Promoter regulation by distinct mechanisms of functional interplay between lysine acetylase Rtt109 and histone chaperone Asf1

Abstract

The promoter activity of yeast genes can depend on lysine 56 (K56) acetylation of histone H3. This modification of H3 is performed by lysine acetylase Rtt109 acting in concert with histone chaperone Asf1. We have examined the contributions of Rtt109, Asf1, and H3 K56 acetylation to nutrient regulation of a well-studied metabolic gene, ARG1. As expected, Rtt109, Asf1, and H3 K56 acetylation are required for maximal transcription of ARG1 under inducing conditions. However, Rtt109 and Asf1 also inhibit ARG1 under repressing conditions. This inhibition requires Asf1 binding to H3-H4 and Rtt109 KAT activity, but not tail acetylation of H3-H4 or K56 acetylation of H3. These observations suggest the existence of a unique mechanism of transcriptional regulation by Rtt109. Indeed, chromatin immunoprecipitation and genetic interaction studies support a model in which promoter-targeted Rtt109 represses ARG1 by silencing a pathway of transcriptional activation that depends on ASF1. Collectively, our results show that ARG1 transcription intensity at its induced and repressed set points is controlled by different mechanisms of functional interplay between Rtt109 and Asf1.

The acetylation state of nucleosomal histones has a profound influence on the initiation, elongation, and termination phases of transcription. Much of the regulation of transcription impinges on the proteins responsible for histone acetylation—the histone-directed lysine acetylases (KATs). One recently discovered KAT being intensively studied from the viewpoint of its regulation is Rtt109. This yeast protein catalyzes K9, K23, K27, and K56 acetylation of histone H3. All of these reactions depend, to a greater or lesser extent, on the conserved H3-H4 chaperone Asf1. Specifically, Asf1 stimulates H3 K9, K23, and K56 acetylation by Rtt109 on its own, and K27 acetylation by Rtt109 in complex with histone chaperone Vps75 (1–4). In current models, transcriptional regulation by Rtt109 is ascribed to its ability to acetylate H3, and functional interplay between Rtt109 and Asf1 in the regulation of transcription is limited to Asf1 stimulation of Rtt109 KAT activity.

Here we examine the role of Rtt109 and Asf1 in the regulation of ARG1, a well-studied metabolic gene of budding yeast. ARG1 is repressed in arginine-replete cells by the ArgR/Mcm DNA binding complex consisting of Arg80, Arg81, Arg82, and Mcm1 (5–7). Upon arginine limitation, ARG1 is activated by the transcription factor Gcn4 (8, 9). Chromatin reconfiguration, in particular, acetylation of residues in the amino-terminal tails of H3 and H4, makes an important contribution to the physiological regulation of ARG1 promoter activity. The enzymes implicated in this regulation include the KATs Gcn5 and Esa1 (10, 11).

We extended these findings by exploring the contributions of Rtt109 and Asf1 to ARG1 regulation. In part our results support the evidence that Asf1-dependent acetylation of H3 K56 by Rtt109 is important for high transcription (12–15). We also find that Asf1 and Rtt109 control ARG1 promoter activity under repressive conditions by an unprecedented mechanism likely involving Rtt109 inhibition of transcription stimulation by Asf1.

Results and Discussion

H3 K56ac Favors High Transcription of ARG1.

We studied the mechanism of ARG1 transcriptional regulation under two steady-state conditions: repression in arginine-replete medium (yeast extract, bactopeptone, dextrose, YPD), and induction (or activation) in arginine-free minimal medium (composition in Table S1, M1D) (Fig. 1A). Compared to repression, the induced configuration of ARG1 promoter chromatin is characterized by lower H3 content and enrichment of H3 K56ac (Fig. 1 B and C). H3 K56ac occupancy is sensitive to deletion of RTT109 and ASF1 in cells cultured in either arginine-replete or arginine-free medium (Fig. 1D), whereas H3 occupancy has little dependence on ASF1 (Fig. 1B). Therefore, (i) ARG1 promoter nucleosomes are marked by H3 K56ac whether the gene is active or repressed, (ii) high H3 K56 acetylation is a hallmark of the induced state, and (iii) Asf1 is not uniquely required to maintain H3 promoter occupancy under repressing or inducing conditions. Consistent with published evidence that H3 K56 acetylation is favorable for transcription, ARG1 expression is dampened under inducing conditions by the H3 K56R mutation which mimics deacetylation (Fig. 1E). Conversely, repression is dampened (ARG1 is “induced”) by the K56Q and K56A mutations, which mimic the charge state conferred by lysine acetylation (Fig. 1F).

