Role of the pre-initiation complex in Mediator recruitment and dynamics

[Editors’ note: the author responses to the first round of peer review follow.]

[…] In view of the considerable amount of new experimental work that seems necessary, the extended time-frame this would likely entail, and the largely uncertain consequences of the new results for your final model, it was decided that the manuscript should be rejected rather than returned to you for revisions. The following list summarizes the most important criticisms, some of which were already mentioned above.

1) Repeat the Taf1 anchor away experiment in a WT KIN28 strain to determine whether Med15 occupancy at UASs is increased (as in Tbp3 depletion) or decreased, because if the latter was observed, this would implicate Taf1 in the initial recruitment of Mediator to promoters in addition to, or instead of, the transition of Mediator from UAS to PIC.

We have included Taf1 anchor away experiments as requested; these are included in the revised Figure 3 (formerly Figure 2). These experiments clearly show that there is no decrease in Mediator occupancy at UAS regions upon depletion of Taf1.

2) To confirm the differences seen between the Tail Med15 and Head Med18 occupancies on depletion of TBP in the kin28-AA strain in Figure 4C, examine the genome-wide behavior of additional Head subunits.

We have included new experiments (Figure 5—figure supplement 4) examining the occupancy of two additional Mediator subunits, Med17 from the head module and Med2 from the tail module, after simultaneous depletion of Kin28 and TBP. The results corroborate those obtained with Med15 and Med18.

3) Compare the effects of depleting TBP, Rpb3, or Taf1 on PIC assembly to rule out the trivial possibility that the different outcomes on depletion of each factor represent different extents of reducing PIC assembly rather than indicating distinct functions in Mediator transfer or retention at the promoter. This would also address whether depletion of one factor by anchor away is simply more efficient than another.

We have included data on the effect of depleting PIC components on occupancy of other PIC subunits, and also cited relevant literature on this point (we had cited a bioRxiv paper on this previously, which has now been published in Genes & Development). Specifically, we document depletion of TBP, Rpb3, and Taf1 (which is less complete) in the corresponding anchor away strains in Figure 3—figure supplement 4 and 5. We also show the effects on occupancy of Pol II following depletion of TBP (Figure 3—figure supplement 4D), TBP and Kin28 (Figure 7A), and Taf1 and Kin28 (Figure 7A); on occupancy of Taf1 after depletion of TBP or TBP and Kin28 (Figure 7B); on occupancy of Taf1 after depletion of Rpb3 (Figure 7C); and on occupancy of TBP after depletion of Rpb3, and occupancy of Rpb1 after depletion of TBP (Figure 3—figure supplement 4). The implications of these results on interpreting Mediator occupancy changes after PIC component depletion are considered towards the end of the Results section.

4) Results for all four Mediator subunits, for all groups of genes, must be shown for a comprehensive analysis of the effects of depleting TBP vs. Rpb3 vs Taf1.

The results for the effects on occupancy of four Mediator subunits following depletion of TBP or Rpb3 are shown in Figure 3 (formerly Figure 2) in the revised manuscript, and results for two subunits are shown for Taf1 (all that we have done), as little change was seen for Taf1, and the major point for that PIC component is that recruitment of Mediator is not decreased upon its depletion. These results are now presented in the same way (at “UAS genes”) in all cases.

5) Provide a biological replicate for the results on Med2 in Figure 1A, as well as an input or untagged control.

We conducted a biological replicate for the results in Figure 1A of the original submission, and discovered that the signal we observed for Med2 occupancy was not significantly different to that seen with a control, untagged strain. We therefore have removed this part of Figure 1 from the manuscript.

6) Present time-courses of anchor away to justify the rapamycin treatments chosen for each protein.

We have presented a time course for anchor away of TBP, and also provide justification of our choice of one hour depletion by comparison with other published work using this method to deplete components of the transcription machinery.

In addition, we have included new data showing that Pol II occupancy is reduced genome-wide to about 25% of wild type levels in the med2∆ med3∆ med15∆ mutant, demonstrating that the tail module plays an important role in Pol II occupancy at essentially all genes.

