The parA locus of plasmid R1 encodes a prokaryotic centromere‐like system that mediates genetic stabilization of plasmids by an unknown mechanism. The locus codes for two proteins, ParM and ParR, and a centromere‐like DNA region (parC) to which the ParR protein binds. We showed recently that ParR mediates specific pairing of parC‐containing DNA molecules in vitro. To obtain further insight into the mechanism of plasmid stabilization, we examined the intracellular localization of the components of the parA system. We found that ParM forms discrete foci that localize to specific cellular regions in a simple, yet dynamic pattern. In newborn cells, ParM foci were present close to both cell poles. Concomitant with cell growth, new foci formed at mid‐cell. A point mutation that abolished the ATPase activity of ParM simultaneously prevented cellular localization and plasmid partitioning. A parA‐containing plasmid localized to similar sites, i.e. close to the poles and at mid‐cell, thus indicating that the plasmid co‐localizes with ParM. Double labelling of single cells showed that plasmid DNA and ParM indeed co‐localize. Thus, our data indicate that parA is a true partitioning system that mediates pairing of plasmids at mid‐cell and subsequently moves them to the cell poles before cell division.
Accurate distribution of genetic material to progeny cells at cell division is essential for all organisms. Despite the central role of DNA segregation in the bacterial cell cycle, the molecular mechanism of the process is not well understood. Model plasmids, such as R1, F and P1, contain several different types of systems that prevent plasmid loss at cell division (reviewed by Nordström and Austin, 1989; Hiraga, 1992). One such type encodes a differentiation programme that leads to killing of plasmid‐free cells and thereby confers plasmid stabilization (reviewed by Jensen and Gerdes, 1995; Gerdes et al., 1997). In contrast, true partitioning systems stabilize their replicons by actively distributing the plasmid molecules. In molecular terms, plasmid‐encoded par genes are the best characterized determinants of DNA segregation in bacteria. Bacterial chromosomes contain genes that are homologous to the par genes from plasmids. In Bacillus subtilis (soj and spo0J) and Caulobacter crescentus (parA and parB), such genes are known to be involved in chromosome segregation (Ireton et al., 1994; Sharpe and Errington, 1996; Mohl and Gober, 1997). The recent molecular genetic and cytological analyses of the DNA segregation process in prokaryotes has demonstrated that the par genes specify centromere‐like systems that pair and separate DNA molecules in an ordered sequence (Harry, 1997; Wheeler and Shapiro, 1997; Jensen et al., 1998; Sharpe and Errington, 1999).
In general, plasmid partitioning systems are composed of a cis‐acting centromere‐like site and two genes that encode proteins required for the partitioning process. In the parA system of plasmid R1, the cis‐acting parC site is located upstream of the genes encoding ParM (motor protein) and ParR (repressor) (see Figure 1; Gerdes and Molin, 1986; Dam and Gerdes, 1994). The 160 bp parC site contains the parA promoter flanked by two sets of five direct repeats (iterons) to which the ParR protein binds (Jensen et al., 1994, 1998; Breüner et al., 1996). The presence of all 10 iterons is required for plasmid partitioning and for regulation of the parA promoter (Dam and Gerdes, 1994; Jensen et al., 1994; Breüner et al., 1996). The ParM protein has ATPase activity that is stimulated by ParR when ParR is bound to parC (Jensen and Gerdes, 1997). Single amino acid changes in ParM that inactivate the ATPase activity also abolish in vivo partitioning. Recently, we showed that the parA system mediates efficient pairing of plasmid molecules in vitro (Jensen et al., 1998). Pairing required the presence of parC bound by ParR. The pairing reaction was stimulated further by supercoiling and by ParM in the presence of ATP. This suggested that replicon pairing is an essential step in the partitioning process (Jensen et al., 1998). Thus post‐replicational pairing of DNA molecules occurs in prokaryotes as well as in eukaryotes.
