or part of Salmonella pathogenicity island 7 (SPI7), a 134-kb island which has the viaB operon for Vi exopolysaccharide (47). The sizes of all of the fragments not listed are not distinguishable from the sizes of the normal fragments.

Enzymes and chemicals.

Endonucleases were obtained from New England Biolabs (AvrII [= BlnI], I-CeuI, and SpeI) and Boehringer-Mannheim (XbaI). Taq polymerase and deoxynucleoside triphosphates were obtained from Amersham. Other chemicals, including Luria-Bertani medium and agarose, were obtained from Sigma Chemicals.

Endonuclease digestion and PFGE methods.

Preparation of high-molecular-weight genomic DNA, endonuclease cleavage of DNA in agarose blocks, separation of the DNA fragments by PFGE, and double-digestion techniques were performed as described previously (32, 35, 37). For digestion by I-CeuI, including partial digestion, we used the methods described previously (39).

Primers.

The primers were designed based upon the sequence of DNA flanking each of the seven rrn operons in S. enterica serovar Typhi CT18 (50) (Gen Bank accession no. NC_003198) by using the Primer3 program (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi/), and they were synthesized by the University Core DNA Services (Health Sciences Centre, University of Calgary). The sequences of the primers used in this study and their locations relative to rrn operons are shown in Table S1 in the supplemental material. The locations of the primers in the genomes of S. enterica serovar Typhimurium LT2 and S. enterica serovar Typhi CT18 are shown in Fig. 1. These primers were used in different combinations to amplify the rrn operons.

FIG. 1.

Location and order of the seven I-CeuI fragments on the chromosome (the rrn skeleton). oriC is the site of initiation of bidirectional replication; Ter is the termination site. The numbers with arrows represent the different primer combinations used to amplify the seven rrn operons (indicated by arrows outside the circles). The numbers outside the circles indicate the sizes of the I-CeuI fragments (in kilobases) based on the previously published sequences (45, 50). (A) S. enterica serovar Typhimurium LT2. (B) S. enterica serovar Typhi CT18.

PCR amplification and agarose gel electrophoresis.

Chromosomal DNA used in PCR was isolated with a Wizard genomic DNA purification kit (Promega) used in accordance with the manufacturer's instructions.

Each PCR was carried out by using a HotStart storage and reaction tube (Gordon Technologies Inc.) in an Eppendorf gradient thermal cycler. In each 50-μl (total volume) PCR mixture, 20 μl was the lower mixture and 30 μl was the upper mixture. The lower mixture contained 250 ng of template DNA, 1 μl of each primer (0.4 μM), and 2 μl of deoxynucleoside triphosphates (0.4 mM) (the concentrations of the components of the lower mixture were calculated based on the 50-μl reaction mixture), and the final volume was adjusted with double-distilled water. Denaturation was done at 90°C for 30 s to melt the wax pellet. After the mixtures were cooled to room temperature, the reactions were initiated by addition of 5 μl of 1× PCR buffer, 3 μl of MgCl₂ (1.5 mM), and 2.5 U of Taq DNA polymerase (0.25 μl), and 21.75 μl of double-distilled water was added (upper mixture) (the concentrations of the components of the upper mixture were calculated based on the 50-μl reaction mixture). PCR amplification was performed with 30 cycles of denaturation at 96°C for 1 min, annealing at 57°C for 1 min, and extension at 72°C for 10 min, followed by a final extension at 72°C for 10 min.

The PCR product was electrophoresed at 65 V on a 1% agarose gel in 0.5× Tris-borate-EDTA buffer (45 mM Tris, 45 mM boric acid, 10 mM EDTA [pH 8]) with 1 μg of ethidium bromide per ml. Following electrophoresis the gel was photographed under UV light.

Computer methods.

The individual fragment sizes were estimated by using the S. enterica serovar Typhi CT18 genome sequence (50) (GenBank accession no. NC_003198) and the BLAST Program produced by National Center for Biotechnology Information, Bethesda, Md. (www.ncbi.nlm.nih.gov/BLAST).

