Heritable gene expression variability and stochasticity govern clonal heterogeneity in circadian period

Abstract

A ubiquitous feature of the circadian clock across life forms is its organization as a network of cellular oscillators, with individual cellular oscillators within the network often exhibiting considerable heterogeneity in their intrinsic periods. The interaction of coupling and heterogeneity in circadian clock networks is hypothesized to influence clock's entrainability, but our knowledge of mechanisms governing period heterogeneity within circadian clock networks remains largely elusive. In this study, we aimed to explore the principles that underlie intercellular period variation in circadian clock networks (clonal period heterogeneity). To this end, we employed a laboratory selection approach and derived a panel of 25 clonal cell populations exhibiting circadian periods ranging from 22 to 28 h. We report that a single parent clone can produce progeny clones with a wide distribution of circadian periods, and this heterogeneity, in addition to being stochastically driven, has a heritable component. By quantifying the expression of 20 circadian clock and clock-associated genes across our clone panel, we found that inheritance of expression patterns in at least three clock genes might govern clonal period heterogeneity in circadian clock networks. Furthermore, we provide evidence suggesting that heritable epigenetic variation in gene expression regulation might underlie period heterogeneity.

Fig 1. Both heritable and nonheritable stochastic components contribute to clonal circadian period heterogeneity.

(A) Divergence of circadian period distributions of short-period (red) and long-period (blue) clones from a common founding culture (gray) across multiple assay generations. Dashed black lines depict the mean of respective period distributions. The gray dashed lines extended from assay generation 1 depict mean period of the founding culture (assay generation 0) for visual assessment of the period divergence. Red arrows (short-period clone) and blue arrows (long-period clone) indicate the periods of representative clones selected for the successive assay generation. (B) Detrended bioluminescence traces of representative clones from founding culture (dashed line), SCL (red), and LCL (blue). Detrending results in truncation of 12-h data at the beginning and end of time series. (C) Mean circadian periods of three representative SCLs (SCL1–3) and LCLs (LCL1–3) across four assay generations. Error bars are SD (n = 3 experiments). Letters a–c depict statistical significance for both interassay and intra-assay generation comparisons. Same letters indicate lack of statistical significance between the average periods of SCLs and LCLs, whereas different letters indicate statistically significant differences in period (ANOVA followed by Tukey’s HSD; p < 0.001). Underlying data for this figure can be found in S1 Data. ANOVA, analysis of variance; HSD, honestly significant difference; LCL, long-period clonal line; rel., relative; SCL, short-period clonal line.

Fig 2. Rhythm stability is associated with period, and clonal period heterogeneity does not stem from polymorphisms.

(A) Correlation of SD of intercycle (peak-to-peak) period with clonal circadian period indicating a reduction in rhythm stability with increasing clock period. (B) Venn diagrams depicting overlap of SNPs and CNVs identified among SCLs and LCLs with the 243 period modifier genes reported by Zhang and colleagues [46]. For SNPs, numbers within brackets indicate the total number of SNPs identified, and numbers otherwise represent the total number of genes harboring the identified SNPs. (C) Correlation of rhythm parameters—average bioluminescence (“biolum.”; green), relative amplitude (“rel. amp.”; blue), and damping rate (red)—with clonal circadian period. (D) Regression of average bioluminescence (“avg. biolum.”) of parental clones over progeny clones, indicating that parental average bioluminescence is a very good predictor of progeny bioluminescence. Green solid line is the linear regression fit with its 95% CI (green dotted line). Underlying data for (A), (C), and (D) can be found in S1 Data. Data used for (B) are available at https://zenodo.org/ (DOI: 10.5281/zenodo.3876533). CNV, copy number variant; cps, counts per second; LCL, long-period clonal line; SCL, short-period clonal line; SNP, single nucleotide polymorphism.

Fig 3. Inheritance of clock-gene expression patterns might govern clonal period heterogeneity.

(A) Scree plot depicting the percentage of variance explained by the 19 PCs (black bars) and the expected values based on the Broken Stick model (red line). (B) Factor map of individual clones plotted across PCs 1 and 2 reveals that the first two PCs cluster the clones in three clusters of short- (red), intermediate- (black), and long-period (blue) clones. (C) Cos² values (a measure of the extent of influence of a gene on the PC) of the 19 genes for PCs 1–5. The color and size of circles represent the magnitude of cos² value. (D) Hierarchical clustering based on the expression of five genes selected from PC2. With the exception of one clone, all others clustered into three groups of short, intermediate, and long clones (red, black, and blue dashed rectangles, respectively). (E) Hierarchical clustering based on the expression of five genes from PC1 resulted in two clusters: (1) intermediate period (black dashed rectangle) and (2) short and long period (green dashed rectangle). Color coding of clones in (D) and (E) is the same as in (B). Underlying data for this figure can be found in S1 Data. PC, principal component.

Fig 4. Epigenetically regulated expression of E-Box-associated factors may govern clonal period heterogeneity.

Period change (compared to nonsilencing scrambled shRNA control) upon knockdown of the (A) five PC2 genes and (B) three PC1 genes for the short- (red), intermediate- (gray), and long-period (blue) clones. Bars with different letters indicate significant differences (p < 0.05), and bars with the same letter are not significantly different from each other (mixed-model ANOVA followed by Tukey’s HSD). Numbers within square brackets indicate the knockdown efficiency of the respective gene. (C) Averaged absolute (“abs.”) period change across all clones upon knockdown of genes from PC2 and PC1. (D) Period change (compared to vehicle control) upon treatment of short- (red), intermediate- (gray), and long-period (blue) clones with HDAC inhibitor SAHA (1.6 μM). (E) Methylation percentage of CpG islands proximal to the transcription start site of the three genes—ARNTL2, NR1D2, and BHLHe49. For all panels in this figure, n = 3–4 experiments and error bars are SD (*p < 0.05; **p < 0.001; ****p < 0.0001). Underlying data for this figure can be found in S1 Data. ANOVA, analysis of variance; HDAC, histone deacetylase; HSD, honestly significant difference; LCL, long-period clonal line; PC, principal component; SAHA, suberoylanilide hydroxamic acid; SCL, short-period clonal line; shRNA, short hairpin RNA.