Mechanistic Interpretation of Promoter-Proximal Peaks and RNAPII Density Maps

Genome-wide RNA polymerase II (RNAPII) density profiles provide averaged snapshots of transcription but are difficult to interpret in the context of dynamic gene expression. We performed computational modeling to simulate RNAPII density profiles from individual transcription parameters and constructed simple mathematical models to explore general relationships between transcription parameters and density profiles. The density of RNAPII on genes will depend mainly on three parameters: elongation rate (bases added per unit time), initiation frequency (number of start events per unit time that result in productive elongation), and processivity (fraction of polymerases remaining on the template after each catalytic event). Differences in elongation rate will affect the density via changes in the average spacing between polymerases (Figure 1A, top). It can be shown that this results in the density being inversely proportional to the elongation rate when the initiation frequency remains unchanged (see Data S1 available online for all mathematical derivations). A computational model, ChIPMOD, that simulates RNAPII density patterns based on the basic transcription parameters independently confirms this relationship, both for the total density across a gene and for density peaks that result from local pauses, defined here as regions of slow elongation (the reader is encouraged to test the relationships mentioned throughout the text by using the online version of the program (http:// www.chipmod.org.uk; seeData S2 for details). The initiation frequency is a key regulated step in the transcription cycle. In agreement with the intuitive expectation that a higher initiation frequency will result in a higher density of polymerases (Figure 1A, bottom), the average RNAPII density across a gene can be shown to be directly proportional to the average initiation frequency on that gene, assuming constant elongation rates. Even though density patterns are affected by processivity, we assume here that the processivity of RNAPII is high, in linewith the observation that chromatin immunoprecipitation sequencing (ChIP-seq) read densities are typically largely flat across most of the coding region of a gene. Nevertheless, as outlined below, imperfect processivity at the beginning of genes may contribute to specific density peaks observed at metazoan genes. We note that, although the relationships between elongation rate, initiation frequency, and RNAPII density may appear intuitive, it can often be difficult to determine the impact of changes. This is in part because initiation frequency and elongation rate have similar but opposite effects on the density pattern, making it difficult to distinguish between the two. For example, Spt5 knockdown resulted in a significant increase of polymerase density across the body of several genes (Rahl et al., 2010). The authors concluded that this was due to increased release of RNAPII from the promoter. However, given that Spt5 is also an elongation factor, the higher RNAPII density across the coding region could arguably be equally well explained by a lower elongation rate. It is worth emphasizing that an increased mRNA output can only be achieved by increasing the initiation frequency, but not by increasing the elongation rate: at an initiation frequency of ten polymerases per minute, for example, the mRNA synthesis rate can never exceed ten molecules per minute, regardless of the elongation rate. Indeed, only when the elongation rate is very low relative to the initiation frequency will it affect expression levels, as elongating polymerases may then saturate the beginning of a gene and prevent new rounds of initiation