ABSTRACT
The promiscuous plasmid pAM
[beta]
1 from Gram-positive bacteria encodes a resolution system which differs from that of
Tn
3
in that (i) it requires a histone-like protein and an unusual resolvase-DNA interaction to promote recombination and (ii) it mediates
in vivo
DNA inversion in plasmid substrates. In this
in vivo
analysis, the pAM
[beta]
1 resolution site is narrowed down to a 99 bp segment, the strand exchange is
mapped within 10 bp and the serine residue at position 10 of the resolvase is
shown to be essential for enzyme activity. In addition, data showing that the
resolution system does not promote DNA inversion in the
Bacillus subtilis
chromosome are presented. Implications of this observation are discussed.
Site-specific recombinases are well-characterized enzymes classified into two families (see
1
for review). Recombinases of the Integrase family promote integration,
resolution and inversion on supercoiled and linear substrates, use a tyrosine
residue located close to the C-terminus to cleave and join DNA and display a weak homology at the C-terminal region. Recombinases of the Resolvase-Invertase (Res-Inv) family are specific for either resolution
(resolvases) or inversion (invertases), are highly conserved ( >=30% identity) and recombine supercoiled substrates through a cut and seal
reaction mediated by a serine residue located at the N-terminus of the protein.
Recently, the characterization of three highly (>90%) related, functionally
interchangeable site-specific recombinases originating from Gram-positive bacteria has been initiated (
2
-
5
). These enzymes designated Res[beta], [beta] and ResIP are encoded by plasmids pAM[beta]1, pSM19035 and pIP501, respectively. They are thought to be
related to the resolvases of the Res-Inv family since they are ~30% identical to these proteins, they contain the putative catalytic
serine, they carry out resolution and they require supercoiled DNA to mediate
resolution (
2
-
5
). However, these enzymes differ from this class of recombinases in three
respects. First, their interaction with the resolution site (
res
) is atypical. Usually, resolvases of the Res-Inv family bind three regions within
res
, each containing two repeats in opposite orientation (
6
,
7
). On the contrary, Res[beta] and [beta] interact with only two regions of
res
and one of them is thought to contain a single repeat (
5
,
8
,
9
). Second, Res[beta] and [beta] require a host-encoded factor to promote resolution
in vivo
and
in vitro
while resolvases of the Res-Inv family do not (
1
,
5
-
8
,
10
). This cofactor, the histone-like protein Hbsu of
Bacillus subtilis
, may interact with the resolvase-DNA complex to facilitate the joining of distant recombination sites (
8
,
10
,
11
). Third, Res[beta] and [beta] allow efficient DNA inversion in plasmid substrates
in vivo
while Res-Inv enzymes do not (
1
,
6
,
7
,
10
,
12
,
13
). This activity, also detected
in vitro
with [beta] (
5
,
10
) but not with Res[beta] (
8
), is 2-4-fold lower than the resolution activity (
10
). In this report, a further
in vivo
characterization of the pAM[beta]1 resolution system is presented. The
res
site was delimited, the strand exchange was located within a 10 bp segment and
the serine residue at position 10 was shown to be essential for enzyme
activity. Additionally, it is reported that Res[beta] does not promote DNA inversion in the
B.subtilis
chromosome.
The bacterial strains used were (i)
Escherichia coli
DH5[alpha], JM105, TG1 and a derivative of JJC40 overexpressing the LacI repressor
from a pACYC184-related plasmid (
14
,
15
) and (ii)
B.subtilis
HVS495 and its isogenic RecA- strain HVS567 (
16
). Induction of competence, transformation and growth conditions were as
previously described (
14
,
17
). The plasmids used in this report and described elsewhere are: pUC9
res
+, pMTL500E (
2
), pHV1461, pHV1461* (
4
), pIL253, pDH32 (
18
), pHV1436 (
19
) and pDG148 (
20
). To localize the resolution site, a
B.subtilis
-
E.coli
shuttle vector developed by N. Minton and T.-J. Swinfield (unpublished data) designated pMTL511E was used. This plasmid
is a hybrid between pMTL20E, a pUC-like vector containing an erythromycin (Em) determinant, and the
replication region of pAM[beta]1 extending from the
Hpa
I to the
Acc
I sites (coordinates 2278-4866;
21
). The replication region was inserted at the
Nhe
I site of pMTL20E in such a way that the pAM[beta]1
RepE
gene faces the Em determinant. pMTL511E does not contain a functional
res
site (
8
,
12
) but it still carries the repeat R1 at its native position. The segments tested
for resolution activity are represented in Figure
1
A. Segments #1-3, 8 and 9 were PCR products. The oligonucleotides used to generate the 5' end of these segments were 5'-GGCCGAATTCCTTTTAATTTTCTATC (segment #1), 5'-ATGAATTCTCTATCTTTTATAGGTCATTAG (#2 and
8), 5'-ATGAATTCATAGGTCATTAGAGTATACTTA (#3 and 9). Primers at the 3' ends were 5'-GCCGAATTCAAACCACGTAACC (#1-3), 5'-ATGAATTCTTTTCCTCCTCTAATATGCTCA (#8) and 5'-ACTTAAATGACCTATTCAATA (#9). Segment #4 corresponds to the
Acc
I-
Eco
RI restriction product represented at the top of Figure
1
A. Segments #5-7 extend from the
Ase
I site to various endpoints generated by
Exo
III treatment (
18
). Segments #1-3 and 4-9 were cloned in
E.coli
at the
Eco
RI and
Stu
I sites of the polylinker region of pMTL511E, respectively. They are thus
located ~3 kb downstream of the pAM[beta]1 origin, which prevents any effect of Res-DNA interactions on pAM[beta]1 replication (
12
; L.J., unpublished data). The absence of mutations in the cloned segments was
confirmed by sequence determination.
