Intron length increases oscillatory periods of gene expression in animal cells

Abstract

Introns may affect gene expression by increasing the time required to transcribe the gene. One way for extended transcription times to affect the behavior of a gene expression program is through a negative feedback loop. Here, we show that a logically engineered negative feedback loop in animal cells produces expression pulses, which have a broad time distribution that increases with intron length. These results in combination with mathematical models provide insight into what may produce the intron-dependent pulse distributions. We conclude that the long production time required for large intron-containing genes is significant for the behavior of gene expression programs.

Figure 1.

An engineered network with delayed autoinhibition exhibits pulses of protein expression. (A) Schematic diagram of the reporter used for the study of intron delays in a gene network with delayed autoinhibition (sequences and assembly strategy are detailed in Supplemental Table 1). The mouse β-actin promoter (Pactin) that includes its first intron drives the expression of the reporter. Fluorescence detection is possible because of YFP, the Venus fast-maturing variant of yellow fluorescent protein. TetR is the tetracycline repressor that, when delivered to the nucleus because of a fusion with a nuclear localization signal (NLS), binds to operator sequences (o) that have been engineered into the promoter to drive repression by steric hindrance and promoter looping-out. The PEST sequence (rich in proline, glutamic acid, serine, and threonine) is a protein degradation element originating in the mouse ornithine decarboxylase gene. AU-rich elements present in the 3′-untranslated region of our gene were included to reduce the stability of the mRNA. Destabilization of both the protein and mRNA increases the responsiveness of the negative feedback loop. (B) Fluorescence micrographs of YFP signal and phase contrast micrographs from a 3T3 cell of a clone that contains a Flp-In, 3-kb version of the gene depicted in A. Images were taken every 6 min, but only images for every 10th exposure are presented. (C) Time-lapse trajectory of the quantified average fluorescence intensity in arbitrary units (a.u.) from the cell in B. The data presented were acquired in a 28-h window that encompasses a single cell division.

Figure 2.

Time between pulses of protein expression increases with intron length. (A,C,E) Representative expression trajectories from single cells of clones (each color is a different cell) containing either a 3-kb (A), 10-kb (C), or 19-kb (E) version of the gene from Figure 1. (B,D,F) One-hundred-fifty-two cells were examined and presented are histograms of pulse lengths (n = 97–200) from single cells within the clones represented in A–C. Pulses are interrupted by the mitotic constraint as in E and G (red arrow highlights the mitotic event). (G) After a cell division, sister cells behave similarly for a window of time (here ∼8 h) before the reporter’s behavior begins to diverge in the sibling cells. In the bottom panel, the area of the dividing cell is present to illustrate several hallmarks that indicate a mitotic event: rounding (sharp decrease in area) and division (area becomes about half of predivision area). (H) Over longer periods of time, cells without autoinhibition exhibit relatively constant expression that fluctuates slowly.

Figure 3.

The impact of the mitotic constraint increases with intron length. (A) Cells without autoinhibition were observed at the higher time resolution of 2 min as they went through mitosis. Here, the presented fluorescence values are total fluorescence because segmented object (the nucleus) changes size rapidly during and shortly after mitosis. In early G1, after two nuclei are resolvable, there is a dip in expression (gray stripes, A) that persists for longer lengths of time as the size of the gene increases, quantified for multiple cell division in B (n = 10,10, and 8 for 3-kb, 10-kb, and 19-kb genes). Error bars mark the standard deviation.

Figure 4.

Simulations of delayed autoinhibition suggest that bursting accumulates during transcription elongation. (A) Schematic of delayed autoinhibition that highlights bursting originating from either transcription initiation or transcription elongation involving polymerases that transcribe with a broad distribution of velocities (v_i). (B) Trajectories from simulations of delayed autoinhibition without bursting exhibit little variation in period (shown in C). (D) Trajectories from simulation of delayed autoinhibition that incorporate transcription initiation bursts exhibit a broad distribution of period lengths (shown in E). (F) When simulations incorporate transcription bursts due to congestion and traffic jams during elongation, broad distributions of the pulse period are observed (G,H). (H) As gene length increases in simulations, only those that incorporate elongation bursts exhibit a similar length sensitive increase in the standard deviation of period lengths observed experimentally. (I) In simulations that involve slow leading RNA polymerases, the standard deviation of the pulse periods increases rapidly when the standard deviation of polymerase velocities exceeds 0.2 kb/min.