INTRODUCTION

Top Abstract Introduction Materials & Methods Results Discussion References

The term pathogenicity island was coined by Hacker et al. to describe two large, unstable DNA regions of the chromosome of uropathogenic Escherichia coli (23). This term refers to a usually large (35-kb) chromosomal segment that carries genes involved in pathogenicity. Characteristically, pathogenicity islands are bordered on one side by a tRNA gene and, less frequently, may be flanked by insertion sequences. These islands are often unstable, and their deletion occurs at frequencies of 10⁴ to 10⁵. Their GC content is usually different from that of the rest of the host chromosome, suggesting that they originate from horizontal transfer between different bacterial genera (24). The number of gram-negative bacterial species shown to harbor pathogenicity islands has grown steadily and includes uropathogenic (3, 28, 48) and enteropathogenic (33) E. coli, Helicobacter pylori (9), Salmonella typhimurium (35, 44), Dichelobacter nodosus (10), and Vibrio cholerae (29).

Pathogenicity islands have also been identified in the genus Yersinia. In Yersinia enterocolitica, such an island is found in high-pathogenicity strains of biotype 1B only and not in low-pathogenicity strains (5). Since the presence of this region determines the level of pathogenicity, it was termed a high-pathogenicity island (HPI). The HPI of Y. enterocolitica Ye8081 is 45 kb long and is bordered on one side by an asn tRNA gene. It also carries a single copy of four different sequences repeated elsewhere in the chromosome (RS.3, RS.4, IS1400 [5], and IS1328 [41]) and genes involved in siderophore-mediated iron acquisition: the yersiniabactin biosynthetic genes irp2 and, probably, irp1 (5) and the fyuA locus, which codes for the yersiniabactin receptor (42).

In Y. pestis, a 102-kb region called the pgm locus was first identified by Fetherston et al. (17). This region is deleted en bloc at a frequency of 10⁵, probably by homologous recombination between its two flanking IS100 copies (39). The 102-kb region can be divided into two functionally and physically distinct parts. One carries the hms (for hemin storage) locus (36, 38), which confers a pigmented phenotype on colonies grown on Congo red-agar plates and enhances transmission of the microorganism by its flea vector (26). This region has been designated the pigmentation segment in this study. The other region carries the same irp2, irp1, and psn/fyuA genes involved in iron acquisition as those of the Y. enterocolitica HPI (25), the recently identified siderophore biosynthetic loci ybtT and ybtE (1), and the activator gene ybtA (14). This region is called here the iron acquisition segment.

We wished to determine whether the whole 102-kb region is a true pathogenicity island. Both segments of this region carry genes important for either virulence or disease transmission, and they are deleted en bloc, suggesting that the whole 102-kb region forms a pathogenicity island (5). However, a previous analysis of 43 Y. pestis strains revealed that 16% of these strains harbored the irp2 gene (located on the iron acquisition segment) but were nonpigmented (27). Fetherston et al. also identified a single strain of this type (M23) out of 43 examined (17) and demonstrated that the M23 mutants resulted from different genetic alterations such as a mutation in the hmsR gene and at least two different-size deletions (17, 31). To determine whether internal deletion of the pigmentation segment could occur independently from that of the iron acquisition segment, the 102-kb region of strain 6/69 was characterized and used to analyze the different mutants.

We show that the 102-kb region of Y. pestis is composed of a ca. 35-kb HPI homologous to the HPI of Y. enterocolitica and of a ca. 68-kb adjacent pigmentation segment. We also demonstrate that the Pgm Irp2⁺ type observed in nine Y. pestis isolates results from several distinct phenomena such as variable deletions involving the pigmentation segment, point mutation in the hms locus itself, or mutation in a gene outside the hms locus but contributing to the pigmented phenotype. Finally, we show that despite the various rearrangements observed in vitro, the 102-kb region is highly conserved among Y. pestis strains of different geographical origins and biotypes.

