Chromosomal Position Effects Are Linked to Sir2-Mediated Variation in Transcriptional Burst Size

Abstract

Gene expression noise varies with genomic position and is a driving force in the evolution of chromosome organization. Nevertheless, position effects remain poorly characterized. Here, we present a systematic analysis of chromosomal position effects by characterizing single-cell gene expression from euchromatic positions spanning the length of a eukaryotic chromosome. We demonstrate that position affects gene expression by modulating the size of transcriptional bursts, rather than their frequency, and that the histone deacetylase Sir2 plays a role in this process across the chromosome.

Cells display considerable variability in gene expression due to fluctuations in the rates of gene activation, transcription, and translation. In eukaryotes, slow promoter kinetics can result in transcriptional bursting and high cell-to-cell variability (noise) in gene expression (1). This phenomenon has been linked to gene position through spatial variation in the recruitment and retention of transcription factors, nucleosomes, and chromatin remodeling complexes (2–6). However, high-throughput studies of endogenous gene expression in yeast have failed to provide strong support for this hypothesis (7,8), presumably due to the masking of position effects by gene- and promoter-specific variables.

To characterize chromosome position effects independently of gene- and promoter-specific variables, we integrated reporter cassettes at 128 different euchromatic loci along the length of chromosome III in Saccharomyces cerevisiae (Fig. 1 a). We measured reporter expression driven by two promoters with contrasting architectures (9). The ADH1 promoter (P_ADH1) is a covered promoter where gene expression is facilitated by SAGA (10) and a consensus TATA-box occupied by nucleosomes (11). Both features are linked to high transcriptional noise (7,12). The ACT1 promoter (P_ACT1) has a contrasting open promoter architecture where nucleosome deposition is inhibited by the presence of a Poly(dA::dT) tract (9) and through Reb1-mediated DNA bending (13).

Figure 1.

Mapping position effects. (a) Position measured and chromosome III landmarks. (b) Population-average expression. (c) Relative standard deviation (noise). (d) Variation in polymerase II occupancy and H4K16 acetylation across chromosome III (see Supporting Material).

We quantified reporter gene expression by flow cytometry, and calculated the population-average expression and expression noise from measured fluorescence intensity distributions (see Supporting Material). For P_ADH1, the distribution was bimodal at two positions located between the heterochromatic regions at the left telomere and the silenced HML mating type locus (see Fig. S1 in the Supporting Material). A similar effect was recently observed when a fluoresent reporter gene was flanked by artificial Sir-mediated silencing gradients (14). For P_ACT1 at the same positions (Fig. S1), and at all other positions, the intensity distributions were unimodal. In the following, we focus on the unimodal expression distributions.

We observed that the average reporter expression and expression noise vary considerably across the chromosome for both promoters (Fig. 1, b and c). As expected, P_ADH1 is more sensitive to position effects than P_ACT1, and displays the highest variation in both expression level and expression noise. Nevertheless, expression noise is significantly correlated between the two promoters (p = 4 × 10⁻⁹, Table S1), suggesting that the observed position effects are linked to common, promoter-independent factors.

To identify these factors, we compared our data to polymerase II occupancy and histone modifications across chromosome III (Fig. 1 d, (15)). In this analysis, we mitigated potential effects of gene disruption by excluding experimental outliers and averaging over nearest-neighbor positions (see Supporting Material). We observed significant correlations between our data and regions depleted in polymerase II binding and acetylation of histone lysines H3K9, H3K14, and H4K16 (Table S1), targeted by the histone deacetylase Sir2 (16). Notably, expression noise is negatively correlated with polymerase binding (p = 5 × 10⁻⁸ for P_ACT1, p = 3 × 10⁻⁴ for P_ADH1) and H4K16 acetylation (p = 4 × 10⁻¹⁵ for P_ACT1, p = 2 × 10⁻¹⁵ for P_ADH1). Moreover, regions with low polymerase binding and high apparent Sir2 activity are robustly enriched in low expression and high noise positions (Fig. S2). As expected, these positions are predominantly located adjacent to heterochromatin.

