The role of Mediator and Little Elongation Complex in transcription termination

Mediator is a coregulatory complex that regulates transcription of Pol II-dependent genes. Previously, we showed that human Mediator subunit MED26 plays a role in the recruitment of Super Elongation Complex (SEC) or Little Elongation Complex (LEC) to regulate the expression of certain genes. MED26 plays a role in recruiting SEC to protein-coding genes including c-myc and LEC to small nuclear RNA (snRNA) genes. However, how MED26 engages SEC or LEC to regulate distinct genes is unclear. Here, we provide evidence that MED26 recruits LEC to modulate transcription termination of non-polyadenylated transcripts including snRNAs and mRNAs encoding replication-dependent histone (RDH) at Cajal bodies. Our findings indicate that LEC recruited by MED26 promotes efficient transcription termination by Pol II through interaction with CBC-ARS2 and NELF/DSIF, and promotes 3′ end processing by enhancing recruitment of Integrator or Heat Labile Factor to snRNA or RDH genes, respectively.

potential understanding of how specific transcription complexes interact to regulate transcription termination of snRNA genes and replication dependent histone (RDH) genes. The authors use a range of established methods to establish the requirement for LEC in non-poly(A) transcription termination. LEC was shown to co-purify with proteins required for the non-poly(A) termination of specific genes in particular snRNA and RDH genes. These experiments outline a role for specific LEC components in the termination of RDH genes. The findings are relevant for understanding the relationship between mediator and transcriptional termination of snRNA and RDH genes and point to unique roles for LEC protein components in RDH gene regulation. Nevertheless, the manuscript falls short of providing compelling evidence that demonstrates a significant step in understanding the transcriptional consequences of the proposed mechanism termination at poly(A) versus upstream termination sites. Data pointing to the functional consequence of failure to terminate transcription at non-poly(A) sites would add considerably to this manuscript. Comments 1) What are the functional consequences of transcription termination producing polyadenylated mRNAs of genes that would otherwise produce a stem-loop structure instead of poly(A) tails? Could the authors speculate why cells evolved to terminate these genes upstream of polyA sites? Does it affect cell cycle, nucleosome structure, chromatin remodelling? Replication dependent Histone genes are expressed predominantly through S-phase and supply cells with requisite histones necessary for the DNA replication. Changes Hitone transcripts may translate into changes in nucleosome structure which can have profound effects: transcription kinetics, the association of key regulatory proteins, the association with chromatin remodelling complexes. The polyadenylation sites downstream of the of HDEs are suggested to act as a mechanism to prevent PolII from reading through into regions downstream of the RDH genes. Is this the likely the only role of these downstream polyA sites? 2) Is this a widespread mechanism? Similar and/or supporting data using other cell lines would be of enormous benefit. Equally how evolutionary conserved would this process be? Can the authors comment from literature on similar observations and mechanisms from yeast, files or worm models?
3) The authors performed RNA seq from polyA-selected libraries and ribo depleted libraries prepared from HEK293T cells. RNA seq of poly(A) mRNA libraries to show enrichment of genes upregulated in cells in which MED26 transcripts have been reduced (siRNA). RDH genes were among those genes upregulated when MED26 levels were reduced. Complementary RNA seq from ribo-deplted libraries showed either no change or even a reduction in in transcripts that show increases from the poly A library analysis. Regarding the later the authors comment on the unexpected finding that those transcripts upregulated in the polyA libraries were either unchanged or reduced. Could the authors comment further? I agree the reduction of transcripts is unexpected however one would have expected to see no difference between control and MED26siRNA treated cells. 4) Immuno-precipitation of the complexes using epitope (FLAG)-tagged proteins is powerful and has been exploited extensively by these authors in the past, they are clearly well positioned to interrogate this kind data with caution. Nevertheless, proteomic analyses that affinity purify endogenous proteins would eliminate artefacts induced by expression of epitope tagged proteins versions. For example, the authors affinity purify FLAG-tagged ICE 1 from 293 cells and identify interacting proteins that are enriched for the pull down with anti-flag antibodies. Interacting proteins from mediator were shown (Fig 5) however MED26 was absent from this analysis? Given the preceding data focus on the interaction with MED26 and point to the NTD of MED 26 as the domain responsible for the recruitment of factors necessary to carry out 3'-termination could the authors comment on the absence of MED26 in this analysis? The authors use Ab's raised against endogenous versions of MED26, ICE1, ELL, CPSF73 etc in ChIP assays, have the authors considered proteomic analyses on these rather than FLAG-tagged versions? 5) The authors have published many proteomics data sets using NSAF as a method of providing relative quantitation in their pull-down experiments. I have few issues with spectral counting as a means of relative quantification and the authors are experts, it nevertheless remains a rather imprecise method for protein quantification. The field of relative proteomic quantification is improving rapidly both in technical developments at an instrument level and in software used for analysis. Could the authors provide details of the instrumentation and methods used to generate this data. The references point to two publications from 2001 and 2006, an update in the main body of text or supplementary section would be appreciated. 6) How does NSAF compare with rapidly evolving quantitative methods such as isobaric labelling using tandem mass tags and where multiplex capacity now currently stands at 11 samples in a single run. The statistical power, assessment of variability, ability to generate PCA analysis and assess biological vs batch effects is considerable using such methods. Furthermore, potentially unlimited comparisons can be now achieved by the inclusion of reference channels in the experiments. Could data regarding replicates (biological and technical), statistics/variability be included in the main body of text or supplementary materials. 7) The software used (DTASelect and CONTRAST) were first published published in 2002, what versions were used for these analyses? Other software eg SAINT may provide better precision and alternatively, as alluded to above, isotopic labelling methods such SILAC or isobaric labelling would provide better quantitative accuracy with fewer missing values that could improve the sensitivity and quantitative accuracy of the data.
