Silencing of ANKRD12 circRNA induces molecular and functional changes associated with invasive phenotypes

Ovarian cancer cell lines PA-1 (ATCC® CRL-1572), SKOV3(ATCC® HTB-77™),Caov3(ATCC® HTB-75™), NIH:OVCAR-3 (ATCC® HTB-161™), breast cancer cell lines MDA-MB-231(ATCC® HTB-26™), MCF7(ATCC® HTB-22™),T-47D (ATCC® HTB-133™) and breast normal cell line MCF 10A(ATCC® CRL-10317™), Lung cancer cell lines, HCC2935(ATCC® CRL-2869™) and NCI-H226 (ATCC® CRL-5826™) and Lung Normal Fibroblast cell line LL 24(ATCC® CCL-151™) (all from American type Cell Collection, Manassas, VA), APOCC (ovarian primary cell line derived from ascites fluid) (pers. communication Dr. Arash Tabrizi), A27809 (93112519-1VL,Sigma), A2780 CIS (93112519-1VL,Sigma) were used for the current study. Cells were cultured in DMEM (Life Technologies, NY, USA) supplemented with 10% fetal bovine serum (Life Technologies, USA). Low Passage number cells were used for all the experiments. Cell culture were routinely checked for mycoplasma contamination using MycoAlert Mycoplasma detection kit (Lonza, Basel, Switzerland).

The nuclear and cytoplasmic RNA was extracted using The SurePrep™ Nuclear or Cytoplasmic RNA Purification Kit (Fisher Grand Island, NY, USA). Total RNA from whole-cell lysates were isolated by using RNAesay mini kit (Qiagen Valencia, CA USA). For RNase R treatment, 2 μg of total RNA was incubated 60 min at 40 °C with or without 3 U μg^− 1 of RNase R (Epicentre Technologies, Madison, WI), and the resulting RNA was subsequently purified using an RNeasy MinElute cleaning Kit (Qiagen, Valencia, CA USA). cDNA synthesis was carried out using First strand synthesis kit (AMV) from Roche Biosciences or Biorad select cDNA synthesis kit using random primer for circRNA experiments. Fast Start Universal SYBR Green Master mix (Roche, Clovis, CA) was used to amplify the specific gene using cDNA primes obtained from Primer bank (https://​pga.​mgh.​harvard.​edu/​primerbank/​ (Additional file 4: Table S1). Each Real-Time assay was done in triplicate on Step One Plus Real-time PCR machine (Life Technologies, CA, USA).

siRNA transfection was carried out using custom-designed siRNAs for both ANKRD12 circular and linear transcripts (Fig. 1 and Additional file 4: Table S1). The SKOV3, MDA-MB-231, OVCAR3, NCI-H226 cells were grown in 6 well plates for transfection. The cells were transfected at 24 h with 30 pmol concentration of siRNA (VWR, Radnor, PA, USA) or scrambled control (Mission siRNA universal negative control, Sigma, St.Louis, USA) using Lipofectamine RNAi max (Invitrogen MA USA) according to manufacturer’s protocol. These experiments were conducted in three different biological triplicates for subsequent RNA-sequencing.

SKOV3, OVCAR3, NCI-H226, MDA-MB-231 cells transfected with either circANKRD12 or a universal scrambled control were used for RNAseq analysis. RNA-seq library preparation and In silico detection of circRNA candidates from paired-end RNA-seq data was conducted as described earlier [21].

Cells were cultured at a density of 5 × 10³ cells per well in flat-bottomed 96-well plates in DMEM supplemented with 10% FBS with antimycotic antibiotic. The experiment was done at different time points (24 h and 48 h or based on cell doubling time). CellTiter 96® Aq_ueous One Solution Reagent (Promega, Madison, WI) was added and the experiment was conducted according to the manufacturer’s instructions.

CellTiter-Glo assay (Promega Madison USA) for determining cell proliferation was conducted according to the manufacturer’s instructions. Briefly, CellTiter-Glo reagent was added directly to the wells of 96 well plate and luminescence was measured on an Envision reader (PerkinElmer).

