To probe interactions between the Red system and cellular proteins, Redβ was fused at its C-terminus to the SPA affinity tag [16
] on a medium copy number plasmid, under control of the lac
repressor. Redβ-SPA was found to be toxic, inhibiting the growth of wild type E. coli
when induced. This toxicity is surprising, because Redβ itself is innocuous, and hundreds of non-toxic SPA-tagged E. coli
proteins have been made and characterized [17
]. One possible explanation is that the synthetic toxicity of Redβ-SPA results from the delivery of the SPA tag to a critical site, possibly the normal site of Redβ action in the cell. As a hypothetical example, if Redβ normally binds to a replisome protein, it would necessarily have evolved to do so without interfering with replication. However, Redβ with the 69 amino acid residue SPA tag appended might bind to its normal site and clash sterically with another component of the replisome. According to this reasoning, cellular mutants altered in their interaction with Redβ could be selected as suppressors of Redβ-SPA toxicity. Several such suppressor mutants were isolated. Initial mapping showed that three of the mutations mapped to the same area in the E. coli
chromosome. Fine-structure mapping and sequencing revealed that these three mutants each encoded a GreA protein with a single amino acid substitution: D41N (isolated twice), or E44K. The affected residues are particularly significant in GreA's activity. Both participate in the coordination of Mg++
ions in the catalytic core of RNA polymerase [18
]. The greA
-D41N mutation was found not to affect the level of either Redβ-SPA or Redβ as detected by western blot (not shown), ruling out the simplest explanation--that the greA
mutants suppress the toxicity of Redβ-SPA by preventing its synthesis.
To study the unexpected relationship between GreA and Red, E. coli strains were constructed bearing the greA D41N and E44K mutations, as well as a D41A mutation and a simple greA deletion, in a variety of genetic backgrounds. The strains were found to have a number of recombination-related phenotypes.
Consistent with the idea that the greA
mutants might be affected in their interaction with Red, λ behaves as if it had a mutation in either red
), or both, when infecting these mutant bacteria. A λ red gam
double mutant does not form plaques on a recA
host, though red
single mutants do so [21
]. As indicated in Table , a λ gam
single mutant fails to form plaques on a recA
host when the recA
host additionally bears a greA
point mutation (but not a greA
deletion). A deletion eliminating both red
genes reduces λ plaque size; similarly, the greA
point mutations in the host reduce the size of wild type λ plaques. As might be expected if mutating either red
is equivalent in this test, the greA
point mutations have less effect on the plaque size of a λ red
mutant than on the plaque size of λ wild type. This latter observation suggests that the greA
point mutation primarily affects λ red
function, rather than λ transcription, but does not rule out the possibility that there might be some effect on transcription.
The effects of greA
mutants on λ recombination were tested more directly in a marker rescue test, shown in Figure . The crosses were done in a recA
host, to restrict recombination to the Red-promoted, replisome invasion pathway [15
]. Marker rescue was reduced 3-4 fold by the D41N and E44K substitutions, 2-fold by D41A, and very slightly by the greA
deletion. The effects of the greA
mutations on Red-mediated recombination are modest compared to the effects of a red
deletion mutant, which reduces plaque formation in this test to the background of reversion, approximately 40-fold (data not shown).
Figure 1 Marker rescue. A. λ cI857 R-am5, bearing an amber mutation in the endolysin gene R, can form a plaque on the non-suppressing host only if it recombines with the plasmid or reverts. B. Efficiencies of mutants relative to wild type. Titers of the (more ...)
The greA mutations also affect Red-mediated recombination outside the context of a phage infection. As shown in Figure , the efficiency of recombination between the bacterial chromosome and a linear dsDNA introduced by electroporation is reduced in the greA mutants, which exhibit the same rank order of effects as in the marker rescue experiment.
Figure 2 Red-mediated chromosomal gene replacement. A. Strains bearing an alteration of the lac locus in which lacI through the N-terminal coding region of lacZ is replaced by the C-terminal two-thirds of the bla gene and the aph gene of Tn903 can recombine with (more ...)
