Potential contribution of tandem circadian enhancers to nonlinear oscillations in clock gene expression

Abstract

Limit-cycle oscillations require the presence of nonlinear processes. Although mathematical studies have long suggested that multiple nonlinear processes are required for autonomous circadian oscillation in clock gene expression, the underlying mechanism remains controversial. Here we show experimentally that cell-autonomous circadian transcription of a mammalian clock gene requires a functionally interdependent tandem E-box motif; the lack of either of the two E-boxes results in arrhythmic transcription. Although previous studies indicated the role of the tandem motifs in increasing circadian amplitude, enhancing amplitude does not explain the mechanism for limit-cycle oscillations in transcription. In this study, mathematical analysis suggests that the interdependent behavior of enhancer elements including not only E-boxes but also ROR response elements might contribute to limit-cycle oscillations by increasing transcriptional nonlinearity. As expected, introduction of the interdependence of circadian enhancer elements into mathematical models resulted in autonomous transcriptional oscillation with low Hill coefficients. Together these findings suggest that interdependent tandem enhancer motifs on multiple clock genes might cooperatively enhance nonlinearity in the whole circadian feedback system, which would lead to limit-cycle oscillations in clock gene expression.

FIGURE 1:

Mapping of the upstream region of hCry2 required for transcriptional oscillation. (A) Schematic representation of deletion mutants of a 3-kbp 5’-flanking region upstream from the hCry2 transcription start site. Here +1 corresponds to the transcription start site. (B) Cell culture–based luminescent monitoring was performed with the hCry2 constructs indicated. U2OS human osteosarcoma cells were transfected with hCry2-luc constructs and then stimulated with dexamethasone. After the synchronization, in the presence of luciferin, light emission was measured and integrated for 1 min at intervals of 15 min. The maximum bioluminescence was set to 0.6. The warm and cold colored lines represent data from three independent biological replicates for the full-length hCry2-luc (−2925 to +41) construct and its deletion mutants, respectively. Data sets obtained from three experiments were detrended by subtracting the 24-h running average from raw data. (C) Schematic representation of a deletion series of the hCry2-luc (−249 to +41) construct. E1 and E2 represent putative E-boxes. (D) U2OS human osteosarcoma cells were transfected with hCry2-luc constructs. After the synchronization, in the presence of luciferin, light emission was measured. The maximum bioluminescence was set to 0.6. The warm and cold colored lines represent data from two independent biological replicates for the hCry2-luc (−249 to +41) construct and its deletion mutants, respectively. Data sets obtained from two experiments were detrended by subtracting the 24-h running average from raw data.

FIGURE 2:

Interdependent E-boxes enhance nonlinearity in the rhythm-generating system. (A) Schematic representation of deletion mutants of the hCry2 (–249) construct. The deleted sequences are shown at the bottom. (B) Luciferase assay. NIH3T3 cells were transfected with the indicated constructs. Data show relative firefly luciferase activity (Fluc RLU), which was normalized by Renilla luciferase activity (Rluc RLU). Data are represented as mean ± SE for triplicate samples. Student’s t test was performed for statistical analysis (BMAL1-CLOCK [–] versus BMAL1-CLOCK [+]; *p < 0.001, **p < 0.01, ***p < 0.05). (C) U2OS human osteosarcoma cells were transfected with hCry2-luc constructs. After dexamethasone-induced synchronization, in the presence of luciferin, light emission was measured. The maximum bioluminescence was set to 0.6. The warm and cold colored lines represent data from three independent biological replicates for the hCry2-luc (–2925 to +41) construct and deletion mutants of hCry2-luc (–249 to +41), respectively. Data sets obtained from three experiments were detrended by subtracting the 24-h running average from raw data. (D) Transcription function represented by the Hill equation, f(X₃, K) = k₁/{1 + (X₃/K)ⁿ} (k₁ = 1.5, n = 4.2 or 8; dotted line) or its square form, f(X₃, K) = [k₁/{1 + (X₃/K)ⁿ}]² (k₁ = 1.5, n = 4.2; solid line). Left, CLOCK-BMAL1 level is varied as K ∈ [0, 4], with inhibitor level fixed to X₃ = 1. Right, inhibitor level varied as X₃ ∈ [0, 2], with CLOCK-BMAL1 level fixed to K = 1.