Scratch assay was conducted as described earlier [23]. Cells were plated into a 6-well plate with complete medium and grown to 80% confluence. Cells were transfected with respective siRNA in OPTIMEM medium, and the medium was replaced with serum-free DMEM after transfection. After cells were grown to 100% confluence, a wound was created by scraping the confluent monolayer cells with a p200 pipette tip. Cells were then grown either in serum-free medium or medium containing 3 mM Thymidine. The distance between the two sides of the cell-free area was photographed using 10x objective in AXION Zeiss epifluorescence microscope. The distance is measured using Zeiss Zen software (Carl Zeiss Carpenteria, CA, USA).

Cellular migration and invasion were determined using a Transwell Boyden chamber assay as described previously [24].

3D anchorage-independent spheroids were developed in SKOV3 cancer cell lines; initially, cells were seeded on ultra-low attachment plate (Corning, NY, USA) for three days. Transfection of the spheroids with circANKRD12 siRNA was conducted using the reverse transfection method.

Cells were seeded at a density of 300,000 cells/well into ultralow attachment plates. After 72 h, cells were transfected either with scrambled siRNA or circANKRD12 siRNA. After 48 h the diameters of at least 50 spheroids were measured and the spheroid area was measured Zeiss Zen Software.

Ten thousand cells were seeded on each well of ultra-low attachment 96 well plate with OPTIMEM medium. After three days, once spheroids were formed, cells were transfected with either circANKRD12 siRNA or scrambled control. After 24 h and 48 h of transfection, MTS reagent was added to the medium. Measurements were according to the manufacturer’s instructions.

3D organotypic models of either circANKRD12 transfected or scrambled control in SKOV3 cells expressing GFP were placed on the top of jellified collagen matrix (Rat tail collagen1, 1 mg/ml). The invasiveness analyzed after 24 h to 10 days.

Cellular protein was extracted after 48 h of transfection. The cells were lysed in 100ul of RIPA buffer with protease inhibitor cocktail. Then 40 micrograms of protein were resolved in SDS PAGE gel and transferred to a nitrocellulose membrane. The primary antibodies used were anti-Cyclin D1, Anti- Cyclin B1, Anti CyclinD2, anti-Cyclin Phospho B1 and β-actin (Cell Signaling, USA). The blots were visualized by ECL detection (Amersham, N J, United States). The Western blot experiments and analysis were done as described earlier [25]. Briefly, cell lysate protein (25 μg) was separated on an SDS-Polyacrylamide gel and transferred onto nitrocellulose membranes. Membranes were then blocked with 5% (w/v) non-fat dry milk tris-buffered saline (TBS) containing 0.1% (v/v) Tween20 and incubated at 4 °C with the primary antibody (1:1000 dilution). The excess primary antibody was washed off in TBS-T wash buffer, and the membranes were incubated with HRP-linked secondary antibodies (1:2500) for one h at room temperature. The excess secondary antibody was then washed off in TBS-T, and the protein levels were detected using enhanced chemiluminescence reagent (Sigma-Aldrich, Inc., MO, USA) and imaged on a Geliance P600 gel documentation system (PerkinElmer, Inc. MA, USA). β-Actin was used as the loading control.

siRNA longevity was calculated based on cell doubling time. The cell doubling time of SKOV3 cells was calculated based on the mathematical equation.

Cells were seeded on 60 mm petri plates and transfection was done with siRNA of circANKRD12, or ANKRD12 mRNA. Cells were harvested on an interval of 48 h (doubling Time). The experiment was extended till 10th doubling time (20 days). The silencing efficacy was measured by gene expression analysis (qRT-PCR).

The mitochondrial oxygen consumption rate (OCR) in the MDA-MB-231 and SKOV3 cells was assessed by using a Seahorse Bioscience XFe96 analyzer (Massachusetts, USA). Seahorse Bioscience XF Cell Mito Stress Test assay kit was used for the study. In this assay, subsequent additions of the ATP synthase inhibitor oligomycin, the mitochondrial uncoupler carbonyl cyanide 4-(trifluoromethoxy) phenylhydrazone (FCCP) and the complex I + II inhibitors rotenone + antimycin A were injected into the assay medium containing cells with different treatments as per manufactures protocol. 10,000 cells were seeded on assay plate and the assay was completed in 48 h after the transfection.