Phage typing and serotyping.

Bacteriophage typing was performed at the National Microbiology Laboratory (formerly the Laboratory Centre for Disease Control), Health Canada, as reported previously by Khakhria et al. (29), by using the methods and scheme described by Anderson and Williams (3) and the Vi phage of S. enterica serovar Typhi. Serotyping was performed at the National Microbiology Laboratory, Health Canada, by the methods described in a report on the Kauffmann-White scheme (53).

RESULTS

Partial digestion by I-CeuI in S. enterica serovar Typhi.

Partial digestion yielded the seven bands expected from complete digestion plus other bands resulting from a failure to cleave between adjacent fragments; representative data are shown in Fig. 2. All strains produced seven fragments (fragments A to G), which ranged from 44 to about 2,400 kb long, and usually the lengths were indistinguishable in different strains (with the exception of fragment G in strain SARB64 [Fig. 2, lane 2], which was about 130 kb smaller than the normal fragment G). However, the fragments resulting from partial digestion were different in different strains, indicating that the fragments in the chromosome are in different orders. For example, the following partial digestion bands were observed in Fig. 2, lanes 1 and 6: DF, EF, and DEF. These data indicate that the order is EFD. The order of these three fragments is different in other strains. For example, the partial digestion bands DF, DE, and DEF were observed in Fig. 2, lane 4; these data indicate that the order is fragment FDE. The order of many of the other fragments could be determined from the same gel. Some of the expected partial digestion fragments could not be recognized in other lanes because the gel was overloaded; in order to determine the order of all fragments, several gels with different loading and electrophoresis conditions were run if necessary.

FIG. 2.

Partial digestion of DNA of strains of S. enterica serovar Typhi with endonuclease I-CeuI, separation by PFGE, and staining with ethidium bromide. The gel is shown on the left. The fragments are shown on the right, and the inferred composition and sizes (in kilobases) are indicated. Lane 1, strain SARB63 (fragment order, CBEFDG; genome type 25); lane 2, SARB64 (fragment order, BDCEFG; genome type 19); lane 3, 27566 (fragment order, ECBFDG; genome type 26); lane 4, SA4864 (fragment order, BCFDEG; genome type 6); lane 5, SA4665 (fragment order, GFCEDB; genome type 16); lane 6, Ty2 (fragment order, GCEFDB; genome type 9).

A set of 136 strains of S. enterica serovar Typhi which were assembled from a variety of sources (Table 1) was analyzed by partial digestion with I-CeuI, as shown in Fig. 2. Many different arrangements of I-CeuI fragments were detected; these arrangements apparently resulted from inversions and translocations following recombination between rrn operons, as illustrated in Fig. 3. Figure 4 shows I-CeuI fragments A, B, C, D, E, F, and G as a contiguous block arranged in different orders; 27 different genome types are shown, some of which were not detected in the 136 strains. This linear unit is joined at both ends through fragment A to produce a circular chromosome, as shown in Fig. 1.

FIG. 3.

Proposed model of genomic rearrangements due to homologous recombination between rrn operons resulting in inversions or translocations. oriC is indicated by a shaded circle in fragment C (which corresponds to I-CeuI-C), and Ter is indicated by a shaded square in fragment A (I-CeuI-A). pro (proline requirement) and his (histidine requirement) indicate the positions of standard genes. (A) Both the A and C fragments are in the normal, uninverted orientation (A+C+), and the fragment order is I-CeuI-ABCDEFG. (B and C) Inversion. Recombination between rrnH and rrnG results in inversion of fragment A. (D to F) Translocation. Recombination between rrnC and rrnA deletes fragment D, which is reinserted by recombination with rrnE; this results in translocation to produce the fragment order ABCEFDG.

FIG. 4.