Reviewer #3 (Remarks to the Author): In their manuscript entitled "The role of Mediator and Little Elongation Complex in transcription termination", Takahashi et al report that the Mediator component, Med26, recruits the Little Elongation Complex (LEC) to replication dependent histone genes (RDH) and regulates their 3' end processing. The LEC is shown to exist in a complex with both 5' cap-binding factors and 3' end processing factors and to localize to Cajal bodies. Depletion or mutation of Med26 or depletion of LEC results in alterations in expression of both RDH and snRNA genes. The current extend previous observations from this group and others. It has been shown previously that the LEC, which is recruited to snRNA genes by Med26, regulates their initiation and elongation. It was also known that 3'end processing of snRNA transcripts is affected by factors that regulate their transcription: replacing an snRNA promoter with an actin promoter in drosophila alters 3'end processing. Thus, the association between recruitment of transcription initiation factors, such as LEC, and 3' processing factors was already predicted. The localization of Med26 and ICE to Cajal bodies was also reported. The novel aspects of this study are: 1) the demonstration that the LEC is recruited to RDH genes, 2) 3' processing of RDH genes is LEC dependent and 3) the LEC is in a complex with 5'cap-binding factors and 3' end processing factors. The authors have provided extensive evidence in support of their model that Med26 recruits the LEC to RDH and snRNA gene promoters, which in turn brings in the 3' processing factors; in the absence of Med26 or LEC, transcription reads through the normal RDH termination sites, leading to aberrant polyA+ addition to the RDH. Overall, the studies are convincing. However, the authors need to address a number of issues, as detailed below. Fig. 1 -• The depletion of Med26 results in an increase in unprocessed transcripts which is statistically significant. However, the actual fraction of UT, relative to CT, is extremely small, less than 1% in many cases. The increases are even less. As reported previously, Med26 results in decreased gene expression. Could those small variations reflect non-specific effects? Internal controls need to be provided, such as the effect on read-through of non-LEC genes, to demonstrate specificity of the effect. The authors argue that this suggests that Med26 recruits LEC which regulates 3' termination. Why is the effect of ICE1 less than Med26 in all cases? Since Med26 also recruits SEC, what is the effect of Med26 KD on read-through of an SEC dependent gene (which is a different question than the effect of AFF4 on an RDH gene).
• KD of ZC3H8 also increases the UT/CT ratio, but except for H1H3E, the ratios are less than 0.1% . Even if statistically significant, what is the biological significance of such a small increase in polyA+histone mRNA? This needs to be discussed in terms of the overall biological significance of the LEC control of 3' end processing. • In Fig. 2c,d the Venn diagram and table presumably summarize the status of polyA+ RDH RNA following KD. However, neither the figure nor the legend so indicate. If this is not the case, the results are inconsistent with previous results. • The majority of ICE1 regulated genes that show an increase in polyA+RNA do they not overlap with Med26. Why? Fig. 3 -• The Figure shows that published ChIP-seq data of ELL and ZC3H8 overlap with Med26 CHiPseq, consistent with Med26 recruiting the LEC to RDH genes. However, the number of genes associated with ZC3H8 is much greater than with ELL. Since ELL is in both SEC and LEC, while ZC3H8 is only in LEC, why are the number of genes associated with ZC3H8 so much greater? Why isn't there a greater overlap of Med26 with ELL and ZC3H8? The authors need to discuss these points.  fig. 4D, clearly shows that the read counts in the gene body are also increased such that the overall distribution of reads does not appear to be different between WT and Mut in histone genes. This increase is clearly apparent in browser views in a) and c). The UT/CT ratio, based on number of reads, should be calculated from the metadata to determine whether there is a significant difference between WT and mutant Med26. The CT calculation should be based on reads downstream of the promoter, not upstream (see below). • In the supplementary figure, a metadata analysis of UT/CT has been done. However, the ratio determined was comparing reads 100 bp upstream to the downstream reads. This would not distinguish between increases of readthrough alone from increases across the gene body. The data should be recalculated comparing reads 50bp downstream of the promoter with readthrough.
