Intramolecular transposition of insertion sequence IS91 results in second-site simple insertions

Construction of a series of plasmids carrying a kanamycin-resistant (Km^r) transposable cassette

A series of plasmids was constructed carrying an IS91-derived transposable cassette of the type IR_L-kan-IR_R and an external transposase gene (tnpA) in different relative locations. The cassettes carried a kanamycin resistance gene (kan ), which allowed the detection of direct transposition events. Each of the four structural ‘modules’ of these series of plasmids, i.e. the kan and tnpA genes, IR_R and IR_L, was flanked by unique restriction sites, which allowed separate manipulation of each module, including exchange by modules from related ISs. Figure 1 explains the construction of plasmid pSU2620, which carried the basic IR_L-kan-IR_R cassette. In pSU2620, IR_R is flanked by unique EcoRI and KpnI restriction sites, the kan gene by BamHI sites, and IR_L by SphI and XbaI restriction sites. Transposition of IR_L-kan-IR_R occurred after complementation with the transposase gene, which was provided in trans by either of the two helper plasmids shown in 2Fig. 2B. In pSU2637, transposase expression could be induced by an upstream lac promoter (Plac ), whereas in plasmid pSU2638, the transposase was cloned in the opposite orientation to Plac. Alternatively to the trans configuration for the transposase, a series of plasmids was also constructed that carried the IR_L-kan-IR_R cassette and tnpA in cis. There were four configurations, which were represented by plasmids pSU2631 to pSU2634, as shown in Fig. 2A. In plasmids pSU2631 and pSU2632, tnpA was expressed under the control of Plac, as in pSU2637, while in plasmids pSU2633 and pSU2634, tnpA was reversed, as in pSU2638. In pSU2631 and pSU2633, IR_R is proximal to tnpA, while in pSU2632 and pSU2634, IR_L is proximal to tnpA.

Transposase overexpression led to an increase in the transposition frequency

Transposition frequencies obtained for the series of plasmids carrying tnpA in cis are shown in Table 1. When tnpA expression was driven by Plac, transposition frequencies increased by 200-fold (pSU2631) or 15-fold (pSU2632) upon Plac induction by the addition of IPTG to the medium. Conversely, when tnpA was reversed with respect to Plac (pSU2633 and pSU2634), transposition frequencies did not increase upon Plac induction. They were very similar to the basal levels of expression of the equivalent set-ups in pSU2631 and pSU2632. Basal levels of transposase activity were higher in pSU2632 and pSU2634 than in pSU2631 and pSU2633, as judged by transposition frequencies in the absence of induction. Thus, induction seems to be stronger for pSU2631, although transposition frequencies in induced cultures were identical in both cases. We think that basal transposition levels could be dependent on a number of parameters, such as the proximity of IR_R or IR_L, external transcription effects or even an effect of tnpA transcription on plasmid copy number. In any case, the increase in transposition frequencies upon Plac induction indicated that the transposition frequencies of the cassettes were dependent on the level of IS91 transposase, which could be correlated with the expected activity of Plac.

Table 1. . Transposition frequencies of IR_R-kan-IR_L cassettes when IS91 transposase gene was present in cis^a. a. Transposition was carried out by a mating-out assay. Donors were derivatives of strain D1210 carrying R388 plus the plasmid shown in the first column. They were mated with strain DH57α on solid medium for 90 min at 37°C, and transconjugants were selected on appropriate selective media.b. Before mating, donor strains were streaked on selective media containing either 0.5 mM IPTG (+) or 0.5% glucose (−) to either induce or repress transcription from the lac promoter respectively.c. The frequency of transposition of IR_R-kan-IR_L to R388 was calculated as the ratio of Nx^r Km^r transconjugants (transposition products) to Nx^r Tp^r transconjugants (recipients that received R388). Frequencies shown are the average of a number of experiments [n]. When data from three or more experiments were available, the standard deviation of the means are expressed in brackets.d. The percentage of transposition products that were co-integrates or fusions was calculated as the ratio of Nx^r Cm^r transconjugants to Nx^r Km^r transconjugants.e. Plasmid DNA was obtained from individual Nx^r Cm^r colonies and characterized by restriction endonuclease digestion as a co-integrate or as a fusion product, as explained in the text.

