ABSTRACT
The regulatory region of the Mycobacterium fortuitum plasmid pAL5000 was studied in vivo and in vitro by mutational analysis. This region comprises the origin of replication for the plasmid and the start point of transcription for the repA/B genes, which encode the two replication proteins RepA and RepB. In this region there are two binding sites for RepB: a low-affinity site which is probably the origin of replication and a high-affinity-site which overlaps the promoter and implies an autoregulated expression of RepB. The high-affinity site contains two 8 bp palindromes, as well as an inverted repeat structure. By introducing point mutations into each of these motifs and monitoring changes to RepB binding in a gel-retardation assay, it was shown that the central, GC-rich palindrome (the GC-box) is the most important motif for protein binding. Mutations in the second, AT-rich palindrome (the AT-box) had no effect on protein binding and the inverted repeat structure per se was not needed, though some single-base changes affected binding to one or other of the DNA strands. These mutations were subsequently tested in vivo for their effects on plasmid replication in Mycobacterium smegmatis. Any change to the GC-box abolished replication, but changes to the other motifs were dependent on the position of the changed base, again indicating that the inverted repeats are not essential and that the AT-box is part of the promoter and not primarily recognised by RepB. The mutated plasmids did not show any changes in copy number to that of the wild-type. The expression of RepB was boosted by introducing a stronger promoter upstream of the repA/B genes. The resulting plasmid was capable of increasing to a degree in trans the copy number of other plasmids carrying the ori region, but was unstable when present on its own in M.smegmatis.
The plasmid pAL5000 (1 ; GenBank Accession Number M23557) was first isolated from Mycobacterium fortuitum (2 ). It is the most studied mycobacterial replicon and constructs based on the pAL5000 origin are able to replicate in a wide range of mycobacteria. Several Escherichia coli-Mycobacterium shuttle vectors have been constructed based on this replicon (3 -7 ).
We have previously shown (8 -9 ) that the minimal functional replicon of pAL5000 comprises a cis-acting site, presumably the origin itself, and two genes, repA and repB, coding for replication proteins (Fig. 1 A). The repA and repB genes overlap by 1 bp and are transcribed as a single RNA species (9 ). Plasmids carrying the origin but lacking the repA/B genes are able to replicate if those two genes are present in trans.
Escherichia coli strain DH5[alpha] (17 ) was used throughout to manipulate plasmid DNA. Constructs created in this study are shown in Table 1 . Escherichia coli cells were grown in TY medium (16 g tryptone, 10 g yeast extract/l) with or without the addition of kanamycin (Km; final concentration 50 [mu]g/ml), ampicillin (Ap; 50 [mu]g/ml), or chloramphenicol (Cm; 40 [mu]g/ml). Mycobacterium smegmatis strain mc2155 (18 ) was grown in Lemco medium (Difco) or on Lemco agar plates.
Plasmid DNA was isolated from E.coli cells by standard procedures (19 ). For large-scale plasmid preparations, Wizard midipreps (Promega) were used. Mycobacterial plasmid DNA was extracted through electroduction (20 ) into E.coli cells followed by standard DNA preparations.
Competent M.smegmatis cells were prepared as described by Snapper et al. (19 ). Transformation with 0.1-1 [mu]g DNA was performed using 300 [mu]l aliquots of cells.
Relative plasmid copy numbers were determined as single cell resistance (SCR) to Km (21 ) as described earlier (8 ).
