Mitotically heritable, RNA polymerase II-independent H3K4 dimethylation stimulates INO1 transcriptional memory

Read the full article See related articles


For some inducible genes, the rate and molecular mechanism of transcriptional activation depends on the prior experiences of the cell. This phenomenon, called epigenetic transcriptional memory , accelerates reactivation and requires both changes in chromatin structure and recruitment of poised RNA Polymerase II (RNAPII) to the promoter. Memory of inositol starvation in budding yeast involves a positive feedback loop between transcription factor-dependent interaction with the nuclear pore complex and histone H3 lysine 4 dimethylation (H3K4me2). While H3K4me2 is essential for recruitment of RNAPII and faster reactivation, RNAPII is not required for H3K4me2. Unlike RNAPII-dependent H3K4me2 associated with transcription, RNAPII-independent H3K4me2 requires Nup100, SET3C, the Leo1 subunit of the Paf1 complex and, upon degradation of an essential transcription factor, is inherited through multiple cell cycles. The writer of this mark (COMPASS) physically interacts with the potential reader (SET3C), suggesting a molecular mechanism for the spreading and re-incorporation of H3K4me2 following DNA replication.

Article activity feed

  1. Evaluation Summary:

    Transcriptional memory is a phenomenon via which certain genes are activated more robustly in response to repeated stimulation and in this manner, are able to "remember" previous experiences. This report dissects the molecular mechanism of inositol-driven transcriptional memory and highlights the key role of the histone mark H3K4 di-methylation, which is deposited independently from RNA Polymerase II activity. This memory-specific H3K4 di-methylation is found to be inherited over multiple cell divisions and to require specific transcription factors and chromatin machinery components to be established and maintained. The work will be of interest to those studying transcriptional regulation and epigenetics.

    (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. The reviewers remained anonymous to the authors.)

  2. Reviewer #1 (Public Review):

    This paper presents a large number of significant findings that extend understanding of the molecular mechanism of transcriptional memory operating at the yeast INO1 gene, wherein an episode of derepression by inositol starvation sets in motion a series of events that poise the INO1 promoter, following its repression by re-supplying inositol (the memory phase), for more rapid derepression in a subsequent starvation (re-activation) episode. The Brickner lab has dissected this mechanism extensively in the past. Here they conduct an extensive program of well-executed experiments that lead to several new insights. (i) They confirm that memory leads to more rapid INO1 derepression during reactivation, but observing this requires anchor-away (AA) of the repressor Opi1, and they show that Nup100-dependent memory increases competitive cell fitness during reactivation conditions. (ii) They show that memory (as assayed by peripheral nuclear localization of INO1) does not require ongoing transcription by examining the rpb1-1 Pol II mutation and deletion of the INO1 TATA element. They show by ChIP experiments that deposition of H3K4me2 at the INO1 promoter during memory requires nuclear pore factor Nup100 and the memory-specific TF Sfl1, and that Sfl1 binding to the cis-acting MRS signal at INO1, required for memory, is dependent on Nup100, the Set3 subunit of the HDAC Set3C, and the Swr1 subunit of the SWR complex required for H2AZ deposition, that both Sfl1 and Set3 are required for H2AZ deposition at INO1, and that AA of COMPASS (required for H3K4 methylation) impairs Sfl1 binding at INO1. These findings support a positive feedback loop wherein Sfl1/Nup1 dependent association with the NPC leads to H3K4me2 by COMPASS, which supports H2AZ deposition in the INO1 promoter, and these chromatin modifications, in turn, reinforce Sfl1 occupancy during memory. They also suggest a role for a reader of H3K4me2, Set3, in memory in addition to writer, COMPASS. (iii) The TF Hms2, related to Sfli, is also required for poised Pol II and H3K4me2 during memory, but unlike Sfl1 binds to the promoter during conventional activation as well as during reactivation, and is required for Sfl1 binding during reactivation. Interestingly, Hms2 is even more important than Sfl1 for retention of INO1 at the nuclear periphery during memory. (iv) Using an analog-sensitive allele of the Ssn3 kinase of the Cdk8 complex, they show that Ssn3 kinase activity is crucial for rapid INO1 induction and increased cell fitness during reactivation, but not during conventional activation, acting to enhance poised Pol II binding but being dispensable for Set3-dependent H3K4me2 deposition. Interestingly, AA of TBP does not impair H3K4me2 during memory, unlike during conventional activation where it is co-transcriptional. (vi) The Leo1 subunit of Paf1C is identified as the only subunit of this complex that is specifically required for H3K4me2 deposition during memory but not conventional activation, and its deletion leads to reductions in poised PolII, INO1 mRNA induction kinetics, cell fitness, and nuclear periphery localization, during memory. (vii) Using an SFL1-AID allele for regulated depletion of Sfl1 during different stages of memory formation, they provide evidence that once memory has been established, H3K4me2 can persist for up to two hours in the absence of Sfl1, implying that Sfl1 is needed primarily to establish the memory state. Other experiments carried out using a URA3 gene with an MRS inserted in the promoter, which is sufficient for localization to the nuclear periphery and attendant H3K4me2 deposition but not for other aspects of transcriptional memory, they provide evidence that H3K4me2 can persist through about 4 mitoses following depletion of Sfl1, suggesting its epigenetic inheritance; and they demonstrate physical association of COMPASS and Set3, which they propose underlies the mechanism of epigenetic inheritance.

