Dynamic epistasis analysis reveals how chromatin remodeling regulates transcriptional bursting

Transcriptional bursting has been linked to the stochastic positioning of nucleosomes. However, how bursting is regulated by the remodeling of promoter nucleosomes is unknown. Here, we use single-molecule live-cell imaging of GAL10 transcription in Saccharomyces cerevisiae to measure how bursting changes upon combined perturbations of chromatin remodelers, the transcription factor Gal4 and preinitiation complex components. Using dynamic epistasis analysis, we reveal how the remodeling of different nucleosomes regulates transcriptional bursting parameters. At the nucleosome covering the Gal4 binding sites, RSC and Gal4 binding synergistically facilitate each burst. Conversely, nucleosome remodeling at the TATA box controls only the first burst upon galactose induction. At canonical TATA boxes, the nucleosomes are displaced by TBP binding to allow for transcription activation even in the absence of remodelers. Overall, our results reveal how promoter nucleosome remodeling together with Gal4 and preinitiation complex binding regulates transcriptional bursting.

The transcription of many genes occurs in stochastic bursts of transcriptional activity, interspersed by periods with no transcriptional activity. To achieve the correct transcriptional output of a bursting gene, the regulatory factors may modulate when a burst starts (burst frequency), when it ends (burst duration) and the rate of polymerase loading during a burst (initiation rate) [1][2][3][4][5][6][7] . Yet, it remains elusive how each of these steps is controlled. Previous studies from budding yeast suggested a link between the bursting and chromatin structure 4,[8][9][10][11][12] . Inference of transcriptional parameters from mRNA or protein distributions suggests that bursting is affected by mutations in chromatin regulators 8,13,14 . In addition, single-cell mapping of nucleosome conformations at the PHO5 promoter suggests stochastic transitions between different promoter configurations, of which only some may be permissive for transcription 12,15 . However, how different promoter nucleosome configurations affect the dynamics of transcription in living cells is unexplored.
The positioning of nucleosomes throughout the genome is controlled by chromatin remodeling enzymes. For the promoter regions, the most important remodelers are RSC and SWI/SNF, which together maintain a promoter architecture consisting of a nucleosome-depleted region (NDR) flanked by two nucleosomes referred to as the +1 and −1 nucleosome [16][17][18] . Upon depletion of RSC, the +1 nucleosome shifts into the NDR for 70% of all genes 19,20 . At highly expressed genes, SWI/SNF acts redundantly with RSC to maintain the NDR 17,18 . In addition, RSC also regulates partially unwrapped unstable nucleosomes, referred to as fragile nucleosomes, that are often found in the promoters of highly expressed genes, such as GAL10 (refs. 21-25). The interplay between different remodelers at the same promoter is a dynamic process. Individual remodelers remain chromatin bound for just a few seconds and multiple remodelers can sequentially occupy the same promoter within minutes 26 .
Nucleosomes affect different steps in the transcription activation process. For example, nucleosomes inhibit transcription factor Article https://doi.org/10.1038/s41594-023-00981-1 To understand how this perturbed chromatin structure upon RSC depletion affects GAL10 transcription dynamics, we used the PP7 technology to directly monitor the transcription in individual cells in real-time ( Fig. 1c and Supplementary Video 1). In short, 14 PP7 repeats were introduced endogenously in the 5ʹUTR of GAL10 and, upon transcription, the PP7 RNA stemloops were bound by PP7-coat protein fused to GFP-Envy 35 . Using widefield fluorescence microscopy, the accumulation of RNAs at the transcription site (TS) was visualized as a bright spot in the nucleus ( Fig. 1d and Supplementary Video 1), of which the intensity was tracked over time (Fig. 1e,f). From these intensity traces, the parameters of transcriptional bursting were determined: the active fraction (the fraction of cells that shows a TS within 1 h after galactose addition), the induction time (the time between galactose addition and the first burst), the burst duration, the time between bursts (as a measure for burst frequency) and the burst intensity. The burst size, defined as the total number of RNAs produced during a burst, is dependent on the burst duration and the burst intensity. Upon RSC depletion, we observed an increased induction time (Fig. 1g) and time between bursts (Fig. 1h), whereas the active fraction, burst duration and burst intensity showed minor or no changes (Extended Data Fig. 3a,h-j). Using single-molecule fluorescence in situ hybridization (smFISH), we validated that the effect of RSC depletion on steady-state transcription did not vary between the different cell-cycle stages (Extended Data Fig. 2c). The RSC remodeling of the fragile GAL10 promoter nucleosomes at the UASs and the TSS is thus correlated with changes in the induction time and the start of each burst.

SWI/SNF remodeling at the TATA regulates the induction time
Promoter nucleosomes at highly expressed genes are remodeled redundantly by RSC and SWI/SNF 17,18 . Therefore, we addressed how SWI/SNF remodeling affects the nucleosome positioning and transcription dynamics at GAL10 by nuclear depletion of Swi2, the catalytic subunit of SWI/SNF (Extended Data Fig. 2a,b). Consistent with previous reports, MNase-seq upon SWI/SNF depletion showed no changes in the genome-wide promoter nucleosome architecture (Extended Data Fig. 1h-k). In contrast, for GAL10, the coverage of stable nucleosomes within the NDR and specifically at the TATA was increased (Fig. 2a). No change in GAL10 fragile nucleosomes was observed (Extended Data Fig. 3k,l). At the level of transcription, depletion of SWI/SNF resulted in an increased induction time ( Fig. 2d) but had a minor effect on other GAL10 transcriptional parameters ( Fig. 2g and Extended Data Figs. 2c and 3b-e,h-j). The higher nucleosome coverage within the NDR and at the TATA upon SWI/SNF depletion thus affects induction time, but once the cells are activated, this increased coverage does not affect transcription.

