Dysregulation of Transcription Factor Networks Unveils Different Pathways in Human Papillomavirus 16-Positive Squamous Cell Carcinoma and Adenocarcinoma of the Uterine Cervix

Abstract

Squamous cell carcinoma (SCC) and adenocarcinoma (ADC) are the most common histological types of cervical cancer (CC). The worse prognosis of ADC cases highlights the need for better molecular characterization regarding differences between these CC types. RNA-Seq analysis of seven SCC and three ADC human papillomavirus 16-positive samples and the comparison with public data from non-tumoral human papillomavirus-negative cervical tissue samples revealed pathways exclusive to each histological type, such as the epithelial maintenance in SCC and the maturity-onset diabetes of the young (MODY) pathway in ADC. The transcriptional regulatory network analysis of cervical SCC samples unveiled a set of six transcription factor (TF) genes with the potential to positively regulate long non-coding RNA genes DSG1-AS1, CALML3-AS1, IGFL2-AS1, and TINCR. Additional analysis revealed a set of MODY TFs regulated in the sequence predicted to be repressed by miR-96-5p or miR-28-3p in ADC. These microRNAs were previously described to target LINC02381, which was predicted to be positively regulated by two MODY TFs upregulated in cervical ADC. Therefore, we hypothesize LINC02381 might act by decreasing the levels of miR-96-5p and miR-28-3p, promoting the MODY activation in cervical ADC. The novel TF networks here described should be explored for the development of more efficient diagnostic tools.

Keywords: RNA-Seq; cervical adenocarcinoma; cervical cancer; cervical squamous cell carcinoma; lncRNA; miRNA sponge; transcriptional regulatory network.

Figure 1

General design and descriptive results of cervical cancer transcriptome analysis. (A) Pipeline applied in cervical squamous cell carcinoma (SCC) and cervical adenocarcinoma (ADC) samples from our study, followed by inclusion of publicly available HPV-negative non-tumoral cervical tissue (non-CC) samples [10] and cervical SCC and ADC samples from The Cancer Genome Atlas (TCGA). Gene expression of SCC, ADC, and non-CC samples was evaluated. Expression levels of selected long non-coding RNA (lncRNA) genes differentially expressed between cervical ADC and SCC were validated by quantitative real-time PCR. Gene expression networks were predicted using gene expression profiles. (B) Number of genes expressed in SCC and ADC samples from this study and non-CC samples. (C) Frequency of genes expressed in cervical SCC, ADC, and non-CC samples, according to protein-coding, lncRNAs genes, and pseudogenes categories (x-axis). LncRNA genes were subdivided in antisense and long intergenic non-coding RNA (lincRNA), according to Ensembl database annotation (GRCh37/hg19 version). (D) Bar chart of top 10 enriched Gene Ontology biological processes based on gene expression profiles of cervical ADC and SCC samples from our study and non-CC samples, according to combined ranking values (y-axis) calculated by Enrichr (11).

Figure 2

Differential gene expression profiles among cervical squamous cell carcinomas (SCC), cervical adenocarcinomas (ADC), and non-tumoral cervical tissue and HPV-negative (non-CC) samples. (A) Number of genes differentially expressed in each group. Transcription regulatory networks (TRN) of genes upregulated in SCC (B) and ADC (D), including transcription factors (purple squared), other protein-coding genes (green circle), long non-coding RNAs (lncRNA) (light blue circle), and pseudogenes (light orange circle). Ten most enriched Gene Ontology Biological processes involved in cervical SCC (C) and ADC (E) TRNs. Biological processes were grouped by enriched Gene Ontology (GO) Biological processes (y-axis) and sorted by lower logarithm scale of p-values (x-axis, p < 0.05), calculated by Enrichr (11).

Figure 3

Gene expression levels of six differently expressed long non-coding RNA (lncRNA) genes between cervical adenocarcinoma (ADC) and squamous cell carcinoma (SCC). RNA levels were evaluated by quantitative reverse transcriptase-polymerase chain reaction (qPCR) and normalized to RPLP0 gene levels in SCC (n = 25) and ADC (n = 5) samples. Red square indicates median value. Gene expression levels were significantly different using Mann–Whitney test (p ≤ 0.05). TINCR (A), CALML3-AS1 (B), IGFL2-AS1 (C), and DSG1-AS1 (D) lncRNA genes presented higher expression levels in cervical SCC as compared with ADC, whereas LINC02381 (E) and LINC01833 (F) lncRNA genes were upregulated in ADC as compared with SCC samples. p-values are indicated as **p < 10⁻², ***p < 10⁻³, ****p < 10⁻⁴.

Figure 4

Gene set enrichment analysis and transcriptional regulatory network (TRN) of upregulated genes in cervical adenocarcinoma (ADC) samples from our study. (A) Fourteen genes enriched in ADC that participate in maturity-onset diabetes of the young (MODY) pathway were not expressed neither in SCC nor in non-tumoral cervical tissue and HPV-negative (non-CC) tissue samples, calculated by GSEA software (12). (B) Among transcription factors (TF) related to MODY pathway, hepatocyte nuclear factor 4 (HNF4G/A), pancreatic and duodenal homeobox 1 (PDX1), and forkhead box protein A2 (FOXA2) (red circles) are related to expression of several upregulated genes in ADC samples in TRN analysis, including protein-coding genes and of LINC02381 long non-coding RNA (lncRNA) gene (in red).

Figure 5

Area under the receiver operating characteristic (AUROC) curve for cervical ADC classifier.

Figure 6

Proposed scheme of cervical adenocarcinoma (ADC) gene regulation by long non-coding RNA (lncRNA) acting as a sponge of microRNA (miRNA). (A) in silico prediction analysis of miRNA repressing all transcription factor (TF) genes upregulated in cervical ADC as compared with squamous cell carcinoma (SCC) in maturity-onset diabetes of the young (MODY) KEGG pathway. (B) TF genes HNF4G and PAX6 might upregulate LINC02381 lncRNA, which could act as a miRNA sponge, as positive feedback for all TF genes pathway expression. Thicker lines on (A) represent more positive predictions of miRNA targeting to a given mRNA (Supplementary Table 5).