Silencing of Genes Flanking the P1 Plasmid Centromere

See allHide authors and affiliations

Science  22 Jan 1999:
Vol. 283, Issue 5401, pp. 546-549
DOI: 10.1126/science.283.5401.546


Partition modules stabilize bacterial plasmids and chromosomes by actively promoting their segregation into daughter cells. The partition module of plasmid P1 is typical and consists of a centromere site,parS, and genes that encode proteins ParA and ParB. We show that ParB can silence genes flanking parS (to which ParB binds), apparently by polymerizing along the DNA from a nucleation site at parS. Wild-type ParB contacts an extensive region of P1 DNA; silencing-defective ParB proteins, which were found to be partition-defective, are less able to spread. Hence, the silenced structure appears to function in partitioning.

In eukaryotes, transcriptional silencing of DNA regions, which range in size from short segments to entire chromosomes, is an essential feature of development (1). In fission yeast and in Drosophila, extensive regions of silencing that spread from centromeric sites appear to be necessary for full centromere functioning (2). We consider the possibility that silencing may have a role in the prokaryotic equivalent of mitosis—the process of partitioning plasmids and chromosomes (ensuring their orderly segregation to daughter cells). Knowledge of partitioning has been mainly derived from studies of plasmid-encoded partitioning genes and their chromosomal homologs, which have been recognized to be present in diverse bacteria (3–5). Although the few reports of silencing in bacteria are generally of silencing over short regions, studies have shown that genes several kilobases distant from the centromere of the bacterial plasmid F can be silenced by high levels of the F partition protein, SopB (6, 7). Here, we characterize silencing by comparable elements of the P1 plasmid, provide evidence concerning its mechanism, and assess its relevance for partitioning.

P1 partitioning requires two P1 proteins (ParA and ParB) and a DNA site (parS) (8). ParB binds to parS(9, 10); ParA is an adenosine triphosphatase (11). These factors, together with unknown host components, ensure plasmid stability. We previously observed that, dependent on the location of parS, ParB can either destabilize aparS-bearing plasmid (12) or prevent such a plasmid from conferring antibiotic resistance (13), perhaps by gene silencing in each case. To further study silencing, we insertedparS and reporter genes lacZ and catat the λ attachment site (attλ) ofEscherichia coli, close to the biotin biosynthesis genes (bioA, B, F, C, andD) (Fig. 1A). We made constructs with parS in opposite orientations (I and II) and a construct without parS (0). ParB was supplied from an inducible plasmid source.

Figure 1

(A) Map of region flanking parS in tester strains (28). Within the parS sequence (shown oriented as in construct II), heptameric “A” and hexameric “B” sites to which ParB binds (10) are boxed; the binding site of IHF is underlined. (B) Relation between ParB concentration and expression of lacZ, cat, and phoA in construct I. Specific activities of chloramphenicol acetyltransferase (CAT), β-galactosidase (β-Gal), and constitutive alkaline phosphatase (Pho) were determined by spectrophotometric methods (27) in cultures of strains carrying an inducible source of ParB, pOAR32, or the corresponding vector (29). Results of typical experiments are shown; remaining specific activity is relative to specific activity in “no ParB” (vector) controls. The ParB concentration in construct 0 (no parS) carrying a P1 marked with a kanamycin-resistance determinant (30) is included for comparison (represented as the P1 bar). ParB that was supplied at concentrations sufficient to reduce lacZ expression by 90% in construct I caused no reduction in construct 0 (499 ± 5 compared to 517 ± 18 Miller units in “no ParB” controls). (C) Effect of an ihfA mutation on the silencing of lacZ and cat in constructs I and II. ParB was supplied to constructs I and II (or to their ΔihfA::Tn10 derivatives) from pBR327Ptrp-parB (pMLO102) or from the corresponding plasmid bearing the Δ263-333 variant (12). Bacteria were grown in the presence of 3-β-indoleacrylic acid (100 μg/ml) as inducer (31) for about six generations. Results are the average of three determinations.

Expression of the cat, lacZ, and bio genes was markedly decreased by the presence of wild-type ParB, in contrast to their expression in controls that lacked parS or in which ParB was binding-defective. With increasing ParB, the expression oflacZ diminished rapidly (Fig. 1B). ParB was required in larger amounts to diminish expression of the more distantcat. At no tested concentration did ParB diminish expression of phoA, which is a gene much farther from parS(Fig. 1B). The amount of ParB generated by a wild-type P1 plasmid was sufficient to reduce the expression of lacZ in construct I by ∼50%, which suggests that ParB-mediated gene silencing might have a physiological function. Silencing extended in both directions for several kilobases, but the reduction in gene expression was greatest to the side of parS that, in P1, would face away from the partition genes, which lie upstream (Fig. 1C). A decrease in the efficiency of ParB-mediated silencing was expected to occur in the absence of the architectural protein IHF (integration host factor), because IHF strongly promotes the binding of ParB to parS in vitro and, to a lesser extent, assists in P1 partitioning (9,14). Our expectation was confirmed, although the decrease in silencing due to an ihfA mutation was largely to one side ofparS (Fig. 1C).