Order and orientation of I-CeuI fragments in 136 independent wild-type strains of S. enterica serovar Typhi. The sizes (in kilobases) of the fragments based on the sizes in CT18 (50) are indicated at the top, shown approximately to scale. The order of I-CeuI fragments B to G was determined by PFGE (Fig. 2) and was confirmed by PCR (Fig. 5). The I-CeuI-A fragment (2,422 kb) is inferred to join the left end to the right end of a fragment to form a circle. The orientation of I-CeuI fragments B, D, E, F, and G was inferred from the polarity of the rrn genes and was confirmed by PCR. The order and sizes of fragments for E. coli K-12 and S. enterica serovar Typhimurium LT2 (STM LT2) and the orientation of rrn operons are indicated at the bottom. The chromosomes of the different genome types are shown in the A+C+ orientation (with both the A and C fragments uninverted); the open square in fragment A indicates pro (proline utilization), and the open triangle indicates his (histidine requirement). Since both I-CeuI-C and I-CeuI-A are flanked by inverted rrn operons, these fragments could be inverted. The number of strains of each genome type that fall into each of the four sets of orientation of A and C fragments was determined from the PCR data (see Fig. 5). The dot in the I-CeuI-C fragment indicates the location of oriC; T indicates the terminus. The sizes of the fragments (in kilobases) were calculated from previously published sequences of S. enterica serovar Typhimurium LT2 (45) (GenBank accession no. NC_003197) and E. coli K-12 (5) (GenBank accession no. NC_000913); these fragments are shown at the bottom.

Genome types 1 to 6 had all possible rearrangements of the three small fragments, fragments D, E, and F; all of these rearrangements were detected, although some were much more frequent than others. Genome type 1 (I-CeuI-ABCDEFG) is most common in the enteric bacteria; e.g., it has been observed for S. enterica serovar Typhimurium LT2 (39), for 17 wild-type strains of S. enterica serovar Typhimurium (41), for S. enterica serovar Enteritidis (36), and for S. enterica serovar Paratyphi B (33), as well as for the following strains of E. coli whose complete sequences have been deposited in the GenBank database: E. coli K-12, E. coli CFT073, and E. coli O157:H7 (strains EDL933 and Sakai). However, only 4 of 136 strains of S. enterica serovar Typhi are genome type 1. Genome type 3 (I-CeuI-BCEFDG) is by far the most common genome type, represented by 59 strains. Translocation of the fragments to new locations could result from deletion of a fragment due to homologous recombination between rrn operons, followed by insertion of the fragment in a different rrn operon (Fig. 3). Genome types 7 to 12 involve the same arrangements of the D, E, and F fragments, but fragments B and G are inverted, presumably due to a crossover between rrnD and rrnE; these types are less common than genome types 1 to 6. Strain Ty2, which is widely used as a wild-type strain, is a genome type 9 organism; the detailed genomic cleavage map for the enzymes XbaI, BlnI, SpeI, and I-CeuI for Ty2 (38), the partial I-CeuI digestion data (this study), and the complete nucleotide sequence (12) all confirm the same order. Genome types 13 to 16 are types in which I-CeuI-F has been translocated to a position to the left of I-CeuI-C; genome types 17 to 20 and 21 to 24 represent equivalent translocations of I-CeuI-D, and I-CeuI-E, respectively. Genome types 13 to 24 are uncommon, and some of these types were not detected in the sample studied, although they might be found in a larger sample. All combinations in which I-CeuI-B and -G are adjacent to fragment A are represented in genome types 1 to 24; genome types 25 to 27 are three of the rarely encountered types in which a different fragment is adjacent to fragment A.

PCR to detect chromosomal rearrangements in independent wild-type strains of S. enterica serovar Typhi.

