We have recently performed a comparative analysis on modulation of ISW1a’s sliding activity by the main S. cerevisiae GRFs [17], finding that Rap1 has the property of hindering ISW1a’s sliding activity. We also found that this ability correlates with binding strength, particularly in terms of dwell time. In addition, we found that the binding strength of Rap1 correlates with its ability to maintain low nucleosome occupancy and hinder histone deposition [17]. In light of these findings, we aimed to find whether there are Reb1 and Abf1 binding site variants that could confer binding strengths higher than those found in our previous study and whether these higher binding strengths endow these GRFs with the property of hindering ISW1a’s nucleosome sliding activity.

Sequence variation at flanking or less conserved positions of Reb1 and Abf1 consensus binding sites involve major differences in dwell time

As detailed in the Introduction, recent findings point to a linkage between defined GRFs’ binding site variants and their preferential occupancy at gene promoters, relative to gene bodies, while other findings show that sequence variations in flanking positions of TFBSs could deeply affect TFs’ binding strength [11,12,13]. In light of these evidences, we focused our search in subtle variations from the canonical binding sites. These subtle variations were defined in terms of flanking sequences and less conserved positions within Reb1 and Abf1 consensus binding sites. Four different variants were tested for each GRF.

We analyzed binding strength of Reb1 and Abf1 for these variants in terms of overall affinity (apparent Kd) and dissociation kinetics (Koff). These analyses were performed using reconstituted mononucleosomes as probes, in order to use the same conditions subsequently used to test their influence on ISW1a’s nucleosome sliding activity. The probes harbor the 147 bp 601 nucleosome positioning sequence and 80 bp of extranucleosomal DNA (from here referred to as linker DNA) downstream the 601 sequence. In each probe, the binding site variant is located in the linker region, 10 bp downstream the 601 sequence (Figs. 1 A and 2 A). For both GRFs, the mobility shift generated on by these factors on the probes at the form of naked DNA was very similar to that generated on nucleosome probes, being only distinguishable in the case of Abf1 (Figs. S1A and S2A).

Fig. 1

Variations at flanking or less conserved positions of Reb1 consensus binding sites involve major changes in affinity and dwell time. (A) Upper panel: Depiction of Reb1 binding site variants tested. The name of each variant is based on the specific nucleotides present at positions −4, +4 and +5. Displayed on top is the consensus binding site, according to the JASPAR database [18]. Lower panel: Schematic representation of the nucleosome probes used in the assays. 601 NPS = nucleosome positioning region of the 601 sequence (gray bar). The oval represents the translational position adopted by the nucleosome core upon reconstitution, which spans the 601 region. Probe names indicate length of linker downstream (right) of the core and binding site variant located in the linker DNA region. (B, C) Apparent Kd (B) and dissociation kinetics (C) determinations for Reb1BS variants. The gel images correspond to electrophoresis in a non-denaturing polyacrylamide gel; each one is representative of three independent assays. The probe used in each reaction and Reb1 concentrations are depicted at the top of gel images; migrations of free DNA probe (DNA), nucleosome probe (Nuc), DNA probe bound by Reb1 (Reb1-DNA) and nucleosome probe bound by Reb1 (Reb1-Nuc) are indicated at the right. In addition, for the dissociation kinetics analysis, the different time points and use of an unlabeled double-stranded oligonucleotide harboring a Reb1BS for Reb1 removal (chaser) are depicted at the top of the gel image. The graphs at the right of each gel image correspond to densitometric quantification of binding percentages used to calculate Kd (B) and Koff (C) values, which are displayed in Table 1