IS 91 transposase is active in trans

The fact that the relative position of IR_R had a minor effect on the transposition frequencies of the IR_L-kan-IR_R cassette suggested that positional effects may not be important for transposase function. In fact, the results obtained in Table 1 can be compared with those obtained when the transposase gene was provided in trans (Table 2). pSU2637 is equivalent to pSU2631 and pSU2632, in that tnpA is in the sense orientation with respect to Plac. The transposition frequency obtained when transposase was provided in trans was only three times lower than that obtained when transposase was provided in cis (3 × 10⁻⁴ for pSU2637 versus 9 × 10⁻⁴ for both pSU2631 and pSU2632). These results showed that IS91 transposase is also trans-acting. It has been shown previously that transcription from Plac is unregulated in the pUCs, even in the presence of a lacI ^q gene product, because of the high copy number of these plasmids of over 100 copies per cell (Stewart et al., 1986; Stark, 1987). Therefore, it was not surprising that the frequency of transposition of IR_L-kan-IR_R in plasmid pSU2637 was the same irrespective of the presence or absence of Plac induction.

Table 2. . Transposition frequencies of the IR_R-kan-IR_L cassette in plasmid pSU2620 in the presence of helper plasmids carrying the IS91 transposase gene^a. a. Experiments were carried out and results expressed as in Table 1, except that donors were derivatives of strain D1210 containing R388, pSU2620 and the helper plasmid shown in the first column.

The frequency obtained for pSU2638, in which tnpA was reversed with respect to Plac, was 100-fold lower than that obtained for pSU2637 (4 × 10⁻⁶ versus 3 × 10⁻⁴). In both cases, IR_L-kan-IR_R was complemented by the transposase encoded by a gene carried in another plasmid. Therefore, the observed difference must be caused by differences in transposase level. Basal transposase activity in plasmid pSU2638 was about the same as that of its equivalent pSU2633, which provided tnpA in cis (4 × 10⁻⁶ versus 3 × 10⁻⁶).

Transposition products of IS 91

Transposition of IR_L-kan-IR_R was detected by the presence of Km^r transconjugants produced by transposition of the cassette into the conjugative plasmid R388. As shown in Fig. 2, donor plasmids contain a cat gene outside the cassette, as well as the kan gene. Cm^r transconjugants occurred at a frequency of roughly 10–20% of the frequency of Km^r transconjugants for all donor plasmids (Tables 1 and 2). The Cm^r transconjugants were the result of co-integrate formation or multiple tandem insertions of donor plasmids into R388. DNAs from both Km^rCm^s and Km^rCm^r transposition products between R388 and donor plasmids carrying tnpA both in cis and in trans were analysed with the restriction enzyme AvaI. Digestion of plasmid pSU2620 with AvaI resulted in three fragments of 274 bp, 435 bp and 3602 bp (Fig. 1). In the case of plasmids pSU2631 to pSU2634, digestion with AvaI resulted in three fragments of 274 bp, 435 bp and 4997 bp (Fig. 2A). R388 also produced 15 AvaI fragments, ranging from 0.54 kb to 6.9 kb (Avila and de la Cruz, 1988). AvaI digests of plasmid R388 DNA were compared with DNAs from R388 recombinants in which the transposable cassette had inserted. One R388 DNA band disappeared, while new fragments appeared as a result of the insertion. All the Km^rCm^s transconjugants analysed were simple insertions of the IR_L-kan-IR_R cassette into R388. Km^rCm^r transconjugants were the sum of co-integrates between R388 and donor plasmids, and multiple tandem insertions of donor plasmids into R388 (plasmid fusions). The relative proportion between co-integrates and plasmid fusions ranged from 6:4 in the case of plasmid pSU2631 to 3:7 in plasmid pSU2632 (Table 1). In the case of plasmid pSU2631, 87% of transposition products were simple insertions of the transposable cassette resulting from direct transposition; the remaining 13% were the sum of co-integrates and plasmid fusions. We assume that co-integrates and plasmid fusions are the result of the failure in IR_L recognition when the transposase encounters IR_L during rolling-circle transposition (Mendiola et al., 1994).