Partial digestion with I-CeuI determines the order of fragments, but not their orientation. The orientation of rrn operons limits the types of rearrangements which can be formed. The chromosome is composed of two replichores, and replication begins at oriC and proceeds bidirectionally to the termination of replication (Ter). For example, in S. enterica serovar Typhimurium LT2 (Fig. 1A) replichore 1 contains rrnCABEH and replichore 2 contains rrnDG, and all of these genes are oriented so that they are transcribed from oriC toward Ter. Homologous recombination between rrn operons to produce inversions or translocations can occur only between rrn operons in the same orientation; thus, the orientation of fragments which have two similarly oriented rrn operons at their ends, such as I-CeuI-B, -D, -E, -F, and -G, can be predicted. However, I-CeuI-C (which contains oriC) has two rrn operons at its ends transcribed away from oriC, and I-CeuI A (which contains Ter) has the two rrn operons at its ends transcribed toward Ter. Thus, both fragments can be inverted by recombination, which can occur between rrn operons at their ends (Fig. 1 and 3); therefore, the orientation of these fragments cannot be predicted by PFGE methods. Recombination between other rrn operons in different replichores results in inversions (Fig. 3).

Therefore, we used PCR analysis (as first described by Helm and Maloy [20]) to confirm the order of the seven I-CeuI fragments analyzed by PFGE and to determine their orientations. All 14 primers were designed from the S. enterica serovar Typhi CT18 sequence (Fig. 1B; see Table S1 in the supplemental material) and are located inside the genes which are adjacent to each of the seven rrn operons on either side. These primers were used in different pairwise combinations, based on the order of I-CeuI fragments predicted from PFGE analysis (Fig. 2); appropriate pairwise combinations resulted in ca. 6-kb amplicons containing the rrn operons. Since the orientations of fragments A and C were unknown, two primer combinations might work for these two fragments. PFGE indicated that the order of I-CeuI fragments for S. enterica serovar Typhi 26.047 was ABCEFDG (genome type 3) (Table 1). Template DNA from this strain gave successful amplification with the following primer pairs, indicating that specific I-CeuI fragments are adjacent: primers 3 and 4 (AB), primers 15 and 12 (BC), primers 14 and 13 (CE), primers 6 and 5 (EF), primers 10 and 9 (FD), primers 16 and 11 (DG), and primers 8 and 7 (GA) (Fig. 5A). Primer pairs 7-4 and 3-8, as well as primer pairs 15-14 and 12-13, would have produced an amplicon if the A and C fragments, respectively, were inverted; all these combinations failed. This indicates that both fragment A and fragment C are in the orientation present in most strains of Salmonella and E. coli. This was illustrated by the locations of the pro and his genes in I-CeuI-A and the location of oriC in fragment C at the clockwise end; we called these orientations A+C+ (Fig. 5A).

FIG. 5.

PCR analysis of the rrn skeleton of S. enterica serovar Typhi genome type 3 strains. The primer pairs are indicated above the gel showing the PCR products. Lane M contained the marker (HindIII-digested lambda). The inferred rrn skeleton is shown on the left. (Upper set) Template DNA of strain 26.047 (genome type 3, A+C+). (Lower set) Template DNA of strain 425Ty (genome type 3, A−C+).

When the same primer pairs were used with genomic DNA of S. enterica serovar Typhi 425Ty as the template (genome type 3) (Table 1), seven PCR amplicons were again produced (Fig. 5B). Primer pairs for amplification of the rrn operons between the following fragments produced the same pattern that was observed with S. enterica serovar Typhi 26.047 (Fig. 5A): B and C, C and E, E and F, F and D, and D and G. However, primer pairs 7-4 and 3-8 yielded PCR amplicons, whereas primer pairs 3-4 and 7-8 did not, indicating that the A fragment was inverted. The lack of amplicons for primer pairs 15-14 and 12-13 indicated that fragment C was in the normal orientation (uninverted); this structure is represented by A−C+ (Fig. 5B).

The order and orientation of the seven rrn operons were determined in the same way by using template DNA from all 136 wild-type strains of S. enterica serovar Typhi previously tested by PFGE. In all cases the template DNA yielded seven PCR amplicons, and the order agreed with the order of fragments determined by PFGE and with the orientation of fragments B, D, E, F, and G (inferred from the polarity of the rrn operons). The genome type and orientation of fragments A and C are indicated for each strain in Table 1 and are summarized in Fig. 4, which shows the number of strains for each genome type and for each of the four orientations of fragments A and C. Fragment I-CeuI-A was frequently inverted, and I-CeuI-C was rarely inverted. Thus, strains of S. enterica serovar Typhi showed many different rearrangements of the gene segments between rrn operons (the I-CeuI fragments).

