Fluorescence activated cell sorting followed by small RNA sequencing reveals stable microRNA expression during cell cycle progression

Abstract

Background: Previously, drug-based synchronization procedures were used for characterizing the cell cycle dependent transcriptional program. However, these synchronization methods result in growth imbalance and alteration of the cell cycle machinery. DNA content-based fluorescence activated cell sorting (FACS) is able to sort the different cell cycle phases without perturbing the cell cycle. MiRNAs are key transcriptional regulators of the cell cycle, however, their expression dynamics during cell cycle has not been explored.

Methods: Following an optimized FACS, a complex initiative of high throughput platforms (microarray, Taqman Low Density Array, small RNA sequencing) were performed to study gene and miRNA expression profiles of cell cycle sorted human cells originating from different tissues. Validation of high throughput data was performed using quantitative real time PCR. Protein expression was detected by Western blot. Complex statistics and pathway analysis were also applied.

Results: Beyond confirming the previously described cell cycle transcriptional program, cell cycle dependently expressed genes showed a higher expression independently from the cell cycle phase and a lower amplitude of dynamic changes in cancer cells as compared to untransformed fibroblasts. Contrary to mRNA changes, miRNA expression was stable throughout the cell cycle.

Conclusions: Cell cycle sorting is a synchronization-free method for the proper analysis of cell cycle dynamics. Altered dynamic expression of universal cell cycle genes in cancer cells reflects the transformed cell cycle machinery. Stable miRNA expression during cell cycle progression may suggest that dynamical miRNA-dependent regulation may be of less importance in short term regulations during the cell cycle.

Keywords: Cell cycle; DNA staining; Dynamic expression; Fluorescence-activated cell sorting (FACS); miRNA.

Fig. 1

Fluorescence activated cell sorting (FACS) analysis of stained cells and validation of cell cycle sorted populations in HDFa, NCI-H295R and HeLa cells. Panels a-c: FACS analysis shows cell cycle distribution of stained cells before (first image) and after sorting to G1, S and G2 phases (second, third and fourth images, respectively; Panel a: HDFa, Panel b: NCI-H295R, Panel c: HeLa. Intervals of fluorescence intensity (shown as vertical bands) were defined to gate G1, S and G2 phases, respectively. Panels d-e: Western blot analysis confirms the high efficacy of cell cycle sort (Panel d: NCI-H295R, Panel e: HeLa). Density values of phospho(Tyr15)-CDC-2 were first normalized to β-actin loading control and were further normalized to G1 phase (displayed above each band)

Fig. 2

High throughput gene expression profiling, qRT-PCR validation and functional bioinformatics analysis of cell cycle dependent transcription in HDFa, NCI-H295R and HeLa cells. Panels a-b: Heat map of mRNA transcripts with significantly different expression between cell cycle phases (Panel a: NCI-H295R, Panel b: HeLa). Above the heat maps, hierarchical clustering upon differently expressed genes are also shown (additional information, including the list of gene symbols and descriptions can be found in Additional file 1: Table S3, panels B and C). Panels c-e: qRT-PCR validation of six genes chosen upon microarray analysis (Panel c: HDFa, Panel d: NCI-H295R, Panel e: HeLa). In each case ΔCt value was normalized to G1 phase (ΔΔCt) and was subjected to FC = 2^-ΔΔCt transformation. Error bars show standard deviation. Asterisks mark statistical significance (p < 0.05). Panels f-h: Molecular and cellular functions concerned by gene expression alterations in cell cycle phases. Δ(G2-G1) gene expression changes of significantly differently expressed genes (Panel g: NCI-H295R, Panel h: HeLa) or genes with fold change > 2 expression (Panel f: HDFa) were subjected to IPA core analysis. In each cell, the five most significantly concerned networks are shown. Significance threshold (p = 0.05) corresponds to –log(p-value) = 1.301. For additional molecular and cellular functions see Additional file 2: Figure S2

