Correlation of circular RNA abundance with proliferation--exemplified with colorectal and ovarian cancer, idiopathic lung fibrosis, and normal human tissues

Abstract

Circular RNAs are a recently (re-)discovered abundant RNA species with presumed function as miRNA sponges, thus part of the competing endogenous RNA network. We analysed the expression of circular and linear RNAs and proliferation in matched normal colon mucosa and tumour tissues. We predicted >1,800 circular RNAs and proved the existence of five randomly chosen examples using RT-qPCR. Interestingly, the ratio of circular to linear RNA isoforms was always lower in tumour compared to normal colon samples and even lower in colorectal cancer cell lines. Furthermore, this ratio correlated negatively with the proliferation index. The correlation of global circular RNA abundance (the circRNA index) and proliferation was validated in a non-cancerous proliferative disease, idiopathic pulmonary fibrosis, ovarian cancer cells compared to cultured normal ovarian epithelial cells, and 13 normal human tissues. We are the first to report a global reduction of circular RNA abundance in colorectal cancer cell lines and cancer compared to normal tissues and discovered a negative correlation of global circular RNA abundance and proliferation. This negative correlation seems to be a general principle in human tissues as validated with three different settings. Finally, we present a simple model how circular RNAs could accumulate in non-proliferating cells.

Figure 1. Circular RNA expression from RNA-seq data of normal colon mucosa (Normal) and colorectal carcinoma (Carcinoma) samples.

(a) Expression levels of all biuniquely annotatable and reliably quantifiable 85 circular RNAs in twelve matched normal and carcinoma samples; normalised log₂ values were averaged among the samples (p<2.2e-16). (b) Expression levels of reliably quantifiable 260 circular RNAs in pooled normal and carcinoma samples; log₂ values were normalised to total sequencing read numbers (p<2.2e-16). Significance was assessed by paired student's t-tests (2.2e-16 is the minimum value of the used test in R).

Figure 2. Resistance of circular RNAs to RNase R treatment.

RNase R enriched circular compared to linear isoforms in pooled normal colon mucosa and tumour tissue samples of colorectal cancer patients. Enrichment correlates negatively with length of the circular RNA (cf. Table 1). CircRNA 7780 was normalised to housekeeping genes due to lack of the corresponding linear isoform. (Y-axis log₂ scale of enrichment of the circular compared to the linear isoform).

Figure 3. Ratio of circular to corresponding linear RNAs in 31 matched normal mucosa (No) and colorectal tumour (Ca) tissues and in eleven colorectal cancer cell lines (CL).

Expression of (i) five exemplarily chosen circular RNAs (circ0817, circ3204, circ6229, circ7374, and circ7780), (ii) four corresponding linear RNAs (CUL5, USP3, METTL3, and TNS4), and (iii) ratios of circular to linear RNAs (circRNA/linRNA); determined by RT-qPCR. Circ7780 was not detectable (n.d.) in cancer cell lines. Y-axes represent relative expression values normalised to housekeeping genes for the upper two panels and the ratios of circular RNAs to linear RNAs in the lower panel. Significance for multiple comparison of No versus Ca versus CL was assessed by Kruskal-Wallis rank sum tests (given p-values) followed by pairwise comparisons applying the post-hoc Nemenyi test with Chi-squared approximation (significant pairs indicated by blue brackets).

Figure 4. Correlation of ratios circRNA/linRNA to cell proliferation.

The histograms in each of the subpanels show the distribution of Ki-67 proliferation indices (upper left) and the distribution of (a) the ratios of circRNA/linRNA and of (b) relative circ7780 expression (lower right) in matched normal mucosa and colorectal tumour tissues. The dot plots illustrate the correlation between the Ki-67 index (x-axes) and the (a) ratios of circRNA/linRNA and the (b) relative circ7780 expression (y-axes) together with linear (dashed) and loess regression (continuous) lines. The axes of the dot plots apply also to the x-axes of the histograms: the x-axes of the dot plots correspond to the Ki-67 histograms; the y-axes of the dot plots correspond to the histograms of (a) circRNA/linRNA ratios or of (b) relative circ7780 expression. The corresponding Spearman's rank correlation values are depicted in the upper right. Significance: °p<0.1, *p<0.05, **p<0.01, ***p<0.001. (c) Representative Ki-67 immunohistochemical stainings of two example tissues, quantified with an automated analysis pipeline using the ImmunoRatio ImageJ plugin (lower panel). Scale bar, 100 µm.

Figure 5. Validation of the correlation of the circRNA index with MKI67 expression.

(a) Normal lung tissues (four with low MKI67 expression, N_low, and three with high MKI67 expression, N_high) and tissues from idiopathic pulmonary fibrosis (IPF), a non-cancerous neo-proliferative disease (Supplementary Table 4a) are compared concerning MKI67 expression (p = 0.011, Kruskal-Wallis rank sum test) and percent circRNA reads (trend p-value = 0.056, Kruskal-Wallis rank sum test, R-package compareGroups). (b) Ascitic ovarian cancer cells (Cancer) and (immortalised) normal ovarian surface epithelium cell lines ((i)OSE CL) (Supplementary Table 4b) are compared concerning MKI67 expression (p = 1e-04) and percent circRNA reads (p = 0.047, both Kruskal-Wallis rank sum tests). Yellow indicates normal tissues or cell lines and blue indicates tissues or cells from the diseased state. (c) Correlation of MKI67 expression and global circRNA abundance (the circRNA index) in 13 human tissues (red, twelve human tissues from study SRP033095 and blue, monocytes from study E-MTAB-2399). A significant strong negative correlation (Pearson correlation coefficient ρ = -0.573, p = 0.041) is seen (Supplementary Table 4c).

Figure 6. Schematic model of circular and linear RNAs in proliferating and non-proliferating cells.

Assumption: Circular and linear RNAs are synthesized by specific splice events in a gene (cell type and condition) dependent ratio. While linear RNAs are in a steady state of regulated synthesis and degradation, circular RNAs are much more stable. During proliferation both RNA species are evenly distributed to daughter cells. The steady state level of linear transcripts is accurately controlled and maintained by active transcription and degradation also leading to new expression of circular RNAs and resulting in constant ratios of circular to linear isoforms (lower panel). In contrast, in non-proliferating cells stable circular RNAs accumulate whereas linear transcripts are in a regulated steady state of transcription and degradation (upper panel).