• snRNA genes more clearly show increased readthrough. However, the meta • effects are modest. The analysis should be of polyA+ histones, as in Fig. 1. Fig. 5-• Both pull-down gel and mass spec analyses document that ICE1 is recovered in a complex with other LEC, CBCA-NELF and termination factors. However, it is surprising that Med26 is not recovered in either the pull-down gel or mass spec. This brings into question the entire premise of the study that Med26 recruits the LEC. The authors need to explain this. • NSAF is defined in the supplementary figure, but it should be defined in the legend to fig. 5. Fig. 8 -• The role of ICE1 in recruiting HLF and Integrator is assessed by determining the effect of ICE1 knock-down on CPSF73 and Ints9, respectively. However, neither of those proteins appears in the mass spec results shown in Fig. 5. Therefore, the authors need to explain how those two proteins were chosen. Were others that did show up in the mass spec tested? In conclusion, the authors report an interesting study in which 3' processing factors and capbinding factors are in a complex with the LEC, which is recruited to RDH and snRNA genes by Med26, resulting in proper 3' end processing. However, to firmly document their model, the authors need to address two issues: 1) is the observed increased read-through with Med26 KD due to defective 3' end processing or to increases across the gene body which spill over to the 3' end and 2) what is the effect of replacing an RDH promoter with a promoter dependent on the SEC, to determine the effect on 3' processing.
Reviewer #4 (Remarks to the Author): In this report, Takahashi et al reported that Med26 and the Little Elongation Complex (LEC) function in replication-depenent histone (RDH) and snRNA transcription termination. Mechanistically the authors provided evidence that the Med26 recruits LEC, which forms a large complex with various termination factors. Overall the study is well performed and presents potentially novel insights into RDH and snRNA termination as well as Mediator and LEC functions. However, there are some technical issues and their model is not fully substantiated by the data:

1
The reviewers' comments are listed below in boldface black characters, and our pointby-point responses to the comments are shown in regular green characters. I would like the authors to address the following concerns.

Major concerns 1) In figure 7a-d the authors study colocalization of different markers by confocal
microscopy. I have a major concern about the AFF4 staining reported in Fig.7g. As it appears from this figure, the staining is mainly cytoplasmic with no or little nuclear signal. This data supports their conclusion that SEC is not colocalized with Cajal bodies: "but not AFF4, which is an SEC component ( Fig. 7a-g) We thank the reviewer for this comment. To address this suggestion, we quantified the number of particles of ICE1 and MED26 and calculated the intensity of the particles in nuclei of HeLa cells. As shown in the new Fig. 7h, we observed nuclei of 33 HeLa cells and found that 43 of 132 ICE1 particles were colocalized with coilin at Cajal bodies and that the size of ICE1 particles colocalized with coilin was larger than that of 89 particles not colocalized with coilin. In addition, the signal intensity of ICE1 particles colocalized with coilin was much higher than that of particles not colocalized with coilin. Furthermore, as shown in the new Fig. 7i, we found that 59 of 176 MED26 particles in 18 HeLa cell nuclei colocalized with coilin and that the size of MED26 particles colocalized with coilin was larger than that of 117 MED26 particles not colocalized with coilin. In contrast to the ICE1 particles, the intensity of MED26 particles colocalized with coilin was similar to that of MED26 particles not colocalized with coilin. In light of our previous evidence that MED26 plays a role in transcriptional regulation of SEC-targeted genes that encode polyadenylated transcripts, including cmyc, in addition to LEC-targeted genes encoding non-polyadenylated transcripts, it is possible that the MED26 particles not colocalized with coilin are involved in transcriptional regulation of SEC-targeted genes. We agree with the reviewer that future studies addressing the functional consequences of changing the balance between non-polyadenylated and polyadenylated forms of RDH transcripts will be of great interest. However, as the reviewer points out these are likely to be complex, and unraveling these consequences is, we believe, beyond the scope of the present study, which defines new and unanticipated roles of the Mediator and LEC in 3'-processing and termination of non-polyadenylated transcripts.
As the reviewer notes and as we point out in the manuscript (Page 4, lines 21-Page 5, lines 10), one proposed function for polyadenylation of RDH transcripts is to provide a fail-safe mechanism to ensure that Pol II does not read-through into downstream regions. This is potentially important, since RDH loci are gene-dense regions with ample opportunity for read-through transcription from one RDH gene to alter or interfere with expression of nearby genes, including other RDH genes.