Fig. 1.

ARG1 regulation by H3 K56ac. (A) ARG1 transcription in wild-type cells in inducing minimal medium, relative to transcription in repressive YPD medium (latter set to one). (B) ChIP analysis of H3 cross-linking to the promoter of ARG1 in wild-type and asf1Δ cells, under repressing and inducing conditions. Occupancy in wild-type cells subject to repression is set to one. Average of two experiments; the error bar shows the range. (C) ChIP analysis of H3 K56ac at the promoter of ARG1 under repressing and inducing conditions. All data points are normalized to H3 occupancy, and occupancy under repression is set to one. (D) ChIP analysis of H3 K56ac dependency on RTT109 and ASF1. ARG1 promoter chromatin was probed under repressing and inducing conditions. Analysis as in C. (E) ARG1 transcription in H3 K56 mutants relative to wild type (H3 K56K), under inducing conditions. Average of two experiments; the error bar shows the range. (F) As in E, under repressing conditions. (G) Effect of RTT109 and ASF1 deletion (alone and in combination) on ARG1 transcription under inducing conditions. (H) ARG1 transcription in asf1Δ and asf1^V94R cells under inducing conditions. Wild-type transcription is set to one. (I) Effect of ASF1 deletion and H3 K56 mutation on ARG1 transcription under inducing conditions.

The effects of RTT109 and ASF1 deletion on activated transcription are consistent with regulation of ARG1 by a mechanism that involves H3 K56 acetylation, as described for other genes (12, 13). Specifically, in arginine-free minimal medium ARG1 transcription is lower in rtt109Δ and asf1Δ cells than wild type (Fig. 1G), and the Asf1 V94R mutation which compromises binding to H3-H4 and H3 K56 acetylation (16, 17) phenocopies the ASF1 null (Fig. 1H). These results suggest that the shared function of Rtt109 and Asf1 in the regulation of H3 K56 acetylation is important for activated transcription of ARG1. Consistent with this interpretation, simultaneous deletion of RTT109 and ASF1 has no greater effect on activated transcription than either deletion alone (Fig. 1G), and the effect of ASF1 deletion is similar in magnitude to, and nonadditive with, the H3 K56R mutation (Fig. 1I).

ARG1 Regulation by Rtt109 and Asf1 Under Repression.

The H3 K56R mutation which mimics deacetylation has no effect on ARG1 transcription under repressing conditions (Fig. 1F). Therefore, deletion of RTT109 and the resulting global deacetylation of H3 K56 were not expected to affect ARG1 transcription in YPD-grown cells. Surprisingly, however, deletion of RTT109 causes sixfold induction of ARG1 mRNA (Fig. 2A, Left; compare to dampening of induction under arginine limitation, Right). ARG1 induction is likely due to elevated transcription initiation because RNA polymerase (RNAP) II cross-linking to its promoter is elevated in rtt109Δ in arginine-replete medium (Fig. 2B). Constitutive DNA damage signaling in rtt109Δ cells (12) cannot account for ARG1 induction because ARG1 is not controlled by the DNA damage sensor kinase Mec1 (18). Possible induction by oxidative stress (19) is also unlikely. Deletion of RTT109 does not confer sensitivity to exogenous oxidants (20). It follows that rtt109Δ cells are not under a higher than normal level of endogenous oxidative stress, and that oxidative stress signaling pathways are not constitutively activated in rtt109Δ cells. Although ARG1 is normally induced in G2 (21) and rtt109Δ cells accumulate in G2/M (1), G2/M arrested rtt109Δ cells (Fig. 2C) support higher ARG1 transcription than arrested wild-type cells (Fig. 2D). Overall, we conclude that Rtt109 can repress ARG1 independently of cell cycle cues, by a mechanism that regulates the transcription process prior to elongation and does not involve Rtt109 acetylation of H3 K56.

Fig. 2.

