Comparison of Two Major Forms of the Shigella Virulence Plasmid pINV: Positive Selection Is a Major Force Driving the Divergence

Abstract

All Shigella and enteroinvasive Escherichia coli (EIEC) strains carry a 230-kb virulence plasmid (pINV) which is essential for their invasiveness. There are two sequence forms, pINV A and pINV B, of the plasmid (R. Lan, B. Lumb, D. Ryan, and P. R. Reeves, Infect. Immun. 69:6303-6309, 2001), and the recently sequenced pINV plasmid from Shigella flexneri serotype 5 is a pINV B form. In this study we sequenced the majority of the coding region of the pINV A form from S. flexneri serotype 6 other than insertion sequence or related sequences and compared it with the pINV B form. More than half of the genes sequenced appear to be under positive selection based on their low ratio of synonymous to nonsynonymous substitutions. This high proportion of selected differences indicates that the two pINV forms have functional differences, and comparative studies of pathogenicity in different Shigella-EIEC strains could be informative. There are also genes absent in the S. flexneri serotype 6 plasmid, including the sepA gene encoding serine protease, the major secreted protein of S. flexneri serotype 2a, and the stbAB genes, which encode one of the two partition systems found in S. flexneri serotype 5. The incompatibility of the two pINV forms appears to be due to either small differences in the mvpAT postsegregational killing system or the presence of an unknown system in pINVA.

Shigella strains are the causative agent of shigellosis or bacillary dysentery, a diarrheal disease of humans (1). Molecular evidence derived from studies involving DNA hybridization, multilocus enzyme electrophoresis, and sequencing of housekeeping genes indicates that Escherichia coli and all members of the genus Shigella belong to the same species (8, 30, 35). Enteroinvasive E. coli (EIEC) and Shigella have long been known to be similar (28) and can in fact be treated as comprising a single pathovar of E. coli (21). Most Shigella strains fall into three clusters (clusters 1, 2, and 3) based on the sequence of housekeeping genes (36), but Shigella dysenteriae serotypes 1, 8, and 10 and Shigella sonnei are not in any of the clusters while Shigella boydii serotype 13 is outside E. coli.

The pathogenesis of EIEC and Shigella involves invasion of mucosal epithelial cells of the large intestine (1, 34). Virulence in Shigella strains is dependent on the presence of a large 210- to 230-kb plasmid. The virulence plasmids pWR100 in S. flexneri serotype 5, pMYSH6000 in S. flexneri serotype 2a, and pSS120 in S. sonnei, together with those of other Shigella bacteria, have been shown to carry determinants for invasiveness and the ability to cause disease. These large plasmids are collectively termed pINV plasmids (13), which are also present in EIEC strains. The cell invasion capacity of Shigella-EIEC is determined by a cluster of 38 genes within a 32-kb segment of the pINV plasmid, often referred to as the entry or invasion region, which includes genes for invasins, molecular chaperones, motility, regulation, and a specialized type III secretion apparatus (32).

In a previous study three pINV genes, ipgD, mxiC, and mxiA, were sequenced from strains representing a range of Shigella isolates and two EIEC isolates and showed that the plasmid exists in two related but clearly distinct sequence forms (pINV A and pINV B) (20). The phylogenetic relationships of the plasmid and chromosomal genes of Shigella strains are largely consistent. The cluster 1 and cluster 3 strains tested have pINV A and pINV B plasmids, respectively. However, of the three cluster 2 organisms, S. boydii serotypes 9 and 15 have pINV A while S. boydii serotype 11 has a pINV B plasmid. S. dysenteriae serotypes 8 and 10 and S. sonnei and also EIEC strains, none of which fall into the three clusters, were all found to have either pINV A or pINV B, except for S. dysenteriae serotype 1, which has a distinct pINV form. The outlier organisms other than S. dysenteriae serotype 1 must have obtained the plasmid relatively recently, after divergence of the two forms.

The invasion plasmid from S. flexneri serotype 5 strain M90T has been sequenced by two groups (9, 50). The two sequences are essentially identical but use different numbering and often different gene names. We use the numbering and gene names of Buchrieser et al. (9). The plasmid contains 93 segments, totaling 58 kb, which are homologous to known or putative insertion sequences (ISs), suggesting a remarkable history of IS-mediated acquisition of DNA. Analysis of the GC content, position, and function of non-IS-related genes indicates that the plasmid contains blocks of genes of various origins. There are three partition systems, two functional and one remnant, and the GC content of the replication system genes is different from that of the partition genes, indicating yet another source for this region (9). It appears that pINV was assembled from several different plasmids (9).

In this study we sequenced the non-IS-related coding regions of an A form of pINV from S. flexneri serotype 6 (designated the F6 plasmid) to compare with the sequenced S. flexneri serotype 5 pINVB form (designated the F5 plasmid). The divergence ranged from 0 to 5.56%, with most genes differing by 0 to 1.0%, consistent with divergence levels in housekeeping genes. However, one group of genes and a few isolated genes showed much higher levels of divergence. The comparison indicated that divergence in these genes is driven by selection pressure, which aids our understanding of virulence variation and mechanisms of pathogenesis.

MATERIALS AND METHODS

Bacterial strains.

An S. flexneri serotype 6 isolate (M1382) was used for the study, with S. flexneri serotype 5 isolate M1356 included as a control for PCR amplification. Both isolates were used in previous studies (20, 36).