One-ended transposition of kan-IR_R occurs at similar frequencies to transposition of IR_L-kan-IR_R

To analyse one-ended transposition, the IR_L fragment of the IR_L-kan-IR_R cassette was excised from plasmids that carried tnpA in cis. In this way, plasmids pSU2640, pSU2645 and pSU2641 were obtained, which represent one-ended derivative versions of pSU2631, pSU2632 and pSU2633 respectively (Fig. 2C). Transposition frequencies obtained with these constructions are shown in Table 3. One-ended transposition occurred at nearly the same frequency as transposition of the complete cassettes (2 × 10⁻⁴ for pSU2640 versus 9 × 10⁻⁴ for pSU2631). When Plac was activated by the addition of IPTG to the medium, most recombinants (65% for pSU2640 and approximately 90% for both pSU2641 and pSU2645) were multiple tandem insertions of the donor plasmid, as described previously (Mendiola et al., 1994), and they were selected as Km^rCm^r transconjugants; the remaining 35–10% were Km^rCm^s transconjugants corresponding to incomplete plasmid fusions, which did not carry the cat gene of donor plasmids. On the contrary, when there was no induction, most recombinants were incomplete plasmid fusions.

Table 3. . Transposition frequencies of IR_R-kan cassettes (lacking IR_L) when the IS91 transposase gene was carried in cis^a. a. Experiments were carried out and results expressed as in Table 1.b. A total of 100 individual Km^r transconjugants were replica plated on Cm-containing plates.

Two mutants of IR_L were constructed in which 2 bp or 3 bp of IR_L were deleted. The DNA sequences of wild-type IR_L in plasmid pSU2600 and of the mutant IR_Ls of plasmids pSU2635 and pSU2636 are shown in Table 4, together with the transposition frequencies obtained with these plasmids. IS91 transposition from these mutants resulted in multiple tandem insertions of the donor plasmid in the recipient, indicating that the mutant IR_Ls were inactive. These results imply that the tetranucleotide CGAG is essential for IR_L function. Transposition frequencies obtained for these one-ended derivatives were equal to that of pSU2600 (2–3 × 10⁻⁴ versus 4 × 10⁻⁴), further indicating that the lack of IR_L function does not impair IS91 transposition.

Table 4. . Effect of 2 bp and 3 bp deletion of IR_L in IS91 transposition and its products^a. a. Donor strains were DH5α derivatives carrying pSU4053, which has the TRA_W region of plasmid R388 cloned in pHG327 (Stewart et al., 1986), and the plasmid listed in the first column. They were mated with strain UB1637 for 90 min at 37°C, and transconjugants were selected on either Sm + Cm (which measures the formation of co-integrates or fusions between pSU4053 and the plasmid in each case) or Sm + Ap (conjugation frequency of pSU4053).b. Nucleotides from wild-type IR_L sequence are indicated by upper case letters, and the mutated nucleotides by lower case letters.c. Transposition frequencies are the relation between Sm^rCm^r and Sm^rAp^r transconjugants. Frequencies shown are the average of a number of experiments [n], with the standard deviation shown in brackets.d. Plasmid DNA was obtained from individual Sm^rCm^r colonies and analysed by digestion with restriction enzymes Bgl I, ApaLI and AvaI.

Specificity of insertion of IS 91 derivatives lacking ORF121

It was known that IS91 inserts specifically 3′ of the sequences CTTG or GTTC, both of them used at roughly the same frequency (Mendiola and de la Cruz, 1989; Table 5). The same specificity of insertion is maintained in one-ended transposition (Mendiola et al., 1994; Table 5). However, when we sequenced the junctions of eight recombinants produced by simple insertion of IR_L-kan-IR_R in R388, we found great specificity for the sequence CTTG-3′ (Table 5). Wild-type IS91 contains an additional open reading frame (ORF), named ORF121, which precedes the tnpA gene (Mendiola et al., 1992). Three Ω-interposon insertions in ORF121 were available from previous work (Mendiola et al., 1992). IS91 derivatives with Ω inserted in ORF121 were transposition proficient, so these insertional mutants were transposed to R388 to determine whether the lack of ORF121 expression was causing the change in IS91 insertion specificity. We sequenced the junctions between IR_R and R388 DNA in 18 of the resulting transposition products. The results (Table 5) confirm that ORF121 mutants give very similar specificity to IR_L-kan-IR_R. The χ² test was applied to ascertain if the results obtained with wild-type IS91 were significantly different from those obtained with the mutants. A χ² value of 15.2 was obtained, corresponding to a P < 0.001 for a set of data with two degrees of freedom, implying that the two sets of experiments differ very significantly, i.e. target selection was altered in these mutants. It is of special interest that two of the mutants inserted in the sequence 5′-CTCG (or its complementary 5′-CGAG), which corresponds to the last four nucleotides of IS91 IR_L. These results imply that ORF121 could be involved in target selection and specificity of insertion of IS91.