Table 1 Kd, Koff and half-life values for Reb1 binding site variantsFig. 2

Variations at less conserved positions of Abf1 consensus binding sites involve major changes in dwell time. (A) Upper panel: Depiction of Abf1 binding site variants tested. The name of each variant is based on the specific nucleotides present at positions −9 and +8 or present at positions −1 to +2 in the case of the A3G variant. Displayed on top is the consensus binding site, according to the JASPAR database [18]. Lower panel: Schematic representation of the nucleosome probes used in the assays. 601 NPS = nucleosome positioning region of the 601 sequence (gray bar). The oval represents the translational position adopted by the nucleosome core upon reconstitution, which spans the 601 region. Probe names indicate length of linker downstream (right) of the core and binding site variant located in the linker DNA region. (B, C) Apparent Kd (B) and dissociation kinetics (C) determinations for Abf1BS variants. The gel images correspond to electrophoresis in a non-denaturing polyacrylamide gel; each one is representative of three independent assays. The probe used in each reaction and Abf1 concentrations are depicted at the top of gel images; migrations of free DNA probe (DNA), nucleosome probe (Nuc), DNA probe bound by Abf1 (Abf1-DNA) and nucleosome probe bound by Abf1 (Abf1-Nuc) are indicated at the right. In addition, for the dissociation kinetics analysis, the different time points and use of an unlabeled double-stranded oligonucleotide harboring an Abf1BS for Abf1 removal (chaser) are depicted at the top of the gel image. The graphs at the right of each gel image correspond to densitometric quantification of binding percentages used to calculate Kd (B) and Koff (C) values, which are displayed in Table 2

Table 2 Kd, Koff and half-life values for Abf1 binding site variants

In the case of Reb1, all binding site variants tested comply with the consensus sequence NTTACCCKN found in the JASPAR database (central position underlined; [18]), differing at positions −4, +4 and +5; the variants tested are named based in these positions (Fig. 1A). Position +5 stands as a flanking position, as it is commonly not included within the consensus Reb1 target sequence in mainstream TFBS databases, such as JASPAR, CIS-BP, YeTFaSCo and MotifMap [18,19,20,21]. Nevertheless, this position appears as part of the consensus binding sequence in a number of genome-wide analyses [11, 22,23,24]. Our Kd determinations show that the presence of T and A at positions −4 and +4, respectively (T-AG probe), results in a Kd nearly 5 times higher than that obtained by the presence of G in these positions (G-GG probe, Figs. 1B and S1B; Table 1). Regarding position +5, G and C (G-GG and G-GC probes) gave the same Kd value, at least in the sequence context tested in our analysis (Figs. 1B and S1B; Table 1). Consistent with our Kd determinations, our dissociation kinetics analyses showed a very short residence time of Reb1 binding for the T-AG variant (Fig. 1C), with a complex half-life below 30 s (Fig. S1D; Table 1), while the other three variants displayed markedly longer dwell times (Fig. 1C; Table 1). Differences in dwell time were also observed within these three variants. In this regard, the presence of C, A or T at position +5 has been associated with strong Reb1 binding sites in a previous genome-wide study performed by Rossi and co-workers [11]. Consistently, we found that C at position +5 (G-GC variant) confers a binding strength higher than that conferred by G at this position (G-GG variant), since the former variant displays a lower Koff value and a longer half-life than the latter one (Figs. 1C and S1C; Table 1). Interestingly, this long dwell time given by the presence of C at position +5 is significantly reduced when switching from G to T at position +4 (Figs. 1C and S1C; Table 1; compare G-GC to G-TC probe).

Regarding Abf1 binding site variants tested, they comply with the consensus sequence NCGTNNNNNRNKMBNN found in the JASPAR database (central position in bold; [18]). Three of them are derivatives of the sequence TATCGTATTGCATGAT (the underlined section of these sequences corresponds to their equivalent stretch), which was deducted as a consensus sequence from a SELEX approach [25] and used in our previous study [17]. These three variants differ in positions −9 and +8 (T-T, A-T and T-G variants; Fig. 2A). Position −9 is present in the consensus sequence of the SELEX study, but commonly absent in other studies and databases. On the other hand, the presence of G or C in position +8 has been associated to a biased enrichment resulting from a preferential cross-linking reaction generated by these nucleotides in that position, rather than resulting from contribution to binding strength [11]. An additional variant harbors the sequence AAAG from positions −1 to +2 (A3G variant, Fig. 2A). Although the region is not part of the most conserved positions of the consensus Abf1 binding site (Fig. 2A), we aimed to test this sequence variant as it is present in a higher frequency in databases and recent studies, relative to the SELEX consensus sequence [15, 24]. Under our assay conditions, all variants displayed similar Kd values, ranging from 2 to 2.5 nM (Figs. 2B and S2B; Table 2). The A3G variant displayed a slightly lower Kd value. Although this difference was statistically non-significant, this variant consistently displayed the longest dwell time among the variants tested in our dissociation kinetics analyses (Figs. 2C and S2C; Table 2). Indeed, for the other variants we needed to test much shorter time points, in order to determine Koff and half-life values (Fig. 2C). Differences in complex half-life were also observed between these variants. In this regard, the T-G variant displayed a dwell time shorter than that displayed by the T-T variant (Fig. 2C; Table 2), indicating that the presence of G at position +8 has a negative effect on binding strength. This result is consistent with previous studies showing that its enrichment in assays based on immunoprecipitation of cross-linked material does not reflect contribution to binding strength [11].