The second family of prokaryotic partitioning systems includes par of P1 and sop of F (Ogura and Hiraga, 1983; Abeles et al., 1985). In the P1 par and F sop systems, the centromere‐like sites parS and sopC are located downstream of the genes encoding the partitioning proteins (Ogura and Hiraga, 1983; Abeles et al., 1985). ATPase activity and DNA‐binding activities have also been described for the partitioning proteins of P1 and F (Davis and Austin, 1988; Mori et al., 1989; Davis et al., 1992; Watanabe et al., 1992). The P1 and F plasmids and the SopB partitioning protein are positioned at mid‐cell or at 1/4 and 3/4 positions (Gordon et al., 1997; Niki and Hiraga, 1997; Kim and Wang, 1998).
The parA system of plasmid R1 shows no sequence similarity at the protein or nucleic acid levels to the systems of the par/sop family. Elucidation of the molecular mechanism of plasmid partitioning by parA may therefore provide new insights into the process of DNA segregation. Here we examine the intracellular localization of the ParM and ParR proteins using fusions to green fluorescent protein (GFP) and immunofluorescence microscopy (IFM). We also use a GFP–LacI fusion protein to visualize the intracellular position of a plasmid partitioned by the parA system. Fluorescence and phase‐contrast microscopy reveal that the ParM protein and the parA‐containing plasmid co‐localize at specific sites near the cell poles or at mid‐cell. These results demonstrate that parA of R1 specifies a true partitioning system and suggest a process that involves pairing of newly replicated plasmids at mid‐cell followed by separation, and active movement to the poles of the pre‐divisional cell.
Properties of ParM–GFP fusion proteins
To study the subcellular localization of ParM, we constructed gene fusions to a bright variant (Mut1) of GFP from Aequorea victoria (Cormack et al., 1996). Full‐length ParM was fused to both the N‐ (ParM–GFP) and the C‐terminus of GFP (GFP–ParM) (Figure 2A). Protein fusions were also made to truncated or mutated ParM. The GFP fusion proteins were expressed from the lac promoter, rendering expression inducible by isopropyl‐β‐d‐thiogalactopyranoside (IPTG). Using Western analysis, we confirmed that all fusion proteins were expressed and had the expected sizes (data not shown). With the induction conditions used here (see Materials and methods), the amount of ParM–GFP fusion protein in the cell was close to the amount of ParM expressed from a wild‐type parA system resident in R1 (1.5‐ to 2.5‐fold overproduction).
We then tested the functionality of the fusion proteins. The mini‐R1 test plasmid pDD1509K, that contains a parA system with a deletion in parM, was not stabilized when the fusion proteins were expressed in trans from a co‐resident plasmid. However, expression of ParM–GFP destabilized the parA+ mini‐R1 test‐plasmid pAB1503 500‐fold, corresponding to active destabilization. Thus, ParM–GFP exhibits trans‐dominance since it overrides wild‐type ParM.
Visualization of the subcellular localization of ParM using GFP fusion proteins
The subcellular localization of fusions between ParM and GFP was examined using combined fluorescence and phase‐contrast microscopy. Figure 3A shows cells expressing the ParM–GFP protein. No other components of the R1 parA system were present. The majority of the cells (80–90%) showed localization of ParM–GFP to specific positions in the cell. Three different types of cells were observed (Figure 3B). Small cells contained two GFP foci specifically located to regions close to but not at the cell poles (Figure 3A, a–c). Larger cells contained three GFP foci; two foci were located close to the cell poles, and the third focus was located at mid‐cell (Figure 3A, d–f). Pre‐divisional cells contained four GFP foci, two of which were located close to the cell poles and two close to mid‐cell (Figure 3A, g–j). In some cells, the two mid‐cell foci were located very close to each other (Figure 3A, g and h) and in others the mid‐cell foci were more separated (Figure 3A, i and j). The ParM–GFP foci were found to be located symmetrically in the vast majority of the cells (all of the 340 cells examined). The positions of the ParM–GFP foci relative to the cell poles were measured for cells of all types and different lengths. The polar foci were found to be located at a fixed distance from the cell pole that was independent of the cell length. The distribution of the ParM–GFP foci as a function of cell length is shown in Figure 3C.
To examine whether cell division was required for the specific subcellular localization of ParM, we treated the cells with cephalexin. This antibiotic inhibits FtsI and causes the cells to form filaments that are depleted of FtsZ ring structures (Pogliano et al., 1997). Figure 3A, k, shows a typical ParM–GFP‐expressing cell after treatment with cephalexin. Fluorescent foci were seen spaced fairly evenly along the filament, and some of them appeared to be in the middle of the duplication process.