Might parS act as a nucleation site for ParB polymerization along the DNA? Although the SopB protein of plasmid F, by itself, was found to be incapable of forming nucleoprotein filaments with linear DNA in vitro (7), it seemed likely that SopB and its homolog (ParB), in view of their abundance, could be primary protein components of such filaments. A demonstration that ParB is specifically associated with the silenced region of the DNA was achieved by formaldehyde-mediated DNA-protein cross-linking followed by immunoprecipitation and polymerase chain reaction (PCR) analysis of the DNA that was released by heating the precipitate (Fig. 2A). Mutant ParB (Δ263-333) that was unable to bind to parS did not associate with DNA in any of the tested regions.

Figure 2

(left).Identification of immunoprecipitated DNA that was cross-linked with formaldehyde in vivo to ParB (construct I) (A) Association of DNA with wild-type (WT) ParB. Cells of the tester strain were fixed with formaldehyde, lysed, and sonicated to reduce the average size of the DNA to 0.5 to 1 kb; the DNA that was associated with ParB was immunoprecipitated as described (5), except that the lysis buffer contained ribonuclease. Cells containing binding-defective ParBΔ263-333 served as a negative control. DNA fragments released from immunoprecipitates and from whole cell extracts (input) were identified by PCR amplification with primer pairs from the indicated (numbered) locations along the DNA (32). The templates for primer pairs “1,” “6,” and “7” were located 50 kb to the left of parS in the sdhA gene ofE. coli, 50 kb to the right of parS in an open reading frame of unknown function, and at the chromosomal antipode in the E. coli aroC gene, respectively. PCR products were electrophoresed through a 3.5% agarose gel, stained with ethidium bromide, and photographed, and these images were processed. Lane numbers correspond to those of the relevant primer pairs; M, marker lanes. (B) Association of DNA with ParB of silencing-defective ParB mutants. Experimental conditions were as in (A). Results obtained with mutants Q88L, I94T, T145P, Q148R, V201Q, E204G, E204K, and D250V (33) were indistinguishable and are represented in the panel labeled “Typical mutant.”

Cross-linking experiments were also performed with nine ParB mutants that had been isolated as being unable to destabilize the ParB-sensitive plasmid pMLO6 and that had been selected as being capable of binding parS in vitro (12). These mutants were totally silencing-defective (15), and in addition, each of them had been previously found to be partitioning-defective. Upon in vivo treatment with formaldehyde, each of the mutant proteins showed the expected, if somewhat attenuated, cross-linking to parS (Fig. 2B). No mutant exhibited DNA binding outside parS, with the exception of Ile105 → Thr105 (I105T). These results suggest that ParB polymerization beyond parS is necessary, although not sufficient, for the ParB-mediated silencing. Because the mutants are partitioning-defective, ParB polymerization could also be important in the partitioning process.

In principle, the growth of a filament nucleated at parS and spreading outward might be blocked by a protein that is tightly bound to a DNA locus in its path. We investigated whether a “roadblock” interposed between parS and a reporter gene would alleviate silencing of the reporter. A set of tandem sites to which the P1-encoded replication initiator protein (RepA) can bind tightly was tested for this capacity. How growth of the nucleoprotein filament might be stopped by RepA protein, acting as a repressor, is shown schematically (Fig. 3A). Strains that are similar to those in Fig. 1were used; however, the promoter-operator region of the P1repA gene (controlling lacZ) included all five RepA binding sites rather than just one. RepA was supplied constitutively. The bound repressor completely alleviated silencing of the cat gene, which is distal to the roadblock and is 4 kilobases (kb) away, but did not prevent ParB from silencing thebio locus, as judged by continued auxotrophy for biotin (Fig. 3A). A specific association of ParB with the silenced regions flanking parS (but only a minimal association with DNA distal to a roadblock, including DNA within lacZ) was shown by the formaldehyde cross-linking technique (Fig. 3B). Controls in which an antibody to RepA was used showed that RepA could be cross-linked to DNA that included its operator sites and to no DNA of other regions tested. The largely unilateral alleviation of silencing in an ihfA mutant (Fig. 1C) might be interpreted as a consequence of a facilitation of ParB binding to parS while blocking propagation of silencing from the side of parS that contains most of the ParB-binding boxes (Fig. 1A) and that is essential and sufficient for ParB binding and partitioning (16).