Table 1
Templates for DNA-binding assays with mutated H-sites were prepared in PCR reactions using forward oligonucleotides based on the sequence 5" CGTGTCGGACCATACACCGGTGATTAA 3" (oligonucleotide OKDH; 9 ) but with mutations corresponding to those shown in Figure 2 . The two exceptions were the oligonucleotides used to introduce the mutations +7 (5" CGTGTCGGACCATACACCGGTGATTAATGGTGG 3") and +11 (5" CGTGTCGGACCATACACCGGTGATTAATCGTGCTCT 3"). The 5" ends of the oligonucleotides were radioactively labelled using [[gamma]-32P]ATP (Amersham) and T4 polynucleotide kinase (Promega). The oligonucleotide OFP1 (5" GAGCAGATCGTCGCTTGCCA 3"; 9 ) was used as reverse primer; all templates are 118 bp in size. Pfu DNA polymerase (Stratagene) was used in all PCR reactions. The expression of recombinant RepB protein as an MBP-fusion protein and the demonstration of specific binding by the purified RepB protein to the H-site have been reported previously (9 ). The binding assays were carried out as described earlier (a 15 min incubation at 37oC in a buffer consisting of 100 mM Tris-HCl pH 8.0, 10 mM dithiothreitol, 20 mM MgCl2; 1 [mu]g salmon-sperm DNA per reaction; complexes separated by electrophoresis on a 6% non-denaturing polyacrylamide gel) and the calculation of KdDNA was done as described. All binding experiments were carried out in triplicate with a wild-type titration in parallel to each assay, in order to minimise dilution errors. Templates for the L-site were prepared as described previously (9 ).
The construct pDQ31, carrying the L-site but not the H-site, was created by amplifying the region between bp 3871and 4589 of pAL5000 using the primers ORF1E (5" GCGCGATATCGAGCCGAGAAC 3") and OKDB (5" ACGAGCTCCAAGTCAGATAT 3"; 9 ), which were treated with 5 U T4 polynucleotide kinase (PNK; Promega) and ATP (1 mM) for 30 min prior to the PCR reaction. The PCR product was ligated into SmaI-digested pUC18.
The wild-type control region of pAL5000 coding for the repA/B genes and the H-site, but lacking the L-site, was amplified in a PCR reaction using the oligonucleotides O[Delta]L (5" CAGCGAGATATCTGACTTGGAGCT 3") and ORF2B (CGACACCGGATCCCCAATTGCGTTA; 9 ). These oligonucleotides introduce an EcoRV and a BamHI site, respectively and the PCR product was cloned in pBSKS- (Stratagene) in the EcoRV-BamHI sites, creating pDQ11.
To introduce mutated binding-sites into the regulatory region, the oligonucleotides described for the gel-retardation assay were used as forward primers in PCR reactions with ORF2B as reverse primer to amplify the mutated H-sites together with the repA/B genes. The forward primers were kinased with ATP and PNK and the products cloned in the EcoRV-BamHI sites of pBSKS-.
DNA sequencing of the mutated constructs was done with the modified dideoxy-chain termination method (22 ) using the Auto Read sequencing kit and the Automatic Laser Fluorescent DNA sequencer (Pharmacia).
To add the L-site to these constructs, the cloned regions were ligated to the vector pDQ31 as blunt/BamHI fragments (after cutting with HindIII, filling in with the Klenow polymerase [Promega] and cutting with BamHI) in the BamHI site and the filled-in XbaI site of pDQ31.
The KmR gene from Tn903 (23 ,24 ) was added to all the final constructs on a BamHI-BglII fragment ligated to the BamHI site in the replicons (downstream of the repA/B genes) in such an orientation that transcription from the KmR gene was away from repA/B as not to interfere with the replicon. The resulting series of mutated replicons is listed in Table 1 .
The repA/B-overexpressing construct pDQ66 was made by introducing the Hsp60 promoter from M.bovis BCG (25 ) which was cut out from the plasmid pMV261 (5 ) on a PvuII-EcoRV fragment and ligated to the EcoRV site of pDQ11. The cassette was then cut out from the vector with XbaI and placed in the XbaI site of pDQ31 in the same orientation as in the wild-type plasmid. Adding the KmR gene on a blunt-ended BamHI-BglII fragment in the blunted HindIII site downstream of repA/B created pDQ71. Again, transcription from the KmR gene was away from the repA/B genes.