RSC and SWI/SNF remodeling synergically affect transcription
Because of their redundancy 17,18 , the effect of simultaneous RSC and SWI/SNF depletion on both the chromatin structure and transcription dynamics is expected to be larger than the combined effect of their individual depletions. To test this, we simultaneously depleted both RSC and SWI/SNF (Extended Data Fig. 2a,b). As expected, a larger effect on stable nucleosomes was observed than for either single depletion, both genome wide (Extended Data Fig. 1l,m) and at the GAL10 locus (Fig. 2b,c), where the TSS nucleosome moved further into the NDR than for either single depletion. However, the coverage at the GAL10 TATA was similar to the coverage in the single SWI/SNF depletion, and the effect on fragile nucleosomes mimicked the effect of single RSC depletion (Extended Data Fig. 1n,o). RSC and SWI/SNF are thus partially redundant and have different functions to remodel the promoter nucleosomes.
To interpret the effect of RSC and SWI/SNF double depletion, we performed epistasis analysis on each dynamic parameter of transcription and called this method dynamic epistasis analysis. Dynamic epistasis analysis is based on classical epistasis analysis, a method to interpret phenotypic growth defects from genetic interactions that (TF) binding and change the residence time of TFs on nucleosomal DNA 4,[27][28][29][30] . TF binding to nucleosomal DNA is facilitated by remodelers, as a loss of RSC reduces TF occupancy and binding frequency, and increases TF residence time 8,22 . Additionally, movement of the +1 nucleosome in RSC-depleted cells increases nucleosome coverage of the TATA element (TATA) and transcription start site (TSS), which affects TBP binding, preinitiation complex (PIC) assembly and transcription 19,20,31,32 . However, how each of these mechanisms influences transcriptional bursting is unclear.
In this study, we dissected the role of promoter nucleosome remodeling in regulating transcriptional bursting. We acutely depleted RSC and SWI/SNF and used single-molecule live-cell imaging to measure the changes in transcription dynamics at the GAL10 gene in budding yeast. To decipher the regulatory mechanisms of remodeling at specific promoter nucleosomes, we combined remodeler depletion with perturbations of the TF Gal4, PIC components and histones, and analyzed the effect of these single and combined perturbations on each dynamic parameter of transcription using dynamic epistasis analysis. We found that the fragile nucleosome at the Gal4 binding sites is controlled by the redundant action of RSC and Gal4 binding to facilitate consecutive bursts of transcription. The nucleosomes around the TATA and TSS are remodeled in a partially redundant manner by RSC and SWI/SNF to allow for the first burst of transcription after activation. In addition, our results revealed that TBP competes with nucleosomes at the TATA to enable transcription in the absence of chromatin remodelers. Overall, our study exposed how remodeling at different promoter nucleosomes controls the accessibility of the DNA for binding of TFs and the PIC, and how this affects different kinetic parameters of transcriptional bursting.

Promoter nucleosome remodeling by RSC affects each burst
Upon addition of the sugar galactose to yeast cells, nucleosomes in the promoter region of GAL10 are remodeled to activate transcription 33 . Micrococcal nuclease digestion with deep sequencing (MNase-seq) with high and low MNase concentrations showed the coverage of stable and fragile nucleosomes, respectively, in inactive and active conditions 23 (Extended Data Fig. 1a-c). As reported previously 4,22,33 , in transcriptionally inactive conditions (raffinose), the GAL10 promoter region showed three nucleosomes: a fragile nucleosome at the Gal4 upstream activation sequences (UASs) and two stable nucleosomes at the TATA and TSS. Upon activation with galactose, the TATA nucleosome was evicted, and the TSS nucleosome was moved downstream, creating an NDR. Consistent with previous findings at wide NDRs, the fragile nucleosome at the UASs remained present 21-23 .
To dissect how remodeling at each of these three nucleosomes regulates transcriptional bursting, we conditionally depleted the chromatin remodelers from the nucleus, mapped the effect on promoter nucleosome positioning and linked this to the effect on transcription dynamics. First, we depleted RSC, an important nucleosome remodeler controlling both the stable and fragile nucleosomes in promoter regions 19,20,22 . To map the changes in nucleosome positions, we performed MNase-seq in a galactose-rich media in cells where Sth1, the essential catalytic subunit of RSC, was depleted from the nucleus for 60 min using anchor-away 34 . Sth1 depletion was confirmed by a lack of growth on the rapamycin-containing plates (Extended Data Fig. 2a) and imaging of the mScarlet-anchor-away-tagged Sth1 (Extended Data Fig. 2b). In line with previous studies, the RSC depletion led to a fill-in and shortening of NDRs genome wide (Extended Data Fig. 1d,e) 17,19 . We also observed a lower coverage of the fragile nucleosomes in the promoter regions (Extended Data Fig. 1f,g), in agreement with the fragile nucleosomes representing RSC-bound, partially unwrapped nucleosomal intermediates 22 . At the GAL10 locus, RSC predominantly regulated fragile nucleosomes, with a lower coverage over the Gal4 UASs and TSS (Fig. 1a). In addition, a small shift in the TSS nucleosome was observed (Fig. 1b).
Article https://doi.org/10.1038/s41594-023-00981-1 allows one to determine whether perturbed factors act in the same or different pathways 36 . Dynamic epistasis analysis compares, for each dynamic parameter of transcription, the observed effect of a double perturbation to the expected effect (the product of the fractional changes in this parameter observed in the individual perturbations). If two perturbations act independently, the double perturbation follows the expected effect (Fig. 2e). Observing a larger effect than expected indicates that the perturbed factors act on the same process and are (partially) redundant, whereas a smaller effect than expected indicates that the factors act in the same pathway or have opposing biochemical functions 36 . For RSC and SWI/SNF double depletion, a larger effect than expected was observed for GAL10 induction time and time between bursts (Fig. 2d,f,g). In addition, the burst duration and intensity was shorter than expected (Extended Data Fig. 3h-j). As the redundancy between RSC and SWI/SNF was previously established 17,18 , the observed synergy for RSC and SWI/SNF depletion validates our approach to identify functional epistatic relationships using transcription dynamics. Moreover, we observed a variation between strains that should theoretically behave the same, such as longer induction times for SWI/ SNF and RSC and SWI/SNF than for RSC in nondepleted conditions (Fig. 2d). Such variation could be caused by the anchor-away tags affecting protein function, experiment-to-experiment variation, or additional off-target mutations, although the effects of the latter source were minimized by including 2-3 independent biological replicates. Importantly, dynamic epistasis analysis circumvented this variation and allowed extraction of the specific effects of the perturbations. Overall, these results showed that remodeling of the fragile nucleosomes at the UASs by RSC and nucleosome displacement around the TATA and the TSS by RSC and SWI/SNF are associated with synergistic changes in the induction time, time between bursts and burst size.