	MATERIALS AND METHODS

Top Abstract Introduction Materials & Methods Results Discussion References

Bacterial strains and growth conditions. The characteristics of the Y. pestis strains used in this study are listed in Table 1. Strain 6/69 is a derivative of strain 6/69 that has spontaneously lost the irp2 gene and the surrounding region (7). The geographical origin of the strain corresponds to the name of the country or city in use at the time the strain was isolated. The other Yersinia strains used in this study were Y. pseudotuberculosis IP32954 (serotype I) and IP32938 (serotype III); Y. enterocolitica Ye 8081 (biotype 1B, serotype O:8), IP864 (biotype 4, serotype O:3), and IP383 (biotype 2, serotype O:9); Y. mollaretii IP21081; Y. bercovieri IP21531; Y. intermedia IP21356; Y. kristensenii IP21577; and Y. frederiksenii IP21689. The E. coli strains used were DH1 (supE44 hsdR17 recA1 endA1 gyrA96 thi-1 relA1), LE392 [hsdR574 supE44 supF58 lacY1 or (lacIZY)6 galK2 galT22 metB1 trpR55), and XL1Blue {supE44 hsdR17 recA1 endA1 gyrA96 thi relA1 lac [F' proAB, lacI^qZDM15 Tn10 (Tet^r)]}. Yersinia strains were grown for 24 h (peptone broth) or 48 h (Trypticase soy agar plates) at 28°C. E. coli strains were grown at 37°C for 24 h. The pigmentation phenotypes of the Y. pestis isolates were determined on Congo red-agar plates after 4 days of growth at 26°C, as described by Surgalla and Beesley (47).

DNA techniques and cloning methods. Isolation and digestion of genomic DNA were performed as previously described (8). Plasmid extractions were done by the method of Birnboim and Doly (2). For PCR-amplified DNA, the products of the reactions were purified by elution from an agarose gel with the Geneclean kit (Bio 101, Inc., La Jolla, Calif.) or by using the Wizard PCR preps DNA purification system (Promega). Double-stranded DNA labeling was performed either radioactively with [³²P]dATP (Amersham) or nonradioactively by the enhanced-chemiluminescence reaction (ECL; Amersham) or the digoxigenin (DIG) random primed labeling system (Boehringer). Single-stranded oligonucleotides were labeled with either the 3' oligonucleotide labeling system (ECL) from Amersham or the DIG-oligonucleotide tailing kit from Boehringer. To screen the cosmid library, a Y. pestis-specific probe was constructed. A 14-kb EcoRI chromosomal restriction fragment (E14) from Y. pestis 6/69, which contains the irp2 gene (6), was cloned into a bacteriophage EMBL4 vector (Promega). The ligated DNA was packaged by using the Packagene system (Promega) and transfected into E. coli LE392. Recombinant E. coli colonies were screened by colony blotting with the Y. enterocolitica Ye8081 8-kb ClaI fragment (Cl8) carrying the irp2 gene (6). The hms locus was obtained by insertion of a 10-kb BglII-SalI fragment extracted from cosmid peH64 into the BamHI-SalI sites of plasmid PSU18. The recombinant plasmid was designated pSUhms. The other cosmid restriction fragments of interest were eluted from the agarose gel with the Geneclean kit (Bio 101, Inc.) and used directly as probes or cloned into the corresponding sites on the polylinker of the pBluescript II KS⁺ plasmid (Stratagene, La Jolla, Calif.). To obtain a portion of the sequence inserted into the pigmentation segment of strain K169, the genomic DNA of this strain was digested with EcoRI and subjected to electrophoresis. Restriction fragments of approximately 3 kb were eluted from the gel and ligated into the corresponding site on the polylinker of the pBluescript II KS⁺ plasmid (Stratagene). Following electroporation, transformed colonies of E. coli XL1Blue (Bio 101, Inc.) were selected on ampicillin (100 µg/ml)-containing agar plates. Recombinant colonies were screened by colony blotting with the EH2 probe from strain 6/69.

Cosmid library. Preparation of high-molecular-weight DNA from pYV-cured strain 6/69 was done mainly as described in reference 5. Briefly, chromosomal DNA was partially cleaved with Sau3A (Janssen Biochimica) and sized on a 10 to 40% sucrose gradient. DNA restriction fragments ranging from 35 to 50 kb were ligated into BamHI-digested and alkaline phosphatase-treated cosmid vector pHC79 (BRL, Cergy Pontoise, France). Recombinant cosmids were packaged into Gigapack II Plus packaging extracts (Stratagene) and used to infect E. coli DH1. Approximately 800 recombinant colonies were spotted on nylon filters.