Of note, and as also pointed out in the text of both the original and revised manuscript (Page 5, lines 10-12), recent evidence also supports the idea that polyadenylated RDH transcripts contribute to low-level expression of replicationdependent histones outside of S phase and in long lived, terminally differentiated cells, where they have been proposed to serve as a source of replacement histones 1 . Hence, it is tempting to speculate that Mediator-dependent changes in the balance between nonpolyadenylated and polyadenylated forms of RDH transcripts could contribute to the decision to synthesize polyadenylated RDH transcripts in terminally differentiated tissues.
Finally, there is substantial evidence that excessive accumulation of replicationdependent histones outside of S-phase is toxic, and use of non-polyadenylated 5 transcripts is a key mechanism for ensuring that the high levels of Histone mRNAs needed during S-phase disappear during other phases of the cell cycle. Nonpolyadenylated Histone mRNAs are only stable during S-phase, because stem-loop binding protein (SLBP), which is only present during S-phase, binds to the conserved stem loop structure of RDH transcripts to prevent non-polyadenylated RDH mRNAs from undergoing degradation during S-phase 2 . At the end of S-phase, SLBP is phosphorylated by cyclin-dependent kinase and degraded by the ubiquitin proteasome pathway. Degradation of SLBP outside of S-phase is tightly associated with disappearance of Histone mRNAs outside of S-phase 3 . Thus, SLBP plays an important role in ensuring high levels of Histone mRNAs and Histone proteins at S-phase to prevent harmful accumulation of free Histone proteins in cells outside of S-phase 2, 4, 5, 6 .
Considering that SLBP also copurified with FLAG-tagged ICE1 (Revised Fig. 6a) and was previously identified as a component of CBCA-NELF-DSIF 7 , our results are consistent with the possibility that in proliferating cells, LEC recruited by MED26containing Mediator to RDH genes helps to restrict synthesis of RDH transcripts to prevent harmful production of free Histone proteins in cells outside of S-phase. We comment on this possibility in the Discussion section (Page 27, lines 6-22). In addition, auxin treatment led to a significant increase in levels of unprocessed RDH transcripts in MED26-AID-expressing HCT116 cells, but not in parental HCT116 cells (new Fig. 1e). This result suggests that regulation of 3'-processing and termination by MED26 and LEC is at least conserved in other types of human cells. (Page 10, lines 5- 6

15)
Furthermore, MED26 is a metazoan-specific subunit in the Mediator complex and conserved from Drosophila melanogaster to humans, but does not exist in yeast, indicating that the machinery is metazoan-specific mechanisms. In accordance with the reviewer's suggestion, we have mentioned this issue in the Discussion section and stated: "Considering that MED26 is a metazoan-specific subunit of the Mediator complex and conserved from Drosophila melanogaster to humans, but is not present in yeast, we expect that the mechanism uncovered in our study machinery is metazoan- We appreciate the reviewer for raising this issue. As the reviewer pointed out, RDH transcripts upregulated in polyA-selected libraries were either unchanged or reduced in ribo-depleted libraries following knockdown of MED26 (new Fig. 1a). We also found that majority of RDH transcripts upregulated in polyA-selected libraries were reduced in ribo-depleted libraries following acute loss of MED26 in MED26-AID expressing cell line or in MED26 hypomorphic mutant cells (new Fig. 1d, new Fig. S5a). In addition, we found that induction of RDH genes after release of the cells from S-phase was decreased in MED26 hypomorphic mutant cells (new Fig. S5c). These results suggest that MED26 contributes to multiple processes of RDH gene expression, likely including not only transcription termination and 3'-end processing, but also transcription initiation and elongation. We now make this point explicitly on Page 12, lines 10-Page 13, lines 7.  Fig. 5c and Supplemental Fig. S4, respectively) and complemented with sequences from usual contaminants (human 9 keratins, IgGs, proteolytic enzymes). To estimate false positive discovery rates (FDRs), each sequence was randomized keeping amino acid composition and length the same, and the resulting 'shuffled' sequences were added to the 'normal' database (doubling its size) and searched at the same time. Peptide/spectrum matches were sorted and selected NSAF and its newest implementation dNSAF, which takes into account peptides shared between multiple proteins, have been used in hundreds of published manuscripts (from our group and others) as useful abundance measures for straightforward studies of affinity purifications analyzed by mass spectrometry (APMS). In the present manuscript, proteins interacting with ICE1 were readily identified and quantified relatively to the bait protein by comparison against the proteins detected in FLAG negative controls (as shown in the expended version of new Fig. 5c). Costly isobaric labeling, using tandem mass tags for example, is not necessary in cases such as this one, where we are not trying to accurately and precisely measure small changes in protein levels between samples or when multiplexing the LC/MS analysis is not necessary to reduce technical variations. In APMS studies, proteins that copurify with the bait should be among the most abundant proteins that are enriched relative to negative controls (i.e., in negative controls they are either not detected or fewer peptides/spectra are detected).