The RNA-seq data for cell-lines has been submitted to Sequence Read Archive (NCBI SRA) under Bio project accession number PRJNA526399.

SRA records will be accessible with the following link: https://​www.​ncbi.​nlm.​nih.​gov/​sra/​PRJNA526399

We characterized the circular RNA expression using RNA sequencing (RNA-seq) of total RNA in ovarian cancer cell lines (SKOV3 and CAOV3) undergoing Epithelial to Mesenchymal Transition. Using our circRNA detection pipeline [21], we identified a total of ~ 18,700 circRNAs in these cell lines. The circRNA from the ANKRD12 gene having a backsplice junction between exons 8 and 2 (circANKRD12) was one of the highly abundant circular RNAs in both SKOV3 and CAOV3 cell lines. The circANKRD12 originates from 18p11.22 chromosome locus with the backsplice junction forming between exons located ~ 39.3 kb apart (Fig. 2a). By carefully designing two sets of divergent primer pairs in Exon2 and Exon8 and Sanger sequencing of the amplified products, we were able to identify two circular isoforms of 286 bp and 925 bp lengths sharing the same backsplice junction (Additional file 3: S2:1–3), Additional file 4: Table S2). The first isoform of 286 bp is comprised of only two exons (Exon2 and Exon8), while the 925 bp isoform is comprised of 6 exons (Exons 2,3,5,6,7 and 8). All subsequent functional studies do not distinguish between these two circular isoforms as they targeted the junction sequence common to both circular forms but distinct vis-à-vis linear forms.

Multiple validation experiments were used to confirm circANKRD12 expression (Fig. 2b-d).  Divergent primers were designed to amplify the backsplice exon junction. As expected, each primer pair produced a single distinct band of expected product size in PCR assay indicating the presence of the circular junction. The backsplice junctional sequence was confirmed by Sanger sequencing. The divergent primers, with respect to the genomic sequence, only amplified when cDNA was used as a template. The same primers did not produce a product from genomic DNA (gDNA). Conversely, PCR using convergent primers with respect to genomic sequence could amplify both cDNA and gDNA templates from ANKRD12 gene (Fig. 2b). This strongly indicates that a head-to-tail backsplicing junction in circANKRD12 only exists in the RNA form.

To examine whether circANKRD12 is resistant to exonucleases, we treated the total RNA extracted from the SKOV3 cell line with RNase R. This exoribonuclease enzyme digests all linear RNA forms with a 3′ single-stranded region of greater than 7 nucleotides [26]. As circRNAs are devoid of any 3′ single strand overhangs, they are expected to show resistance to digestion by RNase R. Indeed, circANKRD12 was resistant to RNase R digestion compared to linear forms of ANKRD12, HIF1 alpha, and GAPDH. Resistance to digestion with RNase R exonuclease confirmed that the circANKRD12 is a stable circularized transcript (Fig. 2c). Further, cDNA created by priming with oligo (dT) primers failed to produce any amplification products for circANKRD12 in the PCR assay confirming the circular form does not have a typical polyA tail as do linear mRNAs. On the other hand, priming with random hexamers – which can also amplify non-poly-adenylated RNAs – resulted in distinct PCR bands for the backsplice junction (Fig. 2d). The PCR and qRT-PCR analysis of nuclear and cytoplasmic fractions of RNA demonstrated that ANKRD12 circRNA is predominantly localized in the cytoplasm (Additional file 1: Figure S1:Fig. 2a,b). The purity of cytoplasmic and nuclear fractions was confirmed by amplifying the fractions using nuclear and cytoplasmic specific markers (Additional file 1: Figure S1: Fig a, b, Additional file 3: S2:8).