Establishment of the physical maps of the pgm locus and analysis of adjacent regions. To characterize the HPI of strain 6/69, recombinant E. coli colonies containing the cosmid library were hybridized with the E14 probe. Positive clones were isolated, and their cosmid DNA was extracted and digested with the restriction enzymes BamHI, EcoRI, HindIII, SpeI, and NotI. Restriction maps of the different cosmids were obtained after combinations of single and double digestions with these enzymes. The restriction fragments located at each extremity of the cosmids were used to screen the cosmid library in search of new recombinant clones. The same strategy was applied to the new clones for further chromosome walking. The DNA segments were considered to be inside the 102-kb region when they hybridized with the DNA of strain 6/69 only and outside it when they hybridized with the DNAs of both strains 6/69 and 6/69. The borders of the locus corresponded to DNA fragments which hybridized with both strains but with a different restriction pattern. The same criteria were used to define the limits of the deleted fragments in other strains of Y. pestis. To determine whether the chromosomal region flanking the left border of the pgm locus was the same in the different Y. pestis strains studied, two probes located in the vicinity of the border, of which one is internal (a 3-kb EcoRI-SpeI fragment designated ES3 [see Fig. 1A]) and the other is external (a 3.3-kb BamHI segment designated B3.3 [see Fig. 1A]) to the iron acquisition segment, were hybridized with the SpeI-digested DNAs of the different strains. Recognition of the same 12-kb SpeI restriction fragment by both probes suggested that the ES3 and B3.3 fragments were adjacent, while recognition of different-size SpeI fragments indicated that B3.3 did not flank the iron acquisition segment. The same strategy was used for the right border of the pigmentation segment with the internal 3-kb EcoRI-ClaI (EC3) and the external 2.4-kb EcoRI-HindIII (EH2.4) probes (see Fig. 1A).

PCRs. Several of the sets of primers used in this study were described previously (5). In addition, new sets of primers were used. The sequences of the sense (SP) and antisense (ASP) primers, the sizes of the amplified fragments (S), and the annealing temperatures (A) for the PCR were as follows. (i) IS285 (18): SP, 5'-TGGACGAAAAGAAAC-3'; ASP, 5'-AACAATGGGATACAG-3'; S, 488 bp; A, 50°C. (ii) irp1 (18, 37): SP, 5'-AGAAACCGATGCTCACCC-3'; ASP, 5'-TCCTCTCCTGACGTAGCC-3'; S, 526 bp; A, 57°C. (iii) psn (16): SP, 5'-CTTTCCACCAACACCATCC-3'; ASP, 5'-AAACCGCCACTTCGCTTC-3'; S, 1,062 bp; A, 57°C. (iv) ybtA (14): SP, 5'-ACAGAGTCACCGCAAACG-3'; ASP, 5'-CAGATCAGCCAGCAGCAG-3'; S, 810 bp; A, 57°C. (v) ybtE (18, 37): SP, 5'-CCCTTACCCATTGCCGAAC-3'; ASP, 5'-TCCCCACCTCATCCAGCC-3'; S, 1,189 bp; A, 57°C. (vi) ybtT (37): SP, 5'-CCGCTCAGAAGCATTACACAAC-3'; ASP, 5'-TCGCCGTCAATCACCACC-3'; S, 500 bp; A, 57°C. The template used for all PCRs was the genomic DNA of strain 6/69.

Pulsed-field gel electrophoresis. Genomic DNA was prepared in agarose plugs as previously described (4). Following digestion with SpeI or NotI, macrorestriction fragments were resolved by contour-clamped homogeneous electric field electrophoresis using a CHEF-DRIII apparatus (Bio-Rad Laboratories), an electric field of 6 V/cm, and an angle of 120°. Migration of the DNA fragments was determined in 0.5× Tris-borate-EDTA buffer and in 0.9% agarose gels maintained at 17°C. Pulse times were ramped from 1 to 10 s or from 1 to 18 s over 29 h for NotI- and SpeI-generated restriction fragments, respectively.