Reviewer #1:

[…] While some of their results can be viewed as being largely confirmatory of previous findings from the Struhl and Robert labs, other findings increase our understanding of how Mediator is recruited to genes, and the transition of Mediator from the UAS to the promoter. Examination of the triple Mediator tail subunit deletion, in which the tail is completely eliminated provides stronger evidence that obtained previously with just two subunits deleted, that the tail module is important for Mediator recruitment at most active genes; but also reveal a tail-independent pathway that allows appreciable Mediator recruitment and transcription to continue, which they are able to argue involves Mediator recruitment by Pol II. Their findings that TBP is required for the transition of Mediator from UAS to promoter, even when the CDK module is depleted, whereas Pol II is important for retention of Mediator at promoters, and might contribute only indirectly to enhancing the UAS-promoter transition by enhancing TBP occupancy at the promoter, are signficant. The role of Taf1 is clearly distinct from that of TBP, but it remains unclear whether Taf1 is enhancing Mediator recruitment to the UAS (like TBP) or stabilizing Mediator at the promoter after the UAS-promoter transition (like Pol II).

To resolve this last issue, the authors should conduct the additional experiment of depleting Taf1 alone to clarify its role in Mediator recruitment vs stabilization at promoters.

We thank the reviewer for a thoughtful and thorough critique, and are pleased that s/he finds our results significant. The reviewer states that the experiments using the med2∆ med3∆ med15∆ are “largely confirmatory” of previous work from the Struhl and Robert labs, but also states that this result “provides stronger evidence than obtained previously”. We agree with both of these comments, mostly; although the effect of the triple deletion is not unexpected, the results establish definitively that the tail module triad is non-essential, contributes considerably to Mediator recruitment at most active genes, and demonstrates that alternative modes of Mediator recruitment and gene activation, independent of the tail module triad, exist. Prior to this work, the possibility existed that the tail module triad was essential, and that residual function of Med2 was sufficient to allow viability in the double mutant; this is now conclusively ruled out. We have also added a new experiment showing that Pol II occupancy is reduced to ~25% of wild type levels in the med2∆ med3∆ med15∆ mutant across the genome.

With regard to the last point, that it is unclear whether Taf1 is important for Mediator recruitment to the UAS or for stabilizing Mediator at the promoter after transit from the UAS, we have included in the revised manuscript experiments that provide evidence that Mediator continues to occupy UAS regions after Taf1 depletion, in the presence of active Kin28. These data support the idea that Taf1 is important for stabilization of Mediator at promoters after transit from the UAS.

Reviewer #2:

[…] First section entitled "Mediator is recruited to gene promoters in the absence of the tail module triad" is simply confirming data from Struhl and Robert labs. The only difference here is the use of a triple deletion of the triad (instead of a double deletion that leads to the loss of the third component). The data is very nice but constitutes a very small increment over published data.

We are pleased that the reviewer considers our data to be solid. We disagree that the first section, regarding the knockout of the complete tail module triad, is “simply confirming data from Struhl and Robert labs.” Our finding that Mediator occupancy at UAS regions in med2∆ med3∆ med15∆ yeast is nearly completely eliminated can fairly be regarded as confirmatory, as a similar result was reported for the double mutant (med3∆ med15∆) and the clear expectation would be that loss of Mediator occupancy at UAS regions would be seen in the triple mutant also. However, it could not be predicted with confidence that the triple mutant would be viable, nor what the effect would be on Mediator occupancy at promoters. Our results provide the first definitive evidence that the Mediator tail module triad is not essential, and that Mediator can be recruited to gene promoters in its absence. We also show in the revised manuscript that association of both Mediator and Pol II is substantially decreased in the triple mutant, considerably more than the less than two-‐fold effect reported previously (Jeronimo et al., 2016), showing that the tail module triad is broadly important for recruitment of these components.

We agree with the reviewer that based on the in vitro results of Carey and co-workers, it seemed likely that Mediator association with gene promoters would depend on PIC integrity. But in vitro results do not always accurately predict in vivo interactions, and this prediction had not previously been tested experimentally.

The second section entitled "Mediator association at UAS regions is stabilized by loss of TBP or Pol II" shows that Mediator (Med2, Med15 and Med14) detection at UAS is increased upon disruption of the PIC by Rpb3 or TBP anchor-away. This observation is interesting and the authors interpret this data in the context of the "transition model" proposed by Struhl and Robert labs and suggest that interfering with the PIC leads to the stabilisation of Mediator at UAS. Unexpectedly, however, no change is observed with the Head subunit Med18. If the whole Mediator was stabilised at UAS as proposed by the authors, then all subunits should be affected. The Med18 data may suggest more complex interpretations. For instance, it may be that after its transient interactions with the PIC, Head/Middle dissociates, leaving the Tail stably associated with the UAS. In subsequent rounds, the Head/Middle would re-attach on the UAS-bound Tail and so on. This section of the manuscript has probably the most potential for novelty, but more experiments are needed. I would also suggest reorganizing Figure 2 (and Figure 4). The data for all subunits tested should be represented the same way. Why are Med18, Med2 and Med14 not fully represented as for Med15? Also, why are Med2 and Med14 mapped on SAGA and TFIID genes while the other subunits are mapped on "UAS genes".