DNA sequencing strategy.

pINV plasmid DNA was prepared as described previously (42). Primers were designed to amplify overlapping 1-kb segments for each non-IS region based on S. flexneri serotype 5 pINV (9).

PCR and DNA sequencing.

The following cycling profile was used: initial denaturation at 94°C for 2 min and then 35 cycles of 94°C for 15 s, 55 to 60°C for 30 s, and 72°C for 1 min with a final extension at 72°C for 5 min. The annealing temperature was between 55 and 60°C depending on primer pair. PCR products were purified with the Wizard purification system (Promega). Samples were sequenced using dye terminator technology (Perkin-Elmer) through the Sydney University and Prince Alfred Hospital Macromolecule Analysis Center and with an automated 377 DNA sequencer (Applied Biosystems).

DNA sequence editing and analysis.

Sequences were edited using the PHRED, PHRAP, and CONSED programs (http://bozeman.mbt.washington.edu) (12). Sequence comparisons were done using MULTICOMP (39). Open reading frame (ORF) analysis and database searches were done using ORF finder and BLAST search facilities at the National Center for Biotechnology Information (http://www.ncbi.nlm.nih.gov). The synonymous and nonsynonymous substitution rates were calculated using a program provided by Li (23).

RESULTS AND DISCUSSION

Comparison of pINV B and pINV A coding regions.

The pINV plasmid genome has a high proportion of IS DNA, and the non-IS DNA was treated as 35 regions for which we give the start position of the first gene and the end position of the last gene in Table 1. We have used the gene annotation of Buchrieser et al. (9), and their genetic map will be of assistance in reading this paper. We attempted to sequence from S. flexneri serotype 6 all non-IS coding regions defined by Buchrieser et al. (9), a total of 103,727 bp. Both PCR and sequencing depended on the use of primers based on S. flexneri serotype 5 pINV, and for a number of regions we failed to obtain a PCR product, even after attempts with alternative primers. We obtained a sequence of 79,364 bp as 25 segments (Table 1), and six genes (see below) were shown by Southern blotting to be absent, accounting for a further 6,986 bp. We considered this to be more than sufficient for a comparison of the two plasmids and did not attempt to determine the nature of the changes that led to the failure to amplify the remaining 17 kb. We are naming the plasmid pINV_F6_M1382 in the GenBank entries to include both serotype and strain name, as we anticipate further sequencing of pINV genes, but in this paper refer to the two sequences as pINV F5 and pINV F6.

TABLE 1.

Gene coding regions of pINV F6

Region	First gene	Starta	Enda	Length (bp)	Sequenced in F6	Status in F6b	Accession no.
1	icsP	991	1938	948	948	Complete	AY206427
2	ospB	3485	5420	1,936	1,936	Complete	AY206428
3	ospD2 (part a)	9501	13971	4,471	2,860	9501-12361	AY206429
	ospD2 (part b)				581	13393-13971	AY206430
4	ospD1	20964	21641	678	678	Complete	AY206431
5	orf22	22153	23315	1,163	1,163	Complete	AY206431
6	parA	29020	31199	2,180	2,180	Complete	AY206432
7	virF	38511	39299	789	789	Complete	AY206433
8	ospE2	40311	40577	267	267	Complete	AY206450
9	ipaH2.5	43257	44948	1,692		Southern positive
10	orf47	46738	47513	776	776	Complete	AY206434
11	ospC2	49695	51149	1,455	1,455	Complete	AY206435
12	sepA	54594	58688	4,095		Absent
13	ipaH7.8	64062	65759	1,698	1,698	Complete	AY206436
14	ipaH4.5	66187	67911	1,725		Poor sequence
15	ospD3	74911	78350	3,440	3,440	Complete	AY206437
16	orf81	80644	82400	1,757		No amplification
17	orf85	85586	86343	758		Absent
18	ospC3	90851	92305	1,455	1,455	Complete	AY206438
19	orf94	94791	95567	777		Absent
20	ipaJ	99403	131402	32,000	32,000	Complete	AY206439
21	orf136	136624	137028	405	405	Complete	AY206440
22	orf137	137580	138356	777	777	Complete	AY206440
23	virA (part a)	144615	149654	5,040	1,203	144615-145817	AY206441
	virA (part b)				3,309	146346-149654	AY294290
24	ushA	151839	153491	1,653	1,653	Complete	AY206442
25	phoN1	155789	156538	750		Poor sequence
26	orf157	157598	163394	5,797		Absent
27	orf169a	169678	170550	873	873	Complete	AY206443
28	orf171	171912	173950	2,039	526	Not attempted	AY206444
29	ipaH9.8	174343	177299	2,957	1,638	174343-175980	AY206445
30	orf182	181974	184886	2,913		No amplification
31	orf185	185369	189266	3,898	3,815	185452-189266	AY206446
32	trbH	191768	199128	7,375	7,375	193484-199128	AY206447
33	orf201	201725	203963	2,239	3,836	201725-204398	AY206448
34	ipaH1.4	206811	209194	2,384	1,728	206811-208538	AY206449
35	orf212	212330	212896	567		No amplification
Total				103,727	79,364

^aStart and end of pINV F5 gene coding sequences are as in the work of Buchrieser et al. (9). The first gene of the region is indicated for easy identification.