The plasmids used to test DNA inversion were pHV1436-1, pHV1436-2 and pDH32-1. pHV1436-1 and -2 were constructed by inserting at the
Eco
RV site of pHV1436, a DNA fragment containing an inverted repeat of the
res
site (pHV1436-2) or of
res
plus some flanking sequences (pHV1436-1; see Fig.
2
A). Plasmid pDH32-1 is an integrative vector derived from pDH32 which carries at the
Eco
RI site the same inverted structure as pHV1436-1. It was used to insert the inverted structure in the
amy
locus of the
B.subtilis
chromosome. To localize the site of strand exchange, a segment encompassing the
res
site of pIP501 (corresponding to segment #8 Figure
1
A and generated by PCR using as left and right primers the oligonucleotides 5'-ATGAATTCGCTATCATTTATAGGTCAATAG and 5'-ATGAATTCTTTTACTCCTCC
T
TATTATGCC, respectively) was inserted at the
Sma
I site of pMTL500E in
E.coli
. The resulting plasmid, pMTL500EIP, carries the pAM[beta]1 and pIP501
res
sites in opposite orientation, 556 bp apart. The bold T in the right primer is
a +1 mutation located immediately downstream of the Res[beta]-R3 interaction (Fig.
1
A, bottom). This mutation does not alter significantly the activity of the
res
site (L.J., unpublished data). The integrity of the pIP501
res
site was controlled by sequence determination using the universal primers of
the
Taq
Dye Primer Cycle Sequencing kit from Applied Biosystems which flank the pIP501
res
site in the parental configuration of pMTL500EIP.
Plasmid and total DNA were extracted as previously described (
18
). In recombination assays, the DNA was extracted after 30-40 generations unless stated otherwise. These generations include growth
on tranformation plate and in liquid medium. Fragment purification, gel
electrophoresis, DNA labelling and hybridisation were carried out using
standard procedures (
14
). Restriction and modification enzymes were used as recommended by the
suppliers. PCR reactions were carried out using the
Thermus aquaticus
DNA polymerase (Promega) and a Perkin Elmer-Cetus apparatus. Sequences were determined using the Taq Dye Primer Cycle
Sequencing Kit and the 370A Sequencer from Applied Biosystems. In some
instances, manual sequence determination was carried out using the Sequenase
kit from USB. A PhosphorImager SI apparatus from Molecular Dynamics was used
for quantifications. Oligonucleotides were synthesized either by Eurogentec
(Seraing, Belgium) or with an Oligo 1000M DNA Synthesizer from Beckman
Instruments (Gagny, France).
Res[beta] was purified as described previously (
8
).