We have remade Figure 2 (Figure 3 in the revised manuscript) so that it shows heat maps and line graphs at “UAS genes” for all four Mediator subunits. Because depletion of Taf1 showed little effect on Med15 and Med18, we did not examine the effect of Taf1 depletion on Med2 and Med14. As to the possibility that the tail module might interact with UASs in the absence of the rest of Mediator, we do mention this in the text but do not favor this possibility. First, the increased occupancy seen for Med14 upon depletion of Rpb3 or TBP (Figure 3D) is consistent with results for Med15 and Med2, and suggests that Mediator as a whole shows increased occupancy under these conditions. We believe this interpretation is most consistent with results from the Struhl and Robert labs, and in particular with the model shown in Figure 7 of Jeronimo et al., 2016. We show in Figure 3—figure supplement 5B that, as in the Jeronimo et al. paper, we observe ChIP signal at some genes after Kin28 depletion at both UAS and promoter regions for Med15, but only at promoter regions for Med17 and Med18. When considered together with the re-ChIP experiments in Petrenko et al., 2016, Figure 1, these results seem most consistent with Med18 exhibiting weaker ChIP signal at UAS regions than does Med15.

The third and fourth sections "Mediator association with promoters depends on PIC components" and "Depletion of TBP, and to lesser extent Pol II, stabilizes Mediator occupancy at UAS regions" examine the effect of TBP, TAF1 and Rpb3 depletion on Mediator occupancy at promoters and UAS in kin28AA cells. I found the description of this data somewhat confusing and some of the conclusions not to be supported by the data. First, both sections really describe the same data so I do not see why separating them. It would be simpler to go through the data systematically and (as suggested for Figure 2) represent all datasets the same way (Figure 4). Currently, the third section describes TAF1 and Rpb3 and says nothing about TBP. It is not clear why. Second, it seems to me that all three depletions lead to the same phenotype, although with different severity. The TBP depletion is described as having fundamentally different effects on Mediator compared to Rpb3 and TAF1 depletions but one could argue that TBP depletion simply has a more pronounced effect on Mediator as a consequence of a more profound effect on the PIC. The authors should show the effect of their depletions on a few PIC components. One would expect TBP depletion to abolish the PIC altogether, whereas TAF1 and Rpb3 depletions may leave partial PIC behind. Alternatively, the differences in magnitude may be due to variable efficiencies (or timing) of depletion for TBP, Rpb3 and TAF1, a possibility not acknowledged. On a related note, the authors should show the depletion of their AA strains over time (easily done by ChIP-qPCR) in order to justify the time point used and relativize the effect observed for each of them. Third, I do not see what data allows the authors to conclude about timing of events as claimed in their discussion: "[…] TBP being most important for transit of Mediator from UAS to promoter and Rpb3 and Taf1 being more important for stabilizing occupancy at the promoter following transit".

We agree that depletion of TBP is likely to have a more profound effect on the PIC than depletion of Rpb3 or Taf1, but we are puzzled by the statement that all three depletions lead to the same phenotype, but with different severity. The data in Figures 4C and D (Figure 5 in the revised manuscript) show that both Med15 and Med18 peaks shift upstream upon depletion of TBP at SAGA-dominated genes and genes having upstream Mediator occupancy in wild type yeast (“UAS genes”); Med15 can be seen to shift upstream at TFIID-dominated genes as well. Depletion of Rpb3 or Taf1 leads to a partial upstream shift or no upstream shift, respectively, but a strong decrease in signal. If TBP depletion simply had a more pronounced effect than depletion of Taf1, for instance, it should result in a stronger decrease and no upstream shift, or Taf1 should give rise to a partial upstream shift. These results indicate that TBP is necessary for Mediator occupancy to shift from UAS to promoter, but cannot determine whether it is sufficient, as Rpb3 occupancy is abolished by loss of TBP (Figure 3—figure supplement 4D and Figure 7A). In contrast, Mediator occupancy does not shift upstream upon depletion of Taf1, and only partially upon depletion of Rpb3, suggesting that these PIC components are not necessary for Mediator to transit from UAS to promoter region. Hence their behavior is qualitatively different. I believe the other two reviewers agree with this point.