Such proteins will be ranked as enriched by either label-free (dNSAF) or labeled (TMT) quantitation methods. In summary, the LC/MS analysis performed here didn't need to be quantitative to accomplish the goals of this work, namely, to identify potential interactors that we have followed up on and confirmed by other methods (see Western blots in new Fig. 6a-c).  We appreciate the reviewer's comment. Nevertheless, we think it is worth mentioning that while we agree that the fraction of UT is in many cases is very small even after MED26 knockdown or in MED26 mutant cells, in others, it can reach as high as 10-40% of CT after knockdown.
In accordance with the reviewer's suggestion, we tested the effect of MED26 knockdown on read-through (UT/CT ratio) of two genes that encode polyadenylated transcripts, GAPDH and the well-characterized SEC target, c-myc. As shown in the revised Fig. 1c and Fig. 2a, knockdown of MED26 did not affect the level of unprocessed, read-through transcripts at either GAPDH or c-myc (Page 10, lines 23-Page 11, line 3). Finally, as shown in new Fig. 4h and S6e, we observe that transcription read-through at non-RDH protein-coding genes is the same in wild type and MED26 hypomorphic mutant cells when measured using PRO-seq or Pol II ChIP-seq readthrough ratios. (Page 14, lines 1-13)

2) Knock-down (KD) of ICE1 and Med26, but not the SEC component AFF4, increases UT/CT ratio. The authors argue that this suggests that Med26 recruits LEC which regulates 3' termination. Why is the effect of ICE1 less than Med26 in all cases?
We thank the reviewer for raising this issue. We agree with the reviewer's comments that the effect of ICE1 knockdown on the read-through transcripts of RDH genes is less than the effect of MED26 knockdown. There are several potential explanations for this observation. First, siRNA mediated knockdown may reduce levels of different targets to different extents. Although we have not performed rigorous quantitation of our western blots, it is worth noting that there appears to be more residual ICE1 than MED26 after siRNA treatment (Fig. S2, compare panels a and b). Second, Mediator could in principle affect processes other than LEC recruitment. To address this issue, we performed several new experiments. We found that depletion of MED26 or N-terminal deletion of MED26 decreased the total RDH transcripts in ribo-depleted RNA-seq libraries and the induction of RDH genes at S phase ( Fig. 1a and d, Fig. S5a, Fig. S5c), while ICE1 knockdown did not affect total levels of RDH transcripts (new Fig. S7d). Consistent with these results, knockdown of MED26 decreased the occupancy of LEC component ELL, Pol II, and HLF component CPSF3 at RDH genes ( Fig. 3d and new Fig. 9c). In contrast, knockdown of ICE1 decreased the occupancy of CPSF3 at RDH genes, but did not affect the occupancy of Pol II (new Fig. S7b).
These results suggest that knockdown of MED26 interferes with multiple transcription processes including transcription initiation, elongation, and termination. In contrast, it is likely that knockdown of ICE1 interferes with transcription termination, but not initiation and elongation, of RDH genes. Based on our results, we propose a model that (i) MED26 recruits LEC to the snRNA and RDH genes and plays a role in transcription processes including initiation and elongation, (ii) LEC subsequently binds to CBCA-NELF-DSIF and cooperatively inhibits read-through by Pol II at these genes, and (iii) LEC finally promotes 3'-processing of RDH genes and snRNA genes through recruitment of 3'-end processing factors for RDH genes or Integrator complex, respectively (new Fig. 10). 13 Taken together, it is understandable that the effect of MED26 knockdown was greater than that of ICE1 knockdown, because MED26 knockdown likely affects multiple transcription events including transcription initiation, LEC recruitment, Pol II elongation, CBCA-NELF assembly with LEC, and HLF recruitment around the transcription termination sites. This could be the reason why the transcription termination defect after MED26 knockdown was greater than that after ICE1 knockdown in the RNA-seq analyses shown in Fig. 2a and e. We have added relevant comments to the Results section and stated: "These results suggest that knockdown of MED26 interferes with multiple transcription processes including transcription initiation, elongation, and termination, while knockdown of ICE1 is likely to interfere with transcription termination only, consistent with the result of the polyA-selected RNA-seq showing that the transcription termination defect by MED26 knockdown was greater than that by ICE1 knockdown (Fig. 2a and e)." (Page 21, lines 3-8)

3) Since Med26 also recruits SEC, what is the effect of Med26 KD on read-through of an SEC dependent gene (which is a different question than the effect of AFF4 on an RDH gene).