Bacillus subtilis
crude extracts were prepared from cultures (10 ml) grown in LB medium
supplemented with kanamycin (5 [mu]g/ml) +- IPTG for 4.5 h. At OD
650
= 1, cells were collected, resuspended in 200 [mu]l of 10% sucrose, 20 mM Tris-HCl, pH 8, 10 mM MgCl
2
and 5 mM DTT and treated with lysozyme (2 mg/ml, 10 min at 4oC and then 10 min at 37oC). They were then frozen and thawed three times by incubation in
liquid nitrogen and water bath at 37oC successively and centrifuged at 18 000
g
for 30 min at 4oC. The supernatants, corresponding to the crude extracts, were finally
collected and stored at 80oC. Protein concentration in the crude extracts was routinely 0.2 mg/ml as
determined with the Bradford Assay kit from Bio-Rad Laboratories. Res[beta] concentration was estimated using a western blot analysis. For
this, crude extracts (7.5 [mu]l) were electrophoresed on SDS-PAGE gel according to standard procedures (
14
) along with various amounts (18, 36 and 78 ng) of purified Res[beta] protein. The proteins were then transferred by electroblotting to a
Millipore PVDF membrane and the membrane was incubated successively with a
blocking agent (milk 1% for 1 h), polyclonal rabbit antibodies directed against
Res[beta] (dilution 1/10
4
, overnight) and protein G-HRP conjugate (dilution 1/5000, 1 h). The antibodies were detected by
chemiluminescence using the ECL kit of Amersham. The Res[beta] concentration was 10 ng/[mu]l in crude extracts prepared from cells harbouring pHV1460 and
pHV1460SA, and 30 ng/[mu]l in the case of pHV1460YF after growth in 250 [mu]M IPTG. In all the constructed plasmids, the spac-I promoter is not fully repressed by the plasmid encoded LacI
protein and the remaining synthesis of active resolvase is sufficient to allow
site-specific recombination. Without induction, cells harbouring pHV1461
contain ~3000 Res[beta] protomers, and the addition of 250 [mu]M IPTG causes an ~100-fold increase in Res[beta] concentration, as estimated by immunological
assay.
Gel shift assays were carried out as described previously (
8
) except that the labelled fragment was segment #8 (Fig.
1
A) and the crude extracts were diluted in 10% sucrose, 20 mM Tris-HCl, pH 7.5, 20 mM MgCl
2
and 1 mM DTT before incubation with the labelled DNA.
In vitro
data showed that Res[beta] interacts with two sequences located ~250 bp upstream of the
res
[beta] gene (see Fig.
1
bottom;
5
,
8
,
9
). These sequences encompass three conserved repeats [R1-R3, 5'-TAGG(T/A)CANNNNAGT] which allow Res[beta] binding and are organized as displayed in Figure
1
A (
8
). To delimit the resolution site
in vivo
, various segments cloned into the
res
-vector pMTL511E were tested for resolution activity in
B.subtilis
cells. In the presence of pHV1461, a compatible plasmid encoding Res[beta] under the control of the IPTG inducible
spac
-I promoter, hybrids carrying inserts #1-3, 5, 6 and 8 did not accumulate multimers while those containing
inserts #4, 7 and 9, did (Fig.
1
B). In contrast, in the presence of pHV1461* encoding a truncated, inactive form
of Res[beta] (Res[beta]*;
2
,
4
), all the constructs accumulated large amounts of plasmid multimers (not
shown). These results, summarized in Figure
1
A, show that the
res
site is included within a 99 bp segment extending from repeat R1 to a position
located 21 bp downstream of R3 (Fig.
1
A, bottom, coordinates 4852-4951 according to
21
). Interestingly, a segment delimited by the external ends of the distal repeats
(segment #9) was not proficient for resolution indicating that a short sequence
located downstream of R3 plays an important role in DNA resolution. This
sequence might be 5'-ATATTA located 6 bp downstream of the 5' end of R3 (filled triangles Fig.
1
A bottom), as it is strongly protected against attack with hydroxyl radicals by
the related [beta] resolvase (
9
). Since this region does not share any significant homology with repeats R1-R3, we propose that the pAM[beta]1 resolution site carries, in addition to repeats R1-R3, a fourth highly divergent sequence, located downstream
of R3, with which Res[beta] interacts. Furthermore, the
res
site does not include all the nucleotides protected against DNase I cleavage
(the open box Fig.
1
A, bottom;
8
), as the stretch of four Ts located at the left boundary of the Res[beta]-R1-R2 contact is not required for resolution. This suggests
that the protein-DNA interactions at these positions are not specific.
We previously observed inverted structures in high (~100) copy number pAM[beta]1 derivatives propagated in a Rec+
B.subtilis
strain producing Res[beta] (L.J., unpublished data). However, inversion was not detected
in vitro
(
8
). To shed some light on these apparently conflicting results, we first searched
for DNA inversion in a low (~10) copy number plasmid unrelated to pAM[beta]1. The plasmid used, pHV1436-1, contains the [theta] replicon of the
B.staerotermophilus
plasmid pTB19 (
19
) and an inverted repeat including
res
and some pAM[beta]1 flanking sequences (see Fig.