We have included new data and citations to show the effect of individual PIC components on each other both with and without Kin28 depletion (Figure 3—figure supplement 4 and Figure 7). As the reviewer suggests, depletion of TBP eliminates Pol II association, while depletion of Rpb3 only causes partial loss of TBP, consistent also with recently published work (Joo et al., 2017). Perhaps surprisingly, Taf1 occupancy was only modestly reduced upon depletion of TBP or Rpb3 (Figure 7). We have added text to the Results that considers the results seen on Mediator occupancy in light of cooperative assembly of the PIC.

With regard to our rationale for using a one hour treatment with rapamycin, we chose this as a reasonable balance that would allow near complete depletion but not allow much time for indirect effects, as reviewer #3 suggests. Our ChIP data for anchoring away TBP indicated that depletion was essentially complete after 30 minutes, and we have now included this experiment as supplemental data. Other studies using anchor away to deplete components of the transcription machinery have used similar times of rapamycin treatment (30 min for Taf proteins (Warfield et al., 2017), 1 hr for Spt7 (Baptista et al., 2017) and for Mediator subunits, TBP, and Rpb1 (Wong and Struhl, 2014; Petrenko, 2016, 2017); 90 min for Mediator subunits (Jeronimo et al. 2016); 120 min for Mediator subunits (Anandhakumar et al., 2016)). Interestingly, the latter study found only about 80% depletion of Mediator subunits after 1 hr of rapamycin treatment, whereas the Struhl laboratory found depletion to background levels in 1 hr (Wong and Struhl, 2014), although the two studies used different criteria to monitor depletion. It is possible that the GFP tags also employed in the former study affected kinetics of depletion, but this has not been rigorously investigated.

Finally, with regard to organization, we prefer to keep the two sections mentioned separate. The first reports the evidence that Rpb3 and Taf1 are required for stable association of Mediator at promoters, while the second reports on the shift to the UAS region seen upon depletion of TBP. These distinct outcomes are best emphasized, in our opinion, by being discussed in separate sections.

Reviewer #3:

[…] 1) It seems that single replicates were performed for some "main figure" experiments, including Figure 1A (Med2 TAP). Please confirm Figure 1A with a second biological replicate for Med2, since this is a main figure and drives the rationale for the rest of the manuscript. Also, is there an untagged/input control for Figure 1A? This would be helpful to get a sense of the recruitment of the tailless mutant over background.

We repeated the experiment of Figure 1A and included an untagged control. The signal seen in this experiment for Med2 at UAS regions was small (as in the original figure), but importantly not different from that seen with the untagged control. We have therefore not included this figure in the revised manuscript. However, we do not (and did not) consider this to be essential as a rationale for examining the complete deletion of the tail module triad. Prior to the work reported here, it remained possible that Med2 could contribute to Mediator recruitment to UAS regions and/or promoters in med3∆ med15∆ yeast; we have now formally excluded this possibility, while also defining rigorously the extent to which the tail module triad contributes to Mediator and Pol II recruitment genome-wide.

With regard to other experiments, in all cases where exact replicate experiments were not performed, we have performed corroborating experiments by doing ChIP against multiple Mediator subunits under the same conditions (e.g. Figure 3), or have relied on published results that corroborate our experiments (e.g. decreased occupancy of TBP upon depletion of Pol II). We have also added new experiments in which occupancy of two additional Mediator subunits, Med2 and Med17, is examined after depletion of Kin28 and TBP. The results are consistent with those observed for Med15 and Med18, with the tail module subunit, Med2, exhibiting greater signal at the UAS following depletion, while the head module subunit, Med17, shows strongly decreased signal relative to depletion of Kin28 alone, but very little upstream signal.

2) There needs to be a correlation coefficient for Figure 1C to get a quantitative measure of difference in the gene expression in mutant relative to WT. "dispensable for most gene expression" would be more convincing with showing the numerical value for the graph.