In cells that expressed the GFP–ParM fusion protein, specifically located foci were only observed in a minor fraction of the cells (10–20%) and the foci were less clear. When the GFP–ParM foci were visible, the subcellular localization was identical to that of the ParM–GFP fusion protein. Cells that expressed GFP itself or fusion proteins containing truncated ParM (Figure 2A) showed a uniform distribution of fluorescence (not shown). We also introduced the D170E mutation, which severely reduces the ATPase activity of ParM (Jensen and Gerdes, 1997), into the ParM–GFP fusion protein. ParM D170E–GFP did not localize (Figure 3A, l). Thus, ATPase activity seems to be required for localization of ParM.
Localization of ParM using IFM
To confirm that the localization of the ParM–GFP fusion protein observed here reflects the intracellular localization of the native ParM protein, we performed IFM using antibodies against ParM (Figure 3D). The ParM protein was expressed from a wild‐type parA system present in the low copy number plasmid pRBJ460. IFM revealed a pattern of ParM localization that was clearly reminiscent of that observed using the ParM–GFP fusion protein (compare Figure 3D and A, a–j). Thus, it seems that ParM–GFP localizes to the same intracellular sites as native ParM, even though the GFP fusion protein is not fully functional.
Properties of ParR–GFP fusion proteins
We also constructed gene fusions between the second partitioning protein ParR and GFP (Figure 2B). The fusion to both the N‐ (ParR–GFP) and the C‐terminus of GFP (GFP–ParR) exhibited biological activity (measured as repression of the parA promoter). Both types of fusion proteins showed a uniform distribution (Figure 3A, m). The presence of a parA‐containing plasmid or expression of ParM did not lead to formation of specifically located fluorescent ParR–GFP foci (data not shown).
Visualization of the subcellular localization of plasmids using the GFP–LacI assay
We then examined the intracellular localization of plasmids using the GFP–LacI system (Robinett et al., 1996; Straight et al., 1996). An array of repeated lac operators was inserted into a mini‐F plasmid that contained either the parA system of R1 or no partitioning system. Thus, pRBJ460 is a mini‐F plasmid that contains the lacO array and parA. Plasmid pRBJ461 is the corresponding par− control. The lacO‐carrying plasmids were as stably maintained as plasmids that did not contain the lacO array (data not shown).
We expressed GFP–LacI from a second plasmid and subjected the cells to combined phase‐contrast and fluorescence microscopy. Fluorescent dots that were localized to specific subcellular positions were observed in 85–95% of the cells (Figure 4A and B). The vast majority of the cells (all of the 711 examined) had one or two fluorescent foci. The size of the plasmids used here was ∼28 kb, but since the plasmids most likely exist in a compact, supercoiled form in vivo, the positions of the fluorescent foci are expected to indicate the cellular localization of the replicons.
Figure 4A shows cells that contained the parA plasmid pRBJ460. In cells with one GFP focus only, the focus was located close to one of the cell poles (Figure 4A, a–c) or close to mid‐cell (Figure 4A, d–f). In cells with two foci, the fluorescent dots most often were located towards both cell poles (Figure 4A, j–l) and less often close to each other at mid‐cell (Figure 4A, g–i). The proportions of the different cell types are summarized in Figure 4C. Most importantly, the majority of the two‐foci cells had symmetrically located foci, and asymmetrically located foci were rare (6% of the two‐foci cells, type IV in Figure 4C.
We measured the positions of the GFP–LacI foci relative to the cell poles for cells of all types and all lengths. The polar GFP–LacI foci were found to be located at a fixed distance from the pole that coincides with the position of the polar ParM–GFP foci. The distributions of the GFP–LacI foci as a function of cell length are shown in Figure 4D. Whether the fluorescent foci in the one‐dot cells were located towards the cell pole or towards mid‐cell was independent of cell length. In a few of the two‐dot cells, the foci were located symmetrically at positions in the cells where no foci were observed in the one‐dot cells (e.g. the cell in Figure 4A, j, where the foci are located at 1/4 and 3/4 positions). They probably reflect intermediates in the DNA segregation process (i.e. replicons under migration towards the cell poles).