As We appreciate the reviewer for raising this issue. Poly A-selected RNA-seq analysis using ZC3H8 knockdown cells revealed that knockdown of ZC3H8 increased the polyadenylated RDH genes (new Fig. 2b and c), but the number of RDH genes affected by ZC3H8 knockdown was much less than that affected by knockdown of MED26 or ICE1.
Although ZC3H8 was reported to be a novel component of LEC, our results suggest that ZC3H8 is present in only a fraction of LECs. As shown in revised Fig. 5c, we found that F-ICE1 copurified lower amounts of ZC3H8 than LEC core components including ICE2, ELL, and EAF1. This result raises the possibility that ZC3H8 may act at only some loci or may be less important for LEC function than the more stoichiometric components of the complex. We have added relevant comments to the Results section and stated: "As expected, ICE1 copurified with other LEC components including large amounts of ICE2, ELL, and EAF1. It also copurified with smaller amounts of ELL2, EAF2, and ZC3H8, suggesting that (i) the majority of LEC in these cells is associated with ELL and EAF1 rather than ELL2 and EAF2 and (ii) ICE1, ICE2, and ELL/EAF1 are core components of LEC, while ZC3H8 is associated with only a subfraction of the LEC we have isolated (Fig. 5c). This observation, together with the fact that the number of genes associated with ZC3H8 was much greater than the number associated with ELL in ChIP-seq analyses (Fig. 3c), raises the possibility that ZC3H8 has functions outside of LEC. " (Page 15, lines 17-24) The reviewer's question about the overall biological significance of LEC control in 3'-end processing of RDH genes is an important one that will require further study. We would argue that even if the effect of MED26 or LEC subunit depletion on termination and processing is small, at others it is quite substantial -for example, ZC3H8 depletion increases the level of unprocessed HIST1H3E from ~10% -~20% of total. It has been shown that non-polyadenylated Histone mRNAs are only stable during S-phase to prevent harmful production of free Histone proteins in cells outside of S- In support of this notion, the results of our plasmid-based assay suggest that the promoter of RDH genes is required for proper transcription termination as well as inhibition of Pol II read-through at RDH genes and aberrant Histone protein production.
In our plasmid-based assay, we exchanged the HIST1H1C promoter with the Cytomegalovirus (CMV) promoter, a well-characterized viral promoter that supports transcription of polyadenylated transcripts. As shown in the new Fig. S8a, we generated two kinds of plasmids that contained the HIST1H1C promoter or CMV promoter, the HIST1H1C gene region from TSS to TES, and the region from TES to 100 bp downstream of TES that has a polyadenylation site (PAS). In addition, we added a FLAG-tag sequence at the C-terminus of the HIST1H1C gene to distinguish the protein from endogenous Histone H1. We transiently transfected the plasmids and compared the levels of total transcripts, read-through transcripts, and proteins. As shown in the new To discuss this issue, we have added relevant comments to the Results section and stated: "Our results provided hints toward answering the important question of the overall biological significance of Mediator and LEC regulation in 3'-end processing of 16 RDH genes. Non-polyadenylated Histone mRNAs were shown to be regulated to be stable only during S-phase to prevent harmful production of free Histone proteins in cells outside of S-phase, because excess histone levels can lead to cytotoxicity through multiple mechanisms 9,47,48,49,50 . Degradation of SLBP by the ubiquitin proteasome pathway outside of S-phase is associated with disappearance of Histone mRNAs outside of S-phase 9,51 . Considering that SLBP also copurified with ICE1 (Fig. 6a)  6) The majority of ICE1 regulated genes that show an increase in polyA+RNA do they not overlap with Med26. Why?
We thank the reviewer for raising this issue. There is a report that ICE1 functions in not only transcriptional regulation, but also other cellular functions including nonsensemediated decay (NMD) of mRNAs 14 . Thus, ICE1 depletion could affect not only transcription of RDH and snRNA genes, but also NMD of mRNAs from other genes.