^bIf a region was only partially sequenced, the start and end positions of the sequenced segment are given. For regions where no sequence was obtained, the reasons are given: Southern positive means that PCR failed but that Southern hybridization indicated that the region is present. No amplification means that PCR failed but that Southern hybridization was not done. Absent means that one or more genes of the region were shown to be absent by Southern hybridization.

There were 511 polymorphic nucleotides, 311 (60.86%) of the substitutions involving amino acid substitution. The average percentage difference is 0.74%, ranging from 0 (11 genes-ORFs) to 5.56% for individual genes (Table 2; Fig. 1). There are 65 genes with differences in the range of 0 to 0.99% with an average of 0.35%, resembling mxiA and mxiC, sequenced previously from multiple Shigella strains (20) with 0.15 and 0.56% differences, respectively, and 15 genes with differences in the range of 1.0 to 5.56% and an average of 2.43%, including ipgD of our previous study, with a 3.4% difference.

TABLE 2.

Sequence variation of pINV between F6 and F5 and between F2A and F5 plasmids

Gene	Region	Positiona	Length (bp)	No. of polymorphisms			Ksc	Kac	Ratio	F2A-F5 comparison
				Total polyb	Nonsynb	%				Total polyb	Nonsynb	%
icsP	1	991	948	5	4	0.527	0.003	0.006	0.51	0	0	0
ospB	2	3485	867	1	1	0.115	0	0.002	0	0	0	0
phoN2	2	4680	741	2	2	0.27	0	0.003	0	2	2	0.27
ospD2	3	9501	1,710	1	1	0.058	0	0.001	0	1	1	0.058
ospF	3	11642	720	1	1	0.139	0	0.002	0	2	2	0.278
orf13	3	13393	579	5	5	0.864	0	0.012	0	1	1	0.173
ospD1	4	20964	678	6	5	0.885	0.004	0.011	0.4	2	1	0.295
orf22	4	22153	315	3	1	0.952	0.019	0.005	3.78	1	0	0.317
ipgB2	4	22749	567	3	1	0.529	0.022	0.003	8.28	0	0	0
parA	6	29020	1,200	2	0	0.167	0.008	0	NAd	2	0	0.167
parB	6	30219	981	1	0	0.102	0.003	0	NA	2	0	0.204
virF	7	38511	789	1	0	0.127	0.004	0	NA	0	0	0
ospE2	8	40311	267	8	7	2.996	0.039	0.033	1.182	0	0	0
orf47	10	46738	483	3	2	0.621	0.006	0.006	0.9	0	0	0
orf48	10	47211	303	0	0	0	0	0	NA	0	0	0
ospC2	11	49695	1,455	3	3	0.206	0	0.003	0	3	3	0.206
ipaH7.8	13	64062	1,698	6	2	0.353	0.007	0.002	4	4	0	0.236
ospD3	15	74911	1,698	5	3	0.294	0.003	0.003	1.28	0	0	0
ospC1	15	76938	1,413	3	2	0.212	0.002	0.002	0.9	2	2	0.142
ospC3	15	90851	1,455	8	4	0.55	0.012	0.004	3.15	56	24	3.849
ipaJ	20	99403	780	0	0	0	0	0	NA	0	0	0
virB	20	101128	930	3	0	0.323	0.009	0	NA	2	0	0.215
acp	20	102274	237	0	0	0	0	0	NA	1	1	0.422
ipaA	20	102514	1,902	6	5	0.315	0.001	0.004	0.37	1	1	0.053
ipaD	20	104424	999	24	19	2.402	0.022	0.023	0.96	2	1	0.2
ipaC	20	105473	1,092	28	15	2.564	0.048	0.019	2.57	3	3	0.275
ipaB	20	106584	1,743	29	15	1.664	0.029	0.011	2.68	2	1	0.115
ipgC	20	108332	468	7	0	1.496	0.045	0	NA	1	0	0.214
ipgB1	20	108856	627	3	2	0.478	0.005	0.004	1.08	2	1	0.319
ipgA	20	109498	390	1	0	0.256	0.007	0	NA	0	0	0
icsB	20	109900	1,485	30	19	2.02	0.033	0.016	2.12	4	4	0.269
ipgD	20	111698	1,617	55	38	3.401	0.05	0.033	1.52	11	9	0.68
ipgE	20	113324	363	9	4	2.479	0.039	0.018	2.17	0	0	0
ipgF	20	113686	459	10	6	2.179	0.04	0.018	2.25	1	1	0.218
mxiG	20	114153	1,116	27	16	2.419	0.052	0.017	3.08	0	0	0
mxiH	20	115276	252	14	9	5.556	0.121	0.054	2.22	0	0	0
mxiI	20	115540	294	2	1	0.68	0.01	0.004	2.48	0	0	0
mxiJ	20	115839	726	3	3	0.413	0	0.005	0	0	0	0
mxiK	20	116561	528	3	1	0.568	0.011	0.002	4.99	0	0	0
mxiN	20	117030	696	2	1	0.287	0.004	0.002	2.47	1	0	0.147
mxiL	20	117718	408	1	1	0.245	0	0.004	0	1	1	0.245
mxiM	20	118109	429	0		0	0	0	NA	0	0	0
mxiE	20	118667	633	1	0	0.158	0.014	0	NA	0	0	0
mxiD	20	119286	1,701	9	5	0.529	0.007	0.004	1.7	3	1	0.176
mxiC	20	121005	1,068	6	4	0.562	0.011	0.006	1.97	1	1	0.094
mxiA	20	122085	2,061	3	1	0.146	0.003	0.001	3.55	1	0	0.049
spa15	20	124158	402	2	1	0.498	0.007	0.004	1.78	0	0	0
spa47	20	124563	1,293	9	5	0.696	0.012	0.006	2.01	2	0	0.155
spa13	20	125948	339	1	0	0.295	0.008	0	NA	0	0	0
spa32	20	126273	879	0	0	0	0	0	NA	0	0	0
spa33	20	127145	882	4	2	0.454	0.006	0.003	2.43	1	1	0.134
spa24	20	128031	651	11	3	1.69	0.057	0.007	8.77	0	0	0
spa9	20	128696	261	3	1	1.149	0.022	0.005	4.84	0	0	0
spa29	20	128956	771	4	1	0.519	0.017	0.002	8.1	1	1	0.13
spa40	20	129733	1,029	6	2	0.583	0.017	0.003	6.37	0	0	0
orf131a	20	130774	243	2	0	0.823	0.053	0	NA	0	0	0
orf131b	20	131133	270	0	0	0	0	0	NA	0	0	0
orf136	21	136624	405	0	0	0	0	0	NA	0	0	0
orf137	21	137580	777	4 (5)e		0.515				2	1	0.279
virA	23	144615	1,203	7	4	0.582	0.011	0.004	2.54	0	0	0
icsA	23	146346	3,309	9	8	0.272	0.001	0.003	0.33	2	1	0.06
ushA	24	151839	1,653	4	2	0.242	0.009	0.002	4.62	0	0	0
orf169a	27	169678	483	1	0	0.207	0	0.003	0	0	0	0
orf169b	27	169912	639	3	1	0.469	0.009	0.002	4.62	1	1	0.156
ccdA	28	173425	219	0	0	0	0	0		0	0	0
ccdB	28	173624	327	0	0	0	0	0		0	0	0/PICK>
ipaH9.8	29	174343	1,638	4	3	0.244	0.002	0.003	0.64	3	3	0.183
orf186	31	186214	1,089	4	2	0.367	0.005	0.003	2.03	0	0	0
virK	31	187307	951	32	18	3.365	0.077	0.026	2.94	3	3	0.315
msbB2	31	188319	945	11	6	1.164	0.021	0.009	2.32	0	0	0
trbH	32	191768	720	4	3	0.556	0.004	0.006	0.61	1	0	0.139
mvpA	32	191777	399	1	1	0.251	0	0.004	0	0	0	0
mvpT	32	192175	228	2	1	0.877	0.013	0.007	1.95	1	0	0.439
traI	32	192484	5,278	21 (35)e		0.4				12 (22)e		0.23
traX	32	197767	747	1	0	0.134	0.004	0	NA	3	3	0.402
finA	32	198568	561	3	2	0.535	0.01	0.005	2.01	1	1	0.178
orf201	33	201725	591	3	2	0.508	0.005	0.004	1.12	1	0	0.169
repB	33	202555	261	0	0	0	0	0	NA	0	0	0
repA	33	203106	858	0	0	0	0	0	NA	0	0	0
ipaH1.4	34	206811	1,728	12	7	0.694	0.008	0.006	1.41	3	1	0.174