ARG1 regulation by Rtt109 and Asf1 under repressing conditions. (A) Effect of RTT109 deletion on ARG1 transcription under repressive and inducing conditions. Under each condition, transcription in the mutant is relative to wild type (set to one); this normalization highlights the fold-effect on steady-state transcription under each condition. (B) ChIP analysis of RNAP II cross-linking to the promoter of ARG1 under repressing conditions. RNAP II occupancy in wild-type cells is set to one. Average of two experiments; the error bar shows the range. (C) Wild-type and rtt109Δ cell cycle profiles in repressive medium without or with nocodazole. (D) ARG1 transcription in rtt109Δ cells relative to wild type in mixed populations of cells and G2/M-arrested cells. (E) ARG1 transcription in asf1Δ and asf1^V94R cells under repressive conditions (wild-type transcription is set to one). (F) ChIP analysis of TBP cross-linking to the promoter of ARG1 under repressing conditions. Occupancy in wild-type cells is set to one. PGK1 is a control gene not regulated by arginine (10). (G) ChIP analysis of RNAP II cross-linking to the promoter of ARG1 under repressing conditions in the presence and absence of Asf1. RNAP II occupancy in wild-type cells is set to one. Student’s t test was used to assess significance (^∗P < 0.05).

Rtt109 function in transcriptional activation of ARG1 requires Asf1 (Fig. 1G). Accordingly, we hypothesized that ARG1 repression by Rtt109 also involves Asf1. Under this hypothesis, deletion of ASF1 is expected to dampen ARG1 repression. A microarray study provided evidence in favor of this possibility (5.9-fold relief of repression in asf1Δ cells) (12), and targeted mRNA analysis revealed twofold increased ARG1 expression in asf1Δ and asf1^V94R cells grown in YPD (Fig. 2E), associated with increased occupancy of the ARG1 promoter by both the TATA binding protein and RNAP II (Fig. 2 F and G). We conclude that Rtt109 and Asf1 have a dual role at ARG1: They both promote transcription when steady-state physiological conditions trigger high ARG1 expression and dampen promoter activity under physiological conditions of low steady-state transcription. Repression and activation of ARG1 both require robust binding of Asf1 to H3-H4 (Figs. 1H and 2E). Overall, our results reveal an unprecedented role for Rtt109 in stimulation and inhibition of promoter activity, and demonstrate that the ability of Asf1 to promote functionally opposite states of chromatin architecture at an individual locus is not restricted to elongation-coupled events in coding regions (22): At an individual promoter, Asf1 can also exert positive and negative affects on chromatin that impact on transcription.

Functional Interplay Between Rtt109 and Asf1 in ARG1 Repression.

Our analysis of ARG1 suggests previously unknown roles for Rtt109 and Asf1 in dampening of transcription. This unexpected outcome prompted us to consider conventional but indirect mechanisms that might account for the effects of RTT109 and ASF1 mutations on ARG1 repression. We sought to explain three key observations, starting with dampening of ARG1 repression in the absence of either Rtt109 or Asf1. Loss of Rtt109 or Asf1 could induce ARG1 if these proteins normally support expression of the ArgR/Mcm repressor (an equivalent mechanism could explain ARG1 induction in mutants of the SWI/SNF chromatin remodeling complex) (23). If this model is correct, then deletion of ARG80 which encodes an essential subunit of ArgR/Mcm should have the same effect on ARG1 transcription as deletion of either RTT109 or ASF1. Our results do not support this prediction: arg80Δ has a stronger inducing effect on ARG1 than either rtt109Δ or asf1Δ, and deletion of either ASF1 or RTT109 has an additive effect on depression caused by loss of ARG80 (Fig. 3A). It follows that Rtt109 and Asf1 do not regulate ARG1 under repressing conditions by modulating previously described mechanisms of ARG1 regulation by ArgR/Mcm. Consistent with this interpretation, ASF1 deletion is not associated with altered mRNA expression of any ArgR/Mcm component in budding yeast (microarray analysis of YPD-grown cells) (12) and ARG1 induction in an RTT109 null mutant of Candida albicans is not associated with misregulation of ArgR/Mcm subunits (24).

Fig. 3.