This could be a reason why knockdown of ICE1 affected the expression of mRNAs that are not regulated by MED26. In the Results section of the revised manuscript, we state: "We note that many transcripts upregulated by ICE1 knockdown were not affected by MED26 knockdown. It was recently shown that ICE1 has a role in nonsense-mediated decay (NMD) of mRNAs outside the context of LEC 38 ; thus, it is possible that ICE1 knockdown affects not only transcription of RDH and snRNA genes but also NMD, and 18 perhaps other functions, at other genes. " (Page 11, lines [11][12][13][14][15] Our microscopic analyses provided data consistent with the possibility that ICE1 could have LEC-independent functions outside of the nucleus. In particular, when we quantified the number of ICE1 particles stained by anti-ICE1 antibodies, we found that 43 of 132 ICE1 particles in 33 cells were colocalized with coilin at Cajal bodies, and 89 ICE1 particles were not colocalized at Cajal bodies in nuclei, while 130 ICE1 particles in these cells were present in the extranuclear area and had smaller sizes than the particles colocalized with coilin in nuclei (new Fig. 7h). We have added the following sentences to the Results section: "Intriguingly, we observed the extranuclear regions of 33 cells and found that 130 ICE1 particles were present in the extranuclear area and that these particles had a much smaller size than particles colocalized with coilin in nuclei. Considering a recent report that ICE1 plays a role in NMD of mRNAs 38 , it is possible that ICE1 has a role in NMD of mRNAs in the extranuclear region. " (Page 19, lines 4-8) As the reviewer pointed out, the number of genes associated with ZC3H8 peaks was much greater than the number associated with ELL. Whether this is because ELL is really associated with a surprisingly small number of genes or because the available ELL antibodies are of relatively poor quality for ChIP remains to be determined. We think that what is more noteworthy is that snRNA and RDH genes are among the population of genes that indeed do show good overlap.
We also note that the ChIP-seq data for MED26 were obtained from HEK293T cells and we performed this analysis ourselves, while the ChIP-seq data for ELL and ZC3H8 were from published ChIP-seq analyses done by others and derived from HCT116 cells 15 . To compare the ChIP-seq peaks of MED26, ELL, and ZC3H8 using the same cell line, we performed ChIP-seq analysis of ELL and ZC3H8 using HEK293T cells, similar to our experiments for MED26 ChIP-seq. We found ChIP-seq peaks of ELL in a subset of snRNA and RDH genes, and the results were similar to those obtained with the published data from HCT116 cells. However, it was difficult for us to carry out ChIP of ZC3H8 using the commercially available antibody (ab113260; Abcam). Thus, we had difficulty in comparing the ChIP-seq peaks of MED26, ELL, and ZC3H8 using HEK293T cells. As mentioned in our response to Comment #2) of Reviewer #2, levels of polyadenylated read-through transcripts of RDH genes were significantly increased by depletion of MED26 in both HCT116 and HEK293T cells (revised Fig. 1a-e). This result indicates that regulation of transcription termination by MED26 and LEC is at least conserved in different types of human cells. Taken together, it is convincing to compare the ChIP-seq data derived from HEK293T cells and HCT116 cells.
As mentioned in our response to Comment #4), ZC3H8 was reported as a novel component of LEC, although our result suggests that ZC3H8 is one of the subcomponents of LEC. As shown in revised Fig. 5c, we found that F-ICE1 copurified with a smaller amount of ZC3H8 than LEC core components including ICE2, ELL, and EAF1. This result raises the possibility that ZC3H8 is a subcomponent of LEC and it is possible that ZC3H8 has a different role from LEC in cells. We have added relevant comments to the Results section and stated: "As expected, ICE1 copurified with other LEC components including large amounts of ICE2, ELL, and EAF1. It also copurified with smaller amounts of ELL2, EAF2, and ZC3H8, suggesting that (i) the majority of LEC in these cells is associated with ELL and EAF1 rather than ELL2 and EAF2 and (ii) ICE1, ICE2, and ELL/EAF1 are core components of LEC, while ZC3H8 is associated with only a subfraction of the LEC we have isolated (Fig. 5c). This observation, together with the fact that the number of genes associated with ZC3H8 was much greater than the number associated with ELL in ChIP-seq analyses (Fig. 3c We thank the reviewer for raising this issue. In accordance with the reviewer's suggestion, we calculated a "PRO-seq read-through ratio". The "PRO-seq read-through ratio" was defined as "sum of reads from 500 bp to 1000 bp downstream of transcription end site (TES)" divided by "sum of reads from TES to 50 bp upstream of We thank the reviewer for raising this issue. As mentioned in our response to Comment #8), we calculated the "PRO-seq read-through ratio". The "PRO-seq read-through ratio" was defined as "sum of reads from 500 bp to 1000 copurified with Mediator subunits including MED26 and MED1 (Revised Fig. 6a), consistent with previous data showing that MED26 interacts with SEC 8,9 . These results showed that ICE1 interacts with MED26-containing Mediator. As the reviewer pointed out, we did not detect MED26 in F-ICE1 immunoprecipitates in our proteomics analysis ( Fig. 5b and revised Fig. 5c). Because protein modifications, especially phosphorylation, interfere with ionization of peptides and identification of proteins by mass spectrometry, one possibility is that protein modifications such as phosphorylation make it difficult to detect MED26 by mass spectrometric analysis of F-ICE1-associated proteins.  14) In conclusion, the authors report an interesting study in which 3' processing factors and cap-binding factors are in a complex with the LEC, which is recruited to RDH and snRNA genes by Med26, resulting in proper 3' end processing.