We have added correlation coefficients to the graphs in question and calculated Fisher’s z-score for the difference between correlation coefficients of the two graphs, yielding a p-value < 10^-10. However, we have revised our conclusion regarding the role of the tail module in gene expression based on the new data showing that Pol II association is substantially (about 4‐fold) reduced in the tail module deletion mutant: evidently the tail module is broadly important for Pol II recruitment, but the effect of this reduced recruitment is likely offset by altered RNA decay, as has been observed for other perturbations in PIC components.

3) Figure 2B, why isn't occupancy of Med 18 changed with AAR?

We do not have a definitive answer for this question, but speculate that it may reflect Med18 being less proximate to UAS regions and correspondingly exhibiting a weaker ChIP signal there, as proposed by Jeronimo et al., 2016; see also our response to reviewer #2. Alternatively, Mediator may exist in a configuration that decreases accessibility to Med18 under the conditions of TBP depletion. Reviewer #2 suggests it could reflect loss of the Mediator head module (or perhaps only of specific subunits), but we do not favor this explanation at present as there is little evidence to support it.

4) Anchor away was performed for 1 hr. Please provide an explanation for why this time frame, as opposed to less time like 20-30 min. Obviously, there is a trade-off between completeness of the sequestering and the potential for indirect effects to creep in. Was this condition chosen for a reason or was it arbitrary? Is there GFP localization experiment to verify the anchor away?

We chose 1 hr as a reasonable balance that would allow near complete depletion but not allow much time for indirect effects, as the reviewer suggests. Our ChIP data for anchoring away TBP indicated that depletion was essentially complete after 30 minutes, and we have now included this experiment as supplemental data. Other studies using anchor away to deplete components of the transcription machinery have used similar times of rapamycin treatment (30 min for Taf proteins (Warfield et al., 2017), 1 hr for Spt7 (Baptista et al., 2017) and for Mediator subunits, TBP, and Rpb1 (Wong and Struhl, 2014;Petrenko 2016, 2017); 90 min for Mediator subunits (Jeronimo et al., 2016); 120 min for Mediator subunits (Anandhakumar et al., 2016)). Interestingly, the latter study found only about 80% depletion of Mediator subunits after 1 hr of rapamycin treatment, whereas the Struhl laboratory found depletion to background levels in 1 hr (Wong and Struhl, 2014). It is possible that the GFP tags also employed in the former study affected kinetics of depletion, but this has not been rigorously investigated.

5) The results in Figure 4 are striking, and show that when Kin28 is depleted Mediator occupancy at TFIID promoters is highly dependent on Pol II and Taf1. I'm a bit puzzled as to why there needs to be a dependence on Kin28. Or is that a product of the experimental design, in that Taf1 or pol II AAR was not conducted in a WT KIN28 strain? Why are opposite effects seen for Med 15 and Med 18 in tbp+kin28 AAR?

Depletion of kin28 in the experiment of Figure 4 was used to allow detection of Mediator at promoters of active genes, as shown by the Robert and Struhl labs (Jeronimo and Robert, 2014; Wong and Struhl, 2014). We have edited the text to clarify this point. We did include an experiment in which Pol II was depleted (“pol II AAR”), shown in Figure 2 (now Figure 3), and have added data for Taf1 depletion in the presence of wild type KIN28 to Figure 3.

I believe that what the reviewer is referring to as “opposite effects seen for Med15 and Med18 in tpb + kin28 AAR” is the apparent increased signal at SAGA-dominated genes for Med15 and decrease for Med18. We believe that this reflects Mediator being stranded at the UAS region after TBP depletion; because the contact with Med15 is much closer, ChIP efficiency is high and the signal is strong, whereas the Med18 subunit is more proximate to the promoter in kin28-‐AAR conditions than to the UAS in kin28-tbp-AAR conditions, resulting in decreased ChIP efficiency and lower signal. This interpretation is consistent with the model proposed by the Struhl and Robert labs, and specifically with the detailed structural considerations discussed in Jeronimo et al., 2016. We have added text to clarify this interpretation. See also Figure 3—figure supplement 5B, and our response to reviewer #2.

[Editors' note: the author responses to the re-review follow.]

Reviewer #2:

The manuscript by Knoll et al. describes how yeast Mediator occupancy at UAS and promoters is affected by PIC components. Using combinations of anchor-away experiments coupled to ChIP-seq, they propose that TBP is required for the transit of Mediator from UAS to promoters while Pol II and Taf1 are required for stabilising Mediator at promoters. They also provide data showing that the Tail triad is not required for the association of Mediator with promoters.