Figure 4B shows cells that contained the control plasmid without parA. In this case, localization was clearly less well defined (compare Figure 4D and E), and one‐dot cells with a polar focus were more abundant (Figure 4C). The difference in plasmid localization for the two‐dot cells was striking. Cells with asymmetrically located foci were abundant for the control plasmid (57% of the two‐foci cells, type IV in Figure 4C). Thus, the presence of the parA system influenced the localization of the plasmid. The localization of the control plasmid, however, was not random. In this case, plasmid molecules were located mainly in the regions of the cells not occupied by the nucleoids. Random localization of par− plasmids to cytosol spaces but not the nucleoid spaces was also observed by Niki and Hiraga (1997).
Effect of cephalexin on the subcellular localization of plasmids
Figure 4A, m–o, shows typical localization patterns in cephalexin‐treated filamentous cells containing the parA‐carrying plasmid. In most cells, two to four foci were observed and, in the majority of the cells, the foci were located symmetrically. In the case of the control plasmid, the foci appeared to be located at random positions, and very few filaments contained symmetrically located foci (Figure 4B, j). These results confirm that the presence of the parA system influenced the intracellular localization of the plasmid and showed that lack of cell division and depletion of FtsZ ring structures did not abolish parA‐mediated subcellular localization of the plasmid.
Co‐localization of ParM and a parA‐containing plasmid
Measurements of the intracellular positions of the ParM–GFP foci and the parA plasmid indicated that ParM and the plasmid co‐localized. To show this directly, we simultaneously examined the intracellular localization of ParM using IFM (red signals in Figure 4F) and that of the parA‐containing plasmid using GFP–LacI (green signals in Figure 4F). Overlays (i.e. combining the two signals from the same cell) of the red and green signals showed that the GFP–LacI foci fully or partially coincided with the ParM foci (the yellow colour in the overlay panels in Figure 4F shows coincidence of the ParM IFM and GFP–LacI fluorescence). Thus, within the resolution limits of light microscopy and the assays used here, the parA‐containing plasmid appears to co‐localize with the ParM partitioning protein.
We examined the intracellular localization of the three components of the parA partitioning system of plasmid R1, i.e. ParM, ParR and parC‐containing plasmids. We found that the ParM protein and parA‐containing plasmids co‐localize to specific sites close to the poles and close to mid‐cell. This specific localization of the components strongly indicates that parA is a true partitioning system that stabilizes plasmids by securing ordered segregation of the plasmid molecules prior to cell division.
We observed a regular and dynamic localization pattern of ParM using both a ParM–GFP fusion protein and IFM (Figure 3).Newborn cells contained two ParM foci close to but not at the cell poles. Upon cell growth, a new ParM focus appeared at mid‐cell. Later in the cell cycle, the mid‐cell focus duplicated. Concomitant with cell elongation, the two new foci at mid‐cell migrated in opposite directions. Cell division between mid‐cell foci then resulted in daughter cells with two polar foci. Localization is an intrinsic property of ParM, since ParM–GFP localized in the absence of other components of the parA system. Concomitant production of ParM or ParR from a second plasmid or the presence of a plasmid containing the parA system had no effect on the localization pattern of ParM–GFP or on the fraction of the cells that had clear foci (data not shown).
The ParM D170E–GFP mutant protein did not localize. We have shown previously that the D170E mutation reduces the ATPase activity of ParM and concomitantly abolishes its ability to support plasmid partitioning (Jensen and Gerdes, 1997). Thus, the ATPase activity of ParM is required both for partitioning and for its intracellular localization. This result is consistent with the inference that localization of ParM is essential for partitioning.
The intracellular localization of a parA‐containing plasmid was examined using the GFP–LacI system. In cells that contained one GFP–LacI focus, the fluorescent focus was located either at mid‐cell or close to one of the poles (Figure 4). In the cells that contained two fluorescent foci, the dots were located almost exclusively close to the poles. A few two‐dot cells also had the foci at mid‐cell. Most importantly, almost all of the two‐foci cells exhibited a symmetrical pattern of localization (94%), whereas the control plasmid exhibited a more random distribution (compare Figure 4D and E). In a few of the two‐foci cells, the plasmids were located at intermediate positions where no ParM–GFP foci were observed (Figure 4A, j). Since the foci in these cells were located symmetrically, they may reflect plasmids migrating to the cell poles. The low abundance of such intermediates is consistent with rapid plasmid movement towards the cell poles (discussed further below).
