Linking stochastic fluctuations in chromatin structure and gene expression

Abstract

The number of mRNA and protein molecules expressed from a single gene molecule fluctuates over time. These fluctuations have been attributed, in part, to the random transitioning of promoters between transcriptionally active and inactive states, causing transcription to occur in bursts. However, the molecular basis of transcriptional bursting remains poorly understood. By electron microscopy of single PHO5 gene molecules from yeast, we show that the "activated" promoter assumes alternative nucleosome configurations at steady state, including the maximally repressive, fully nucleosomal, and the maximally non-repressive, nucleosome-free, configuration. We demonstrate that the observed probabilities of promoter nucleosome configurations are obtained from a simple, intrinsically stochastic process of nucleosome assembly, disassembly, and position-specific sliding; and we show that gene expression and promoter nucleosome configuration can be mechanistically coupled, relating promoter nucleosome dynamics and gene expression fluctuations. Together, our findings suggest a structural basis for transcriptional bursting, and offer new insights into the mechanism of transcriptional regulation and the kinetics of promoter nucleosome transitions.

Figure 1. Different modes of gene regulation predict distinct expression noise profiles.

(A) The “two-state model” of stochastic gene expression. The model simplifies promoter state dynamics into the stochastic transitioning between two states, ON (transcriptionally active), and OFF (inactive). Transitions→ indicate degradation of the gene product. Greek letters refer to transition probabilities per unit time and molecule (“kinetic parameters”); below, a typical time trace (black curve) for the fluctuation in single cell mRNA molecule number about the steady state mean (dashed gray line). (B) “Deterministic model” of a transcriptionally active gene. The black curve beneath the model represents a typical time trace of mRNA fluctuation about the same mean (dashed gray line) as in (A). (C) Steady-state Fano factor values (Fano) were calculated as a function of a single kinetic parameter (the “regulatory parameter”), with all other kinetic parameters held constant. The coloring of the resulting noise profiles refers to the identity of the kinetic parameter that was allowed to float to vary the mean abundance of protein molecules (mean abundance). Thus, blue refers to the bursting frequency α, see (A), etc. The dashed green line indicates the expected Fano profile for the modulation of ε for the deterministic model B. Noise profiles were determined by analytical calculations as described in Materials and Methods.

Figure 2. EM analysis of single gene molecules.

(A) PHO5 gene chromatin rings were formed by site-specific recombination in vivo . Isolated chromatin rings were crosslinked with psoralen, denatured, and analyzed by EM. Positions of UASp1, UASp2, and the TATA box are indicated by a black, gray, and white circles, respectively; gray ovals represent nucleosomes; promoter nucleosomes are in light gray; RS refers to the recognition sequence for site-specific recombination; and lexA refers to a cluster of LexA operators for ring purification. (B) EM images of transcriptionally inactive PHO5 rings (pho4Δ pho80Δ). (C) EM images of transcriptionally active PHO5 rings (PHO4 pho80Δ). Black arrowheads indicate nucleosome-free DNA segments long enough to accommodate one or more nucleosomes. Bars denote 100 nm.

Figure 3. Nucleosome configurations of “activated” promoters.

PHO5 gene molecules are aligned with their 3′ forked end on the top. A bent arrow indicates the position of the transcription start site. The inferred promoter nucleosome configuration is shown in the left lower corner of each image, where the promoter is represented by a box and occupied nucleosome positions by black dots. The top position represents N-1, the middle position N-2, and the bottom position N-3. Nucleosome configurations representing all eight possible combinations of occupied and unoccupied positions N-1 to N-3 were observable. Bars indicate 100 nm. (See also Figure S1).

Figure 4. Probabilities of promoter nucleosome configurations.

Bars indicate the relative frequencies of promoter nucleosome configurations (“probability”) for EM datasets (Table S1). Model predictions are indicated by black dots, connected by gray edges to aid the visual comparison with EM data. Promoter nucleosome configurations are represented as in Figure 3. (A) Configurational probability distribution of PHO5 promoter nucleosomes in activated cells (PHO4 pho80Δ). Numbers above and below horizontal lines refer to the sum of probabilities for 2-nucleosome and 1-nucleosome configurations, determined by EM (above light gray line), or model calculation (below dark gray bar). Predictions were based on the transition topology in (B). (The same predictions were obtained for a model with “symmetric sliding,” which allows for all possible sliding transitions.) (B) Transition topology without nucleosome sliding; nucleosome assembly and disassembly transitions are indicated by gray and black arrows, respectively. (C) Same as (A), however with predictions based on the topology in (D). The statistical support of the topology in (D) against its rival hypothesis in (B) given the EM dataset R was ; i.e., R was -fold more probable given (D) than given (B) (Materials and Methods). (D) Transition topology with unidirectional nucleosome sliding; dashed arrows indicate sliding transitions. (E) Same as (A), with predictions based on the topology in (F); (and hence ). (G) Transition topology for “stable nucleosome retention.” This hypothesis was disproved by R, for , but ; thus, its likelihood, given R, was . (H) Transition topology for all-or-nothing disassembly: . (I) Transition topology for “deterministic cyclical process”; . The transition topologies in (G) to (I) were refuted given the strong support for topologies in (D) and (F) against their rival hypotheses. For parameter values see Table S3.