Figure 3

(right). (A) A protein-DNA complex acting as a roadblock to the spread of silencing. In the schematic diagram, circles represent ParB, and triangles represent repressor (RepA) that is tightly bound to sites at the operator (incC) of the PrepA (34) from whichlacZ is transcribed in the tester strain. Bacterial strains were identical to constructs I and II (Fig. 1) except that PrepA with a single ParB binding site was replaced by the entire P1 incC region encompassing PrepA and five tandem sites to which RepA binds tightly. ParB was supplied from pOAR32 (27) containing parB under Ptac control, and RepA was supplied from pALA162 (35) containing repA under bla-P2 control. Vectors without functional parB and repAwere used to make the strains being compared otherwise isogenic. The presence of kanamycin (25 μg/ml) and ampicillin (100 μg/ml) ensured plasmid retention. Growth with IPTG (1 mM) for about seven generations served to induce ParB synthesis and dilute protein that was accumulated before silencing. Two or three independent transformants of each kind were assayed for the indicated gene functions in triplicate. CAT protein relative to total protein was measured by an enzyme-linked immunosorbent assay (Boehringer-Mannheim). Beta-galactosidase was measured as previously described (Fig. 1B). Dependence of colonial growth in minimal-glucose medium on added biotin (seen only when ParB was supplied to construct II) was scored as an absence of prototrophy (“No”). (B) The protein-DNA complex acting as a roadblock to the spread of ParB along the DNA. Bacterial constructs were as in (A), except that only strains with parS in orientation I were used. DNA associated with ParB or with RepA was identified by PCR (as in Fig. 2) after immunoprecipitation with the indicated antibodies. (M, marker lanes in gels.)

If an association of ParB with DNA outside of parS is to have a role in partitioning, then such an association should be demonstrable in formaldehyde-treated cells carrying a P1 plasmid. Our study showed that ParB could be cross-linked to P1 DNA over at least 11 kb, of which 8 kb are downstream of the par operon (Fig. 4). Possibly, most of the 7000 ParB dimers reported to be present in a bacterium carrying P1 (17) are bound with DNA. The notion that a genetically silenced nucleoprotein filament extending beyond the limits of a plasmid centromere might function in partitioning is supported by two additional considerations. First, the capacity for gene silencing is common to the centromere-binding proteins of P1 and F plasmids (6) and their nonhomologous analog (ParR) of plasmid R1 (18). Second, the capacities of ParB for gene silencing and for partitioning are altered by the same mutations. Thus, prokaryotic centromere function may depend on a capacity to seed a nucleoprotein filament. The filament (shown in Fig. 3A as a linear structure for simplicity) is more likely to be compacted in a solenoid in which the DNA wraps through multiple turns about a protein core, as was suggested by the observed changes in linking numbers upon formation of a SopB-DNA complex (19).

Figure 4

In vivo association of ParB and RepA with P1 DNA. Electrophoresed PCR products that identify the DNA that was specifically cross-linked in vivo to ParB and RepA of a P1 plasmid are shown. Cells of the P1-carrying strain (Fig. 1B) were treated as described (Fig. 2), with antibodies to ParB and antibodies to RepA for immunoprecipitations. The locations of primer pairs along the DNA of P1 in the regions flanking parS are shown with respect to a physical map of known and putative genes (arrows) and the binding sites (solid rectangles) of plasmid proteins ParB and RepA (36). The leftmost primer pair was “1” of Fig. 2A. The unlabeled genes between repL and dam are of unknown function. Genes between parS and repL are under the control of a late promoter. The regions that are cross-linked to RepA correspond to two sets of RepA binding sites: incC, within the P1 plasmid origin of replication, and incA, a separate replication-control locus. Samples taken for PCR amplification of input DNA, of DNA immunoprecipitated with antibodies to ParB, and of DNA immunoprecipitated with antibodies to RepA were diluted 1:1000, 1:250, and 1:25, respectively.

Prokaryotic centromeres are characterized by the presence of arrays of sites to which one of the partition proteins bind. There are 6 sites in P1 (Fig. 1A), 12 in F, 10 in R1, and 12 in pTAR (20). Whereas these plasmid-borne binding sites are tightly clustered, the several chromosomal binding sites for the Bacillus subtilispartition (and sporulation) protein Spo0J are distributed over a region spanning many kilobases (3). The centromere may serve as a handle that is used to tether or orient a large structure, with its several binding sites facilitating a steady grip or the formation of an intermediate that is appropriately paired for partitioning. Evidence for paired intermediates in the partitioning of plasmid R1 and involving its cognate ParB analog has recently been obtained (21). Like the proteins of heterochromatin that spread from centromeres or telomeres of eukaryotic chromosomes, the ParB that spreads from the P1 plasmid centromere can silence genes. In each case, this capacity for gene silencing may be incidental to a primary structural role that is associated with DNA segregation or movement.

  • To whom correspondence should be addressed. E-mail: myarmo{at}

  • * Present address: Department of Microbial Biochemistry, Institute of Biochemistry and Biophysics, Polish Academy of Sciences, ul. Pawińskiego 5A, 02-106, Warsaw, Poland.


View Abstract

Navigate This Article