^aPosition is the first base of the gene as annotated in the work of Buchrieser et al. (9).

^bTotal poly, total number of polymorphisms; Nonsyn, nonsynonymous substitutions.

^cK_s and K_a designate rates of synonymous and nonsynonymous substitutions, respectively.

^dNA, not applicable because of division by zero.

^eValues in parentheses include indels.

FIG. 1.

Plot of percentage nucleotide differences between F5 and F6 pINV genes. The nonsynonymous percentages were plotted above the percentage of total variation. The genes on the x axis are in map order, allowing clustering of variation to be visualized. A bar above a group of gene names indicates that these genes are in the same inter-IS region (Table 1).

There are some systematic patterns in the levels of difference (Fig. 1). The 19 genes with differences above the average of 0.71% are clustered, the majority being in the 32-kb entry region. Ten of these genes are in two groups of four (ipaD, ipaC, ipaB, and ipgC) and six (icsB, ipgD, ipgE, ipgF, mxiG, and mxiH) separated by only two genes. This pattern could have arisen by recombination, However, the ipgD, mxiA, and mxiC genes, which we studied previously for multiple Shigella strains, cover most of the range in level of sequence differences, but the difference between ipgD and the others is seen at all levels, from divergence of pINV A and B forms to divergence between closely related strains (20). This is consistent with these two regions, which include ipgD, having a mutation rate three to four times higher than the average for the plasmid genome. If the clustering of divergent genes were due to recombination, one would expect a long branch length at the point where the recombination occurred, but otherwise similar levels of variation as for other genes. As described below these genes are thought to be under selective pressure which drives the faster divergence. Perhaps it is more efficient in genome organization for these genes to be clustered for both functional and mutational purposes.

Values of K_s, K_a, and the ratio K_s/K_a are also given in Table 2. K_s is a measure of the rate of synonymous substitutions, being base substitutions without amino acid substitution, and K_a is a similar measure for substitutions which cause amino acid substitution. Synonymous substitutions are generally neutral or of very low adaptive value. Many amino acid substitutions are deleterious and some are neutral, but, critically, adaptive mutations are almost always in the latter category (18). The ratio K_s/K_a is often used as an indicator of neutral or adaptive variation. All genes are subjected to purifying selection to remove deleterious mutations, and it appears that this applies to most nonsynonymous substitutions, giving for an E. coli-Salmonella enterica comparison an average K_s/K_a ratio of 10 to 20 for most genes (45). Genes subject to selection for change have a lower value for K_s/K_a, as selection for adaptive amino acid substitutions increases K_a, which lowers the K_s/K_a ratio. In contrast, when strong conservation of amino acid sequences is required for biological function, K_s greatly exceeds K_a.