Normally, we did not use all possible pairwise combinations of primers for testing each of the strains; we used only the primer pairs predicted by PFGE data to be effective. However, in a few cases we tested all possible combinations, and only those primer pairs predicted by the PFGE results were effective (data not shown).

Changes in I-CeuI fragment lengths.

PFGE data indicated that the lengths of I-CeuI fragments are highly conserved, for all seven sizes were indistinguishable from the sizes observed in strain Ty2 (38) for 86 of the 136 strains, as shown in Table 1. PFGE did not detect changes in fragment I-CeuI-C (517 kb), I-CeuI-D (134 kb), or I-CeuI-F (42 kb) in any of the strains (Table 2), although changes of a few kilobases should have been detectable by PFGE. Our methods could not detect changes in the large I-CeuI-A fragment (2,422 kb). The 57 strains with detectable changes included many types (Table 2). The following numbers of insertions were detected: 13 strains had 20- to 90-kb insertions in I-CeuI-B; 25 strains had 20- to 40-kb insertions in I-CeuI-E; and six strains had 15- to 80-kb insertions in I-CeuI-G. The only strains with a fragment with deletions were 12 strains with deletions in I-CeuI-G. These deletions were shown previously to result from a loss of all or part of Salmonella pathogenicity island 7, a 134-kb island which has the viaB operon for Vi exopolysaccharide, due to recombination between genes for tRNA^Phe. All these strains are untypeable by the phages used for Vi typing and are not agglutinated by Vi antiserum (7, 47). Thus, deletions were very rare. Insertions were more common, including the insertions in 25 strains with a 40-kb insertion in I-CeuI-E and 19 other insertions of various sizes in I-CeuI-B, -E, and -G. However, small changes in fragment sizes (1 to 2 kb in small fragments and about 10 to 20 kb in larger fragments), which could be detected by sequencing, were not observed by PFGE.

TABLE 2.

Changes in lengths of I-CeuI fragments in 136 strains of S. enterica serovar Typhi, measured by PFGE

I-CeuI fragment	Fragment size (kb)a	Deletion size (kb) (no. of strains)b	Insertion size (kb) (no. of strains)b	No. of fragments altered
A	2,422			NDc
B	706		20 (4), 80 (7), 90 (2)	13
C	517			0
D	134			0
E	149		30 (1), 40 (25)	26
F	42			0
G	839	70 (1), 134 (10), 170 (1)	15 (1), 20 (2), 50 (1), 80 (2)	18

^aNormal fragment sizes were determined from the sequence of strain CT18 (50); these sizes are very similar to those of strain Ty2 (12). The sizes are slightly corrected from those determined previously by PFGE (38).

^bDeletions and insertions represent decreases or increases in the size of a fragment relative to the normal size; the number of strains with each size class is indicated in parentheses. The specific strains are shown in Table 1.

^cND, not determined, because an accurate size of the I-CeuI-A fragments could not be determined by PFGE.

Determination of phage type by using Vi phage and of flagellar antigens.

The phage types of most of the wild-type strains are shown in Table 1. The data show that many different phage types are associated with specific genome types (for example, the 59 genome type 3 strains have many different phage types). This indicates that the phage type and the genome type are largely independent of each other. In a few cases, the frequency of occurrence of a specific genome type in a phage type is greater than that expected by chance. For example, of 19 phage type E1 strains, 7 are genome type 2, although genome type 2 is found in only 20 of 136 strains; this may represent isolation of similar strains from a clonal population. Flagellar antigens were determined for almost all of the strains; these antigens were usually the d antigen, although a few strains had the alternative j antigen.