Remodeling regulates multiple gene-activation steps
To understand how promoter nucleosome remodeling affects the kinetics of transcription activation, we estimated the number of rate-limiting steps from the induction time distributions from the shape parameter k of a Gamma fit. In the presence of remodelers, we found ~6 rate-limiting steps during activation (Extended Data Fig. 4h,n,t). Single or double depletion of RSC and SWI/SNF did not lead to a consistent change in k (Extended Data Fig. 4i,o,u), suggesting that the assumption that each step has an equal rate (underlying the Gamma distribution) is no longer valid. In addition, the high number of steps suggested that multiple rate-liming steps in the signaling pathway dominate and obscure any remodeler-specific steps of transcription activation. To expose remodeling-dependent activation steps, we eliminated signaling steps by pre-exposure with galactose, repression with glucose-containing media, and subsequent re-induction with galactose (Extended Data Fig. 4a). Consistent with previous reports, this re-induction is faster than the initial induction (Extended Data Fig. 4b,c,j,p,v), due to transcriptional memory 33,37 . The re-induction times showed a single rate-limiting step, even when RSC or SWI/SNF was depleted (Extended Data Fig.  4j,k,p,q,v). However, the simultaneous depletion of RSC and SWI/SNF increased the number of steps to ~4 or ~2 depending on the repression conditions (Extended Data Fig. 4w,  and time between bursts (Extended Data Fig. 4f). However, the lower burst duration and intensity observed in RSC and SWI/SNF depletion during the first induction was not reproduced during re-induction, suggesting that this effect may be context specific (Extended Data Fig. 4d,e,g). For subsequent analysis, we focused on the induction time and time between bursts. Overall, these memory experiments showed that the remodeling of promoter nucleosomes by RSC and SWI/SNF promotes multiple activation steps, perhaps by acting on the different promoter nucleosomes.

Remodeler-free transcription activation
The above experiments showed that simultaneous RSC and SWI/SNF depletion reduces but does not abolish GAL10 transcription (Fig. 2d,f,g and Extended Data Fig. 3h-j). Also, comparison of the GAL10 promoter  Article https://doi.org/10.1038/s41594-023-00981-1 nucleosome coverage upon RSC and SWI/SNF depletion (Fig. 2b,c) to inactive conditions (Extended Data Fig. 1a-c) indicated that even in the absence of both RSC and SWI/SNF, nucleosomes were remodeled upon galactose induction. This remodeling must occur independently of RSC and SWI/SNF and could either be performed by other remodelers or by the transcriptional machinery itself. To test whether other remodelers are involved, we performed smFISH to detect changes in GAL10 nascent transcription upon single anchor-away depletion of the catalytic subunit of each of the seven budding yeast remodeling complexes (Extended Data Fig. 3q,r). Apart from RSC, none of the remodeling complexes had a large effect on steady-state transcription levels of GAL10. If another remodeler regulates nucleosomes in the GAL10 promoter, it must act redundantly with RSC and SWI/SNF. Alternatively, the transcriptional machinery itself may perturb nucleosome architecture.

Remodeling at the Gal4 UASs controls each burst
The depletion experiments above link the nucleosome changes to changes in transcription dynamics, but both RSC and SWI/SNF depletion affected multiple nucleosomes. To decipher the mechanism by which remodeling at individual nucleosomes controls transcriptional bursting, we combined remodeler depletions with perturbations that affect the binding of regulatory factors at specific nucleosomes. First, we focused on the fragile Gal4 UAS nucleosome. This nucleosome appears to be solely remodeled by RSC, as the coverage at the Gal4 UASs changed upon RSC depletion but not upon SWI/SNF depletion ( Fig. 1a,b, Fig. 2a-c and Extended Data Fig. 3k,l). The effect of remodeling of this nucleosome on transcription dynamics was assessed in GAL4/gal4Δ cells, where Gal4 expression is reduced twofold (Extended Data Fig. 3m) and where we expect a reduced Gal4 on-rate for UAS binding (Fig. 3a). In these GAL4/gal4Δ cells, an increase in induction time was observed (Fig. 3b,c) with a modest effect on other transcriptional parameters ( Fig. 3d and Extended Data Fig. 3f,g,n-p). Upon additional depletion of RSC in GAL4/gal4Δ cells, the induction time was within the expected range, suggesting independent roles of Gal4 and RSC during the first transcriptional burst (Fig. 3c). In contrast, we observed a synergistic increase in the time between bursts (Fig. 3d), which indicates that RSC remodeling of the Gal4 UAS nucleosome is redundant with Gal4 binding. Gal4 binding can thus substitute for RSC, possibly by partially unwrapping the nucleosome at the Gal4 UASs. Furthermore, this combined action of RSC remodeling and Gal4 binding at the fragile nucleosome needs to occur before the start of each burst of transcription.

Synergy of nucleosome remodeling and TBP binding at the TATA
Next, we focused on the nucleosomes around the GAL10 TATA and TSS, a region crucial for PIC assembly. One of the first steps in PIC assembly is TBP binding to the TATA, which recruits the rest of the transcription machinery to start transcribing the gene. The TATA is covered by a nucleosome in inactive conditions and is exposed upon activation (Extended Data Fig. 1a-c). The eviction of this nucleosome was reduced in SWI/ SNF-depleted cells. In addition, the movement of the TSS nucleosome into the NDR after RSC depletion affects TBP binding to the TATA 19 . To uncover the mechanisms by which remodeling of these nucleosomes affects transcription dynamics, we combined a remodeler depletion with a partial TBP depletion by introducing the anchor-away tag on one allele of the TBP-encoding gene in diploid cells, which is expected to reduce the on-rate of TBP to the TATA (Fig. 4a). Partial depletion was chosen, because full TBP depletion nearly completely abrogated GAL10 transcription (Extended Data Fig. 5g). Imaging indicated considerable TBP depletion (Extended Data Fig. 2b), but we note that even full TBP depletion was likely incomplete, as evidenced by growth on rapamycin-containing plates of a TBP anchor-away haploid strain (Extended Data Fig. 2a). Partial TBP depletion had a modest effect on GAL10 transcription (Fig. 4b-d and Extended Data Fig. 5a,b,g-j), likely because the strong GAL10 TATA ensured sufficient TBP binding even at a reduced TBP concentration. Surprisingly, partial TBP depletion resulted in faster induction than no depletion (Fig. 4b,c and Extended Data Fig. 5a,b), possibly due to interference of the anchor-away tag with TBP function (Extended Data Fig. 6, Methods).
To study how RSC remodeling at the TATA affects TBP-TATA interaction, TBP and RSC were depleted simultaneously. This combined depletion resulted in a synergistic delay of GAL10 induction  (Fig. 4d). A similar synergistic delay in induction was obtained for the combined SWI/SNF and TBP depletion (Extended Data Fig. 5e,f,k-p). In contrast to the fragile Gal4 UASs nucleosome, which requires remodeling before each burst, remodeling of the nucleosomes around the TATA is required to initiate the first, but not subsequent, bursts of GAL10 transcription. showing a reduced TBP on-rate after the addition of rapamycin in a diploid yeast strain where one of the two copies of TBP was depleted by anchor-away. b, GAL10 induction time increased upon RSC depletion, did not change upon partial TBP depletion and increased more in simultaneous RSC and partial TBP depletion. Significance determined by two-sided bootstrap hypothesis testing 53 ; NS, not significant; ****, P < 0.00005. P values: ±RSC, P < 10 −15 ; +TBP versus partial TBP 0.092; +RSC and TBP versus −RSC and partial TBP, P < 10 −15 . c, The increase in GAL10 induction time when simultaneously depleting RSC and partial TBP was larger than expected based on their individual depletions. Gray bar, expected effect based on dynamic epistasis analysis. d, The time between the consecutive bursts of GAL10 transcription increased as expected in the double depletion of RSC and partial TBP. Gray bar, expected effect based on dynamic epistasis analysis. Data in c and d are presented as the fractional change based on mean values ± s.d. based on 1,000 bootstrap repeats. e, MNase-seq analysis of stable nucleosomes in the GAL10 promoter region showed increased coverage around the TSS and TATA when RSC and partial TBP were depleted. f, MNase-seq analysis of stable nucleosomes in the GAL10 promoter region showed higher coverage around the TATA and TSS when depleting RSC and partial TBP than when depleting only RSC. g, Histogram of shift in +1 nucleosome upon depletion of RSC for both TATA genes and TATA-mismatch genes, showing that TATA-mismatch genes showed a larger shift in the +1 nucleosome than TATA genes. h, Histogram of shift in +1 nucleosome upon simultaneous depletion of RSC and partial TBP for both TATA genes and TATA-mismatch genes. Solid and dotted vertical lines in g and h indicate the median shift in the +1 nucleosome position of the TATA and TATA-less genes, respectively. Plots in e-h show one representative replicate out of two experiments.

