We constructed genes encoding fusion proteins of the form Med-LexA by integrating LexA DNA at the end of genes encoding 17 of 25 known Mediator subunits. Each hybrid gene was expressed from its native promoter at its ordinary chromosomal location. shows that the transcriptional activating function of each fusion measured by β-galactosidase levels expressed from an integrated GAL1-LacZ reporter bearing LexA sites. Only three fusions elicited significant levels of β-galactosidase – those bearing, in addition to LexA, the Mediator subunits Gal11, Med2, and Pgd1. All three of these subunits are found in the tail domain of the Mediator.
Activities of non-classical activators
shows that all three of the fusions that activate transcription − Gall1-, Med2-, and Pgd1-LexA – are unstable as visualized by their degradation following addition of cycloheximide. The figure also shows that, in contrast, fusions bearing either a subunit from the head domain (Srb2) or middle domain (Med1), which do not activate, are not unstable. We did not test the stabilities of other Med-LexA fusions. shows that whereas the native Gal11 protein is unstable, the native forms of Med2, Pgd1 and LexA are not. We chose Med2-LexA for further study.
As noted in the Summary, Mediator proteins fused to a DNA binding domain (DBD) activate transcription by integrating into the Mediator and recruiting the complex to the promoter by virtue of the attached DBD. This model was supported by the finding that LexA-Gal11 complemented the defect caused by deletion of WT GAL11
]. The experiment of Figure S1
extends this observation to the case of Med2-LexA. Thus, as shown in the figure, Med2-LexA corrects the growth defect caused by deletion of MED2
We considered the possibility that proteolysis of Med2-LexA occurred only as it activated transcription. The experiment of shows that this cannot be so. Thus, although the activity of Med2-lexA was strongly inhibited (squelched) by overexpression of either LexA or of Med2 (), in neither of those scenarios was the Med2-LexA rendered stable (). A further result of the squelching experiment can also be seen in : not only did the activity of Med2-LexA become significantly inhibited when Med2 was overexpressed, so too did the level of the fusion protein. Such an effect on the level of the fusion protein was not observed in cells overexpressing LexA (). The results suggest that a negative autoregulatory effect maintains the concentration of Med2 below some specified level (i.e., at high concentrations, Med2, directly or indirectly, turns off expression of its own gene.). Another possibility is that high concentrations of Med2 displace Med2-LexA from the Mediator and thereby, for some reason, render it more sensitive to proteolysis.
Effects of overexpression of LexA and Med2
shows that Med2-LexA was stabilized in cells bearing temperature-sensitive (ts) pre1-1 pre4-1
double mutations of the 20S proteasome subunits and grown at the non-permissive temperature [9
]. shows that the fusion was also stabilized by growth at the non-permissive temperature of a mutant strain bearing the cdc53-1 ts
]. This mutation inactivates, at the high temperature, the scaffold subunit cullin found in the SCF family of E3 ligases.
The activity of a stabilized Med2-LexA
We introduced Med2-LexA, on a plasmid, into strains each of which is deleted for, or carries a temperature sensitive mutation in, one or another of 20 F-box proteins. These F-box proteins are known or predicted to be the specificity determinants of the SCF-type E3 ligases [2
]. In only one such strain, that bearing deletion of DIA2
, was Med2-LexA found to be stable (not shown). When expressed from its native MED2
promoter, Med2-LexA was also stable in the dia2
Δ strain (). Dia2 has not been found to be involved in degradation of any classical transcriptional activator. In this strain, Med2-LexA stimulated production of β-galactosidase from the reporter at least 5-fold less efficiently than it did in a WT strain. The level of LacZ mRNA was also decreased by the dia2
deletion, but the effect was less dramatic than the effect on the protein level (). In a control experiment, induction of GAL1
mRNA by Gal4 was identical in WT and dia2
deleted strains (not shown). Also not shown, the san1
mutation, which eliminates the protein quality control pathway in the nucleus [12
], had no effect on the stability of Med2-LexA or of Pgd1-LexA, the two fusion proteins tested.
We undertook these experiments to challenge the notion that transcriptional activators must be proteolytically unstable to work with full efficiency. We reasoned that were the instability noted with classical activators such as Gal4 and Gcn4 some accidental attribute of their activating regions, then an entirely separate class of activators, that work in a different way, would, in some cases at least, work with full efficiency as stable proteins. But this expectation has been confounded. Of the three Mediator components that work as activators when attached to a DNA binding domain, one (Gal11) is inherently unstable, and the other two (Med2 and Pdg1) become unstable when attached to LexA. We do not know why these latter two fusions are unstable – it is possible that instability is some accidental effect of fusing these otherwise stable subunits to LexA. Nevertheless, the instability is required for full activity: where we eliminated a relevant F-box protein (Dia2) and rendered a fusion protein (Med2-LexA) stable, its activating activity was reduced. Thus our results are consistent with the idea that instability of transcriptional activators contributes to the efficiencies with which they work. Perhaps transcriptional complexes must continually turn over to sequentially bring to the promoter components required for steps leading and subsequent to initiation, and activator instability can help facilitate this turnover. A hint that this idea might be correct was the finding of Muratani et al. [2
] that, stabilization of Gal4 (caused by deleting DSG1
) has a much greater effect on translation of the mRNA read from the promoter than it does on mRNA production itself. They further showed that the Pol II recruited by Gal4 working in the dsg1
mutant was deficient in Ser5 phosphorylation of its CTD and in recruitment of the mRNA capping complex. We find here (see ) that increasing the stability of Med2-LexA (by deleting DIA2
) preferentially decreased the amount of translated protein compared to that of the mRNA elicited by the fusion activator.
An additional finding reported here is that amongst a wide array of Med-LexA fusion proteins tested, only those bearing one of three subunits from the Mediator tail domain work efficiently as activators. An array of previous reports have suggested that components of the tail domain (and in particular Gal11) are targets of transcriptional activating regions [4
]. The results, taken with those presented here, might reflect some stereospecific restrictions on how recruitment can be successfully effected, either by a classical or a non-classical, activator, to build a transcriptional complex.