Fig. 3

Comparison of cell cycle dependent gene expression observed by cell cycle sort and synchronization experiments and analysis of gene expression dynamics during the cell cycle in various cell types. Panel a: Venn diagram of cell cycle dependent genes detected in cell cycle sorted primary fibroblasts (HDFa SORT), in synchronized primary fibroblasts (PF synchr – data from [5]), in cell cycle sorted HeLa cells (HeLa SORT) and in synchronized HeLa cells (HeLa synchr – data from [4]). Intersections present the number of commonly found genes. For gene lists see Additional 1: Table S4. Panels b-c: Correlation analysis of gene expression differences using normalized expression values obtained from microarray experiments. Pearson’s correlation coefficient was calculated from Δ(G2-G1) expression changes detected by cell cycle sort and former synchronization experiments in primary fibroblasts (synchronization method: serum starvation – SS, Panel b) and HeLa (synchronization method: double thymidine block – DT, Panel c) cells. Correlation coefficients are displayed. Asterisks mark statistical significance (p < 0.05). For additional correlation calculations see Additional 2: Figures S3 and S4. Panels d-e: Analysis of normalized gene expression values of different cell types. Normalized gene expression of 127 cycling genes exported from microarray data (Panel d) and 10 cycling genes exported from qRT-PCR data (Panel e) were analyzed. Note the lower ΔCt(ACTB) values mean higher expression values. Error bars show standard deviation. Asterisks mark statistical significance (p < 0.05). Panels f-g: Analysis of mean gene expression fold change between cell cycle phases of different cell types. Absolute values of fold change of 127 cycling genes exported from microarray data (Panel f) and 10 cycling genes exported from qRT-PCR data (Panel g) of cell cycle sort experiments were analyzed. Error bars show standard deviation. Asterisks mark statistical significance (p < 0.05)

Fig. 4

Analysis of cell cycle dependent miRNA expression by high-throughput screenings methods and individual qRT-PCR in HDFa, NCI-H295R and HeLa cells. Panels a-d: Fold change (log2) of miRNA expression in S/G1, G2/S and G2/G1 phases observed by microarray (Panel a: NCI-H295R cells, Panel c: HDFa cells), TaqMan Low Density Array (Panel b: NCI-H295R cells) and Illumina Small RNA Sequencing (Panel d: HeLa cells). Grey background corresponds to fold change ≤ 2 between cell cycle phases. Colored squares represent miRNAs with significantly different expression in cell cycle phases obtained by statistical analysis of high-throughput data. Colored circles represent some members of the hsa-miR-16 family, if available. For further high-throughput data of NCI-H295R cells (Illumina Small RNA Sequencing) see Additional 2: Figure S5. For qRT-PCR measurement data see Additional 2: Figure S6. Panels e-g: QRT-PCR measurements of hsa-miR-16 family members hsa-miR-16, hsa-miR-15a and hsa-miR-503 in HDFa (Panel e), NCI-H295R (Panel f) and HeLa (Panel g) cells. In each case ΔCt values were normalized to G1 phase (ΔΔCt) and were subjected to FC = 2^-ΔΔCt transformation. Error bars show standard deviation. Asterisks mark statistical significance (p < 0.05)

Fig. 5

Schematic presentation of the hypothesis concerning expression dynamics of mRNAs and miRNAs during the cell cycle phases in primary untransformed and cancer cells. Relative expression changes in G1, S and G2 phases of a representative cell cycle gene and miRNA in untransformed primary (Panel a) and transformed cancer (Panel b) cells. Relative expression changes of mRNAs exhibiting cell cycle dependent expression (dark blue) and cell cycle associated miRNAs (green) are shown. Blue double arrows mark the amplitude of dynamic expression range in cell cycle phases. Note the different phase-specific expression values and the different dynamic expression range in the two cell types. Upon our results, genes exhibiting cell cycle dependent expression profile in untransformed cells are characterized with lower expression levels throughout the cell cycle, however they possess greater variance in expression levels than in transformed, cancer cells. MiRNAs, however, do not display cell cycle dependent expression