We performed Real-time PCR analysis on 12 different cell lines from breast, ovarian and lung cancer and normal breast and lung to assess the cell-type specific expression of circANKRD12 (Additional file 1: Figure S1:Fig. 2c). Most of these cell lines show a high abundance of both ANKRD12 circRNA and mRNA. Ovarian cancer cell lines show a higher abundance of ANKRD12 circRNA compared to breast and lung cancer cell lines.

To investigate the role of circANKRD12, we designed siRNAs to target the backsplice junction (Fig. 1a) and transfected multiple cancer cell lines to induce siRNA-mediated knockdown of the circle while leaving the linear RNA unaffected. The circRNA specific siRNA was designed against the backsplice junction spanning exons 2 and 8 of the gene. We observed high knockdown efficiency of greater than 90% of the circular junction when using the siRNA versus the control (scrambled siRNA) in four cell-line transfections (Fig. 1b). Using two siRNA constructs against the circANKRD12, we confirmed that knockdown of the circular RNA is specific and has no significant effect on the linear mRNA expression (Fig. 1c). The two different siRNAs had similar results reducing the likelihood that the observed effects are due to off-target knockdown. We also designed siRNAs targeting exon9 and exon7 of ANKRD12 gene to knockdown the linear mRNA. Indeed, the siRNA designed against exon 9, which lies outside the circANKRD12 locus was successful at knocking down the mRNA and exhibits no effect on circANKRD12 levels (Fig. 1d). However, the siRNA designed on Exon7 which is shared between mRNA and one of the circANKRD12 isoforms shows a remarkable reduction in mRNA levels as well a minimal but significant reduction in circANKRD12 levels. We investigated the specificity of siRNA constructs designed against circANKRD12, Exon7, and Exon9 through a series of knockdown experiments explained in Additional file 4: Supplementary file S2:4–7).

RNA-sequencing was performed in triplicate for two ovarian cancer cell lines (SKOV3, OVCAR3), breast (MDA-MB-231) and a lung cancer cell line (NCI-H226) for both scrambled control and siRNA targeting circANKRD12. Table 1 gives the number of differentially expressed genes in circANKRD12 knockdown samples (Additional file 5: Supplementary file S1) at a false discovery rate of less than 5% with at least 1.5-fold changes in expression. 

The canonical pathways analysis by Ingenuity’s IPA toolkit (IPA®, QIAGEN Redwood City, (https://​www.​qiagenbioinforma​tics.​com/​products/​ingenuity-pathway-analysis/​) revealed an enrichment of differentially regulated genes involved in cell cycle regulation, invasion, migration, and interferon signaling pathways (Fig. 3, Additional file 2: Figure S2, Additional file 4: Table S3). We observed an upregulation of inflammatory pathways such as tumor necrosis pathway, NFkB pathway, interferon signaling and down regulation of the cell cycle pathway (Additional file 3: S2:15). Silencing of circANKRD12 affects cellular bioenergetics by downregulation of key genes involved in oxidative phosphorylation, AMPK pathway and cell metabolism (Additional file 4: Table S4). IPA predicts a significant activation of interferon signaling pathway through upregulation of STAT1, STAT2, IFT1, MX1genes, and downregulation of estrogen mediated S phase entry through downregulating cyclin D1 and cyclin A. Activation of IL8 signaling and inflamasome pathways, closely associated with immune modulation is also predicted. IL8 signaling pathway upregulates IL8, IRAK, ICAM-1, COX-2, and VEGF genes there by upregulating angiogenesis, inflammation, and inhibits cell proliferation by downregulating cyclin D1. A heatmap of differentially expressed genes from RNAseq analysis shows consistent gene expression changes in circANKRD12 silenced cells in cancer cell lines (Fig. 4a). The differentially regulated genes cyclin D1 (CCND1), CBX5, STAU1, and AK4 were found to be consistently downregulated in the circANKRD12 knockdown across multiple cell lines (Fig. 4a). Using qRT-PCR, we validated the differential expression of these selected genes in SKOV3 cells (Fig. 4c). Cyclin D1 (CCND1) was among the genes that showed consistent downregulation in the circANKRD12 knockdown across multiple cell lines (Fig. 4a). To ensure that the effect on CCND1 is related to circANKRD12 knockdown rather than off-target effects we compared the knockdown effects in SKOV3 and LL24 cells (Fig. 4b). Unlike SKOV3 cells, the circANKRD12 level is low in LL24 cell line, attempts at knocking down circANKRD12 should not affect the cyclin D1 level in the cell line. Indeed, transfection of the siRNA against circANKRD12 in LL24 showed a reduction of the already low levels of circANKRD12 but no effect on cyclin D1 expression, suggesting the siRNA effect on CCND1 is likely mediated through circANKRD12. Because cell cycle regulation seems to be one of the most deregulated pathways in circANKRD12 silenced cells, we decided to follow it up with further functional screening using cell based phenotypic assays.