This work is building on prior work from the Struhl and Robert lab and several aspects are largely confirmatory. The most novel aspect concerns how different PIC components contribute to the association of Mediator with promoters. Unfortunately, the data appears to be of variable quality and the conclusions are based on interpreting cherry-picked replicates while ignoring others that would have led to very different conclusions.

We felt this reviewer was unduly harsh in accusing us of “interpreting cherry-picked replicates while ignoring others that would have led to very different conclusions”. This statement is based on the replicate data shown in Figure 3—figure supplement 3 in the critiqued manuscript, which included data (high signal for both Med15 and Med18 in the TBP anchor away strain in the absence of rapamycin) that was inconsistent with data in the main figure. In the text of the previously submitted version we argued that the preponderance of our results were most consistent with the interpretation we provided; it is not in our opinion accurate to state that we ignored the inconsistent results. Nonetheless the reviewer’s concern regarding reproducibility is valid and was a concern to us as well; that is why we chose to include replicate data for this experiment as a supplementary figure. We have now repeated this experiment—ChIP-seq of Med15 and of Med18 in the tbp-AA yeast strain in the absence and presence of rapamycin—and the results are consistent with those we had presented in Figure 3, including results for ChIP against Med2 and Med14 (see Author response image 1). We have revised Figure 3 by presenting data for Med15 and Med18 ChIP averaged from two independent biological replicate experiments (except for the Med18 ChIP in the taf1-AA strain, for which we have only one replicate).

Author response image 1

Download asset Open asset

Reviewer #2 was also concerned about whether data in Figure 5 was based on replicate experiments. The figure as presented in the critiqued manuscript included combined data for two experiments that were consistent with two additional replicate experiments for both Med15 and Med18 ChIP in kin28-tbp-AA yeast (four replicates in total), and data for single experiments that were consistent with replicate experiments for Med15 and Med18 ChIP in kin28-AA yeast. The reason for combining data only for kin28-tbp-AA yeast is that we obtained fewer reads for those experiments, and combining data yielded better signal to noise. For Med18 in kin28-taf1-AA and kin28-rpb3-AA strains, we showed data from single experiments that were qualitatively consistent with replicate experiments. However, the extent to which Med18 signal was affected in the kin28-taf1-AA mutant differed quantitatively in these replicate experiments. We have now performed an additional replicate of this experiment. All three replicates are qualitatively consistent (see Author response image 2): depletion of Kin28 together with Taf1 decreases signal at TFIID-dominated genes, and does not substantially shift Mediator peaks at UAS genes as observed for kin28-tbp-AA yeast (see Author response image 2). The effect of depletion of Kin28 and Taf1 on Med18 occupancy at SAGA-dominated and UAS genes (which are mostly SAGA-dominated) was variable in our three replicate experiments with regard to peak magnitude relative to depletion of Kin28 alone, although it was always a less pronounced decrease than seen for TFIID-dominated genes. We have modified Figure 5 to use the replicate experiment that shows intermediate changes in signal intensity, and make a note of this variability in the figure legend of the revised manuscript. Finally, only single replicates had been performed for Med15 ChIP using kin28-rpb3-AA and kin28-taf1-AA yeast. We have repeated these experiments and obtained consistent results in kin28-rpb3-AA yeast; however, we have seen stronger signals for Med15 in kin28-taf1-AA yeast in the presence of rapamycin than in the experiment, which was used for Figure 5 in the critiqued version (see Author response image 2). Again, we see an upstream shift upon depletion only with TBP and not with Taf1 or Rpb3, and strong decrease in signal at TFIID-dominated genes (see revised Figure 5). We have replaced the data for Med15 ChIP in kin28-taf1-AA in Figure 5 with data from one of the new replicates (the “run40” replicate shown in Author response image 2), to reflect the stronger signal we see for kin28-taf1-AAR relative to kin28-AAR in the new replicates. We have not used combined data for Figure 5 (except as noted above), as the signal to noise and number of reads has varied among replicates, and the combined data is dominated by the “cleanest” data. Information on replicate experiments provided as Supplementary file 3 has been updated to reflect these additional replicate experiments.