Binding site variants featuring long dwell times endow Reb1 and Abf1 with the ability to hinder ISW1a’s sliding activity

We next tested if the binding site variants displaying low Kd values and/or long dwell times confer the ability of hindering ISW1a’s sliding activity to Reb1 and Abf1. To do this, we performed nucleosome sliding assays employing the same probes used in affinity and dissociation kinetics analyses. Our nucleosome probes were reconstituted by octamer transfer, implicating that they contain non-labeled donor oligonucleosomes in a large excess relative to the nucleosome probe. These oligonucleosomes serve as stringency for GRF binding and define the incubation periods used in our remodeling assays (see Methods for details).

It is currently conceived that low-affinity binding sites at gene regulatory regions, displaying short dwell times, are able to act as functional sites under conditions where a high local cognate transcription factor concentration results in binding saturation levels similar to that obtained on high-affinity sites under lower transcription factor concentrations [26, 27]. To determine whether this principle applies to the molecular function of hindering nucleosome sliding activity, for Reb1 we compared the variant displaying the lowest affinity (T-AG) to that harboring the strongest binding site (G-GC). The comparative analyses were carried out using for the T-AG probe a Reb1 concentration higher than that used for the G-GC probe, in order to compare these variants under the same binding saturation levels. The electrophoretic step of the analyses was performed without prior removal of Reb1 binding, in order to assess both sliding and Reb1 binding extent. Importantly, to directly quantify the extent of sliding activity from the bands corresponding to slid nucleosome using this approach, it was critical to determine in advance whether the GRF is able to bind the slid nucleosome in our probes. To do this, we performed assays were the probes were first incubated with ISW1a, adding Reb1 to the reactions afterwards (Fig. 3A). As the probes harbor 80 bp linker in only one side of the nucleosome core and ISW1a slides the core towards a central position [28], in the slid nucleosome the binding site resides inside the less accessible nucleosome core region (Fig. 3A). As expected for this design, no reduction in the intensity of the band corresponding to the slid nucleosome in the presence of Reb1 was observed, even in the case of the probe harboring the binding site variant of the highest affinity, G-GC (Fig. 3A). Consistently, a reduction in Reb1 binding extent upon ISW1a-mediated nucleosome remodeling was observed, with the remaining binding signal originated essentially from binding to non-slid probe and naked DNA (Figs. 3A and S3A). Thus, for this probe design, Reb1 is not able to bind to its cognate sequence in the slid nucleosome. We then tested the ability of Reb1 to hinder ISW1a’s sliding activity, comparing the same low- and high-affinity probes, by performing assays where the probes were first incubated with Reb1, adding ISW1a afterwards. As observed in Fig. 3B, only a slight Reb1-mediated hindering of sliding activity was observed for the low affinity variant, T-AG. In contrast, a significantly stronger hindering of sliding activity was observed for the probe harboring the high affinity binding site variant, G-GC (Figs. 3B and S3B). The same result was obtained by including a step of Reb1 binding removal before the electrophoretic analysis (Fig. S4B). Importantly, the G-GC variant displayed this stronger hindering even though the Reb1 concentration used for it was more than 10 times lower than that used for the T-AG probe (Fig. 3B). In terms of binding patterns, the low-affinity (T-AG) probe displays a significant reduction in the extent of Reb1 binding, while variations in Reb1 binding levels were non-significant for the high-affinity (G-GC) probe (Figs. 3B and S3B).