Of the 80 genes sequenced, 11 have only synonymous substitutions and 9 have only nonsynonymous substitutions, while for the others K_s/K_a ranges from 0.37 to 8.8. For nine genes K_s/K_a is under 1.0. Apart from the genes with only synonymous substitutions, there are only three genes, spa24, ipgB2, and spa29, with a ratio above 8, which is close to the range for housekeeping genes. Spa24 and Spa29 are putative inner membrane proteins (6). For the majority (35 genes) the ratio is between 1 and 8, and it is difficult to give a cutoff that indicates which are under selection pressure for change. For this discussion we use the value of 4.0 as the cutoff, which we consider to be conservative, given that most genes have K_s/K_a ratios between 10 and 20 in E. coli-S. enterica. We categorize the 30 genes with K_s/K_a ratios between 1 and 4 as being under positive selection pressure and the 5 genes with K_s/K_a ratios between 4 and 8 as being intermediate and not easily categorized. Therefore, of the 80 genes, a total of 48, including 9 with nonsynonymous changes only, 9 with K_s/K_a ratios less than 1, and 30 with K_s/K_a ratios between 1 and 4, appear to be under positive selection according to the K_s/K_a ratio. This is a very high proportion, indicating the adaptive nature of invasive virulence and its functional complexity. For many genes the total number of substitutions is very low, but the overall picture is of many genes under selection pressure for change.

Genes for secreted proteins.

The patterns of variation in secreted proteins are of particular interest. Many of these proteins are effector molecules interacting with host cells, and such proteins have been reported to be more diverse in other type III secretion systems (19, 33). There are at least 25 genes encoding proteins secreted by the Mxi-Spa secretion apparatus (9), of which we were able to sequence 23. The average sequence difference is 0.87%, which is not much higher than average for the genes sequenced. However, five genes (ospB, ospC2, ospD2, ospF, and mxiL) have only nonsynonymous substitutions and another four (ipaH9.8, ospC1, ipaA, and ipaD) have a K_s/K_a ratio less than 1.0, an indication that these genes are under strong positive selection pressure for amino acid variation.

The ipa genes (ipaA to ipaD) encode proteins which are invasins (27, 34). They are humoral immunogens, the predominant antigens recognized by sera from human patients and animal models (15, 38, 49), and so it is not surprising to find that all are under positive selection pressure with K_s/K_a ratios for ipaA and ipaD less than 1.0 and for ipaC and ipaB around 2.5. IpaC and IpaB proteins are inserted into the host membranes to form a pore for translocation of other invasins into target cells (5). Variation in these genes may be due to adaptation for optimal interaction with the host cell.

Two of the secreted proteins are toxins. OspD2 is annotated as shET2-2, an enterotoxin, by Venkatesan et al. (50), and OspD3 (SenA) is a known enterotoxin (29). Both genes appear to be under positive selection. ospD2 has no synonymous substitutions with a single amino acid substitution, and ospD3 has a K_s/K_a ratio of 1.28. However, the enterotoxin gene ospD3 (senA) is found in only 75% of EIEC strains and 83% of Shigella strains (29).

However, three genes (ipgC, ipgA, and spa32) encoding secreted proteins have no nonsynonymous differences. Spa32 is involved in regulation or assembly of the secretion apparatus (24, 48). IpgC is a cytoplasmic chaperone for IpaC and IpaB (27). The function of IpgA is not precisely known, but it is a putative chaperone. These proteins interact with other pINV proteins rather than host components, and this probably accounts for the lack of diversifying selection acting on these particular secreted proteins.

Genes for essential virulence functions (nonsecreted proteins).

There are a number of genes for nonsecreted proteins that are essential in invasion and show strong selection pressure for amino acid variation.

Three genes, icsA, virK, and icsP, are needed for Shigella to move within the host cell (31). All three have a low K_s/K_a ratio. IcsA is an outer membrane protein required for actin-based movement within the host cell. There are nine substitutions in icsA with eight changing the amino acid. VirK is required for proper production or localization of IcsA, although its precise function is not known. virK shows high levels of divergence. IcsP is an outer membrane protease for cleavage of IcsA. Four of the five changes in icsP are nonsynonymous.

The Shigella type III secretion system is composed of a basal body and an external needle (6, 47). The needle interacts with the host cell to deliver effector proteins. MxiH and MxiI are the major and minor needle components, respectively. The mxiH gene, only 252 bp in size, has the highest variation (5.556%) with a K_s/K_a ratio of 2.22, while mxiI has two changes with one being nonsynonymous. Variation in these genes may confer a selective advantage because of direct interaction with host cells. It has been observed that, when MxiH was overexpressed, the type III secretion machinery protruded longer needles than did the wild type, and the bacteria invaded the host cells much more efficiently than the wild type did (47). However, genes encoding proteins for the basal body, mxiD, mxiG, and mxiJ, also show low K_s/K_a ratios, for which we have no explanation.

Virulence-associated regulatory genes under purifying selection.