Table 5. . Specificity of insertion of IS91 and derivatives lacking ORF121. a. Mendiola and de la Cruz (1989).b.Mendiola et al. (1994).c. Transposition experiment as described in Table 1. DNAs from eight simple insertions were purified and the R388-IR_R junctions sequenced.d. The three plasmids, containing φ-interposon insertions in ORF121 (in the IS91 nucleotide positions shown in brackets), have been described previously (Bernales, 1997; Table 6). Transposition experiments with these plasmids were carried out as described in Table 1 for the IR_L-kan-IR_R cassettes, and recombinant DNAs purified and their junctions sequenced.

Intramolecular transposition of IS 91

Storage of derivatives of strain DH5α carrying either pSU2631 or pSU2632 in 50% glycerol at −20°C resulted in the loss of Cm^r recovered bacteria. When these bacteria were cultured and plated on Km-containing plates, about 35% of individual colonies were Cm^s. Loss of antibiotic resistance was not detected in strains containing plasmids pSU2633 and pSU2634, which have low transposase levels, or in strains carrying pSU2620 and the helper plasmids pSU2637 or pSU2638. Furthermore, there was no loss of antibiotic resistance when plasmids pSU2631 and pSU2632 were carried in strain D1210, which carries the lacI ^q gene. Inactivation of cat seemed to be a useful result, as it could have been produced by deletion of sequences adjacent to the IS91 cassette resulting from intramolecular transposition, as occurs for other IS elements (see Introduction ). Nine plasmid DNAs resulting from independent cat inactivation events were analysed. All nine Cm^s plasmids contained an extra copy of IR_L-kan-IR_R inserted within cat. Insertions were mapped to three precise locations by endonuclease digestion and DNA sequencing (Fig. 3). The DNA sequence located between the original and the duplicated copies of the cassette remained intact, without having experienced any rearrangement such as deletion or inversion.

Transposition frequency of IS 91 cassettes is related to transposase transcriptional level

The results shown in Tables 1 and 2 indicate that transposition of IS91 cassettes can be induced to levels of 10⁻³ transposition events/recipient DNA molecule. Such high levels of transposition are convenient for genetic analysis. The fact that a high transposition frequency can be achieved by simply activating a lac promoter, suggests that transposition of IS91 could be regulated mainly at the level of transposase gene expression. This is a common mechanism of transposition regulation (Kleckner, 1990).

IS 91 transposase activity does not show cis preference

Transposition mediated by cis-acting transposases (IS10, IS903 ) is extremely sensitive to the distance between the 3′ end of tnpA and the nearest transposon IR (Morisato et al., 1983; Derbyshire et al., 1990). However, IR_L-kan-IR_R transposition did not show any dependence on the distance between tnpA and IR_R, as transposition of IS91 cassettes occurred at the same frequency in plasmids pSU2631 and pSU2632, the distance between tnpA and IR_R being 1.6 kb in pSU2631 but only 0.4 kb in pSU2632. This result could also imply that IS91 transposase is trans-acting, unlike the majority of IS transposases. In fact, transposition of the IR_L-kan-IR_R cassette when tnpA was carried in a helper plasmid (pSU2637 or pSU2638) occurred at roughly the same frequencies as those obtained when tnpA was carried in cis. Relative transposition frequencies when cis-preferring transposases act in trans are 100-fold lower (Isberg and Syvanen, 1981; Machida et al., 1982; Morisato et al., 1983; Derbyshire et al., 1990). It is unlikely that tnpA gene expression is higher in pSU2637 than in pSU2631, as expression signals are conserved in both constructions.

IS 91 one-ended transposition

After the excision of IR_L, IS91 transposable cassettes undergo one-ended transposition at frequencies similar to the transposition frequency of IR_L-kan-IR_R (Table 3). We can conclude that only the transposase, the 82 bp IR_R region and the adjacent GTTC or CTTG sequences are involved in the interaction between a donor IS element and a target DNA sequence and, consequently, in the initiation of IS91 transposition. On the other hand, in all other transposons analysed, interaction between both IRs and the transposase is essential for transpositional cleavage and strand transfer reactions (Mizuuchi, 1992; Bainton et al., 1993; Sakai et al., 1995). One puzzling result appeared when one-ended transposition products were analysed with respect to their termination sites (Table 3). When one-ended transposition occurred at low levels (no induction), most recombinants (60–70%) were plasmid fusions in which RC transposition had finished before cat, so they do not contain an entire copy of the donor plasmid. Conversely, most fusion products were Cm^r when Plac transcription was induced. It seems that RC transposition does terminate at more proximal sites when there is active transcription from Plac. The explanation cannot be that transposase levels affect site selection, as pSU2640 and pSU2641 both show the same effect, in spite of the fact that Plac induces TnpA expression in pSU2640 but not in pSU2641. One more likely explanation could be that transcription is diminishing the availability of DNA sites for termination directly. How this occurs has yet to be determined.