	RESULTS

Top Abstract Introduction Materials & Methods Results Discussion References

Physical and genetic maps of the 102-kb region of strain 6/69. The 102-kb region of Y. pestis 6/69 was characterized by chromosome walking on different overlapping cosmids (Fig. 1A) as described in Materials and Methods. The size of this region was estimated to be approximately 103 kb, similar to the 102 kb reported by Fetherston et al. (17) for strain KIM6⁺. Its restriction map is shown on Fig. 1A. In addition to irp2 and, probably, irp1, the other genes identified and positioned so far on the 102-kb region of strain KIM6⁺ are the psn (16), hms (36), ybtA (14), ybtE, and ybtT (37) loci. Hybridization with a PCR-amplified portion of these genes indicated that they are also present on the 102-kb region of strain 6/69, on the same restriction fragments, and probably in the same positions as in strain KIM6⁺ (Fig. 1A). It has also been shown that the 102-kb region of strain KIM6⁺ is flanked by single copies of the IS100 insertion sequence (IS) (39). This IS was found in multiple copies on the chromosome of strain 6/69 (Fig. 2). The fact that a PCR-amplified portion of this IS hybridized with an E1.6 fragment in cosmid peH2 and with an E1.5 fragment in cosmid peH96 but with no fragment in the other cosmids (Fig. 1A) indicated that the 102-kb region of strain 6/69 also carries a single IS100 copy at each extremity but that no copy of the IS is present inside this locus.

View larger version (13K): [in this window] [in a new window]   FIG. 1.   Restriction maps of the 102-kb region of Y. pestis 6/69 (A) and the HPI of Y. enterocolitica Ye8081 (B). Plain thick lines, the 102-kb region; broken lines, regions outside the 102-kb region. Restriction sites: E, EcoRI; H, HindIII; B, BamHI; S, SpeI; N, NotI. Horizontal bars below the restriction maps represent the different probes cited in the text. The values below the map indicate the scale in kilobases. Horizontal arrows and bars above the restriction map correspond to the identified genes. peH, cosmid clones used to span the entire 102-kb region. RS, repeated sequences; 2×, two very close BamHI restriction sites.

View larger version (31K): [in this window] [in a new window]   FIG. 2.   Hybridization profiles of four different repeated sequences present in the 102-kb region of strain 6/69. The EcoRI-digested genomic DNA of strain 6/69 was hybridized with the E1.6 probe from Y. pestis 6/69 for IS100 (Fig. 1A), the SCI1 probe from Y. enterocolitica Ye8081 for RS.4 (5), a 50-bp oligonucleotide probe from Y. enterocolitica Ye8081 for the asn tRNA (5), and the Bg3.7 probe from Y. pestis 6/69 for RS.5. The tick marks indicate the following molecular sizes, top to bottom: 23.1, 9.4, 6.5, 4.3, 2.3, and 2.0 kb.

Comparison of the HPI of Y. enterocolitica with the 102-kb region of Y. pestis. The 102-kb region of Y. pestis 6/69 has been represented in the sense opposite to that of the published partial map of the 102-kb region of strain KIM6⁺ to compare its organization with that of the recently described HPI of Y. enterocolitica 1B (5). It has been shown previously that the HPI of Y. enterocolitica Ye8081 extends over 45 kb and is not flanked by repeated elements but harbors the fyuA/psn, irp2, and (probably) irp1 loci (5). Comparison of the physical maps and gene locations of the HPI of Y. enterocolitica and of the 102-kb region of Y. pestis indicates good conservation of the 30-kb segment corresponding to the right part of the Ye8081 HPI (Fig. 1B). Confirmation of this observation was obtained by hybridizing the different probes that spanned the HPI of Y. enterocolitica Ye8081 (5) with the genomic DNAs of strains Ye8081 and 6/69. The probes located on the 30-kb right part of the HPI of strain Ye8081 recognized similar-size fragments in both species with the exception of the E6 and E9 probes (Fig. 1B), which both recognized an 14-kb Y. pestis fragment due to the absence of an EcoRI site in strain 6/69. Upstream of the irp2 locus, a repeated element designated RS.4 was identified in all strains of Y. enterocolitica 1B previously analyzed (5). This repeated sequence was also present on the Y. pestis chromosome, in at least 16 copies (Fig. 2). Of these, only one was present on the iron acquisition segment of strain 6/69 and was located, as for Y. enterocolitica, upstream of the irp2 gene (Fig. 1A and B). We also previously demonstrated that the HPI of Y. enterocolitica, like most of the pathogenicity islands described so far, is bordered on its right side by a copy of a tRNA gene (5). The use of a 50-bp oligonucleotide probe internal to the asn tRNA gene of Y. enterocolitica (5) indicated that this gene is present in three copies on the chromosome of strain 6/69 (Fig. 2) and that one copy is located at the same position as in Y. enterocolitica (Fig. 1A and B).