Nucleosome competition by TBP binding
The observed synergy between reduced remodeling at the TATA and lower TBP concentration suggested a role for TBP in affecting nucleosome positions. To test this, MNase-seq was performed in cells depleted of RSC and partially depleted of TBP. Indeed, a larger shift in the GAL10 TSS nucleosome and more nucleosome density over the TATA were observed in these double-depleted cells than in RSC-only depleted cells (Fig. 4e,f and Extended Data Fig. 7a-f). Thus, in cells with impaired chromatin remodeling, TBP binding at the TATA contributes to positioning of the TSS and TATA nucleosomes.
The function for TBP in nucleosome positioning was supported by genome-wide MNase-seq analysis. For genes with a canonical TATA motif, we observed a smaller shift in +1 nucleosome upon RSC depletion and less coverage at the TATA than for genes lacking the canonical TATA motif (Fig. 4g and Extended Data Fig. 7g). Upon additional partial depletion of TBP, the ability of TBP to compete with nucleosomes is specifically reduced at genes with a strong TATA, but not at TATA-mismatch genes ( Fig. 4h and Extended Data Fig. 7h,i). These findings suggest that a strong TBP-TATA interaction is required for TBP to compete with nucleosomes.

Fractional change
Induction time

TBP nucleosome competition depends on the TBP dwell time
To evaluate whether the strength of the TBP-TATA interaction determines the ability of TBP to compete with nucleosomes, we mutated the GAL10 TATA 38 (Fig. 5a). This mutation reduced the burst duration and intensity (Extended Data Fig. 8a,b,k,l) and corroborated predictions of reduced burst size 38 . In addition, the TATA mutation increased the induction time and time between bursts (Fig. 5b,d,e). Remarkably, a RSC depletion in these TATA-mutant cells resulted in two distinct cell populations, as evidenced by a reduced active fraction (Fig. 5c). One population was not able to activate GAL10 transcription, likely because the TATA nucleosome could not be remodeled by TBP in the absence of RSC. For the active population, there was no additional effect of RSC depletion in terms of induction time (Fig. 5d) and the time between bursts showed the expected effect (Fig. 5e). These cells likely had no nucleosome covering the TATA and thus did not require nucleosome remodeling. TBP could then bind and initiate transcription, resulting in RSC-independent GAL10 induction. These results are thus in line with our prediction that a strong TATA is required for TBP to compete with nucleosomes. The inability of TBP to compete with nucleosomes upon a TATA mutation suggested that TBP nucleosome competition and subsequent transcription activation requires a longer residence time on DNA. To test whether increasing the TBP residence time increases TBP nucleosome competition, we stabilized the TBP-TATA interaction through the depletion of Mot1, the protein facilitating TBP removal 39-41 (Fig. 5f).
Mot1 depletion had a modest transcriptional phenotype (Fig. 5g-i and Extended Data Figs. 2k and 8c-f,m-o), perhaps because the depletion was not complete, as indicated by growth on rapamycin-containing plates and imaging (Extended Data Fig. 2a,b). As predicted, simultaneous depletion of RSC and Mot1 led to a rescue of the effect of RSC depletion on GAL10 induction time and the time between bursts. Stabilized TBP-TATA interaction thus enhances the nucleosome competition ability of TBP, and is sufficient to activate transcription efficiently, even in the absence of RSC remodeling.

Reduced histone levels rescue transcription dynamics
If TBP competes with nucleosomes, it is expected that impaired TBP binding due to a TATA mutation or reduced TBP levels is rescued if histones are less abundant. To verify this, we analyzed how deleting one of the two H3-H4 gene pairs, hht2Δ-hhf2Δ (Fig. 5j) affects bursting in cells with wildtype (wt) or mutant TATA, in the presence and in the absence of RSC. Western blot analysis using a H3K79me3 antibody suggests reduced histone levels (Extended Data Fig. 8t; H3K79me3 is present on 90% of the nucleosomes 42 ), even though a Western blot with the H3 antibody did not detect a change in the histone levels. Measurements of transcriptional bursting revealed that hht2Δ-hhf2Δ affected several transcriptional bursting parameters (Fig. 5j-n and Extended Data Fig. 8u-v). As hypothesized, the inactive population that was present in the TATA mutant upon RSC depletion (Fig. 5c,l) was absent when histones were also deleted (Fig. 5l). The ability of TBP to compete with nucleosomes is thus restored if nucleosomes are less abundant. In addition, both in the absence and in the presence of RSC, the effect of the TATA mutation on several other bursting parameters was partially or fully rescued by hht2Δ-hhf2Δ (Fig. 5j,k,m and Extended Data Fig. 8u-w). To verify these results further, we combined partial TBP and RSC depletion with hht2Δ-hhf2Δ and found a similar rescue (Extended Data Fig. 8x). The only bursting parameter that did not show a rescue by hht2Δ-hhf2Δ was the time between bursts ( Fig. 5n and Extended Data Fig. 8w,x), in line with our model that the TBP nucleosome competition regulates the first but not subsequent bursts. Because hht2Δ-hhf2Δ is not a conditional perturbation, it is important to note that indirect effects may contribute to the observed effects. Nevertheless, these results agree with the model that TBP binding to the TATA competes with nucleosomes to regulate transcription.