FIGURE 3:

The hCry2 tandem E-boxes interact directly with BMAL1 and CLOCK in an interdependent manner. (A) Nucleic acid sequences of a region of hCry2 containing E1 and E2. Asterisks show conserved nucleotides among human, rat, and mouse. The E-boxes are indicated by red squares. Sequences of mutant oligonucleotides used for EMSA and pull-down experiments are shown as mE1, mE2, and mE1 + mE2. A ³²P-labeled, double-stranded oligonucleotide that contained the hCry2 tandem E-boxes was used as an EMSA probe. The purified BMAL1 and CLOCK complex was incubated with the probe. The BMAL1 and CLOCK–bound probe is indicated as BM-CL-probe. The complex was shifted by addition of specific antibodies (Super shift). (B) Liver extracts were incubated with double-stranded biotinylated oligonucleotides (hCry2-ODN), which were immobilized on streptavidin-Sepharose beads. Negative control samples were incubated with streptavidin-Sepharose beads without oligonucleotides (Unconjugated). The resulting precipitates were subjected to immunoblot analysis with anti-BMAL1 and anti-CLOCK antibodies. (C) Competitor experiments. The purified BMAL1 and CLOCK complex was incubated with the labeled probe containing the hCry2 tandem E-boxes in the presence of unlabeled probes. The amount of the unlabeled probes is indicated as relative levels to the labeled hCry2 tandem E-box probe. Middle, autoradiography signals quantified with Typhoon FLA9500. Data are representative of more than two independent experiments.

FIGURE 4:

The multiplicative effect of tandem circadian elements on transcription reduces the Hill coefficient to realistic values. (A) The original and modified Goodwin models. The variables X₁, X₂, and X₃ can be considered as mRNA levels of the clock genes, their cytoplasmic protein concentrations, and nuclear inhibitor concentrations, respectively. k₁, k₂, and k₃ stand for the production constants, and d₁, d₂, and d₃ determine the degradation rates. K represents the level of the CLOCK–BMAL1 complex relative to the inhibitor, and the term 1/{1 + (X₃/K)ⁿ} is known as the Hill equation and describes a sigmoidal binding of transcription factors to gene promoters. The steepness of transcription is determined by the Hill coefficient, n. (B) Left, damped oscillations observed from the original Goodwin model with n = 4.5. From top to bottom, X₁, X₂, and X₃, respectively. The parameter values are set as k₁ = 1.5, k₂ = 1.5, k₃ = 1.5, d₁ = 0.15, d₂ = 0.15, d₃ = 0.15, and K = 1.5. Right, self-sustained oscillations observed from the modified Goodwin model with n = 4.5. From top to bottom, X₁, X₂, and X₃, respectively. The parameter values are set the same as in the original model. (C) Dependence of oscillation amplitude of the X₁ variable on the Hill coefficient. Squares and circles correspond, respectively, to the original and modified Goodwin models.

FIGURE 5:

The multiplicative effect of tandem ROREs further reduces the Hill coefficient. (A) Equations for the interconnected feedback model (top) and the modified model (bottom). X₁, X₂, and X₃ represent concentrations of Per and Cry mRNA, PER/CRY complex in the cytoplasm, and PER/CRY complex in the nucleus, respectively. X₄, X₅, and X₆ represent concentrations of Bmal1 mRNA, cytoplasmic BMAL1 protein, and BMAL1 protein in the nucleus, respectively. X₇ and X₈ stand for concentrations of BMAL1-CLOCK heterodimer and REV-ERB protein, respectively. The parameter values were set as v_1b =18 nM h^–1, k_1b = 1 nM, K₁ = 0.65 nM, c = 0.001 nM, s = 5, k_1d = 0.12 h^–1, k_2b = 0.3 nM^–1 h^–1, q = 2, k_2d = 0.07 h^–1, k_2t = 0.24 h^–1, k_3t = 0.02 h^–1, k_3d = 0.12 h^–1, v_4b = 3.0 nM h^–1, k_4b = 1 nM, K₄ = 0.9 nM, u = 0.5, k_4d = 1.8 h^–1, k_5b = 0.14 h^–1, k_5d = 0.03 h^–1, k_5t = 0.15 h^–1, k_6t = 0.06 h^–1, k_6d = 0.03 h^–1, k_6a = 0.03 h^–1, k_7a = 0.003 h^–1, k_7d = 0.02 h^–1, v_8b = 10.6 nM h^–1, k_8b = 1 nM, K₈ = 1.1 nM, r = 1, v = 2, and k_8d = 1.5 h^–1. (B) Schematic illustration of the mathematical model. BMAL1–CLOCK heterodimer activates the transcription of Per, Cry, and Rev-erb genes. After translation and posttranslational processes, nuclear PER/CRY complex inhibits transcription of Per, Cry, and Rev-erb genes, forming the main negative feedback loop. Through REV-ERB, which represses the transcription of Bmal1, BMAL1 indirectly inhibits its own transcription, forming an additional negative feedback loop. (C) Dependence of the Hill coefficient and the oscillation amplitude of X₁ variable on the interdependence of E-boxes and ROREs. A multiplicative effect was considered for transcription of only E-boxes (purple), only RORα (green), both E-boxes and RORα (blue), or none (red).