The H-site overlaps the promoter region of the repA/B promoter (9 ; Fig. 1 B) and RepB binding here presumably serves to autoregulate its expression. The regions protected from DNaseI cleavage are staggered and DNA-binding assays indicate that two molecules of RepB bind to this site in a cooperative fashion (9 ). There are three notable structural features in the H-site: two 8 bp palindromes and one 5 bp inverted repeat. The palindrome GATTAATC (the AT-box) was suspected on grounds of its high (A+T)-content, to be a feature of the promoter, while the GC-rich and overlapping palindrome CACCGGTG (the GC-box) might be a recognition sequence for RepB.
To test these assumptions, a set of oligonucleotides was constructed, with specific mutations either in one or the other palindrome or in the repeats. In the GC-box, pairs of bases were exchanged, to keep the palindromic structure intact. Some single-base changes were also made to this box. Base changes were always C -> G and A -> T or the reverse. These oligonucleotides were used in PCR reactions with the oligonucleotide OFP1 as reverse primer and the products were used in gel-retardation assays using purified RepB protein.
To score for the relative importance to binding affinity, we titrated RepB onto the target and calculated the KdDNA for the respective templates as described previously (9 ). These values were compared with the binding-constant for RepB to the wild-type H-site in parallel reactions and the relative binding constants determined.
The results of these assays are shown in Figure 2 . The mutation with the most dramatic effect was the changing of the first C of the GC-box to a G (mutation H1). This led to a virtual abolition of binding, with the affinity reduced almost a 100-fold. Changing both the C and the final G of the box, preserving the palindrome, practically abolished binding (mutation H18; Fig. 3 A). The final G of the GC-box, when changed to a C (mutation H8), had a less strong, but still pronounced effect.
To test for the influence of the different motifs on the replication abilities of pAL5000, replicons were constructed with mutated regulatory sites by amplifying the repA/B region including the H-site in PCR reactions. The PCR products were cloned in pBluescript vectors. None of these constructs, which all lacked the L-site and thus the putative ori, was able to replicate in M.smegmatis. The defective replicons were then excised and introduced into a pBluescript vector carrying the L-site but not the H-site (pDQ31) and the resulting constructs were electroporated into M.smegmatis. As a control, we made similar constructs where the amplification was performed using the oligonucleotide O[Delta]L which produces a wild-type regulatory region, still lacking the L-site. This wild-type construct did not replicate in M.smegmatis, but when fused to the L-site in pDQ31, the resulting construct (pDQ51) was viable.
The results of the in vivo assay are summarised in Figure 2 . All the changes to the GC-box which we tested abolished the replicative ability of the plasmid. In contrast, the mutations +2 and +7 to the AT-box produced viable plasmids. The -4 and -5 mutations, however, abolished replication. The mutation +11, which destroys the inverted-repeat motif downstream of the AT-box, yielded a viable plasmid.
The mutation -7 yielded a viable plasmid, although the transformation efficiency of the construct (pDQ57) was consistently ten times lower than that of the other constructs. The other viable mutants transformed with an efficiency comparable with that of pDQ51. There was no difference in growth rate between cells carrying pDQ57 and those carrying pDQ51. Copy-number determinations using the method of Nordström (21 ) showed no significant differences between pDQ51 and the mutant plasmids, including pDQ57 (not shown).
It took 4-5 days for cells transformed with any of these plasmids, including pDQ51, to form colonies on Km plates, in contrast to cells transformed with the shuttle vector pYUB12, which carries all of pAL5000 (5 ). Such cells typically form colonies after 3 days. This indicated that the initial expression of repA/B, immediately upon transformation, was less efficient in the manipulated plasmids. This conclusion was further supported by co-transformation experiments where the construct pUH11 (8 ) was electroporated into M.smegmatis together with pYUB12 or the pDQ series of constructs. The plasmid pUH11 carries the hygromycin resistance gene from Streptomyces hygroscopicus (26 ) and lacks much of repA and all of repB and thus is unable to replicate on its own but can be activated in trans. Cells were readily co-transformed with the pair pUH11/pYUB12, but none of the pDQ series would support replication, not even pDQ51, which carries the wild-type H-site. However, when cells were first transformed with pDQ51 or one of the viable mutant constructs, and then transformed with pUH11 in a second step, this construct could be introduced with high efficiency. This supported the notion that the initial repA/B expression is important for plasmid viability.