Analysis of the mechanisms responsible for the existence of nonpigmented Y. pestis mutants harboring the irp2 locus. In a previous study (27), we found that 7 of the 43 strains of Y. pestis analyzed were nonpigmented but harbored the irp2 locus (Pgm Irp2⁺). In the present study, the availability of cosmid clones spanning the hms locus allowed us to further analyze these mutants. In addition, two new Pgm Irp2⁺ derivatives obtained in vitro from Pgm⁺ Irp2⁺ parental strains (Kenya 169 and Saigon 55-1239) were included in the study (Table 1). Two categories of strains were distinguished. The first category included the two strains from Turkey and the three strains from Kenya whose DNA did not hybridize with a portion of the hms locus (C2.6 probe; Fig. 1A), indicating deletion of at least a part of this locus. The second category was composed of the strains from Germany, Vietnam, and Zaire which did harbor the hms locus although they were nonpigmented. To characterize the mechanisms responsible for the Pgm Irp2⁺ phenotype, we mapped the 102-kb regions of all nine isolates.

Characterization of the two Pgm Irp2⁺ strains from Turkey. Hybridization experiments with probes covering the 102-kb region of strain 6/69 (Fig. 1A) revealed that the same deletion event occurred in the two strains from Turkey and involved not only the hms locus but at least 70 kb of DNA (Fig. 3B).

View larger version (15K): [in this window] [in a new window]   FIG. 3.   Comparison of the 102-kb region EcoRI restriction map of strain 6/69 (A) with those of strains from Turkey (B) and Kenya (C and D). Plain thick lines, the 102-kb region; broken lines, regions outside the 102-kb region; dotted lines, deleted regions; E, EcoRI restriction sites; triangles, DNA sequences that were absent from the 6/69 102-kb region. Horizontal bars below the restriction maps represent the different probes cited in the text. The values below the map indicate the scale in kilobases. Horizontal arrows and bars above the restriction map correspond to the identified genes. RS, repeated sequences; P+I+, Pgm+ Irp2+; PI+, Pgm Irp2+.

Characterization of the three Pgm Irp2⁺ strains from Kenya. In Pgm Irp2⁺ strains K129, K164, and K169.1 from Kenya, the sizes of the deletions involving the hms locus were the same: approximately 38 kb (Fig. 3C). By walking outward from the hms locus, the left limit of the deletion was located on a fragment corresponding to the 2-kb EcoRI-HindIII fragment (EH2) of strain 6/69, and the right limit was on a 3.2-kb HindIII-BamHI fragment (HB3.2; Fig. 3A).

Characterization of the four Pgm Irp2⁺ strains retaining the hms locus. As mentioned above, the four remaining Pgm Irp2⁺ strains, H19, CBL, S55-1239.1, and S55-797, hybridized with the C2.6 probe, indicating that they did not undergo deletion of this region. Further analysis of their 102-kb regions and comparison with that of strain 6/69 were undertaken.

(i) Analysis of strain Hamburg 19. The 102-kb region EcoRI restriction maps and chromosomal flanking regions of strains H19 and 6/69 were almost identical. The only minor differences corresponded to the presence of two DNA insertions of approximately 1.5 kb, both carrying one additional EcoRI and BamHI restriction site (Fig. 4B). The presence of the same two restriction sites in the two additional sequences and their similar sizes suggest that they correspond to the insertion of the same element at two different positions on the pgm locus of strain H19.