Relationship of ARG1 repression by Rtt109 and Asf1 to other pathways. (A) Effect of ARG80 deletion on ARG1 transcription in rtt109Δ and asf1Δ cells cultured under repressive conditions. (B) ChIP analysis of H3 K9ac at the promoter of ARG1 in wild-type and rtt109Δ cells under repressing conditions. Occupancy in wild-type cells is set to one. (C) ChIP analysis of H3/H4 tail acetylation at the promoter of ARG1 in wild-type and asf1Δ cells, under repressing conditions. Occupancy in wild-type cells is set to one. H4 acetylation in asf1Δ cells is significantly different from wild type (Student’s t test, ^∗P < 0.05). (D) Effect of RTT109 and GCN5 deletion (alone and in combination) on ARG1 transcription under repressive conditions.

We next considered a straightforward explanation for the fact that rtt109Δ (Fig. 2A) more substantially induces ARG1 transcription in arginine-replete medium than asf1Δ (Fig. 2E). Because histone chaperone Vps75 can regulate Rtt109 activity (25, 26), we hypothesized that Rtt109 repression of ARG1 in wild-type cells is imposed by parallel nonredundant pathways which separately depend on Asf1 and Vps75. It follows that asf1Δ does not induce ARG1 to the same extent as rtt109Δ because of residual Vps75-dependent Rtt109 repression in the ASF1 null. If this hypothesis is correct, then VPS75 deletion should be associated with partial induction of ARG1 in arginine-replete medium. ARG1 however is not induced in vps75Δ cells cultured under repressive conditions (27). Therefore functional redundancies between Asf1 and Vps75 do not account for the distinct effects of rtt109Δ and asf1Δ on ARG1 repression, and the contribution of histone chaperones to ARG1 regulation under repressing conditions is limited to Asf1.

Finally we sought to explain why we obtained two different answers to a straightforward question: How does H3 K56ac affect ARG1 transcription under repressing conditions? That is, we sought to understand why ARG1 is stimulated by mutations which mimic H3 K56ac (Fig. 1F, H3 K56Q and K56A), and by mutations which eliminate H3 K56ac (Fig. 2A, rtt109Δ; E–G, asf1Δ and asf1^V94R). This discordance suggests that Rtt109 and Asf1 control ARG1 under repressing conditions by a mechanism unrelated to the control of H3 K56ac (installation of H3 K56 mutations which mimic acetylation presumably override another system of regulation by Rtt109 and Asf1).

A likely alternative to regulation of H3 K56ac by Rtt109-Asf1 is regulation of H3 K9ac. This alternative is likely in view of the evidence that tail acetylation of H3 is important for repression of ARG1 (10), and that Rtt109 working in concert with Asf1 catalyzes H3 K9 (4) in addition to H3 K56 acetylation. If this mechanism underlies ARG1 repression by Rtt109, then RTT109 deletion should be associated with low H3 K9 acetylation of ARG1 in YPD-grown cells. We observe no such association (Fig. 3B), and RTT109 deletion does not affect H3 K14, K18, or K23 acetylation (28). Furthermore, we do not observe loss of overall H3 or H4 tail acetylation at ARG1 in asf1Δ cells (Fig. 3C). It follows that neither Rtt109 nor Asf1 controls ARG1 under repressing conditions by a mechanism that depends on H3 K9ac. By extension, we reasoned that repression by Rtt109 is not in the same pathway as repression that depends on Gcn5 acetylation of H3 K9. Consistent with this proposition, rtt109Δ and deletion of GCN5 are additive in their stimulatory effect on ARG1 transcription under the repressing condition (Fig. 3D). We conclude that Rtt109 and Asf1 act in parallel to Gcn5-dependent H3 acetylation to repress ARG1 transcription in arginine-replete medium.

The absence of a compelling conventional explanation for ARG1 repression by Rtt109 and Asf1 prompted our further characterization of this regulation. In current models, regulation of transcription by Rtt109 and Asf1 is mostly ascribed to their activities off chromatin, specifically their ability to collaborate in the acetylation of soluble H3. Because Rtt109-Asf1 acetylation of soluble H3 does not affect ARG1 repression, we turned our attention to the possibility that Rtt109 and Asf1 control ARG1 promoter activity as components of chromatin. We focused on Rtt109, because Asf1 occupancy of ARG1 (Fig. S1) is likely to reflect global, nonspecific association with chromatin (12, 29). Rtt109 can be cross-linked to the upstream activating region, promoter, and coding region of ARG1 under arginine-replete conditions (Fig. 4A). Importantly, Rtt109 promoter occupancy is (i) higher under repression than under induction (Fig. 4B; differential enrichment was not observed in the ORF—Fig. S2), and (ii) dependent under repression on sequence-specific transcription factors that modulate the strength of ARG1 repression (Fig. 4C Left), namely, the leucine zipper protein Gcn4 and the zinc finger protein Arg81, which is assembled specifically on the ARG1 promoter when arginine is not limiting (30). Deletion of GCN4 and ARG81 has little effect on Rtt109 occupancy of POL1, a gene not known to be regulated by arginine (Fig. 4C, Right). These relationships suggest that ARG1 repression depends on promoter-targeted Rtt109 and perhaps Rtt109 regulation of a nonhistone protein directly involved in chromatin metabolism at the ARG1 promoter.