DISCUSSION

Serovars of Salmonella which are pathogens for a wide range of hosts (generalists, such as S. enterica serovar Typhimurium) have very conserved genomes, while serovars which have very limited host ranges (specialists), such as S. enterica serovar Typhi, S. enterica serovar Paratyphi C, S. enterica serovar Gallinarum, and S. enterica serovar Pullorum, show a high frequency of rearrangements among wild-type strains due to homologous recombination between rrn operons (40). Helm et al. (19) showed in laboratory experiments that the frequency of recombination between rrn operons is not distinguishable in strains of S. enterica serovar Typhi and S. enterica serovar Typhimurium. This result suggests that differences in selective value rather than differences in rearrangement frequency are likely to be responsible for the higher frequency of rearrangements found in S. enterica serovar Typhi than in S. enterica serovar Typhimurium. Thus, greater survival of a recombinant might be due to the different lifestyles of generalist and host specialist species.

The following four classes of chromosome rearrangements might be formed due to recombination between rrn operons: deletions, duplications, translocations, and inversions (Fig. 3) (40) (56). Deletions of entire I-CeuI fragments would be readily detectable but were not observed, which is not surprising since all the fragments have essential genes. Deletions of a 9-kb segment between two rrn operons were detected in Bacillus subtilis (27), indicating that this short segment has no essential genes. Duplications would be detected by doubled intensity of the duplicated I-CeuI fragments and in partial digestion data, but these were not seen in these strains. Roth et al. (56) showed that 3% of the cells in cultures of S. enterica serovar Typhimurium LT2 had duplications of the segment between the closest rrn operons (the I-CeuI-F fragment) and found that these duplications are unstable since they revert to the haploid state; our failure to find strains with duplications confirms that duplications of this type are unstable.

Inversions and translocations occur frequently, for not one of the 136 strains tested had genome type 1 A+C+ (Fig. 4) (the genome order normally found in Salmonella and E. coli); even the four genome type 1 strains were A−C+. Thus, all strains tested had at least one translocation or inversion compared with the standard type. Two separate mechanisms could explain translocation. First, there could be deletion of a segment due to recombination between two rrn operons in the same replichore, thus forming a circular fragment, followed by reinsertion of the circle into another rrn operon (Fig. 3D and E). Hill and colleagues observed circles of a size appropriate for the interval from rrnB to rrnE (about 42 kb) (24); this corresponds to the 42-kb I-CeuI-F fragment. Second, translocations could be formed by duplications (e.g., to form DEFEF) followed by two independent deletions of fragments E and F (to form DFE). Genome types 1 to 6 are postulated to result from translocations of the small fragments I-CeuI-D, -E, and -F to form all six combinations. Genome types 13 to 16 involve translocation of I-CeuI-F into rrnD to the left of I-CeuI-C. Inversions are also commonly detected; e.g., genome types 7 to 12 are due to an inversion resulting from recombination between rrnD and rrnE. In addition, these strains have the translocations found in genome types 1 to 6; i.e., genome types 1 and 7 and genome types 2 and 8 have the same translocation, etc.

Several hypotheses were devised by Roth et al. (56) to explain the highly conserved genomes usually found in enteric bacteria. At first glance, it seems that rearrangements in S. enterica serovar Typhi have resulted in total reshuffling of the genome. However, in spite of the many genome rearrangements that we have detected in S. enterica serovar Typhi (Fig. 4), there is still considerable conservation (although not as much as in S. enterica serovar Typhimurium); our data support the gene balance and gene dosage hypotheses for genome conservation in S. enterica serovar Typhi.

Genome balance.

Lengths of the replichores between the oriC and Ter sites on a circular bacterial chromosome must be maintained for balanced bidirectional replication (23). Hill and Gray (22) showed that moving oriC relative to Ter reduced the growth rate of E. coli K-12. S. enterica serovar Typhi CT18 fragment sizes determined from the sequence (50) were used as the standards to calculate the genome balance for all wild-type S. enterica serovar Typhi strains (the sizes calculated from the sequence of strain Ty2 [12] are very similar). The positions of the origin of replication (oriC) and the termination of replication were determined from the E. coli K-12 oriC sequence (46) (GenBank accession no. K01789) and the position of the dif (deletion-induced filamentation) sequence of E. coli (31) (GenBank accession no. S62735), respectively; locations on the chromosome of S. enterica serovar Typhi CT18 were detected with BlastN (2).