Taf1 and RSC remodeling act antagonistically at the TSS
To test if the redundant nucleosome displacement at the TATA by RSC and TBP binding is specific for the TBP-TATA interaction or a general effect from potentially impaired PIC assembly by perturbed TBP-TATA interactions, we measured the effect of Taf1 depletion in combination with RSC depletion. Taf1 is part of the PIC but has no direct interaction with the TATA or the TATA nucleosome, but interacts with the nucleosome around the TSS 43 (Fig. 6a). For Taf1-depleted cells, a delayed GAL10 induction was observed (Fig. 6b-d and Extended Data Fig. 8g-j,p-r).
Rather than the synergistic effect observed between RSC and TBP depletion, we observed a smaller effect on induction time than expected when depleting RSC and Taf1, suggesting opposing roles of RSC and Taf1 in controlling induction time (Fig. 6c). Therefore, in contrast to TBP, Taf1 cannot compete with nucleosomes. Conversely, nucleosome remodeling of RSC at the TSS acts antagonistically with Taf1 binding to control induction time.

Discussion
In this study, we use a dynamic epistasis analysis of single and combined perturbations of nucleosome remodeling, the transcription machinery and histones in combination with nucleosome mapping experiments and single-molecule live-cell imaging at the GAL10 gene in Saccharomyces cerevisiae, to uncover how the remodeling of promoter nucleosomes regulates transcriptional bursting. Based on our findings, we propose a model (Fig. 7) where different promoter nucleosomes have specialized roles in controlling transcription dynamics. Specifically, a fragile nucleosome covering the Gal4 UASs is repeatedly remodeled by RSC, in a manner redundant with Gal4 binding, which regulates the start of each transcriptional burst. Additionally, the nucleosomes around the TATA and TSS are positioned by an interplay of RSC, SWI/SNF and TBP to control the first burst of GAL10 transcription.

Nucleosome at the Gal4 UASs
Similar to observations of the TF Ace1 at the CUP1 array 8 , dynamic remodeling of the fragile nucleosome at the Gal4 UASs by RSC and Gal4 binding is required to allow each burst of transcription to start. The synergy between reduced Gal4 concentration and RSC depletion indicates that Gal4 binding can substitute for RSC in configuring the nucleosome to a state allowing transcription, in a manner similar to what has been observed for general regulatory factors, such as Reb1 or Abf1, although at most genes these general regulatory factors act independently of RSC rather than redundantly, as we observe for Gal4 19,23 . Mechanistically, Gal4 may trap the nucleosome in a partially unwrapped state, as was shown in vitro 30 . The binding of a gene-specific TF and RSC remodeling thus cooperate to enable efficient TF binding at each burst of transcription.

Nucleosome at the TATA
Our study uncovers a role for TBP in competing with nucleosomes covering the TATA. Although TBP depletion in the presence of RSC has a minor effect on nucleosome positioning only at highly expressed genes 19,44 , we show that the effect of TBP on nucleosome displacement becomes more prominent in conditions where nucleosomes cover the TATA, such as after the depletion of RSC or SWI/SNF. Specifically, the simultaneous depletion of TBP and RSC results in a larger increase in the nucleosome density around the GAL10 TATA than single RSC depletion (Fig. 4f). This role for TBP in nucleosome positioning appears more important for genes with a canonical TATA, possibly by supporting longer TBP residence times ( Fig. 4 and Extended Data Fig. 7). The ability of TBP to compete with nucleosomes allows for remodeler-independent GAL10 promoter activation, explaining why a substantial level of GAL10 transcription is still observed in the absence of RSC and SWI/SNF. It was recently shown that TBP is able to bind stably to nucleosomal DNA 32 but cannot efficiently recruit the PIC and activate transcription when nucleosome bound 32,45 . Our experiments where the TBP-TATA interaction is destabilized by a TATA mutation or stabilized by Mot1 depletion reveal that the ability of TBP to compete with nucleosomes may depend on the residence time of TBP at TATA. Although Mot1 depletion could, in principle, rescue GAL10 transcription through its regulation of antisense transcription 39,46,47 rather than by increasing the TBP dwell time 39,48,49 , we find this unlikely, because GAL10 antisense transcription does not affect GAL10 bursting 50 . Rather, we envision a passive competition mechanism where the binding of TBP may perturb the stability of the nucleosome by recruiting PIC components to partially unwrapped nucleosomes intermediates that arise from spontaneous nucleosome breathing 51 . If the residence time of TBP is long enough, the successive binding of multiple PIC components may eventually lead to nucleosome eviction, resulting in a nucleosome-free TATA needed for complete PIC assembly. Alternatively, the bending of the DNA that is introduced by the binding of TBP may change the nucleosome positioning energy landscape 52 . Longer TBP binding may increase the probability of downstream nucleosome movement to energetically more favorable sites. Moreover, because TBP-TATA binding is one of the most stable interactions within the transcription assembly, with a residence time of several minutes 31,46 , we envision that, once bound, TBP can maintain a nucleosome-free TATA, facilitating the initiation of multiple consecutive bursts. This is in line with our observed synergy between TBP and RSC in controlling the first, but not consecutive, transcriptional bursts.