Author response image 2

Download asset Open asset

The reviewer also states that experiments are normally done in duplicate (at least) and some correlation metrics used to insure they are alike. We have found that use of correlation coefficients between ChIP-seq datasets is not informative, as variance in regions outside of the peaks that we focus on (the large majority of the genome) obscures the similarities or differences among the regions of interest in terms of quantitative metrics. Rather, we have used metagene analysis, as in Figures 3 and 5 and the Author response image 1 and 2 to assess similarity between replicate experiments.

Another major concern with the manuscript has to do with the Taf1 anchor-away data. As acknowledged by the authors, the taf1-FRB strain appears to show a phenotype even in the absence of rapamycin. This means that the tagging affected the function of the protein. Because rapamycin did not lead to any addition effect, I do not think that the data can be interpreted as a consequence of conditional depletion. Unless the authors find a way to conditionally affect the function of Taf1 (perhaps auxin degron or tagging FRB in N-terminal would work better), I believe the Taf1 data has to be removed.

The reviewer states that “addition of rapamycin did not lead to any additional effect”, but this is not correct: although adding rapamycin did not affect the ChIP signal for Med15 or Med18 (Figure 3), it did cause lethality (Figure 3—figure supplement 1) and reduced Taf1 occupancy (Figure 3—figure supplement 3 in the revised manuscript); furthermore, anchoring away Taf1 together with Kin28 suppresses the effect of anchoring away Kin28 at TFIID-dominated genes (Figure 5). We therefore believe that the data presented support the main conclusion of experiments on the taf1-AA strain, which is that Mediator association with UAS regions does not decrease upon depletion of Taf1, and therefore that Taf1 is not required for Mediator recruitment to UAS regions. Furthermore, we have observed in three independent experiments that depletion of Taf1 together with Kin28 results in preferential decrease in Mediator occupancy at TFIID-dominated genes. However, the effect of this double depletion on Med15 occupancy at SAGA-dominated genes has been variable in our experiments, as mentioned above, and so we have tempered our conclusions regarding this specific effect.

In sum, I recommend the authors go back to the basics and make sure their experiments are reproducible. For that, they need to show that their replicates are highly correlated. At current state, I suspect that most of the effects observed are batch effects rather than biological effects. For instance, simply inspecting the red tracks in Figure 3—figure supplement 5A clearly shows that rep2 did not work nearly as well as rep1. Not seeing the labels, one could not even tell how to pair them. Also, looking at the blue tracks from the same Figures suggests that the samples cluster by batches rather than by conditions.

We have revised this Figure (now Figure 3—figure supplement 4) so that tracks are normalized across samples for each strain (e.g., all four tbp-AA strain scans are normalized) rather than only between samples within a single experiment (e.g. tbp-AA-1 and tbp-AAR-1 samples were normalized relative to one another but separately from tbp-AA-2 and tbp-AAR-2 samples). This shows more clearly the reduced signal of TBP upon addition of rapamycin.

[Editors' note: further revisions were requested prior to acceptance, as described below.]

Reviewer #2:

The revised manuscript is improved, notably by the addition of datasets completing some that were done only once and by the replacement of non-consistent replicates. However, there are a few issues that are still of concern and should be addressed.

First, some of the anchor-away strains behave strangely. While this is unlikely to drive erroneous conclusions, it should at least be mentioned. For instance, Kin28-TBP-AA Med15-Myc is very sick according to Figure 3—figure supplement 1. More worrisome, tagging various PIC components with FRB greatly affects Mediator ChIP in the absence of rapamycin. This is evident by looking at several figures, notably in Figure 3 (pay attention to the y-axis for all the no rapa traces (black) or look at the intensities of the no rapa heatmaps). This is also directly shown in Figure 3—figure supplement 2B. The authors use this figure supplement to claim that Taf1-AA leads to increased Mediator ChIP signal in absence of rapamycin but they fail to acknowledge that this is also the case for TBP-AA (the average signal is 2 to 3 fold up compared to non-tag). It is not clear whether these differences are batch differences (due to technical variation between experiments performed on different days) or biological differences (caused by partial loss of function of tagged proteins).

We have edited the text to note the effect of the tbp-FRB allele on Mediator occupancy and the slow growth phenotype of the kin28-tbp-AA Med15-myc strain (Results section). The elevated Med15 occupancy observed in taf1-AA yeast in the absence of rapamycin was seen in two out of two replicate experiments, while lower levels of Med15 occupancy, similar to the control lacking any FRB tag, was seen in two out of two experiments with rpb3-AA yeast and two out of three experiments in tbp-AA yeast. One experiment showed elevated Med15 occupancy in tbp-AA yeast in the absence of rapamycin; that is the experiment that led to the request for an additional replicate, which showed low occupancy in the absence of rapamycin in tbp-AA yeast that increased upon rapamycin treatment. We have also edited Figure 3—figure supplement 2 by replacing the data for the “no FRB tag” control with data from a newer replicate (from October this year). The original data had higher levels of artifactual signal over ORF regions (as we and others often see), which made it more difficult to visually compare signal strength over the UAS regions.