Recombination via the RecA-RecBCD pathway is also affected by greA mutations. P1 transduction of the Tn10 tetRA genes inserted into several different locations in the chromosome was tested in the D41N mutant. In most, but not all cases, the mutant gave rise to more transductants than the wild type. A marker consisting of the metB gene replaced by tetRA was typical. As shown in Figure , the efficiency of transduction of metBΔtet is elevated by all the greA mutations. Again, the rank order of effects is the same, though in the opposite direction. The elevated frequency of transduction in the greA mutants could be attributed to an elevated frequency of recombination, rather than to a greater efficiency of P1 infection or tetRA expression, because P1 transduction of tetRA at some other loci, for example proAB or recA, was not elevated (data not shown).
Figure 3 Transduction. Strains were infected with an amount of a P1 lysate grown on a metBΔtet strain sufficient to generate 50-100 transductants in the wild type. Ratios of mutant transductants to wild type transductants were calculated. Means and standard (more ...)
To characterize further the specificity of greA recombination effects, the effect of the greA-D41N mutation on RecA-mediated recombination was tested in a strain in which recBCD was replaced by red. As shown in Figure , the replacement has little effect, in either a GreA+ or a greA-D41N background. In particular, the greA mutation stimulates RecA-Red recombination at least as much as it stimulates RecA-RecBCD recombination.
The GreB protein of E. coli
is very similar to GreA in structure and biochemical activities [22
]. The transcriptional effects of a greA
deletion mutant are often greater in a strain in which greB
is deleted as well [23
]. However, as shown in Table and Figures , , and , deletion of greB
had little or no effect on recombination, either by itself or in combination with greA
The DksA protein of E. coli
, compared to GreB, is more distantly related to GreA, but it shares a number of structural and mechanistic features with the Gre proteins, including an acid-tipped coiled coil which reaches into the secondary channel of RNA polymerase to regulate transcription [24
]. Chromosomal dksA
mutants corresponding to the greA
mutants described above were constructed and tested for recombination phenotypes: a deletion fusing the first three and last three codons, and a double substitution mutation neutralizing both of the acid fingertip residues, D71N/D74N [24
]. In contrast to the greA
mutants, neither dksA
mutant exhibited either depressed or elevated levels of Red-mediated or RecA-RecBCD-mediated recombination, nor was either resistant to Redβ-SPA (data not shown).
The mechanism by which the greA
point mutations suppress Redβ-SPA toxicity was investigated by means of a complementation test. Wild type, D41N, and E44K alleles of greA
, under control of their own promoters, were cloned in multicopy plasmids which are compatible with the Redβ-SPA expressing plasmids. The wild type greA
plasmid was found to restore Redβ-SPA sensitivity to a chromosomal greA
-D41N mutant, indicating that the suppression of Redβ-SPA toxicity does not result from a gain of function by the mutant GreA protein. Consistent with this conclusion, and more remarkably, the mutant plasmids do not confer Redβ-SPA resistance upon wild type E. coli
, indicating that the wild type phenotype prevails even when the mutant gene has a higher copy number. However, interpretation of this latter result is complicated because the mutant plasmids significantly slow the growth of wild type cells, as expected based on the known toxicity of the mutant GreA proteins expressed at high levels [18
The known biochemical properties of GreA substitution mutants suggest that one effect of the D41N or E44K mutation would be to increase the amount of time RNA polymerase spends stalled in certain backtracked transcriptional complexes. In the mutant, it is likely that other cellular proteins which respond to stalled transcriptional complexes will play a greater role in resolving them. One such protein is Mfd, which not only dissociates stalled transcription complexes, but at the same time recruits proteins that repair DNA lesions [27
]. This line of reasoning leads to the question of whether mutant GreA suppression of Redβ-SPA toxicity is mediated by Mfd. Accordingly, wild type and greA
-D41N strains lacking the mfd
gene were constructed. Deletion of mfd
was found to have no effect on either the Redβ-SPA resistance of the D41N mutant, or the Redβ-SPA sensitivity of wild type. It can be concluded that Mfd is not required for greA
-D41N suppression of Redβ-SPA toxicity.