Fig. 3

Binding site variants featuring long dwell times endow Reb1 with the ability to hinder ISW1a’s sliding activity. (A-D) Nucleosome sliding assays performed for the set of probes harboring Reb1BS variants (see Fig. 1A and its legend for a detailed description of the variants tested and probes design). The gel images correspond to electrophoresis in a non-denaturing polyacrylamide gel and each one is representative of three (A, B, D) or four (C) independent assays. The probe used in each reaction, presence of Reb1, ISW1a, ATP and ATP-γ-S, as well as Reb1 concentrations, are depicted at the top of gel images; migrations of free DNA probe (DNA), DNA probe bound by Reb1 (Reb1-DNA) and nucleosome probe bound by Reb1 (Reb1-Nuc) are indicated at the right, where slid and non-slid nucleosome probe populations are represented schematically. The graphs at the right of gel images correspond to determinations of sliding extent and percentage of Reb1 binding; all values used for these determinations were obtained from densitometric analyses of the corresponding gel scans. Bars in the graphs display the average of three or four independent assays for each condition analyzed. Error bars represent one standard deviation. Asymmetric connectors between bars correspond to two-tailed unpaired t-tests, while symmetric connectors correspond to ANOVA with Tukey’s multiple comparisons tests. Asterisks denote statistically significant differences (* p < 0.05; ** p < 0.01); n.s. = non-significant difference. (A) Analysis of Reb1 binding to T-AG and G-GC nucleosome probe variants after nucleosome sliding mediated by ISW1a. Upper left: outline of the steps involved in the assay; N.L.Olig. = non-labeled oligonucleosomes, used for ISW1a removal after nucleosome sliding mediated by this complex. Upper right: Schematic representation of the translational position of the Reb1 binding site resulting upon nucleosome sliding mediated by ISW1a. (B) Analysis of ISW1a’s sliding activity and Reb1 binding to T-AG and G-GC nucleosome probe variants, where ISW1a was added to the reactions after incubation with Reb1. Upper panel: outline of the steps involved in the assay. (C) Analysis of the effect of ISW1a action on Reb1 binding in the presence of ATP or ATP-γ-S. See outline in (B) for the steps involved in the assay. (D) Analysis of ISW1a’s sliding activity and Reb1 binding to Reb1BS variants displaying long dwell times (G-TC, G-GC and G-GG), where ISW1a was added to the reactions after incubation with Reb1. See outline in (B) for the steps involved in the assay. Direct measures of sliding extent and Reb1 binding percentage are presented in Figs. S3 and S5

The result of our analysis of Reb1 binding after ISW1a-mediated nucleosome sliding (Fig. 3A) indicates that the reduction of Reb1 binding observed in Fig. 3B for the T-AG probe arises from generation of the slid nucleosome, where the Reb1 binding site is less accessible. Continuous competition between this GRF and ISW1a for interaction with linker DNA would also result in a similar effect. To test this possibility, we performed an analysis for the low affinity (T-AG) probe equivalent to that shown in Fig. 3B, but including a reaction in the presence ATP-γ-S, a non- hydrolysable analog of ATP. In the presence of this analog we observed no reduction in Reb1 binding owing to the presence of ISW1a (Fig. 3C), supporting the decreased accessibility of the binding site in the slid nucleosome as the mechanism underlying the reduction in Reb1 binding upon ISW1a-mediated nucleosome remodeling.

We next compared the effect on ISW1a’s sliding activity exerted by the Reb1BS variants displaying dwell times in the order of minutes (G-TC, G-GC and G-GG variants, Table 1). These variants displayed no differences in their ability to hinder ISW1a’s sliding activity. Consistently, Reb1 binding was only slightly reduced and in a similar extent for all these variants (Fig. 3D; Fig. S5).This analysis was performed using a higher binding saturation level (near 90%). Under this condition, further hindering of nucleosome sliding is observed for the G-GC variant, as compared to that found using lower binding saturation (Fig. 3B). Importantly, this stronger hindering was obtained using a Reb1 concentration still substantially (8 times) lower than that used for the low affinity Reb1BS (T-AG) variant (compare Fig. 3B and D).