Comparison of the positions of the ParM–GFP and GFP–LacI foci indicates that the parA‐containing plasmid co‐localizes with ParM–GFP (close to the poles or close to mid‐cell; compare Figures 3C and 4D). The co‐localization was shown directly by simultaneous visualization of ParM using IFM and of the parA‐containing plasmid using the GFP–LacI assay in the same cells. As seen in Figure 4F, overlay of the fluorescent signals clearly demonstrates co‐localization. Generally, more ParM foci than plasmid foci were observed (e.g. the cells in Figure 4F, c, have one plasmid and three ParM foci). The function of additional ParM foci could be to prepare for plasmid segregation in the succeeding cell cycle.
It is reasonable to suggest that ParM aggregates close to the cell poles and mid‐cell via the interaction with a host‐encoded factor. Experiments with cephalexin showed that the localization of ParM and of the parA‐carrying plasmid was unaffected by inhibition of cell division and depletion of FtsZ ring structures. Therefore, FtsZ ring structures are probably not involved in the localization process.
Our data are consistent with the proposal that the polarly located ParM protein tethers the plasmids to the polar regions until septum formation has been completed. We have shown previously that ParM interacts with ParR bound to parC (Jensen et al., 1998). Tethering of the parA‐containing plasmid to the polar ParM foci could be achieved if ParR bound to parC interacts with ParM.
A simple model that explains plasmid partitioning
Recently, we demonstrated that ParR mediates specific pairing in vitro of parC‐containing DNA molecules (Jensen et al., 1998). We believe that the symmetric distribution of the intermediate foci in cells with two GFP–LacI foci (Figure 4D) is consistent with the suggestion that parA mediates plasmid pairing at mid‐cell. The observed co‐localization of plasmids and the ParM protein further indicates that ParM tethers parA‐containing plasmids close to the cell poles. These results support a simple model in which plasmid pairing at mid‐cell is followed by active movement to the cell poles. Plasmid pairing at mid‐cell is also consistent with the recent finding that the DNA replication machinery is present at mid‐cell (Lemon and Grossman, 1998). This suggests that the plasmids move from the cell pole to mid‐cell as part of their replication cycle. After replication at mid‐cell, the twin daughter DNA molecules are paired at parC. Subsequently, the paired molecules are moved to the cell poles by the partitioning apparatus. The model is shown schematically in Figure 5 active plasmid movement probably requires host cell components in co‐operation with the plasmid‐encoded factors. Tethering of the plasmids close to the cell poles keeps them in their proper positions until the septum has formed at mid‐cell.
The intracellular localization of the components of other partitioning systems has also been studied. The P1 and F plasmids localize either at mid‐cell or at the 1/4 and 3/4 positions in the cell (Gordon et al., 1997; Niki and Hiraga, 1997). The subcellular localization of SopB–GFP is consistent with the positioning of F in the cell cycle (Kim and Wang, 1998). Directional movement of newly replicated plasmid molecules at mid‐cell to the cell quarter sites may be responsible for plasmid stabilization by the sop and par system. However, the localization of the parA‐containing plasmid is clearly different from that of the P1 and F plasmids. That the difference is real is supported by the observation that an R1 parA‐containing plasmid was segregated efficiently into E. coli minicells, whereas plasmids carrying P1 par or F sop were not (Eliasson et al., 1992).