Fig. 4.

Interplay between Rtt109 and Asf1 in the repression of ARG1. (A) ChIP analysis of Rtt109 cross-linking at the ARG1 locus under repressing conditions. The occupancy measurement obtained in no-antibody control ChIPs is set to one. Average of two experiments; the error bar shows the range. (B) ChIP analysis of Rtt109 cross-linking to the promoter of ARG1 under repressing and inducing conditions. Analysis as in A. (C) Effect of GCN4 and ARG81 deletion on Rtt109 cross-linking to the promoter of ARG1, and coding region of POL1, under repressing conditions. Analysis as in A; results are for three independent experiments. (D) Rtt109 KAT activity is required for ARG1 repression. The averages are for three (RTT109 + and − at left; both strains harbor the empty vector) or two (“WT” and “KAT-dead”) experiments; the error bars respectively show the standard error and range. (E) Effect of RTT109 and ASF1 deletion (alone and in combination) on ARG1 transcription under repressive conditions. (F) Model of ARG1 regulation by Rtt109 and Asf1. See text for details.

To test if ARG1 repression involves protein acetylation by Rtt109, transcription was compared in rtt109Δ cells expressing wild-type or catalytically inactive Rtt109 from a low-copy vector (28) (Fig. 4D). In repressing medium M2D, ARG1 induction associated with deletion of RTT109 is suppressed by wild-type but not catalytically dead rtt109^DD287288AA. Therefore protein acetylation by Rtt109 is important for repression of ARG1. Collectively, our results suggest that ARG1 repression depends on KAT-dependent regulation of a nonhistone protein by chromatin-associated Rtt109.

Inhibition of transcription by Rtt109 and Asf1 is not restricted to ARG1; these proteins also inhibit transcriptional activation of stress response genes (31). There are however important differences between ARG1 and the stress response genes in their regulation by Rtt109 and Asf1. First, Rtt109 and Asf1 control the steady-state set point of ARG1 transcription under repressing conditions, but do not influence this phenotype of the stress response genes. Second, the rtt109Δ and asf1Δ mutations have the same effect on activation of the stress response genes, but significantly different effects on ARG1 repression (Fig. 2 A and E). Third, Asf1 controls H3 dynamics at stress-induced genes but not the promoter of ARG1 (Fig. 1B). The notion that ARG1 and the stress response genes differ in their regulation by Rtt109 and Asf1 was confirmed by a genetic interaction experiment. In this experiment, mRNA expression was measured in rtt109Δ asf1Δ cells and the corresponding single mutants (Fig. 4E). The individual mutations have identical and nonadditive effects on activation of the stress response genes (31). The same mutations have strikingly different effects on ARG1. First, rtt109Δ more strongly affects ARG1 repression than asf1Δ (Fig. 4E; also compare Fig. 2 A and E). Second, deletion of ASF1 suppresses transcription induction associated with deletion of RTT109, such that ARG1 expression in the double mutant is identical to expression in asf1Δ (Fig. 4E). Based on these results, we suggest that Rtt109 and Asf1 are components of a unique system that represses ARG1 independently of mechanisms that control H3 acetylation. Below we refer to this system of regulation using Asf1 not H3Kac yes Rtt109 (ANKYR).