We then proceeded to compare the Abf1BS variants in terms of the ability of Abf1 to hinder ISW1a’s sliding activity. Unlike the case of Reb1, where one of the binding site variants displayed a markedly higher Kd value, the binding affinity of Abf1 for the variants tested appeared to be very similar (Fig. 2A, Table 2). However, as detailed above, marked differences were obtained in terms of dwell time, with the A3G variant displaying the longest time, with a half-life in the order of minutes, followed by T-T and the A-T variants displaying half-life values of only few seconds (Fig. 2B, Table 2). As in the case of Reb1, we first tested whether Abf1 binds to the slid nucleosome probe, by first incubating with ISW1a and adding Abf1 to the reactions afterwards. This analysis was performed for the variant displaying the longest dwell time (A3G; Fig. 4A). Interestingly, Abf1 displayed interaction with the slid nucleosome, reflected by the appearance of a faint band of slightly lower migration than that given by Abf1 interaction with the non-slid nucleosome probe (Fig. 4A). The binding extent of Abf1 to the slid nucleosome was markedly weaker than that displayed by this GRF to the same probe at the form of naked DNA or non-slid nucleosome (Fig. 4A), being therefore discarded in calculations of sliding extent of remodeling assays. Nevertheless, the appearance of this band served as an additional indicator of the differences displayed by the Abf1BS variants in terms of their ability to hinder ISW1a’s sliding activity (see below). The result of the nucleosome remodeling assay showed that Abf1 hinders ISW1a’s sliding activity in all variants, with the binding site variant of the longest dwell time (A3G) displaying the most pronounced hindering and the lowest reduction of Abf1 binding, while the variant of the shortest dwell time (T-G) displayed the weakest hindering of ISW1a’s activity, accompanied by the most pronounced reduction in Abf1 binding (Figs. 4B, S6 and S7). Importantly, the faint band reflecting Abf1 binding to the slid nucleosome appeared for all variants, except in the case of the A3G variant, further confirming the higher ability of Abf1 to hinder ISW1a’s sliding activity in the case of this variant (Fig. 4B and C and S6B). By performing a 30 min nucleosome remodeling incubation, as was the case of the analyses performed for Reb1, differences between the variants harboring intermediate half-life values (T-T and A-T) and the A3G variant were statistically non-significant (Fig. S6B). In light of this result, we reasoned that a longer incubation after adding the CRC in the remodeling assay could result in more marked differences between these variants, considering that more dissociation events would occur during this longer incubation, implying more chances for the CRC to exert its activity. Thus, we performed an additional analysis, extending the incubation after addition of ISW1a from 30 to 60 min. Under this setting, a more marked difference between the longest dwell time variant (A3G) and the other variants was found, being statistically significant relative to all the other variants (Figs. 4B and S7). The results obtained by extending the time of the remodeling incubation additionally suggested that in the shorter remodeling incubation period (30 min) a steady state given by a defined extent of Reb1 binding during continuous action of ISW1a was not present. To confirm this, we performed a time-course analysis from 5 to 30 min, comparing the variants displaying the shortest and longest dwell time (T-G versus A3G, respectively) and using Abf1 concentrations to reach over 90% binding saturation. The result of this analysis shows a continuous increment of the slid nucleosome in the absence of Abf1 during the whole incubation period. Similar to the previous analyses, in each time point the extent of sliding hindering was markedly lower for the shortest dwell time variant (T-G) than for the variant displaying the longest dwell time (A3G). Consistent with the results of the previous remodeling assays (30 and 60 min), in the case of the T-G variant Abf1 displayed a progressively lower sliding hindering and binding extent (Figs. 4C and S8). In addition, the results of this kinetic analysis demonstrate that the extent of sliding hindering and GRF binding in the remodeling assays of our current study do not reflect the reaching a steady state.