The chromosomal partitioning proteins Spo0J of B.subtilis and ParA and ParB of C.crescentus are associated with regions of the nucleoids that usually are located close to the cell poles. In some cells with two separated nucleoids, Spo0J foci near mid‐cell were also observed. The foci were found to replicate and migrate rapidly within the cell (Glaser et al., 1997; Lin et al., 1997; Mohl and Gober, 1997; Sharpe and Errington, 1998; Teleman et al., 1998). The Spo0J protein binds to at least eight sites in the origin‐proximal region of the chromosome (Lin and Grossman, 1998). As expected from this, the origin‐proximal region of the B.subtilis chromosome localized to the same positions as Spo0J and migrated with a similar pattern. This indicates that Spo0J is involved in tethering the origin‐proximal part of the B.subtilis chromosome to positions near the poles (Lewis and Errington, 1997; Webb et al., 1997, 1998; Teleman et al., 1998). The origin‐proximal region of the E.coli chromosome was found to localize and migrate with a pattern similar to that of the B.subtilis chromosome (Gordon et al., 1997; Niki and Hiraga, 1998). Interestingly, the distribution pattern of the ParM foci and the parA‐carrying plasmid observed here was similar to that of the B.subtilis Spo0J protein and the origin‐proximal regions of the B.subtilis and E.coli chromosomes. This suggests that parA may specify a partitioning mechanism that is perhaps related to that of the bacterial chromosomes.
Materials and methods
Bacterial strains and plasmids
The E.coli K‐12 strain KG22 (C600 lacIq lacZΔM15 r− m+; obtained from the Mogens Trier strain collection) was used as host strain in all experiments involving ParM, ParR and GFP fusions. The strain STBL2 [F− mcrAΔ (mcrBC‐hsdRMS‐mrr) recA1 endA1 lon gyrA96 thi‐1 supE44 relA1− λ− Δ (lac‐proAB); Gibco‐BRL] was used for construction of plasmids containing the array of lacO sites. The strain MC1000 [araD139Δ (ara, leu)7697 Δlac X74 galU galK strA; Casadaban and Cohen (1980)] was used as host in all experiments involving GFP–LacI fusions. All plasmids used and constructed are listed in Table I.
Construction of plasmids
To construct fusions between parM, parR and gfp, the following oligonucleotides were used in PCRs as described below:
The parM or parR fusions to the N‐terminus of gfp, which are listed in Table I, were generated by PCR using the following pairs of oligonucleotides as primers: pRBJ430, ParA202B and ParA1122; pRBJ431, ParA1165B and ParA1476; pRBJ451, ParA290 and ParA1122; pRBJ452, ParA280 and ParA1122; pRBJ453, ParA202B and ParA1029; pRBJ454, ParA202B and ParA939; and pRBJ455, ParA202B and ParA1122. The PCR fragments were digested with BamHI and NcoI and inserted into pEGFP. The parM or parR fusions to the C‐terminus of gfp were generated using the following pairs of oligonucleotides as primers: pRBJ432, ParA202C and ParA1124; and pRBJ433, ParA1165C and ParA1478. The PCR fragments were digested with BsrGI and EcoRI and inserted into pEGFP. The parA‐containing plasmid pDD19 was used as template in all PCRs, except for the construction of pRBJ455 where pRBJ338 was used instead.
To construct mini‐F plasmids containing the lacO cassette, the 10 kbp SalI–XhoI fragment of plasmid pAFS59 was inserted into the unique SalI sites of the 18 kbp plasmids pFA10 and pFB10, resulting in the plasmids pRBJ460 and pRBJ461, respectively. Plasmid pRBJ460 is a sop− mini‐F plasmid that contains the parA system and the lacO cassette. The control plasmid pRBJ461 is a sop− mini‐F plasmid that contains the hok/sok post‐segregational killer system and the lacO cassette. The plasmids pRBJ460 and pRBJ461 are stably maintained due to parA and hok/sok of R1, respectively. The hok/sok system stabilizes plasmids by post‐segregational killing and therefore does not influence the subcellular localization of the plasmid.
Growth conditions and media
To express the ParM, ParR and GFP fusion proteins, strain KG22 containing the relevant plasmid was grown for at least eight generations in A + B minimal medium (Clark and Maaloe, 1967) supplemented with 0.2% glucose, 1 μg/ml thiamine, 50 μg/ml casamino acids and 100 μg/ml ampicillin at 30°C. Expression of the fusion proteins was induced by adding 50–100 μM IPTG to the medium. Induction for 3–4 h before microscopy yielded the best results. Under these growth conditions, the generation time of the strain was ∼70 min. Growth at higher temperatures, in rich medium or expression of higher amounts of fusion protein resulted in non‐specific aggregation of ParM–GFP.