    The experiments are very well conceived and executed and the data support the main conclusions of the paper. The findings are significant in providing evidence that a positive feedback loop exists between Sfl1-dependent interaction with the nuclear pore complex and H3K4me2 deposition in the INO1 promoter, which is essential for recruitment of poised Pol II, but which does not require Pol II or transcription initiation to occur. This distinguishes this activity of COMPASS from conventional H3K4 methylation, which is coupled to transcription, and the Pol II-independent mechanism is shown to specifically require Nup100, SET3C, and the Leo1 subunit of the Paf1 complex. It is also significant that this specialized H3K4me2 deposition can persist through multiple cell cycles, which might be enhanced by physical association between the writer of this mark (COMPASS) and a SET3C as a presumptive reader.

    Some additional discussion will be required to explain why nuclear depletion of Opi1 is required to observe that memory stimulates the rate of INO1 derepression during reactivation, and a new experiment is likely needed to show that the persistence of H3K4me2 deposition following depletion of Sfl1 will be diminished in cells lacking Set3 to bolster the model that the proposed epigenetic inheritance of H3K4me2 requires Set3C as a reader of this histone mark.

  3. Reviewer #2 (Public Review):

    The manuscript by Sump et al. investigates multiple molecular aspects of epigenetic transcriptional memory of inositol exposure and characterizes a transcription-independent H3K4 di-methylation pathway that is maintained over cell divisions and that appears to underlie the memory phenomenon.

    Initially, the authors show that transcriptional memory of INO1 gene confers a fitness advantage, which is dependent on a nuclear pore protein Nup100, and that inactivating RNAP II or the INO1 promoter still leads to memory-associated re-localization of the INO1 gene to the nuclear periphery. This Nup100-dependent re-localization occurred very rapidly upon inositol-driven activation, before significant transcription occurred, reinforcing the conclusion that memory mechanisms are separate from transcription. The authors also identify a positive feedback loop between the key players of this memory pathway, such that deposition of H3K4Me2 and histone variant H2AZ during memory depend on Nup100 and transcription factor Sfl1 yet binding of Sfl1 to the promoter during memory, in turn, depends on H3K4Me2 and Nup100. In addition to Sfl1, authors identify an additional transcription factor with a similar DNA-binding domain, Hms2, which is required for INO1 memory and for maintenance of H3K4Me2 and RNAP II during memory.

    The authors construct an elegant system to conditionally inhibit Ssn3, the Cdk8 kinase that has been found to bind RNAP II PIC specifically during memory. Using this inhibition, they show that inhibiting Ssn3 leads to loss of poised RNAP II from INO1 promoter during memory, but not during activation. Importantly, they find that decreasing RNAP II at the promoter during memory or activation does not affect memory-associated H3K4Me2. These results illustrate that the memory H3K4Me2 pathway does not require RNAP II, and that this mark is the more upstream and primary mediator of memory. To further characterize the memory H3K4Me2 pathway, the authors examine the Paf1 complex and identify Leo1 as a critical factor. Loss of Leo1 leads to loss of H3K4Me2 specifically during memory, as well as to loss of RNAP II and peripheral localization, and to a slower rate of reactivation. Interestingly, the loss of peripheral localization is delayed, suggesting that Leo1 is important for maintaining memory but not for establishing it.

    To further understand mitotic heritability of H3K4Me2, the authors use an ectopic insertion of the MRS element in a different location. After Sfl1 degradation, the MRS was found to retain H3K4Me2 for 4 generations, which is much longer than the presence of RNAP II-dependent H3K4 di-methylation or than what is expected of just passive dilution due to replication. These dynamics suggested that there may be an active mechanism of maintaining H3K4Me2 at memory-marked promoters.

    Overall, this is a thorough and elegant analysis of transcriptional memory mechanisms, which introduces significant new insight into our understanding of what serves as the primary memory mark. Multiple factors have been implicated in memory in this and other systems, and this work highlights the importance of H3K4 di-methylation, as opposed to RNAP II poising, H2AZ incorporation, and perhaps most importantly, transcription itself. The existence of RNAP II-independent H3K4 di-methylation pathway is demonstrated convincingly and is an important finding for the field. I have a few suggestions for experiments that can make the mechanistic conclusions more convincing or that expand the mechanism a bit more.

  4. Reviewer #3 (Public Review):

    Sump et al. use elegant time dependent perturbation tools to dissect the molecular mechanisms of INO1 memory. Previous work has uncovered several components that are responsible for both gene activation and spatial repositioning of the INO1 locus to the nuclear periphery. Sequence dependent GRS and MRS sequences that are portable enable the relocalization, activation and the establishment of memory. The system is ideally suited for a mechanistic dissection of how transcriptional memory is both established and maintained. Using a series of anchor away and auxin inducible degradation approaches, the authors demonstrate convincingly that the INO1 system exhibits memory that enhances cell fitness in budding yeast under competitive growth conditions (Figure 1). The authors also show that H3K4me2 is present at the INO1 locus even under conditions where RNA polII is absent highlighting a parallel pathway that contributes to placing this activating mark independent of transcription. The maintenance of H3K4me2 given that it is transcription independent (as per the authors) must arise from an autonomous positive feedback loop that depends on COMPASS-SET3C read-write activity. The manuscript does not provide sufficient data to provide strong support for this claim. Remarkably the authors also highlight the existence of at least six transcription factors that regulate INO1 activation and memory. Whether the interplay between their binding, dissociation and persistence contributes to memory remains a distinct possibility.