2
A). This plasmid was propagated in a
recA
mutant host (HVS567) to avoid a possible involvement of the RecA protein in
inversion, in the presence of pHV1461 or pHV1461* encoding a functional or an
inactive form of Res[beta], respectively. After ~30 generations without IPTG, inverted (I) and parental (P) structures
were observed in the presence of Res[beta], while only the input parental plasmid was detected with Res[beta]*. A quantification of the signals detected in the presence of Res[beta] revealed that the parental and inverted structures were
present in equal amounts. Similar results were obtained with pHV1436-2, a plasmid related to pHV1436-1, but carrying as inverted repeat the
res
site only (segment #8 in Fig
1
A and Fig
2
A). This indicates that DNA inversion can occur in low copy number plasmids
unrelated to pAM[beta]1 and in the absence of the RecA pathway of homologous recombination.
To probe DNA inversion further, the capacity of Res[beta] to invert sequences in the
B.subtilis
chromosome was investigated. For this, the inverted repeat contained in pHV1436-1 was inserted in the
amy
locus of a Rec+ strain harbouring either pHV1461* (Res[beta]*) or pHV1461 (Res[beta]). Several transformants obtained with each strain were grown in the
presence or the absence of 500 [mu]M IPTG. After growth, the total DNA was extracted, cleaved with two
different couples of enzymes and analysed by Southern hybridization. In no case
were inverted structures detected (see as example, Fig.
2
B). The integrity of the inserted structure was shown by Southern blot analysis
and by its capacity to undergo DNA inversion when cloned back into a pTB19- derived plasmid. Furthermore, plasmids pHV1461* and pHV1461 contained in
the transformants did not undergo any DNA rearrangement, as judged from
restriction analysis and from the ability of pHV1461 to maintain pMTL511E
derivatives carrying the
res
site into the monomeric state (not shown). Finally, Res[beta] was observed to efficiently carry out DNA resolution in the
amy
locus (to be published elsewhere). Taken together, these data show that Res[beta] is inefficient in recombining inverted structures in the
B.subtilis
chromosome, even when the resolvase is overexpressed. Res[beta] DNA inversion activity is therefore a plasmid-dependent property.
As shown above, Res[beta] mediates inversion in plasmid substrates. It can also act on the pIP501
res
site, which is slightly different from the
res
site of pAM[beta]1 (see Fig.
1
A, bottom;
4
). We took advantage of these properties to map the site of strand exchange
occurring during recombination. For this, we propagated the high copy number
plasmid pMTL500EIP, which carries the pAM[beta]1 and pIP501
res
sites in opposite direction, in a Rec+ strain encoding Res[beta] from pHV1461. As expected, an equilibrium between the parental and
inverted plasmid forms was observed suggesting that most of the
res
sites have undergone at least one recombination event. The two plasmid forms
were then purified in
E.coli
and the resolution sites were sequenced. In the inverted configuration,
sequence determination revealed that the recombination reaction generated pAM[beta]1-pIP501 hybrid sites with a switch occuring within a 10 bp segment
overlaping repeats R1 and R2 (the grey box, Fig.
1
A). On the contrary, the
res
sites were unchanged in the parental structure. This shows that the strand
exchange directed by Res[beta] occurs within the R1-R2 region.
Res[beta] is highly related to site-specific recombinases of the Res-Inv family and carries the putative catalytic serine at
position 10 (i.e. the serine closest to the N-terminal end of Res[beta]; sequence numbering is as proposed in
22
). However, we also noticed that the C-terminal region of Res[beta] shares some similarity with recombinases of the Int family
including the tyrosine residue which is involved in the cut-and-paste reaction (
12
). In Res[beta], this amino acid is located at position 178. In order to determine which
of the two residues is involved in DNA recombination, these amino acids were
mutagenized as described in Materials and Methods. In the SA10 mutated protein,
the serine at position 10 was replaced by alanine. In the YF178 mutant, a
phenylalanine was substituted to the 178 tyrosine. These two proteins were
encoded by plasmids identical to pHV1460 except for changes introduced at the
mutated codon. The plasmids were designated pHV1460SA and pHV1460YF,
respectively. As shown in Figures
1
B and
2
C, the YF178 protein is efficient in resolution and inversion, while SA10 is
unable to recombine DNA. The deficiency in recombination of the SA10 protein
was observed even when the protein was overproduced or when culture was
prolonged for >100 generations (Figs
1
B and
2
C). To test whether this deficiency results from a strong alteration of the DNA
binding activity of the SA10 protein, gel shift experiments were carried out.
Results presented in Figure
3
show that SA10 binds to the
res
site in a similar way as Res[beta] and YF178. The high DNA binding activity of YF178 is somehow surprising
since the mutated residue maps whithin the helix-turn-helix motif of Res[beta] (located within residues 159-180; E. Le Chatelier, personal communication).