Since the experiments above showed the importance of a sufficient level of RepA and/or RepB in the cells for establishing a plasmid population, it was thought that one possible way of raising the copy number of the plasmid would be to increase the expression of the repA/B genes.
To boost repA/B expression we introduced the Hsp60 promoter from M.bovis BCG (25 ) into pDQ51 between the L-site and the H-site in such an orientation that the repA/B genes would be transcribed. We expected this construct (pDQ66) to have higher repA/B expression, since the Hsp60 promoter is very strong and not regulated by RepB.
This construct was tested for its ability to support replication in trans, by co-transforming M.smegmatis cells with pDQ66 and pUH77. The construct pUH77, which has been described earlier (8 ), carries a 1 kb region comprising the ori, as well as the KmR gene from Tn903 but lacks repA/B. There is only the Ap resistance marker on pDQ66 and so selecting on Km for pUH77 would co-select for pDQ66, since this plasmid is necessary for pUH77 to replicate. As a control, pUH77 was transformed together with pUH56 (8 ), which carries a wild-type ori and repA/B and has wild-type pAL5000 replication characteristics, but lacks a Km resistance marker.
The co-transformation efficiency for the pair pDQ66/pUH77 was 1-5% that of the wild-type pair pUH56/pUH77. The spread in SCR values to Km for pDQ66/pUH77 was greater than that for pUH56/pUH77 (Fig. 4 ), but the values were always higher than for the wild-type. Thus, the increased amount of RepA and/or RepB in the cells carrying pDQ66 seems to have a positive effect on the copy number of the activated pUH77. The increase in copy-number estimated from SCR measurements was not more than 1.4-2-fold.
This paper presents a dissection of the regulatory region of the M.fortuitum plasmid pAL5000. Central to this region is the so-called H-site, where the replication protein RepB binds with high affinity. The structure of the H-site was probed by specific mutations to bases in the three different structural motifs present. These motifs are: a GC-rich palindrome (GC-box), an AT-rich palindrome (AT-box) and a 5 bp inverted repeat. The mutations were tested for changes to RepB binding in vitro as well as for plasmid viability in vivo.
The integrity of the GC-box was shown to be crucial to replication. All changes to this box produced non-replicating plasmids. The initial C of the GC-box was the most important single base in the H-site for RepB binding. If this base was changed to a G, the binding constant dropped by two orders of magnitude. Changing the first and last nucleotides, keeping the palindromic structure of the box intact, virtually abolished binding. Other bases were less important for binding; pairwise changes to the other bases in the box, keeping the palindrome intact, reduced RepB binding between 5- and 25-fold.
Mutations to the AT-box sometimes abolished replication but in no case did they affect RepB binding in vitro, which indicates that this box is indeed part of the promoter structure, rather than of a recognition motif for RepB. The exception was the first G of the AT-box, but this base is also part of the GC-box and thus might have a dual role. Support for this role for the AT-box is given by the observation (9 ) that 11 bp of the repA/B promoter region, including the AT-box, can be found in the promoter region for the Lactococcus lactis dnaE gene (28 ).
If the AT-box is a part of the promoter, single-base pair changes might preserve a functional promoter whilst leaving the RepB binding unaffected. The binding sites for RepB, as defined by DNaseI-footprinting experiments, are staggered (Fig. 1 B; 9 ) and the DNA region protected from DNaseI cleavage is probably larger than the area actually in contact with RepB. Thus, even though the protein would occlude the AT-box, it would not have to be in actual contact with the DNA in this region.