Nucleosome at the TSS
The observed nucleosome competition is specific for TBP, rather than being a common mode of action of all PIC components, because Taf1 depletion does not show the same synergy with RSC depletion as TBP depletion. Even though Taf1 interacts stably with the TSS nucleosome in yeast extracts 43 , Taf1 and other PIC components interact with chromatin only for a few seconds in vivo 31 , contradicting the stable chromatin engagement that may be needed to passively compete away nucleosomes such as we propose for TBP 28 . In addition, we observe a lower induction time than is expected for the simultaneous depletion of Taf1 and RSC, indicating that Taf1 and RSC have opposing functions. In support, recent single-molecule tracking measurements of PIC components revealed that Taf1 binding to chromatin becomes more stable upon RSC depletion, suggesting that RSC promotes TFIID turnover 31 . Our data thus indicate that RSC, SWI/SNF and TBP redundantly are able to displace nucleosomes around the TATA and that RSC inhibits stable Taf1 binding around the TSS nucleosome.
Overall, dynamic epistasis analysis provides a detailed mechanistic insight into how nucleosome remodeling acts in combination with TFs and PIC assembly to control the kinetics of transcriptional bursting. In particular, at the yeast GAL10 gene and other TATA-containing genes, a role for TBP in competing with nucleosomes in vivo is uncovered, which together with chromatin remodelers enables efficient PIC assembly and transcription initiation. Moreover, our work forms a framework for future studies to understand how transcriptional bursting is regulated by the interplay of different transcriptional regulators, of different complex submodules and of complexes that vary in subunit composition or protein isoforms.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41594-023-00981-1.

Yeast strains and plasmids
All strains were derived from BY4741 and BY4742 anchor-away background strains 54 . FRB-tags for anchor-away were introduced either by transformation with a polymerase chain reaction (PCR) product containing the FRB-yEGFP1-hphMX4 cassette (pTL100) or the FRB-mScarlet-hphMX4 cassette (pTL329). Alternatively, a CRISPR-Cas9-based approach was used 55 , and strains were transformed using a plasmid expressing Cas9, a guide RNA and double-stranded PCR repair template from the same plasmids, followed by removal of the Cas9 plasmid by 5-fluoroorotic acid selection. PP7 loops were introduced by transformation with a PCR product containing the PP7 loop cassette and loxP-kanMX-loxP (pTL031), and subsequent removal of the kanMX marker by CRE recombinase expression (pTL014 or pTL191). PP7-coat protein was integrated at URA3 by transformation of a PacI-digested single-integration plasmid (pTL174) 56 . TATA-mut-int238 was introduced in the GAL10 TATA by the CRISPR-Cas9 approach described above using a single-stranded oligo as the repair template. The GAL4/gal4Δ strain was created by mating a BY4741 wt GAL4 anchor-away strain with a BY4742 anchor-away gal4Δ haploid strain that was constructed by the CRISPR-Cas9 approach described above, using a single-stranded oligo as repair template. The GAL4-3V5/GAL4-3V5 strain was created by mating a BY4741 and a BY4742 anchor-away GAL4-3V5 strain that were constructed by the CRISPR-Cas9 approach described above using a repair template created by a PCR product from genomic DNA from a GAL4-3V5 strain (YTL1446) and verified using Sanger sequencing. The GAL4-3V5/gal4Δ strain was created by mating the same BY4741 GAL4-3V5 strain with a BY4742 gal4Δ haploid strain. The hht2Δ-hhf2Δ strains were constructed by the CRISPR-Cas9 approach described above using a single-stranded oligo as the repair template. For all strains, at least two replicates were constructed independently, which were verified by PCR, growth plates (Extended Data Fig. 2a), microscopy (Extended Data Fig. 2b) and, if applicable, sequencing and smFISH (Extended Data Fig. 2c). All strains, plasmids and oligos used in this study are listed in Supplementary Tables 1, 2 and 3, respectively. The yeast strains and plasmids are available on request.

Live-cell imaging of transcription dynamics
Live-cell imaging of transcription dynamics was performed as previously described in detail 4,57 with minor modifications. In brief, cells were treated with 7.5 μM rapamycin or dimethyl sulfoxide (DMSO) for 60 min and subsequently imaged at mid-log (optical density (OD 600nm ) 0.2-0.4) on a coverslip with an agarose pad consisting of 2% agarose and synthetic complete medium containing 2% galactose and 7.5 mM DMSO or rapamycin. Imaging was performed on a setup consisting of an AxioObserver inverted microscope (Zeiss), an alpha Plan-Apochromat ×100 numerical aperture (NA) 1.46 oil objective, an sCMOS ORCA Flash 4v.3 (Hamamatsu) with a 475-570 nm dichroic (Chroma), 570 nm longpass beamsplitter (Chroma) and 515/30 nm emission filter (Semrock), a UNO Top stage incubator (OKOlab) at 30 °C and light emitting diode (LED) excitation at 470/24 nm (Spectra X, Lumencor) at 20% power and an neutral density (ND) 2.0 filter, resulting in a 62 mW cm −2 excitation intensity. Widefield images were recorded for 1 h at 15 s intervals, with z-stacks (nine slices, Δz 0.5 μm) and 150 ms exposure using Micro-Manager software 58 . For each condition, at least two and often three replicate datasets were acquired with at least 80 cells in total.

Analysis of transcription dynamics
For the analysis of the transcription dynamics imaging data, a similar approach was used to that previously described 4 . All analysis was implemented as custom-written Python software (https://github.com/Lenstralab/livecell). First, the images were corrected for xy-drift in the stage using an affine transformation on the maximum intensity projection. Next, the cells were segmented using Otsu thresholding and watershedding. The intensity of the TS was calculated by fitting a two-dimensional Gaussian mask after the local background subtractions as described previously 59 . Initially, a threshold of six times the standard deviation (s.d.) of the background was used. For frames where no TS was detected, a second fit was made in the vicinity of the high-intensity spots detected in that cell, using a threshold of four times the s.d. of the background. For frames where still no TS was detected, the intensity was measured at the location of the previous frame where a TS was detected. The tracking within each cell was inspected visually, and the endpoint of each trace was manually set at the last frame with a visible TS. Cells without a TS, cells that were segmented incorrectly, and cells in which the track contained tracking errors were excluded from analysis. Binarization was performed using a threshold set at five times the s.d. of the background, determined for each cell by fitting a Lorentzian distribution to intensities measured at four points at a fixed distance from the TS in each frame. This threshold reliably distinguished on and off periods at the single-transcript level. Subsequently, the binarization was improved by removing the bursts that lasted a single frame and merging bursts that were separated by a single frame. From these binarized traces, the burst durations, time between bursts and induction time were directly calculated. The burst intensity was measured as the average intensity of all frames in which the cell was on. The fraction of active cells was determined by the manual scoring of whether the cells show a TS during the 1 h acquisition. Reported values for burst duration, time between bursts, induction time and burst intensity were determined by bootstrapping with 1,000 repetitions. Reported error bars are the s.d. from the same bootstrap. Error bars in the number of active and inactive cells are given by the square root of the number of cells. To determine whether the active fractions are significantly different between conditions, a two-sided Fisher's exact test was used. For the other parameters, we have used bootstrap hypothesis testing using equation (4) from ref. 53 to determine the achieved significance level.