In contrast to the genes discussed above that are under selection pressure for change, several genes known to be associated with regulation of virulence are strongly conserved. spa32 has no variation at all while virF, virB, ipgC, and mxiE all have only synonymous substitutions. Spa32 and IpgC have been discussed above. VirF is a transcription activator of the AraC family required for expression of genes of the entry region and icsA (31, 41). It positively controls synthesis of VirB, a DNA-binding protein which binds to promoter regions of virulence genes for activation of transcription (2). MxiE is a transcriptional regulator (17, 26) for at least six secreted pINV proteins (OspB, OspC1, OspE2, OspF, VirA, and IpaH9.8). This observation of regulatory genes being subject to purifying selection also gives indirect support to the conclusion above that the changes observed in other genes are a result of positive selection pressure, rather than accumulation of mildly deleterious mutations.

Genes absent in the F6 plasmid.

A few genes were confirmed as absent in the F6 plasmid by hybridization (data not shown) after PCR failed to generate a product. These include sepA, orf85, orf85b, orf94, stbA, and stbB. The whole stbA and stbB region is likely absent. We discuss stbA and stbB in the replication section and sepA below, but the functions of the other three genes are unknown.

It is interesting that sepA is absent, as an S. flexneri serotype 5 sepA mutant exhibited attenuated virulence in the rabbit ligated ileal loop model (3). SepA is a serine protease of the immunoglobulin A1 protease family and the major protein secreted by S. flexneri serotype 5 (3). SepA is not required for entry into cultured epithelial cells, nor for intercellular dissemination (3), but might be involved in invasion and destruction of the host intestinal epithelium (4). We surveyed some other Shigella strains and found sepA to be variably present. Among pINV A plasmids it is present in those of S. dysenteriae serotype 10 and S. boydii serotype 9 but absent from those of S. boydii serotype 14 and S. dysenteriae serotype 3, while it is absent from the S. sonnei plasmid, the only other pINV B form tested (data not shown).

There are three genes, impCAB, that are known to be absent in S. flexneri serotype 5 (50) but present in some other strains (40). Southern hybridization showed them to be absent in S. flexneri serotype 6 (data not shown).

Comparison of replication, partition, and transfer regions.

Several regions of the plasmid are involved in replication, partition, stable maintenance, or transfer (9, 50). In general genes in these regions have lower levels of difference than do those in other regions.

We sequenced the whole replication region which includes oriR (the origin of replication), repA (required for initiation of replication at oriR), repB (copB) (repressor of transcription of repA), tapA (upstream of repA, encoding leader peptide required for translation of repA), and copA (coding for antisense RNA that binds to leader region of repA for copy number control). The only sequence differences were four single base substitutions in the oriR region of 362 bases. We also sequenced downstream of oriR to the IS. Substantial differences were observed from the start of repA4, a pseudogene. The F5 plasmid has a segment of 220 bp after repA4, followed by IS 91.04 and part of IS 1294, whereas the F6 plasmid has a unique 577-bp segment after repA4 followed by part of IS 1294, which precedes the IS 1294 in the F5 plasmid by 285 bp. Either this part of the sequence has undergone recombination, or one or both have undergone rapid change. However, there appear to be no functional genes in either the F6 or F5 sequence.

(i) Plasmid partition genes.

Low-copy-number plasmids generally have more than one system for stable maintenance of the plasmid (14). These may affect partitioning or cause postsegregational killing (PSK), the latter being a “fail-safe” mechanism (14). pINV F5 has genes for two partition systems, parAB and stbAB, and two PSK systems, ccdAB and mvpTA (9, 50). Note a possible source of confusion through use of the name stb for different genes. The genes now known as mvpA and mvpT were previously called STBORF1 and STBORF2. The stbA and stbB genes referred to in this paper are the second of two partition systems and were first observed by Buchrieser et al. (9) when the plasmid genome was sequenced.

The parA and parB genes of S. flexneri serotypes 5 and 6 are identical, but the stbAB genes are absent in S. flexneri serotype 6. The stbAB genes are also absent in S. sonnei but present in S. dysenteriae serotypes 1 and 4 and S. boydii serotype 1, with presence or absence found in both pINV A and pINV B forms. For the two PSK systems in S. flexneri serotype 5, the ccdA and ccdB genes are identical in S. flexneri serotype 6, while for mvpT and mvpA, which are encoded on the opposite strand of trbH, mvpT has only one substitution (nonsynonymous), and mvpA has two substitutions (one nonsynonymous).

The only reported functional difference between the two pINV forms is incompatibility. An early study (25) showed that the Shigella pINV plasmids belong to two incompatibility groups, later found to correspond to pINV A and pINV B (20). The complete sequence comparison allows us to look for substitutions affecting incompatibility. There are several factors that determine plasmid incompatibility. pINV belongs to the IncFII family (46), in which the major incompatibility determinant is the copA (inc) RNA (14). However, copA is identical in S. flexneri serotypes 5 and 6 and so cannot be the determinant of their incompatibility difference.