IS 91 specificity of insertion can be modulated by ORF121 accessory protein

IS91 uses the target sequences 5′-GTTC and 5′-CTTG indistinctly (Mendiola and de la Cruz, 1989; Table 5). However, IR_L-kan-IR_R, as well as the mutants in ORF121, showed a clear preference for the target sequence 5′-CTTG (Table 5). Thus, the cleavage site of these recombinants was 5′-CTTG/AT, that is the same cleavage site described for the pC194 group of rolling-circle replication plasmids and for phage φX174 (Gruss and Ehrlich, 1989; Novick, 1989). It can be inferred that ORF121 could be involved in target selection by broadening IS91 insertion specificity. As a consequence, the sequence 5′-GTTC could be used efficiently only in the presence of ORF121. Interestingly, ORF121 mutants used 5′-CTCG as a second preference site. This result can be interpreted in connection with the fact that alteration in just two or three of the IR_L terminal nucleotides, that is in the sequence 5′-CTCG, abolished IR_L function (Table 4). Thus, it seems that 5′-CTCG is a secondary cleavage site recognized by TnpA, while 5′-CTTG is the real primary site. This notion will explain why the termination reaction in IS91 transposition is not 100% efficient.

IS 91 intramolecular transposition results in second-site simple insertions

Intramolecular transposition of well-known transposons generates adjacent deletions. In the case of IS91, deletion formation could be analysed conveniently by looking for Cm^s derivatives of plasmids pSU2631 and pSU2632. These derivatives were in fact produced spontaneously upon storage of strains containing these plasmids, presumably because of high transposase intracellular levels. Their analysis showed that, in all cases, antibiotic resistance loss was caused by the insertion of a second IR_L-kan-IR_R element into cat. Insertion of a second cassette could be the result of intermolecular transposition, as IS91 did not show immunity to transposition (unpublished results). However, existing transposition models for other transposons do not predict this outcome as the unique result of intramolecular transposition. Simple secondary insertions occur in sequences adjacent to transposons, but always together with deletions and/or inversions (Galas and Chandler, 1989; Jilk et al., 1993; Craig, 1996). On the other hand, according to the RC transposition model that we proposed for IS91 (Mendiola et al., 1994), IS91-mediated intramolecular transposition would not result in adjacent deletions or sequence inversions. Rather, it would result in the insertion of a second element either as a direct or as an inverted repeat with respect to the pre-existing element, depending on which strand of the target sequence is nicked. The proposed mechanism is explained in Fig. 4. As we always found second-site insertions, but never deletions, when Cm^s derivatives were analysed, we believe that this result further strengthens the validity of our model of RC transposition. It should be noted that second-site insertions were obtained in only one orientation for the plasmids used in these experiments. All pSU2631 recombinants contained the second copy of IR_L-kan-IR_R inserted as an inverted repeat, whereas all pSU2632 recombinants contained the second copy inserted as a direct repeat. That is, irrespective of the orientation of IR_R, insertions always occurred at the same orientation with respect to the plasmid origin of replication. A unique orientation was also produced when we analysed the transposition of a one-ended derivative of IS91 to the pBR322 plasmid derivative pSU4053 (Mendiola et al., 1994). We can offer no explanation for this orientation effect at the moment.

Bacterial strains and plasmids

Bacterial strains used were E. coli K-12 derivatives DH5α [F⁻endA1 recA1 gyrA96 thi-1 hsdR17 supE44 relA1 Δ(arg F-lacZYA)U169 φ80ΔlacZΔM15; Grant et al., 1990], D1210 (F⁻recA lacI ^qrpsL hsdR hsdM ; Sadler et al., 1980), UB1637 (F⁻lys his trp rpsL recA56; de la Cruz and Grinsted, 1982) and UB5201 (F⁻ pro met recA56 gyrA ; de la Cruz and Grinsted, 1982). Plasmids used are listed in Table 6.