Dynamic epistasis analysis
For the dynamic epistasis analysis, the fractional change in each parameter of the transcriptional bursting was determined as the ratio between the bootstrap mean of this parameter in the perturbed population and the unperturbed population. For the effect of the hht2Δ-hhf2Δ perturbations ( Fig. 5 and Extended Data Fig. 7), the fractional change in the presence and absence of RSC was calculated relative to +RSC and −RSC, respectively. The error bars were calculated from the same bootstraps and propagated under the assumption that the measurements are independent between conditions. To calculate the expected effect of a double perturbation on each parameter, fractional changes of the individual perturbations are multiplied, analogous to the way phenotypic growth effects caused by pairwise genetic interactions are assessed 36 . The error bars are calculated by error propagation of the errors of individual perturbations.

Fitting of induction time distributions
To determine whether gene induction depends on a single or multiple rate-limiting steps, a least-squares fit was performed on the histogram of the distribution of induction times, with a binsize of 1 min. Article https://doi.org/10.1038/s41594-023-00981-1 The following parameterization of the probability density function of the Gamma distribution was used: Here, Γ(k) is the Gamma function, defined as: Here t is the variable of integration. The amplitude parameter A was added because there is a dead-time between addition of galactose and actual start of the image acquisition. Free parameters in the fit are A (with lower bound 1 and initial guess 10), k (with lower bound 0.0001 and initial guess 10 for non-memory induction or 1.0001 for re-induction conditions) and θ (with lower bound 0 and initial guess 1). The scale parameter k is a measure for the number of rate-limiting steps.

Testing for subpopulations
Dynamic epistasis analysis relies on the use of the bootstrap mean to describe each parameter in a given condition. This is valid if the cells behave as a single population, but masks the potential effects of the specific subpopulations. To determine whether there are subpopulations with different behavior, we tested whether the shapes of these distributions were well described by a theoretical distribution (Extended Data Fig. 6). For induction time a Gamma function was fit as above. For the time between bursts and burst duration a Gamma function was fit (with A = 1, k with lower bound 1 and initial guess 1.0001 and θ with lower bound 0 and initial guess 1). Theoretically, the burst duration should be described by the sum of a deterministic time (the elongation time for GAL10) and an exponential distribution describing the initiation kinetics, but this was approximated with a Gamma distribution for simplicity. For the burst intensity a log-normal distribution was fit: Free parameters in the fit are σ (with lower bound 0 and initial guess 0.6) and μ (unbounded with initial guess 5.5). The goodness of all fits was determined by calculating the R 2 between the data and the fit according to the following formula, where y 1 …y n are the observed values, with an average value of ȳ, and f 1 …f n are the fitted values: For the time between bursts, Gamma fits consistently showed k between 1 and 1.6 with R 2 > 0.8 (Extended Data Fig. 6d), indicating these distributions are approximately described by single populations with an exponential distribution. The fits to the burst duration and burst intensity distributions showed high R 2 values (>0.97 and >0.94, respectively, Extended Data Fig. 6c,e). In contrast, for the induction time distributions, we found four experiments with R 2 < 0.8 (Extended Data Fig. 6b), which could be caused by the presence of multiple populations. However, careful analysis of these distributions did not reveal signs of subpopulations. We noted that three of the low R 2 experiments describe cells where TBP was tagged for depletion. Together with the finding that TBP tagging resulted in unexpected faster induction (Fig. 4) and showed deviating shapes of the induction time distribution, this suggested that the anchor-away tag may partially interfere with TBP function. For the fourth experiment, GAL4/gal4Δ upon RSC depletion, the low R 2 value appeared to be caused by data sparsity, even though we included 181 cells. This data sparsity also appeared to cause inconsistencies in the k values of the Gamma fits to the induction times (Extended Data Fig. 6b), preventing proper interpretation. In addition, the inconsistent k values could arise because the Gamma distribution assumes different rate-limiting steps with equal rates, which may not be valid in all conditions. Overall, this analysis shows no signs of subpopulations, justifying our dynamic epistasis analysis based on the bootstrap mean.

Analysis of smFISH
Images were analyzed using custom-written Python software (https:// github.com/Lenstralab/smFISH). Here, the cells and nuclei were segmented using Otsu thresholding and watershedding. The spots were localized by fitting a three-dimensional Gaussian mask after local background subtraction 59 . Cells in which no spots were detected were excluded from further analysis, because visual inspection indicated these cells were not properly segmented or not properly permeabilized, such that smFISH probes did not enter the cells. For each cell, the TS was defined as the brightest nuclear spot and the number of RNAs at each TS was determined by normalizing the intensity of each TS to the median fluorescent intensity of the cytoplasmic RNAs detected in all cells. Cells with fewer than five RNAs at the TS were classified as inactive, and cells with five or more RNAs at the TS were classified as active cells. Subsequently, the fraction of active cells and the mean number of RNAs at the TSs of active cells were determined. For each condition, at least three replicate experiments were performed with in total at least 5,000 cells, and the average value and standard error of the mean were determined for both the active fraction and the number of RNAs at the TSs of active cells. The fractional changes of these parameters https://doi.org/10.1038/s41594-023-00981-1 upon nuclear depletion of indicated factors were determined from these mean values.
For the classification of cells into G1, S and G2 cell-cycle stages, the sum of the nuclear DAPI intensity in each cell is calculated from a maximum intensity projection. Subsequently, a histogram of all nuclear DAPI intensities (with 50 equally spaced bins) is fit with a Gaussian mixture model consisting of two peaks. Cells are classified as G1 stage if they are in a window of (s.d. 1 , 0.75 × s.d. 1 ) around the center of the first peak, as G2 stage if they are in a window of (0.5 × s.d. 2 , 1.5 × s.d. 2 ) around the center of the second peak and as S stage if they are in between the two peaks, where s.d. 1 and s.d. 2 are the s.d. of the first and second peak, respectively. Fractional changes in active fraction and number of RNAs at the TSs of active cells are determined as described above for each cell-cycle stage separately.