To express the GFP–LacI fusion protein, strain MC1000 containing the GFP–LacI expression plasmid pSG20 and the relevant mini‐F test plasmid was grown for at least eight generations in LB medium (Bertani, 1951) supplemented with 100 μg/ml ampicillin and 50 μg/ml kanamycin at 20°C. Expression of GFP–LacI was induced by adding 0.2% l–arabinose to the media. After 15–30 min, 0.2% glucose was added to repress the synthesis of GFP–LacI, and growth was continued for 2–3 h before microscopy. The generation time of the strain was ∼2 h under these growth conditions. A short induction of GFP–LacI synthesis followed by several hours of growth gave the brightest signals, probably because it allowed additional GFP to mature. Under these growth and induction conditions, no fluorescent dots were observed in strains that did not contain the lacO cassette, showing that GFP–LacI did not form non‐specific aggregates. Growth at higher temperatures or when more GFP–LacI was expressed resulted in non‐specific aggregation of GFP–LacI. When relevant, 10 μg/ml cephalexin was added to the growth medium and the cells were allowed to form filaments for 2–3 h before microscopy.
GFP‐expressing cells were examined either immediately (living cells) or after chemical fixation. To fix the cells, formaldehyde (to 2%) and glutaraldehyde (to 0.1%) were added directly to a sample of the culture. The cells were incubated at room temperature for 15 min and at 0°C for 30 min, collected by centrifugation and resuspended in 0.9% saline. Living cells were immobilized on microscope slides using a thin film of agarose as described by Glaser et al. (1997) and fixed cells were immobilized using poly‐l‐lysine‐treated slides or SuperFrost Plus slides (Menzel Gläser). The cells were observed with a Leica DMRBE fluorescence and phase‐contrast microscope with a Leica PL APO 100×/1.40 objective. Pictures were obtained with a colour CCD camera connected to a computer. The GFP foci locations were measured using Scion Image 1.62a (Scion Corporation). For all measurements and statistical analysis of GFP foci positions, at least 200 cells from randomly selected fields were analysed.
The described localization pattern of ParM–GFP was observed both in living and in fixed cells. In living cells, the ParM–GFP fluorescent foci were visible only for a short period after mounting the cells on the slide. Mounting the cells on a thin layer of agarose gave more stable fluorescent foci than when the cells were adsorbed to poly‐l‐lysine‐treated slides. The dissipation of the ParM–GFP signals in unfixed cells could be caused by cell death after mounting. Mild chemical fixation made the GFP signal more stable and allowed us to store the cells for weeks without loss of the specifically located fluorescent signals. However, living cells were used in most of the work.
Cells were grown and induced as described for cells expressing the GFP–LacI protein. The cells were fixed using methanol as described by Teleman et al. (1998) and IFM was performed as described by Addinall et al. (1996). Affinity‐purified rabbit anti‐ParM antibodies were used at a 1:20 dilution, and rhodamine‐conjugated goat anti‐rabbit IgG antibodies (Jackson ImmunoResearch) were used at a 1:200 dilution. IFM‐stained cells were observed with a Nikon Eclipse E800 fluorescence and phase‐contrast microscope with a Nikon Plan Apo 100×/1.40 objective. Pictures were obtained with a cooled CCD camera connected to a computer. The images were acquired and processed using Metamorph 3.6A (Universal Imaging Corp.).
Cells that did not express ParM had no detectable IFM signals, and fixation using formaldehyde and glutaraldehyde as described by Addinall et al. (1996) gave ParM localization to the same intracellular sites.
We thank Pia Hovendal for excellent technical assistance, Dr Issinger and Dr Shapiro for use of their microscopes, Rob Wheeler for help with IFM, and Hansjörg Lehnherr and members of the Lucy Shapiro laboratory for critical reading of the manuscript. This work was supported by the Danish Biotechnology Program (CIS‐FEM), the Carlsberg Foundation and the Plasmid Foundation.
- 5′‐GGACGAGCTGTACAAGGCAGCTATGGACAAGCGCA‐ GAACC‐3′;
- Copyright © 1999 European Molecular Biology Organization