View larger version (14K): [in this window] [in a new window]   FIG. 4.   Comparison of the 102-kb region EcoRI restriction map of strain 6/69 (A) with those of strains from Hamburg (B), Ho Chi Minh City (C), and Zaire (D). Plain thick lines, the 102-kb region; broken lines, regions outside the 102-kb region; E, EcoRI restriction sites; triangles, DNA sequences that were absent from the 6/69 102-kb region. Horizontal bars below the restriction maps represent the different probes cited in the text and used to analyze the strains. The values below the map indicate the scale in kilobases. Horizontal arrows and bars above the restriction map correspond to the identified genes. RS, repeated sequences.

(ii) Analysis of the two Pgm Irp2⁺ strains from Ho Chi Minh City. The 102-kb region EcoRI restriction maps and chromosomal flanking regions of Pgm Irp2⁺ strains S55-797 and S55-1239.1 (Fig. 4C) and Pgm⁺ Irp2⁺ parental strain S55-1239 were identical to those of strain 6/69.

(iii) Analysis of the Pgm Irp2⁺ strains from Zaire. The EcoRI restriction map of the 102-kb region of strain CBL closely resembled that of strain 6/69 (Fig. 4D). The only differences were the insertion of two 2-kb DNA fragments containing an EcoRI restriction site. These inserted sequences were located outside the hms locus and therefore could not explain the nonpigmented phenotype. Attempts to complement strain CBL with pSUhms were unsuccessful because the strain did not grow well enough in vitro for preparation of electrocompetent cells.

	DISCUSSION

Top Abstract Introduction Materials & Methods Results Discussion References

In this study, the 102-kb DNA segment constituting the pgm locus of Y. pestis 6/69 (a virulent strain from Madagascar) has been identified and its restriction map has been established. This locus has the same size, restriction map, gene location, and flanking IS100 copies as those previously established for strain KIM6⁺ (16, 17), indicating a high degree of conservation among strains of different origins and biotypes. This observation was reinforced by the fact that the EcoRI physical map of the entire 102-kb region was almost the same in all Y. pestis strains of various geographical origins (Hamburg, Ho Chi Minh City, Zaire, Madagascar, and Kenya) tested and over the remnant of the 102-kb region still present in the strains from Turkey. The minor differences observed concerned the presence in some strains of one or two additional, small DNA segments. However, these differences were limited to one or two positions in the 102-kb region and resulted in a difference in size not exceeding a total of 4 kb.

The 102-kb region of Y. pestis is composed of two physically and functionally distinct portions: a ca. 35-kb segment (left part in Fig. 1A), which is involved in siderophore-mediated iron acquisition, and a ca. 68-kb segment (the right part), which carries genes conferring the pigmentation phenotype. Although this arrangement of pathogenicity-associated genes flanked by ISs superficially resembles a pathogenicity island (24), there is little evidence that both parts have ever formed a horizontally transmitted unit. The fact that the right IS100 flanking copy disrupted the phoE gene in KIM strains (16) indicates that the entire 102-kb locus was not transposed en bloc into this region. Insertion of IS100 into this region of the Y. pestis chromosome seems to be a frequent phenomenon. Sequence data suggest that the two parts of the 102-kb region have different origins: the background GC content of yersiniae is 46 to 50%, and the 7.6-kb hms operon involved in pigmentation has a GC content of 46.9% (36, 38), while the overall GC contents of the fully sequenced Y. pestis iron acquisition genes (psn, ybtE, ybtT, and ybtA) (1, 14, 15) and of the Y. enterocolitica irp2 sequence (20) are 55.8 and 59.2%, respectively. Another piece of evidence that there are two distinct units comes from the fact that most of the iron acquisition segment is well conserved in Y. enterocolitica and Y. pestis, while the pigmentation segment is either absent or highly degenerated in the former species. Furthermore, most of the Yersinia iron acquisition segment is also present as a unit in various pathogenic E. coli strains (43).