Table 6. . Plasmids (Bartoloméet al. (1991); Mendiola et al. (1994); Bolland et al. (1990); Norrander et al. (1983); Datta and Hedges (1972)). Su, sulphadiazine; Tc, tetracycline; Sp, spectinomycin.⁹¹, DNA sequences from IS91.a. These three plasmids derive from a set of three Tn1732 insertion derivatives in ORF121 of IS91 (Mendiola et al., 1992). Tn1732 contains EcoRI sites at its terminal IR. The bulk of Tn1732 was eliminated from each insertion by EcoRI digestion, and substituted by the Ω-interposon (Bernales, 1997).b.Stewart et al. (1986).

Genetic techniques

Plasmid DNAs were introduced in bacteria either by transformation, using the method of Chung and Miller (1988), or by electroporation (Dower et al., 1988). Luria broth (LB; Sambrook et al., 1989) was used for bacterial growth in liquid. LB agar (1.5%) was used for growth on plates. Selective media included antibiotics at the following concentrations: sodium ampicillin (Ap) 100 μg ml⁻¹; chloramphenicol (Cm) 25 μg ml⁻¹; kanamycin sulphate (Km) 50 μg ml⁻¹; nalidixic acid (Nx) 20 μg ml⁻¹; streptomycin sulphate (Sm) 300 μg ml⁻¹; and trimethoprim (Tp) 25 μg ml⁻¹ (in Mueller–Hinton agar; Difco).

Transposition assays

These were carried out by a mating-out assay as described previously (Mendiola et al., 1992, 1994). Donor strains were derivatives of D1210 (Nx^s) containing the conjugative plasmid R388 (Tp^r) plus a non-mobilizable plasmid that carried the transposon. The recipient strain was DH5α. Before mating, donor strains were streaked out on LB agar plates containing antibiotics for plasmid selection and incubated at 30°C for 2 days. IPTG (Boehringer Mannheim) at a final concentration of 0.5 mM or glucose at a final concentration of 0.5% were added to the plates where indicated. Nalidixic acid was used to counterselect donors. Transposition frequency is the number of transconjugants carrying the transposon marker divided by the number of transconjugants carrying the marker of the conjugative plasmid (Tp^r).

Construction and analysis of plasmids DNAs

Plasmid DNA was isolated by a modification of the alkaline lysis method of Ish-Horowicz and Burke (1981), as described previously (Martínez and de la Cruz, 1988). The method of Serghini et al. (1989) was used for small-scale preparations. Molecular cloning techniques were carried out according to Sambrook et al. (1989). Restriction endonucleases, the large fragment of DNA polymerase I (Klenow fragment; Pharmacia Biotech) and T4 DNA ligase (New England Biolabs) were used according to the manufacturers’ recommendations. Construction of plasmid pSU2620 is explained in Fig. 1. Construction of plasmids pSU2637, pSU2638, pSU2631, pSU2632, pSU2633 and pSU2634 is explained in Fig. 2. For the construction of plasmids pSU2635 and pSU2636, 5 μg of pSU2600 digested with XhoI restriction enzyme was treated with 0.8 U of S1 nuclease for 10 min at room temperature; then, plasmid DNA was treated with Klenow polymerase and religated. Construction of plasmids pSU2640, pSU2641 and pSU2645 is explained in Fig. 2. To construct plasmids pSU2930, pSU2931 and pSU2932, the Ω fragment of pHP45 (Prentki and Krisch, 1984), digested with EcoRI, was cloned at the single EcoRI site of plasmids pSU2915, pSU2921 and pSU2922 (Mendiola et al., 1992) respectively.

DNA sequencing

Sequencing was carried out by the dideoxynucleotide chain-termination method (Sanger et al., 1977) with the fmol DNA sequencing System (Promega), which is a polymerase chain reaction (PCR)-directed sequencing kit, and Sequenase version 2.0 (US Biochemical) using plasmid DNA as substrate. Primer oligonucleotides were labelled in their 5′ ends using T4 polynucleotide kinase (New England Biolabs) and [γ-³²P]-ATP (3000 Ci mmol⁻¹; Amersham). The oligonucleotides used were M13 reverse sequencing primer (5′-CAGGAAACAGCTATGAC-3′) (Boehringer Mannheim), and two oligonucleotides corresponding to positions 1761–1777 and 1790–1806 of the upper strand of IS91 respectively (Mendiola et al., 1992) (5′-ACAAGGGGTGTTGAAG-3′ and 5′-GTGGTGCTTTTTTAGTC-3′).