12) NSAF is defined in
However, to firmly document their model, the authors need to address two issues: 1. Is the observed increased read-through with Med26 KD due to defective 3' end processing or to increases across the gene body which spill over to the 3' end?
We thank the reviewer for raising this issue. To address the reviewer's question, we calculated Pol II read-through ratios using Pol II ChIP-seq data from wild-type HEK293T (WT) cells and MED26 hypomorphic mutant HEK293T (MUT) cell lines.
We defined "Pol II read-through ratio" as "sum of Pol II reads from TES to 1000 bp downstream of TES" divided by "sum of Pol II reads from transcription start site (TSS) to 1000 bp downstream of TES". As shown in the right panel of new Fig. 4h

What is the effect of replacing an RDH promoter with a promoter dependent on
the SEC, to determine the effect on 3' processing.
We thank the reviewer for this suggestion. In general, it is technically difficult to replace the endogenous promoter of the HIST1H1C gene with the promoter of, for example, the c-myc gene, as one of the SEC target genes, in cells, even if we take advantage of the CRISPR system. Instead, we used a plasmid-based assay to examine the effect of promoter exchange in cells. We used the Cytomegalovirus (CMV) promoter, a wellcharacterized viral promoter that supports transcription of polyadenylated transcripts, because the c-myc gene promoter is much longer than the HIST1H1C promoter and it is not easy to define the minimum c-myc gene promoter region sufficient to regulate transcription.
As shown in the new Fig. S8a, we generated two kinds of plasmids that each contained the HIST1H1C promoter or CMV promoter, the HIST1H1C gene from TSS to TES, and the region from TES to 100 bp downstream of TES that had a polyadenylation site (PAS). The CMV promoter is generally used to express proteins of polyadenylated genes. In addition, we added a FLAG-tag sequence to the C-terminus of the HIST1H1C gene in plasmids to distinguish the protein from endogenous Histone H1. We transiently transfected the plasmids and compared the levels of total transcripts, read-through transcripts, and proteins. As shown in the new Fig. S8b and c (Revised Fig. 6a), consistent with previous data showing that MED26 interacts with SEC 8,9 . These results showed that ICE1 interacts with MED26-containing Mediator. As the reviewer pointed out, we did not detect MED26 in F-ICE1 immunoprecipitates in our proteomics analysis ( Fig. 5b and revised Fig. 5c). Because protein modifications, especially phosphorylation, interfere with ionization of peptides and identification of proteins by mass spectrometry, our results raise the possibility that protein modifications such as phosphorylation make it difficult to detect MED26 by mass spectrometric analysis.
2) Instead of MED26 recruits heat labile factors to promote processing of RDH transcripts to form unpolyadenylated mRNAs, is it possible that MED26 inhibits the cleavage/polyadenylation pathway?
We appreciate the reviewer for raising this important point. As shown in revised Fig. 5c, FLAG-tagged ICE1 copurified with FLASH, which is a specific component of Heat labile factor (HLF) and is thought to be not present in the protein complex involved in the cleavage/polyadenylation pathway of polyadenylated genes 17 . This result suggests that LEC recruited by MED26 interacts with HLF and promotes 3'-processing of RDH genes, but might not inhibit the cleavage/polyadenylation pathway of other proteincoding genes. In addition, RT-PCR analyses indicate that cleavage at the c-myc and GAPDH genes is the same in cells expressing wild type MED26 or the MED26 hypomorphic mutant (Revised Fig. 2a). PRO-seq results also suggest that read-through 26 transcription at c-myc gene and numerous other protein-coding genes is not affected in MED26 hypomorphic mutant cells (new Fig. 4g and h). This result also supports our notion that MED26 does not inhibit the cleavage/polyadenylation pathway of normally polyadenylated transcripts.
3) Knockdown of ICE1 led to a decrease in the recruitment of CPSF73 and Ints9 (Fig. 8) and the authors argued that this is the evidence that LEC helps to recruit termination factors. Does Med26 knockdown also have the same effect?
We appreciate the reviewer raising this important point. To address the reviewer's question, we performed ChIP assays of CPSF3 (CPSF73) and INTS9 using HEK293T cells in which MED26 was knocked down. As shown in the new Fig. 9c and d, we found that knockdown of MED26 decreased the occupancy of CPSF3 (CPSF73) and INTS9 at RDH and snRNA genes, respectively. This result is consistent with our notion that LEC recruited to RDH or snRNA genes by MED26 plays a role in recruitment of HLF or Integrator to RDH or snRNA genes, respectively.