Second, the Med18 ChIP data displayed in Figure 3 and showing that Med18 does not change in TBP-AAR or Rpb3-AAR does not fit the author's model very well. In subsection “Mediator association at UAS regions is stabilized by loss of TBP or Pol II” the authors say "The increase in ChIP signal observed for Med15, but not for Med18, would be consistent with the most immediate contacts with UAS-bound activators occurring through tail module subunits (such as Med15), while ChIP signal for Med18 at UAS regions would depend on protein-protein cross-links, thus lowering the efficiency of ChIP for this subunit". While I agree that this explains the lower ChIP efficiency of Med18 at UAS, it does not explain why, unlike other subunits tested, Med18 signal does not increase at UAS in TBP-AAR and Rpb3-AAR. The observed Med18 ChIP signal is clearly over the UAS so if Mediator is stuck there, this Med18 signal should increase together with other subunits. The fact that it does not (assuming the data is real), suggests a more complicated model. This result should at least be described more critically.

We agree with this criticism, and have modified the text accordingly. We did not make our point clearly, which is that under some circumstances, such as depletion of Kin28 (Figure 3—figure supplement 4B, and also Jeronimo et al., 2016), Med15 can be detected at UAS regions while Med18 gives rise to virtually no ChIP signal at those sites; in the same sample, signals from both Med15 and Med18 are easily discerned at the proximal promoter. This suggests that Mediator may adopt a configuration in which Med15 signal is robust and Med18 undetectable, or nearly so; such a configuration could result in increased Med15 signal with no increase in Med18 signal. The way we presented this before was poorly constructed at best, and we thank the reviewer for noticing this.

Third, the conclusion that TBP is required for the transition while Rpb3 and Taf1 are stabilising the PIC-associated Mediator is in my opinion a hazardous extrapolation. As explained by the authors in subsection “TBP is required for transit of Mediator from UAS to promoter”, depletion of PIC components has impact on other PIC components. This makes the attribution of specific roles to specific PIC components in Mediator dynamics difficult. Hence, with these limitations in mind, it seems like an over interpretation to say "TBP being critical for transit of Mediator from UAS to promoter, while Pol II and Taf1 stabilize Mediator association at proximal promoters". A more conservative interpretation would be that the PIC is necessary to allow Mediator transiting to the promoter (as shown when PIC is severely compromised; TBP-AA) and that partially formed PICs (without trying to attribute specific roles to subunits or proteins), while allowing the transit, lead to unstable Mediator-PIC association.

These last two points had been raised in previous review rounds but are still not clearly dealt with in the most recent version of the manuscript.

We believe that our statement that TBP is critical for transit of Mediator from UAS to promoter is logically accurate. It does not imply that other factors, in this case other PIC components that depend on TBP for stable association with promoters do not contribute. As we state in subsection “TBP is required for transit of Mediator from UAS to promoter”, we can infer that TBP is necessary for Mediator transit to promoter regions under conditions of Kin28 depletion, but not whether it is sufficient. In contrast, the evidence presented indicates that under conditions of Kin28 depletion, neither Taf1 nor Pol II are critical for this transit. However, the reviewer’s summary statement does provide a succinct summary of these points, and we have edited the text to borrow his/her phrasing (hopefully with permission).

Finally, it is not clear why the authors were expecting a decrease of Mediator at UAS upon depletion of TBP, Rpb3, or Taf1. Isn't an increase what the "dynamic model" proposed by the Struhl and Robert labs predicted?

We had initially anticipated that PIC depletion might reduce Mediator occupancy at both promoters and UAS regions. This could occur, for example, if Mediator contacted both UAS and promoter via DNA looping and reduced PIC occupancy destabilized Mediator occupancy at both regions. This was evidently not the right way to think about it, but it still seems to me not a totally far-fetched possibility in the absence of data. We have edited the text to clarify this point. Clearly what we anticipated was not as important in any event as what we observed.