Rearrangements only in fragments D to F, which result in genome types 1 to 6 (Fig. 4), do not change the genome balance because they are all in the same replichore. When these genome types are A+C+, the size of replichore 1, from oriC in I-CeuI-C through fragments E, F, D, G, and A to Ter, is 2,362 kb; the size of replichore 2, from fragment C through fragments B and A to Ter, is 2,447 kb. Genome balance was calculated by dividing the size of each replichore by the total genome size; the off-balance value, which was one-half the difference between the replichore sizes, was 42.5 kb (Fig. 6 and Table 3). The genome balance was calculated for genome types 1 to 6 for the four different orientations in fragments A and C, which changed the lengths of the replichores, and the numbers of strains of each type were summarized from Fig. 4, as shown in Table 3. Of the 100 strains with genome types 1 to 6, 28 were A+C+ (42.5 kb off-balance), 72 were A−C+ (25.5 kb off-balance), and none were A+C− or A−C− (over 400 kb off-balance). Thus, the fragment orientations which gave the best chromosome balance were the most frequently detected orientations; among independent wild-type strains, strains with inversions of I-CeuI-C, which would have been highly unbalanced, were not detected. It has been noticed previously by Eisen et al. (14) that chromosomal inversions around the origin and termination of replication are usually symmetrical, thus retaining chromosome balance, even when comparisons are done between groups as widely separated as E. coli and Vibrio cholerae.

FIG. 6.

Analysis of genome balance. The chromosome structure of genome type 3 A+C+ illustrates the genome balance. Since the chromosome is bidirectionally replicated from oriC, two replichores are shown. The dot in fragment C represents oriC. (A) Linear form. The cross-hatched line represents calculation of the total fragment sizes to determine the length of replichore 1 and replichore 2. (B) Circular form. Rep1, replichore 1; Rep2, replichore 2.

TABLE 3.

Genome balance analysis of S. enterica serovar Typhi strains belonging to genome types 1 to 6^a

Parameter	A+C+b		A−C+		A+C−		A−C−
	Replichore 1	Replichore 2	Replichore 1	Replichore 2	Replichore 1	Replichore 2	Replichore 1	Replichore 2
Genome balancec	0.491	0.508	0.505	0.495	0.589	0.410	0.604	0.395
Amt off-balance (kb)d	−42.5	42.5	25.5	−25.5	432.5	−432.5	500.5	−500.5
No. of isolatese	28		72		0		0

^aGenome types 1 to 6 have the fragment order I-CeuI-ABC(DEF)G; DEF can be in any order.

^bThe orientation of fragments A and C as shown in Fig. 4.

^cGenome balance was determined from fragment sizes (Fig. 6).

^dSee the Discussion for the method used to calculate the amount off-balance.

^eThe number of strains in each type of orientation, as shown in Fig. 4.

The remaining 36 strains were also grouped into sets, such as genome types 7 to 12, etc., and were analyzed to determine chromosome balance and frequency; these strains are shown in Table S2 in the supplemental material. Most of the genome types occur in small numbers; many of them are off-balance by 100 to 300 kb. The most striking are genome types 25 and 27 (731.5 kb off-balance), represented by two strains, and one strain of genome type 26 (582.5 kb off-balance).

It was surprising to find such a high degree of genomic rearrangement in S. enterica serovar Typhi (and in S. enterica serovar Paratyphi C [21], which causes paratyphoid fever) since the genomes of most enteric bacteria are highly conserved. We propose that the insertion of a large block of foreign DNA into these two organisms (134 kb in Salmonella pathogenicity island 7 containing the viaB genes inserted into I-CeuI-G) (38) may have resulted in chromosome imbalance which triggered a series of rearrangements (32).

Gene dosage.