However, similar data were obtained with some mutants of the Tn
21
resolvase (
23
). Taken together, these observations indicate that the catalytic residue for
Res[beta]-mediated recombination is the serine located at position 10. The
role of the Tyr178 residue, if any, is not known at present.
In this report we show that (i) the pAM[beta]1
res
site is included within a 99 bp segment located ~250 bp upstream of the
res
[beta] gene (Fig.
1
A), (ii) the
res
site contains, in addition to repeats R1, R2 and R3, a fourth highly divergent
sequence located 3' of R3 with which Res[beta] interacts specifically, (iii) the strand exchange is located
within a 10 bp segment which overlaps the spacing region between repeats R1 and
R2 and that (iv) the serine residue at position 10 is essential for Res[beta] recombination activity. We also report that DNA inversion occurs in the
presence of Res[beta] when the
res
sites are carried by a plasmid but not when they are contained in the
B.subtilis
chromosome. Previously published data showed that Res[beta] does not carry out inversion
in vitro
in supercoiled plasmid substrates (
8
) while the related recombinase [beta] does (
5
,
10
). The site-specific recombination system of the Res[beta] family appears, therefore, to be complex and probably depends on as
yet unidentified parameters.
The Ser10 requirement reported here and results presented elsewhere (
2
-
5
) suggest that Res[beta] and related DNA recombinases belong to the group of resolvases of the Res-Inv family. They are thus not expected to carry out DNA inversion (
1
,
6
,
7
). Two general ways to bypass the reaction specificity of enzymes of the Res-Inv family were described. The simplest, and possibly the most likely,
postulates the existence of particular substrates in the assay. One of these
substrates might be knotted molecules since knotting makes plasmids which carry
inverted sites efficient substrates for Tn
3
resolvase
in vitro
and possibly
in vivo
(
24
,
25
). Another substrate might be head-to-head dimers formed by recombination between the left and right
res
sites of two independent molecules. These events might be mediated by a RecA-independent homologous recombination pathway (
26
,
27
) or by Res[beta]. However, the latter possibility is unlikely, as neither Res[beta] nor [beta] were observed to carry out DNA fusion (
5
,
10
; L. J. and S. McG., unpublished data).
The second way to bypass specificity involves amino acid substitutions in the
recombinase. Indeed, it was reported that certain single mutations in several
enzymes of the Res-Inv family (the Tn
3
resolvase and the Hin, Gin and Cin invertases;
28
-
30
; M. R. Boocock, personal communication) lead to relaxation of reaction
specificity. In this line of thinking, it may be envisioned that Res[beta]-related enzymes originate from an ancestral, unrelaxed recombinase,
which has undergone changes altering its specificity. However, this hypothesis is unlikely as enzymes carrying such mutations carry out resolution, inversion
and integration in supercoiled and linear molecules, while Res[beta]-related proteins mediate inversion in some but not all substrates (
5
,
8
,
10
; this work), do not carry out integration (
5
,
10
; L. J. and S. McG., unpublished results), and require supercoiling (
5
).
Another possibility would be that Res[beta]-related recombinases are endowed with authentic inversion activity,
but not with integration activity, as a direct consequence of their requirement
for HBsu and/or their atypical interaction with the
res
site (
5
,
8
-
11
). The fact that Res[beta]-mediated inversion does not occur in all instances suggests a strict
requirement for a specific state of the substrate. A good candidate is
supercoiling, since it influences the recombination activity of enzymes of the
Res-Inv family (
31
-
34
). According to this hypothesis, the supercoiling of the
B.subtilis
chromosome and of the substrate used to test Res[beta]- mediated inversion
in vitro
, might not be appropriate. Alternatively, the requirement for supercoiling may
be more stringent for Res[beta] than [beta], as observed for the related pair of resolvases, Tn
3
and [gamma][delta], respectively (
34
).
Clearly, additional investigations are required to fully understand the
mechanism of generation of inverted structures by resolvases of the Res[beta] family. The characterization of this phenomenon may be relevant for two
other resolvases of the Res-Inv family which also efficiently mediate DNA inversion. These are the
recombinase of plasmid R46 which is functionally interchangeable with the Tn
3
resolvase (
35
), and (ii) the BinR (BinL) protein of Tn
552
which is phylogenetically the closest resolvase to the invertases of the Res-Inv family (
5
,
36
,
37
).
We are grateful to Nigel Minton and Tracy-Jane Swinfield for providing pMTL511E, to Vladimir Bidnenko for making us aware of head-to-head dimer existence and to Costa Agnastopoulos for critical reading of the manuscript. This work was
supported by the EC BIO2-CT91-0268 grant.
REFERENCES
Return