MNase-seq
The preparation and analysis of mono-nucleosomal DNA was performed as described previously 4,54 with minor modifications. Briefly, cells were grown in SC + 2% raffinose or SC + 2% galactose from OD 0.3 to OD 0.75 and then treated with 7.5 μM rapamycin or DMSO for 60 min. Then, cells were fixed in 1% paraformaldehyde, washed with 1 M sorbitol, treated with spheroplasting buffer (1 M sorbitol, 1 mM β-mercaptoethanol, 10 mg ml −1 zymolyase 100T (US Biological, Z1004.250)) and washed twice with 1 M sorbitol. Spheroplasted cells were treated with 0.01171875 U (low MNase) or 0.1875 U (high MNase) micrococcal nuclease (Sigma-Aldrich, N5386-200UN) in digestion buffer (1 M sorbitol, 50 mM NaCl, 10 mM Tris pH 7.4, 5 mM MgCl2, 0.075% NP-40, 1 mM β-mercaptoethanol, 0.5 mM spermidine) at 37 °C. After 45 min, reactions were terminated on ice with 25 mM ethylenediaminetetraacetic acid (EDTA) and 0.5% sodium dodecyl sulfate (SDS). The samples were treated with proteinase K for 1 h at 37 °C and decrosslinked overnight at 65 °C. Digested DNA was extracted with phenol/chloroform (PCI 15:14:1), precipitated with NH 4 -Ac, and treated with 0.1 mg ml −1 RNaseA/T1. The extent of digestion was checked on a 3% agarose gel. For all conditions, two independent experiments were performed, with similar outcomes, except for the SWI/SNF depletion strain treated with high MNase concentration, where one replicate of the DMSO condition was underdigested. For this condition, only one replicate was used for analysis.
Sequencing libraries were prepared using the KAPA HTP Library Preparation Kit (07961901001, KAPA Biosystems) using 1 μg of input DNA, 5 μL of 10 μM adapter, double-sided size selection before and after amplification using 10 cycles. Adapters were created by ligation of Universal adapter to individual sequencing adapters (Supplementary Table 5). Libraries were checked on a Bioanalyzer High Sensitivity DNA kit (Agilent). Sequencing was performed on a NextSeq550. Paired-end 2 × 75 bp reads were aligned to the reference genome SacCer3 ( January 2015) using bowtie2 (ref. 61) with the settings '-sensitive-end-to-end −3 15 −5 5 -X 1980-no-contain-no-discordant -p 40 -x'. The data have been deposited in National Center for Biotechnology Information's Gene Expression Omnibus (GEO) 62 and are accessible through GEO Series accession number GSE190737.

Analysis of MNase-seq
Analysis of MNase-seq data was carried out using custom-written Python software (https://github.com/Lenstralab/MNase_analysis). First, the aligned reads were filtered for length and only reads between 95 and 225 bp were retained for analysis. Subsequently, the read coverage was determined on a chromosome-by-chromosome basis by counting the number of reads covering each base, and normalized to the total coverage on the chromosome. Next, the coverage along each gene was determined using all verified open reading frames in the Saccharomyces Genome Database 63 . TATA and TATA-mismatch genes were identified as 'TATA-containing' and 'TATA-less' as previously described 64 . The coverage in TATA or TATA-mismatch regions was determined as the sum of the coverage in the 8 bp region spanning the TATA or TATA-mismatch sequence.
For metagene plots, genes were aligned at the +1 nucleosome in unperturbed conditions. To determine the position of the +1 nucleosome for each gene in these unperturbed conditions, the (prenormalization) coverage in a 4,000 bp window around the TSS of each gene was extracted from all experiments performed in DMSO using the high MNase concentration (combining the data for all yeast strains, that is YTL524, YTL525, YTL1306 and YTL1584). If the gene was on the Crick strand, the coverage was flipped to facilitate alignment of all genes. For each gene, these coverages were subsequently summed and smoothed using a Gaussian filter with a 40 bp window. The minimum of this smoothed coverage was determined, and a peak-calling function was used to detect nucleosome peaks. The −1 nucleosome was defined as the first peak before the coverage minimum, and the +1 nucleosome as the first peak after the coverage minimum. Genes for which fewer than two peaks were detected or for which the −1 or +1 nucleosome was detected more than 1,000 bp away from the TSS were excluded from the analysis. To generate metagene plots, the normalized coverage of all genes in a window of 2,000 bp centered at the location of the +1 nucleosome in unperturbed conditions of that gene was averaged.
To generate heatmaps of the log2-fold-change of the coverage upon depletions, genes were sorted by the NDR width as determined by the distance between the −1 and +1 nucleosomes in unperturbed conditions. Subsequently, for each gene, the log2-fold-change between the coverage in each depletion (rapamycin) condition and the average coverage between two replicate experiments in the nondepleted (DMSO) condition was calculated. These data were represented as a heatmap.
To determine the shift in the +1 position, the location of the +1 nucleosome was determined in each depletion dataset independently using the same steps as performed on the summed coverage to detect the position of the +1 nucleosome in unperturbed conditions. The shift in +1 nucleosome was then defined as the difference between the +1 nucleosome in depleted conditions and the +1 nucleosome as determined from all unperturbed high MNase datasets.

Reporting summary
Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability
The MNase-seq datasets generated during this study are available in the National Center for Biotechnology Information's GEO 62 through GEO Series accession number GSE190737. For MNase-seq, reads were aligned to the reference genome SacCer3 ( January 2015). MNase-seq metagene plots were generated using all verified open reading frames in the Saccharomyces Genome Database 63 . TATA and TATA-mismatch genes were identified as 'TATA-containing' and 'TATA-less' as previously described 64 . The microscopy data generated during this study are available from the corresponding author on reasonable request. Source data are provided with this paper. The fraction of cells that activated during 1 hour of imaging when depleting TBP fully (-TBP, both copies tagged for depletion) or partially (reduced TBP, one copy tagged for depletion) using anchor-away. Data is the fraction based on the number of active and inactive cells + /-propagated statistical errors in these numbers. Significance determined by two-sided Fisher's exact test; *: p < 0.05; ****: p < 0.00005. p-values: +TBP vs partial TBP 0.0062; +/-TBP 7.7 10 −42 . h-j The fraction of (h) cells that activated during 1 hour of imaging, (i) the burst duration and ( j) the burst intensity, in -RSC&Reduced TBP were as expected based on their individual depletions. k SWI/SNF and simultaneous TBP and SWI/SNF depletion showed increased GAL10 induction time. Significance in (n) determined by two-sided bootstrap hypothesis testing 53 ; n.s.: not significant; ****: p < 0.00005. p-values: +/-SWI/SNF < 10 −15 ; +TBP vs partial TBP < 10 −15 ; +SWI/SNF&TBP vs -SWI/SNF& partial TBP < 10 −15 . l-p SWI/SNF and simultaneous TBP and SWI/SNF depletion showed (k) increased induction time of GAL10.