Apart from the GC-box, where all mutations resulted in loss of replication ability, observations of changes to the DNA-binding properties of a mutated motif in vitro could not be used to predict whether the change would lead to a replicating or a non-functional plasmid in vivo. Thus only two out of three mutations to the upstream sequence of the inverted repeat abolished replication, though the effects on RepB binding were of a similar magnitude. No changes downstream of the GC-box had any effect on RepB binding. The mutation +11 produced a wild-type binding pattern in vitro, while the corresponding mutation in the upstream motif, mutation -5, showed reduced RepB binding and could not support replication. This suggests that the inverted repeat structure is not important as such, but the upstream bases act as part of a binding site. Indeed, the mutation -7, while reducing KdDNA >10-fold and producing a binding pattern indicating that RepB only binds to one strand of this construct, nevertheless produced a replicating plasmid, albeit with reduced transformation efficiency. It is not clear why the other two mutations in this motif, while showing similar binding patterns in vitro, led to non-replicating plasmids.
The mutations H8, -7, -5 and -4 had an effect on the binding pattern of RepB which supports our argument that two copies of RepB bind in a cooperative fashion to the H-site. In particular, the binding pattern to the mutated target H8 was virtually indistinguishable from that to the L-site (Fig. 3 B). From DNaseI-footprinting experiments we have shown that binding to the L-site is to one strand of the DNA helix only (9 ). If the mutations destroyed one binding motif of the H-site, such a pattern of only one protein molecule would be expected. The slope of the binding curve for the H8 mutation is less steep than that for the wild-type H-site (Fig. 5 ), which is also seen for binding to the L-site (9 ). This is another indication that there is no cooperative binding to the mutated site. Different binding patterns for replication proteins in dual roles as autorepressors and replication initiators have been shown for other plasmids; e.g. binding as autorepressor in dimeric form and as initiator as monomer (29 -31 ) but whether this is the case for RepB remains to be determined.
This work was carried out as part of the Glaxo Wellcome Action TB initiative. We thank Ken Duncan for critically reading the manuscript.
*To whom correspondence should be addressed at present address: Division of Molecular Infection Biology, Research Centre Borstel, Parkallee 22, D-23845 Borstel, Germany. Tel: +49 4537 188 486; Fax: +49 4537 188 686; Email: pstolt@fz-borstel.de
Plasmid Characteristics
pUH11 pUH4 with AscI-deletion (152-1687) [Delta]repA/B
pUH12 pYUB12 with AscI-deletion (152-1687) [Delta]repA/B
pUH36 pAL5000 region 3875-752 (repA) in pUC18
pUH52 pAL5000 region 3875-1093 in pUC18
pUH61 pAL5000 region 3875-1093 in pUC18; KmR (Tn903)
pUH77 pAL5000 region 3861-4837 in pUC18; KmR (Tn903)
pDQ31 pAL5000 bp387to 4589 (L-site) in pUC18 SmaI-site
pDQ51 Wild-type H-site + repA/B in pDQ31+ KmR (Tn903)
pDQ52 H-site with mutation (-5) + repA/B in pDQ31+ KmR (Tn903)
pDQ55 pDQ11 with M.bovis BCG Hsp60 promoter from pMV261as
200 bp fragment in EcoRV site upstream of H-site
pDQ57 H-site with mutation (-7) + repA/B in pDQ31+ KmR (Tn903)
pDQ58 H-site with mutation (+2) + repA/B in pDQ31+ KmR (Tn903)
pDQ59 H-site with mutation (+11) + repA/B in pDQ31+ KmR (Tn903)
pDQ60 H-site with mutation (+3) + repA/B in pDQ31+ KmR (Tn903)
pDQ61 H-site with mutation (H1) + repA/B in pDQ31+ KmR (Tn903)
pDQ62 H-site with mutation (+7) + repA/B in pDQ31+ KmR (Tn903)
pDQ63 H-site with mutation (H45) + repA/B in pDQ31+ KmR (Tn903)
pDQ64 H-site with mutation (-4) + repA/B in pDQ31+ KmR (Tn903)
pDQ65 H-site with mutation (H8) + repA/B in pDQ31+ KmR (Tn903)
pDQ66 Hsp60 promoter + H-site and repA/B from pDQ55 as XbaI
cassette in pDQ11
pDQ71 pDQ66 + KmR (Tn903) 
REFERENCES