The most important contribution of this work is the description of previously undiscovered modes of functional interplay between Rtt109 and Asf1 in a system which contributes to ARG1 regulation under repressing conditions. We propose the following working model to explain the architecture of this ANKYR system (Fig. 4F; Fig. S3 explains this model in the context of the transcription phenotypes of the mutants examined). As originally suggested by Tyler and coworkers, we envisage that a population of Asf1 molecules interacts nonspecifically with chromatin throughout the genome (ref. 29; see also ref. 12). This nonspecific interaction accounts for the presence of Asf1 at ARG1 (Fig. S1). When associated with chromatin, Asf1 can potentially reconfigure nucleosomes by one of two mechanisms—one that favors transcription, and one that disfavors transcription (22). We suggest that the pathway of transcription stimulation by ARG1-associated Asf1 is intrinsically more potent than the Asf1-dependent pathway of transcription inhibition. The notion that alternative pathways of promoter regulation by Asf1 can have different strengths is well established for PHO5 (13). In our model, deletion of RTT109 has a stronger inducing effect than deletion of ASF1 because, in wild-type cells, Rtt109 inhibits the positive effect of Asf1 on ARG1 transcription. The latter regulation is likely to occur on chromatin, because (i) both proteins occupy the promoter of ARG1, (ii) Rtt109 occupancy is higher under repressing conditions than under inducing conditions (Fig. 4B), and (iii) ARG1 regulation does not involve H3 K56 acetylation (a reaction that Rtt109 can only perform on soluble H3) (32). Because Rtt109 is present in the promoters of numerous genes (12, 13, 33), and Asf1 is a promiscuous chromatin-binding protein, it is possible that the ANKYR system operates at multiple loci throughout the genome.

In our model of the ANKYR system, under repressing conditions Rtt109 KAT activity dampens transcriptional stimulation by Asf1. How protein acetylation by Rtt109 (Fig. 4D) might control this putative Asf1-dependent stimulation of ARG1 remains a matter of speculation. There are numerous precedents for cellular regulation of nonhistone targets by KATs that also modify histones (34, 35). Although acetylation of cellular Asf1 has not been reported, Rtt109 has weak KAT activity toward Asf1 in vitro (1) and a complex containing Rtt109 and Asf1 can be recovered from yeast cells after chemical cross-linking (36). Perhaps then Rtt109 acetylation directly modulates Asf1 activity. Alternatively, autoacetylation of Rtt109 (37, 38) could affect the function of Asf1. For example, Asf1 stimulation of transcription could be inhibited when Asf1 comes into contact with autoacetylated Rtt109. Finally, Rtt109 might regulate another nonhistone protein involved in Asf1-dependent stimulation of ARG1.

Adkins et al. (29) have proposed that the Asf1 associated with repressed promoters is not able to drive chromatin toward a configuration that is permissive for transcription. How is Asf1 prevented from doing so when it is fully capable of promoting chromatin opening in the context of transcription elongation? This problem remains unsolved. For example, whether the positive pathway of chromatin reconfiguration that depends on Asf1 has a default state of low activity, or a default state of high activity that is restrained under some conditions, has not been studied. Our results suggest that Asf1 stimulation of transcription (at ARG1) is restrained by a mechanism that involves Rtt109. This finding raises the possibility that functional switching of Asf1 between states that close and open chromatin is under physiological control by a pathway that depends on Rtt109.

Materials and Methods

Yeast Strains and Culture.

Yeast strains (Table S2) were obtained from published sources or constructed by methods outlined in Minard et al. (12). The standard repressing medium was yeast extract, bactopeptone, dextrose (YPD). The standard inducing medium (M1D) was based on yeast nitrogen base without amino acids. Strains harboring centromere-based plasmids were cultured in a similar repressing medium with arginine (M2D). Formulations are given in Table S1. Cells were arrested in G2/M with 10 μg/mL nocodazole (Sigma) for 2–4 h.

Cell Cycle Profiling.

Cellular DNA content was measured by flow cytometry (39)

Chromatin Immunoprecipitation.

ChIP was performed as published (12, 40) using antibodies against the indicated proteins or epitopes. Preferential recognition of H3 K9ac by the antibody used for the experiment in Fig. 3B has been reported elsewhere (41). Primers for PCR and antibody suppliers are listed in Tables S3 and S4, respectively. Quantification of PCR data has been described elsewhere (12, 39).

RNA Analysis.

Total RNA isolation and quantification by real-time RT-PCR were performed as described previously (12). Primers including loading controls (SCR1 and ACT1) are listed in Table S5.

Except where indicated, all graphs show the average result of at least three independent experiments and the error bar represents the standard error.