Figure 5. Configurational probability distributions in activator and promoter mutants.

(A) Distribution for molecules isolated from PHO4 pho80Δ cells with a mutated PHO5 TATA box (tata). Theoretical predictions and experimental results (R) are indicated by a chain of dots and histogram bars, respectively, as in Figure 4. Predictions were based on the topology of Figure 4D. The statistical support against its rival of Figure 4F was . (B) Distribution for molecules isolated from pho4[85-99] pho80Δ cells with a mutated PHO5 TATA box. PHO5 expression in pho4[85-99] pho80Δ cells with wild type PHO5 TATA is 0.14 relative to PHO4 wild type (set to 1) . Predictions were based on the topology of Figure 4F, rather than 4D; . (C) Distribution for molecules isolated from pho4Δ pho80Δ cells with a mutated PHO5 TATA box. Expression of PHO5 in pho4Δ pho80Δ cells relative to PHO4 wild type is ∼0.005 (see below and [15]). Predictions were based on the topology of Figure 4D, rather than 4F; . (D) Distribution for molecules isolated from PHO4 pho2Δ cells grown in high phosphate (PHO4^cyt); Pho4 is in the cytoplasm, rather than the nucleus, and PHO5 is repressed, therefore. Predictions were based on the topology of Figure 4D, rather than 4F; . (E) Distribution for molecules isolated from PHO4 cells grown in high phosphate (PHO4^cyt). Predictions were based on the topology of Figure 4D, rather than 4F; . For parameter values see Table S3.

Figure 6. Noise profiles of PHO5 expression.

The results of intrinsic protein noise measurements are indicated by blue dots for strains with mutations in the Pho4 activation domain, and white squares for pho5 UASp1 and UASp2 mutants. Intrinsic noise was measured using the dual reporter system , where cyan fluorescent protein (CFP) and yellow fluorescent protein (YFP) were expressed in diploid cells under control of the PHO5 promoter. (A) Fano factor profile of PHO5 protein noise; Fano factor and mean abundance are indicated in molecule numbers, based on the assumption that the average number of protein molecules expressed under repressing conditions is 1,000 per cell . (The exact number is unimportant to our principle conclusions.) (B) CV ² profile of PHO5 protein noise. Mean (protein) abundance is in units of wild type expression. Curves represent predictions based on the two-state model (Figure 1A), with the burst frequency α as the regulatory parameter, and different probabilities of the ON state in wild type (p _ON). For all calculations, min⁻¹, h⁻¹, h⁻¹ (see Figure 1A, and main text below; like Pho5, CFP and YFP are biochemically stable; the proteins are lost therefore primarily due to dilution by cell division). With , the kinetic parameter for transcription is min⁻¹. The parameter values were determined as described in the main text.

Figure 7. Abundance and half-life of mRNA.

(A) FISH of activated cells (PHO4 pho80Δ) in which the endogenous PHO5 promoter drove expression of CFP. For FISH, we used CFP anti-sense DNA oligonucleotides labeled with Alexa 555. Cells contained 60 CFP-mRNA molecules, on average. The nucleus was stained with DAPI (blue-gray). (B) FISH of repressed cells (pho4Δ pho80Δ) revealed an average of ∼0.3 CFP-mRNA molecules per cell. (C) The PHO5 promoter was induced in PHO80 wild type cells by incubation in phosphate-free medium. RNA samples were taken at different time points following the addition of inorganic phosphate to the medium (+P_i). RNA was fractionated by agarose gel electrophoresis, blotted, and hybridized with ³²P-labeled DNA probes against CFP and ACT1 mRNAs. (D) The natural logarithm of the radioactive signal ratio for CFP and ACT1 mRNAs normalized to 1 for 0 minutes of +P_i (ln[C/A]) was plotted against the time after addition of phosphate; ln2 is reached at about 10 min, the half life of the transcript, which corresponds to a kinetic parameter for mRNA degradation of h⁻¹. A closely similar value for δ was obtained for the PHO5 mRNA (data not shown). The steady state abundance of PHO5 transcripts is therefore expected to be similar to the number of CFP mRNAs per cell.

Figure 8. Integrated model of promoter nucleosome dynamics and gene expression.

(A) Transition topology of the integrated model. Promoter states are represented by boxes, and black dots indicate the nucleosome configuration of the state, as in Figures 3–5. Promoter states with transcriptionally conducive nucleosome configuration are shaded gray, where light gray promoter states are conducive, and dark gray states are active states; inconducive states are represented by white boxes. Nucleosome assembly and disassembly transitions between states 8 and 9, and 10 and 11 were omitted for graphical clarity. (B) The data are the same as in Figure 6 , but the predicted noise profile (blue curve) was calculated (Materials and Methods) using the integrated model with , , and λ as regulatory parameters, which were allowed to float along the line , with , , where the hat and prime mark the parameter values for the PHO4 wild type and pho4Δ mutant, respectively, and t is a real number ≥0. Noise predictions based on the assumption of , , and ε as regulatory parameters, and thus assuming a combination of burst size and burst frequency control, are indicated by the gray curve. Virtually the same result was obtained on the topology of Figure 4F for nucleosome transition (not shown).