We consider next the possibility that mvpAT is responsible for the difference in compatibility. Compatibility was tested by introducing a plasmid (pMSYH6610) with kanamycin resistance and the compatibility region of the S. flexneri serotype 2a plasmid into Shigella strains with selection for kanamycin. The mvpAT system of the S. flexneri serotype 2a pINV B is present in plasmid MSYH6610 and in the compatibility test is retained in all surviving cells. When MSYH6610 is introduced into Shigella strains with kanamycin selection, the pINV plasmids of S. flexneri serotype 2a, 1b, or 2b (and presumably S. flexneri serotype 5) are lost, whereas pINV B of S. flexneri serotype 6 is stably retained. One would expect plasmids with the same copy number system to be lost under these conditions unless there was a PSK system operating on S. flexneri serotype 6 and a sufficient number of cells to retain two copies for these to maintain growth in the presence of kanamycin. MvpT is the toxin and MvpA is the antidote (44), so survival of pINV F6 indicates that it has an MvpT toxin that is not inactivated by the S. flexneri serotype 2a MvpA. It seems at first sight unlikely, as each differs from the S. flexneri serotype 2a form by only one amino acid residue. However, the F plasmid homologue is compatible with S. flexneri serotype 2a pINV (same form as that of S. flexneri serotype 5), and there are only four amino acid differences between MvpA of the F plasmid and that of S. flexneri serotype 5 and two differences in MvpT (37). One or more of these substitutions must determine the difference in incompatibility between F and pINV B. It is also possible that pINV of S. flexneri serotype 6 has a PSK system that is not present at all on pINV from S. flexneri serotype 2a, which would not have been detected by our use of only S. flexneri serotype 5-based PCR primers.

(ii) The plasmid transfer (tra) region.

It is known from tests on several S. flexneri strains that pINV is unable to initiate conjugation (43). However, there is a partial tra region in pINV F5, including the trbH, traI, traX, and finO genes and part of traD. This is only about 25% of the total tra operon as seen in the K-12 F factor, and as virtually the whole is needed for function, the remaining genes presumably have no function. We obtained sequences from trbH, traI, traX, and finO genes. There are four substitutions in trbH, of which three are nonsynonymous, giving a ratio of 0.613, but as trbH presumably has no role in the plasmid, divergence may result from selection on the mvpAT genes on the opposite strand. traI is nonfunctional in both F5 and F6. In comparison with the traI gene of the F plasmid, there are a deletion of seven bases and an insertion of seven bases in F5 and F6, respectively, in a region flanked by two palindromic sequences that disrupted the reading frame. traI also has 21 point mutation substitutions between F5 and F6. The traI gene in F6 suffered further damage with a point mutation of C to T at position 956, resulting in a stop codon. This indicates that the traI gene was inactivated independently after the divergence of the two plasmids. traX has a single synonymous substitution while finO has three substitutions, of which two are nonsynonymous.

ORFs of unknown function.

Ten of the ORFs sequenced have no known function (orf13, orf22, orf131a, orf131b, orf136, orf137, orf186, orf201, orf47, and orf48), and two (orf169a and orf169b) are probably involved in plasmid maintenance (9). orf48, orf131b, and orf136 have no substitutions while all or the majority of substitutions in orf131a, orf169a, and orf169b are synonymous. However, orf13 and orf47 are clearly under positive selection pressure. All five substitutions in orf13 are nonsynonymous, and two of the three substitutions in orf47 are also nonsynonymous with a K_s/K_a ratio less than 1. These ORFs may play an important role in virulence, making them candidates for functional analysis.

The only gene to have suffered a frameshift mutation is orf137, which is flanked by IS elements on both sides. It may be part of a gene that has been disrupted by an IS element, in which case the frameshift observed could be part of a continuing process of gene degradation.

Comparison of two pINV B plasmids.

The pINV B plasmid of S. flexneri serotype 2a (plasmid F2A) has now been sequenced as part of a genome sequence (16). We compared it with the B-form pINV F5 plasmid to enhance our understanding of pINV evolution (Table 2). S. flexneri serotypes 2a and 5 belong to the same cluster based on chromosomal gene trees (36), and divergence is relatively recent. All 80 genes that we sequenced in F6 are all present in F2A. Of those, 37 are identical in F5 and F2A, 17 have a single base substitution, and 24 have from two to four substitutions. traI and ospC3 have much bigger differences. traI has eight single nucleotide substitutions, a 10-bp deletion, and an inversion at a palindromic sequence. ospC3 has 23 substitutions, with part of the sequence being similar to ospC2, most likely as a result of inter- or intrachromosomal recombination. If we consider traI and ospC to have undergone 10 events and 1 event, respectively, we have a total of 76 polymorphisms (65 if traI and ospC are excluded), to be compared with 494 polymorphic nucleotide sites in the F5-F6 comparison. The much greater divergence of F5 and F6 relative to F5 and F2A over 80 genes strongly supports the recognition of two forms of pINV, based originally on the study of only three genes (20).

We also looked at the number of synonymous and nonsynonymous substitutions between F2A and F5. The 11 genes that are identical in F5 and F6 have between them only one (nonsynonymous) substitution in F2A. The 11 genes with synonymous changes only in F5-F6 have 10 substitutions with three being nonsynonymous. The nine genes that have only nonsynonymous F5-F6 substitutions have a total of 10 F5-F2A substitutions, all nonsynonymous. For the genes with a K_s/K_a ratio of 4 or less, we excluded ospC3 from the calculation, but for the 38 other genes in this category 250 of 402 F5-F6 substitutions and 36 of 59 F5-F2A substitutions are nonsynonymous. For the genes with a K_s/K_a ratio above 4 in the F5-F6 comparison, 12 of 37 F5-F6 substitutions are nonsynonymous, but two of two F5-F2A substitutions are nonsynonymous. The pattern of synonymous and nonsynonymous substitutions in the F5-F2A comparison of two pINV B plasmids mirrors the pattern observed in the F5-F6 comparison with the possible exception of the last category where the number of F5-F2A substitutions is too low to be meaningful. It appears that genes thought to be under positive selection in the F5-F6 comparison have similar characteristics in the F2A-F5 comparison.