Due to bidirectional replication, there are extra copies of genes close to oriC, resulting in increased gene expression (57), and since dosage differences may cause the strengths of promoters to be evolutionarily optimized for their specific positions, cells in which genes are a different distance from oriC are at a selective disadvantage (56). Although rearrangements were seen in all 136 strains, some classes of rearrangements were rare or never detected. Three small fragments, I-CeuI-D, -E and -F, are frequently translocated from the normal order DEF into all possible orders in genome types 1 to 6 and also into new sites on both sides of I-CeuI-C in genome types 12 to 24. However, there is not a single case of translocation of any of these three fragments into rrnH between I-CeuI-A and I-CeuI-G, although this translocation within the same replichore would not change the chromosome balance (Fig. 4). We postulate that translocations involving rrn operons occur at random at these positions, as well as at other locations, but that the cells are at a selective disadvantage because, according to the gene dosage hypothesis, such translocations move the genes far from oriC such that the copy number and thus the rate of gene expression are less adaptive; thus, rearrangements of these types occur, but strains with these genome types do not survive in nature.

Gene position might also be conserved because promoters are tuned to the degree of local supercoiling of the DNA, so that rearrangements would be disadvantageous (55). This may partially explain the conservation that we detected. The direction of transcription and replication is normally the same in highly expressed genes, such as those for rRNA, and this may be an additional basis for conservation of the genome (55, 56). It must be emphasized that in the many inversions and translocations observed in strains of S. enterica serovar Typhi, including those between replichores, I-CeuI fragments retain the same orientation with respect to replication, because homologous recombination between rRNA operons, all of which have transcription oriented from oriC to Ter, enforces this.

Genomic rearrangements other than those involving rrn operons are rare in S. enterica serovar Typhi. This was revealed by the fact that the lengths of the I-CeuI fragments in wild-type strains seldom varied, except due to rare insertions or deletions in individual fragments (Table 3), which indicates that the vast majority of genome rearrangements which occur are due to recombination between rrn operons. Recombination might occur between the 25 IS200 elements of S. enterica serovar Typhi CT18 or between the six IS200 elements in S. enterica serovar Typhimurium LT2, for IS200 is 700 bp long and should be a good target for homologous recombination. Recombination between IS200 elements can occur; PCR methods detected an inversion in S. enterica serovar Typhi between two IS200 elements (1) (these were in the same I-CeuI fragment and thus were undetectable by PFGE methods in the present study); in addition, unstable duplications with IS200 endpoints were detected in S. enterica serovar Typhimurium (18). However, there is no evidence that rearrangements involving IS200 or other duplicated regions occurred among the 136 strains in this study, for such events should cause simultaneous changes in the lengths of two I-CeuI fragments if their ends flank an rrn operon.

The genomes of most strains of E. coli and Shigella appear to be stable, like the genomes of most strains of Salmonella, but studies with I-CeuI digestion indicated that rearrangements occur frequently in Shigella dysenteriae and Shigella flexneri strains; originally, this was postulated to be due to rrn-mediated rearrangements (60). However, the complete genome sequences show that although both S. flexneri 2a strain 301 (28) and strain 2457T (61) have large symmetrical chromosomal inversions spanning the replication origin and terminus, most of these rearrangements are due to recombination between insertion sequences, as was also seen in two strains of Yersinia pestis (11). This is quite unlike S. enterica serovar Typhi, in which recombination occurred between rrn operons.

The 136 strains of S. enterica serovar Typhi belong to many different phage types (Table 1). Phage types are very stable and thus have been used a great deal in bacterial typing (49). Genome types are relatively stable but show occasional changes. For example, during 10 years, as we have worked with isolates of S. enterica serovar Typhi strain Ty2 (genome type 9 A−C+) (38), we have detected only rare changes in genome type (less than 1 per 100 single-colony isolates). Hughes (26) summarized the frequency of genomic rearrangements for members of many different genera and within the same species and genus. He noted that rearrangements occurred most frequently in clinical isolates of pathogens of humans and animals. He concluded that orderly and efficient replication of the genome and global regulation of gene expression are both critically important; our data support these conclusions, for we show the importance of genome balance (for genome replication) and gene dosage (for gene expression).