Fig. 4

The binding site variant harboring the longest dwell time endow Abf1 with the strongest ability to hinder ISW1a’s sliding activity. (A-C) Nucleosome sliding assays performed for the set of probes harboring Abf1BS variants (see Fig. 2A and its legend for a detailed description of the variants tested and probes design). The gel images correspond to electrophoresis in a non-denaturing polyacrylamide gel and each one is representative of three independent assays. The probe used in each reaction, presence of Abf1 and ISW1a, as well as Abf1 concentrations, are depicted at the top of gel images; migrations of free DNA probe (DNA), DNA probe bound by Abf1 (Abf1-DNA) and nucleosome probe bound by Abf1 are indicated at the right, where slid and non-slid nucleosome probe populations are represented schematically. The graphs at the right of gel images correspond to direct measures of sliding activity (slid fraction, A and C), determinations of sliding extent in the presence of Abf1 relative to its absence or Abf1 binding upon ISW1a-mediated nucleosome sliding relative to its absence. All values used for these determinations were obtained from densitometric analyses of the corresponding gel scans. Bars in the graphs display the average of three independent assays for each condition analyzed. Error bars represent one standard deviation. Connectors between bars correspond to ANOVA with Tukey’s multiple comparisons tests, with asterisks denoting statistically significant differences (* p < 0.05; ** p < 0.01; *** p < 0.001). Additional analyses related to this figure are presented in Figs. S6, S7 and S8. (A) Analysis of Abf1 binding to the A3G nucleosome probe variant after nucleosome sliding mediated by ISW1a. Upper panel: outline of the steps involved in the assay; N.L.Olig. = non-labeled oligonucleosomes, used for ISW1a removal after nucleosome sliding mediated by this complex. Right panel: Schematic representation of the translational position of the Abf1 binding site resulting upon nucleosome sliding mediated by ISW1a. (B) Analysis of ISW1a’s sliding activity and Abf1 binding to Abf1BS variants, where ISW1a was added to the reactions after incubation with Abf1 and nucleosome sliding incubations were conducted for 60 min. (C) Time-course analysis, ranging from 5 to 30 min of the nucleosome remodeling incubation

The sequence pattern of variants characterized as strong sites is enriched at loci where Reb1 and Abf1 act as strong nucleosome displacement factors

We have previously shown that the binding strength of the GRF Rap1 is inversely correlated with nucleosome occupancy and histone deposition genome-wide. The findings of our current study suggested that Reb1 and Abf1 would display a similar pattern [17]. To confirm this, we analyzed the correlation between binding strength and both nucleosome occupancy and histone deposition for Reb1 and Abf1. To do this, we compared genome-wide datasets of in vivo (ChIP-exo) and in vitro (PB-exo) binding profiles of Reb1 and Abf1 obtained by Rossi et al. [11]. to genome-wide patterns of nucleosome occupancy and histone deposition generated by Kassem et al. [29]. PB-exo consists in transcription factor immunoprecipitation in the presence of only purified genomic DNA, ensuring that binding patterns are not influenced by the chromatin context [11]. Two clusters were extracted from Reb1 binding peaks as well as for Abf1 peaks, corresponding to 20% of highest and 20% of lowest occupancy (termed Top and Bottom, respectively; Fig. S9). These two clusters displayed statistically significant differences in terms of both nucleosome occupancy and histone deposition, for Reb1 and Abf1 (Fig. S9). In addition, these two parameters were inversely correlated with occupancy of these GRFs. Similar results were obtained from both ChIP-exo and PB-exo data (Fig. S9).

As mentioned in the Introduction, recent findings point to a linkage between defined GRFs’ binding site variants and their preferential occupancy at gene promoters, relative to gene bodies, while other findings show that sequence variations in flanking positions of TFBSs could deeply affect TFs’ binding strength [11,12,13]. Considering this, we aimed to determine nucleosome occupancy and histone deposition patterns for Reb1 and Abf1, now comparing their binding sites located at gene bodies (ORFs) to their sites located at gene promoters. These clusters were compared using the same data and methodology described in the previous paragraph. As expected, lower histone occupancy levels are observed for the binding sites located at gene promoters (Figs. 5A and 6A). Regarding histone deposition, it is intimately linked to histone exchange, which is in general higher at gene promoters than at gene bodies [29]. Interestingly, our analysis show that binding sites of both Reb1 and Abf1 located at gene bodies display slightly higher levels of histone deposition than those observed at gene promoters (Figs. 5A and 6A). To ascertain whether these differential nucleosome occupancy and histone deposition levels were correlated to defined sequence patterns of Reb1 and Abf1 binding sites, we determined the consensus binding sites exhibited by the gene body and gene promoter clusters, using the MEME suite [