Variation in the invasion gene homologues in S. enterica.

The invasion ability of S. enterica is determined by a cluster of genes many of which are homologous to the invasion genes of pINV. Eight have been compared in multiple strains of S. enterica (7, 22). The K_s/K_a ratio of invE (pINV mxiC), invA (mxiA), spaP (spa24), and spaQ (spa9) is similar to that of housekeeping genes while invH (no pINV homologue), spaM (spa13), spaN (spa32), and spaO (spa33) have a much lower K_s/K_a ratio. It was concluded (7, 22) that proteins that are membrane bound or membrane associated are relatively conserved whereas those that are exposed to the extracellular environment are hypervariable, reflecting the action of diversifying selection. Our data with a much larger number of genes also show that in general proteins exposed to the extracellular environment are more variable, but it is interesting that, for the seven homologues of the S. enterica genes used, the genes can differ in K_s/K_a ratio in E. coli and S. enterica. In fact only three genes, spaP (spa24), spaQ (spa9), and spaO (spa33), are consistent in the two species. spaM and spaN have a low K_s/K_a ratio in S. enterica, while there is only one (synonymous) change in spa13 and no change in spa32 between F5 and F6 pINV plasmids, and mxiA and mxiC have low K_s/K_a ratios of 3.55 and 1.97, respectively, while those of the S. enterica homologues are similar in this regard to housekeeping genes.

Spa32 and SpaN share only 15% amino acid identity, but the function of Spa32 can be complemented by SpaN (48), whereas there is known functional difference between these homologues (11). Spa32 is implicated in the release of the Ipa proteins but not their surface presentation. In contrast, SpaN is involved in both the secretion and the surface presentation of the Sip proteins (10). However, the earlier observation that Spa32, unlike SpaN, is not secreted to the culture supernatant (51) may not be entirely correct, as it was recently found that Spa32 is translocated through the type III secretion machinery into the medium (48). This may explain the difference in selection pressures between Spa32 and SpaN. It seems likely that there are functional differences for the other three homologues to account for the opposing trends.

Concluding comments.

The pINV plasmid of Shigella-EIEC occurs in two major forms as shown earlier by distribution of sequence forms of three genes sequenced from a range of strains. We have now sequenced 80 genes from the A-form plasmid, pINV F6, enabling us to make a detailed comparison of representative pINV A and pINV B plasmids. The three genes, ipgD, mxiA, and mxiC, studied previously (20) are shown to be representative of the range in level of difference between pINV A and pINV B plasmids. Generally genes for plasmid maintenance vary less than virulence-associated genes, those encoding secreted proteins that interact with host cells being particularly variable.

Many genes appear to be under selection pressure for change. In addition some of the F5 genes are absent in F6. Some of the plasmid-encoded proteins are immunogens, and for them selection may be in part at least for avoidance of the host immune system. More interesting is the possibility that differences may reflect ongoing adaptation to the Shigella niche, or selection for variants of that niche. Shigella strains infect only humans and as a group are a major human pathogen. Their mode of pathogenesis is thought to have become effective only 35,000 to 270,000 years ago, and its antecedents are not known. The high proportion of genes under selection pressure in the pINV plasmid that encodes many of the functions characteristic of this mode of pathogenesis may reflect continuing adaptation to achieve an optimal interaction with the human host. It is also possible that the many serotypes are adapted to different variants of the general niche occupied, and that the selection relates to competition between different Shigella clones which differ in details of their adaptation to the niche. The observation that senA (29), a toxin gene apparently under selection pressure for change, and sepA, known to play a role in virulence in S. flexneri serotype 2a (3, 4), are absent in some Shigella strains supports the hypothesis that Shigella strains differ in the details of their adaptation to their mode of pathogenesis. A similar explanation may account for the observation that genes under selection pressure differ between E. coli and S. enterica.

As noted previously (20), not much has been reported on variation in virulence of Shigella clones, but only a few serotypes are responsible for many of the cases and there are also perceived differences between Shigella and EIEC clones. It could be interesting to relate variation in virulence within and between Shigella and EIEC strains to variation in the genes under selection pressure and genes present or absent in the pINV plasmids. This may help us to better understand the pathogenesis and epidemiology of Shigella-EIEC infections. Comparative functional studies of the genes that differ may point to the factors that give S. flexneri, S. sonnei, and S. dysenteriae serotype 1, for example, their distinctive disease characteristics and global distributions.

Incompatibility between the two plasmids was deduced to be due to very small differences: a single amino acid substitution in mvpA being the only difference likely to be involved. Its role in gene flow-recombination between the two forms and plasmid survival is not clear.

pINV plasmids are nonconjugative (25, 43). The substitution patterns in the incomplete tra region indicate that the remaining genes might be functional. This raises questions of transfer of pINV to other E. coli strains since many such events are presumed to have occurred to give rise to Shigella-EIEC strains of diverse genetic backgrounds (20). The remaining tra region genes may be involved in this process, complemented by genes on other plasmids in the recipient cells, leading to pINV transfer.