To gain insight into the mechanistic origin of chromosomal position effects, we analyzed a stochastic model of gene expression involving transitions between active and inactive promoter states and fluctuations in mRNA and protein abundances (Fig. 2 a). In this model, which is discussed in detail in (1), transcriptional burst size is defined by the ratio of the mRNA synthesis rate (k_M) and the promoter deactivation rate (k_off). Correspondingly, two different scenarios, variation in k_M or in k_off, involve modulation of burst size. We also consider a scenario where position effects arise from variation in burst frequency, which is determined by the promoter activation rate (k_on), and a scenario where transcriptional bursting is absent. Since the four scenarios predict different dependencies of noise on average expression (see Supporting Material), we can use model discrimination techniques to determine which one best explains the experimental data.

Figure 2.

(a) Stochastic model defining position effects in terms of transcriptional bursting. Scenario A, B, and C corresponds to variation in k_M, k_off, and k_on, respectively. Scenario D assumes no transcriptional bursting. (b) Full curves are fit to scenario A. Broken curves indicate the 95% confidence interval. Inset displays all fits to P_ADH1 data. (c) Full and broken curves show the fit of scenario A in the absence (gray) and presence (blue/red) of the nicotinamide, respectively. Inset displays all fits to P_ADH1 data. (d) P_ADH1 expression characterized for five loci in the presence (blue) and absence (red) of Sir2 (A: YCR020C; B: YCL030C; C: YCL037C; D: YCR060W; E: YCL064C). Inset shows the effect of Sir2 deletion on average expression.

The measured dependency of noise on average expression can be fitted well to variation in k_M across the chromosome (r = 0.96 for P_ACT1 and r = 0.91 for P_ADH1, Fig. 2 b). For P_ADH1 (Fig. 2 b, inset), the experimental data is also explained reasonably well by variation in burst size through modulation of k_off (Table S3). Indeed, variation in burst size (through k_M or k_off) explains the experimental data significantly better than variation in burst frequency (p = 6 × 10⁻⁵ for k_M versus k_on, and p = 0.02 for k_off versus k_on). In fact, the burst frequency scenario never performs better than that where transcriptional bursting is absent (Fig. 2 b, inset, Table S3). For P_ACT1, the four scenarios perform equally well, suggesting that its low expression noise is due to low transcriptional bursting.

To establish the involvement of Sir2, we quantified the impact of the Sir2 inhibitor, nicotinamide, on P_ADH1 and P_ACT1 expression across the chromosome (Fig. 2 c). As expected, nicotinamide had the greatest impact at positions associated with high Sir2 activity (Fig. S3). For P_ADH1 expression, the dependency of noise on average expression, measured in the presence of nicotinamide, is captured well by modulation of transcriptional burst size across the chromosome (Figs. 2 c and Fig. S5). As before, variation in burst size (through k_M or k_off) is significantly better at explaining the experimental data comparative variation in burst frequency (p = 3 × 10⁻³⁷ for k_M versus k_on, and p = 4 × 10⁻⁴⁶ for k_off versus k_on, Fig. 2 c).

We confirmed a role of Sir2 by characterizing the effect of Sir2 deletion on P_ADH1 expression at a position adjacent to the heterochromatic HML (YCL064C), and four other randomly chosen positions. The greatest effect was observed adjacent to the HML (Fig. 2 d). For all tested positions, the measured effect is captured by variation in burst size, but not by variation in burst frequency (p = 3 × 10⁻¹¹ for k_M versus k_on, p = 10⁻²² for k_off versus k_on, Fig. 2 d). Interestingly, in these experiments, modulation of mRNA synthesis rates captures the data better than variation in promoter deactivation rates (p = 3 × 10⁻²).

Our finding that chromosomal position modulates burst size rather than frequency is consistent with previous studies of gene expression noise in mammalian cells using randomly integrated viral promoters (4,6). It is also consistent with the finding that the chromatin structure established by Sir2 is permissive to promoter activation, and suppresses mRNA synthesis by blocking a step downstream of transcription initiation (17). Indeed, our sir2Δ mutant data supports a model where Sir2 modulates the rate of mRNA synthesis differentially across the chromosome. However, the mechanisms involved in Sir2-mediated transcriptional repression remain controversial. Notably, P_ADH1 and P_ACT1 are strong promoters, and it is possible that Sir2 has more profound effects on the burst frequency of weak promoters. Additionally, whereas we observe significant Sir2-linked effects at most positons across the chromosome, the activity of Sir2 is typically viewed as being restricted to heterochromatic regions. This view has been challenged by systematic analysis documenting widespread binding of Sir2 across euchromatic genes, including ACT1 and ADH1 and other highly transcribed genes (18). Given our observations, it appears that Sir2 may play a more global role than previously anticipated.