The iron acquisition segment of Y. pestis is downstream of a tRNA locus, carries genes necessary for the expression of a high-pathogenicity phenotype, has a GC content different from that of the remainder of the chromosome, is homologous to the HPI of Y. enterocolitica, and is found in other enterobacterial species. Therefore, this 35-kb left portion of the 102-kb region of Y. pestis can be considered an HPI per se.

We found no evidence for mobility of the HPI independently of the pigmentation segment. Similarly, Fetherston and Perry (16) could not identify any Y. pestis strain with the HPI portion of the 102-kb region alone deleted. These results are consistent with those obtained with Y. enterocolitica, since we were able to identify neither spontaneous HPI deletion mutants in our culture collection nor deleted colonies following repeated subculture of strain Ye8081 (11). Although the HPI of Y. pestis is not deleted independently of the pigmentation segment, the two regions do not display the same degree of in vitro stability. Indeed, several Pgm Irp2⁺ Y. pestis strains were previously identified (27). This study reveals that different mechanisms could cause the nonpigmented phenotype: (i) deletion of an internal 38-kb DNA segment within the pigmentation segment mediated by homologous recombination between two additional internal copies of IS100, (ii) deletion of over 70 kb of DNA involving the right portion of the HPI and probably the adjacent chromosomal region, (iii) mutation in the hms locus, and (iv) alteration of an unidentified gene located in the pigmentation segment and contributing to the pigmented phenotype but different from the hms locus.

Our demonstration of these different types of deletion or mutation within the 102-kb region conflicts with the high degree of conservation of this region overall among strains that differ in origin. However, it should be emphasized that the instability of the 102-kb region and its complete loss on subculture (11) are in vitro phenomena. Actually, all fresh isolates previously studied harbored the irp2 gene (12), indicating that the 102-kb region is maintained under natural conditions, probably because it encodes functions essential for in vivo survival (iron acquisition system) and for transmission to new hosts via flea vectors (hemin storage system [26, 30]). The observation that the geographical origin of the Y. pestis strains correlates with the type of event responsible for the nonpigmented phenotype suggests that specific genomic rearrangements which secondarily led to the nonpigmented phenotype in vitro occurred in each ecosystem. Adaptive genomic modifications generated during the life cycle of a bacterium in the natural environment, followed by clonal expansion, would thus predispose strains from a given geographical origin to certain specific types of rearrangements. Such a clonal expansion of new variants of Y. pestis has recently been demonstrated in the plague foci of Madagascar (22).

Devignat postulated that Y. pestis strains of biotypes Antiqua, Medievalis, and Orientalis were responsible for the first, second, and third plague pandemics, respectively (13). Recent results obtained with molecular typing techniques such as pulsed-field gel electrophoresis or ribotyping of strains from different regions have supported this conjecture (21, 32, 40). The four strains of biotype Antiqua examined in this study (strains K129, K164, and K169 from Kenya and strain CBL from Zaire) had identical characteristic features for their 102-kb regions: (i) similar chromosomal flanking regions, (ii) two IS100 insertions at the same position on the pigmentation segment, and (iii) instability of the region located between these two inserted sequences. These features were not found in the 102-kb region of the biotype Orientalis strains studied, reinforcing Devignat's hypothesis about the clonality of the strains of each biotype. However, these interbiotype differences within the 102-kb region are minor. If we accept the association of different biotypes with different pandemics, then the 102-kb region of Y. pestis is a stable feature of strains associated with outbreaks from the 5th century up to the present day.

In conclusion, this study demonstrates that the 102-kb region of Y. pestis is composed of two clearly distinct parts, an HPI and a pigmentation segment, but these two components are well conserved and stably linked in different strains, probably because they are essential for bacterial survival under natural conditions. Both the originally observed deletion of the entire 102-kb region (17) and the additional mutations we have defined leading to the nonpigmented phenotype are probably in vitro phenomena and are consequences of the presence of numerous ISs in the Y. pestis chromosome. Nonetheless, we believe that these different mutations reflect an underlying specific genomic adaptation of these strains to their local environment, followed by clonal expansion in the ecosystem.