Wound healing assays show that silencing of cirANKRD12 increased cell migration (Fig. 5a, b, Additional file 3: S2:10), in SKOV3 cells after 24 h and 48 h. compared to scrambled control. This is confirmed again by cell migration assays in synchronized (Thymidine incorporation) cells, which show an increased migration rate for circANKRD12 silenced SKOV3 cells (Fig. 5c). Matrigel invasion (inserts coated with matrigel) and migration analysis using Boyden chamber shows circANKRD12 silenced cells undergo significant increase in the invasion and migration compared to scrambled control (Additional file 3: S2:10).

Knockdown of circANKRD12 and ANKRD12 mRNA significantly reduces cell proliferation (Fig. 5d, Additional file 3: S2:9). The results of MTS and ATP assays show a significant reduction in cell proliferation in circANKRD12 silenced cells compared to scrambled control (Fig. 5d, e. The Trypan blue exclusion assays show the silencing of circANKRD12 is not affecting the cell viability significantly. However, silencing of ANKRD12 mRNA significantly reduces both proliferation and cell viability in SKOV3 cells (Fig. 5f). These results indicate that knockdowns of both circular and linear RNA forms of ANKRD12 gene are capable of inducing strong phenotypic changes and modulate the growth or survival of cancer cells.

Since 3D culture is known to have better cell-to-cell interactions and more closely mimics tumors in vivo [27], we investigated the effects of circANKRD12 knockdown in 3D culture experiments. The 3D anchorage independent growth of siRNA transfected SKOV3 cells showed smaller, dense aggregates associated with an invasive phenotype (Additional file 3: S2:12). The 3D organotypic models were efficiently silenced with circANKRD12 siRNA with 74% knockdown efficacy (Fig. 6a). This reduction in spheroid size in circANKRD12-silenced SKOV3 organotypic models results in an alteration of phenotype from less invasive to a more invasive one (Fig. 6b). Cell proliferation MTS assay shows a significant reduction in cell growth within a time period of 24 h and 48 h after knockdown of ANKRD12 circRNA (Fig. 6c). Collagen invasion of the circANKRD12 silenced spheroids shows increased invasion through the collagen gel with highly motile cells after 10 days of transfection (Fig. 6d). These results indicate the phenotypic switching from a highly proliferative phenotype to a less proliferative phenotype with high invasion potential.

circANKRD12 silencing results in cyclin D1 down-regulation and subsequent invasion and reduction in proliferation in ovarian cancer SKOV3 cells. Both real-time PCR and RNA-sequencing showed down-regulation of cyclin D1 in circANKRD12 silenced cells. The qRT-PCR analysis shows cyclin D1 expression is downregulated at least until 48 h after transfection with the siRNA (Fig. 7a). This is also supported by the Western blot analysis, which shows a reduction in cyclin D1 protein level (Fig. 7b). Only cyclin D1 shows differential regulation at the protein level in circANKRD12 silenced cells compared to other cyclin members including cyclin E1, pB1, and D2 (Additional file 3: S2:13). To estimate the duration of the RNAi effect initiated by transfecting siRNA, we performed a longevity assay of circANKRD12 siRNA (si-circANKRD12). The longevity, initially stable for six days (Additional file 3: S2:14) was checked on a cell doubling time basis estimated to be 48 h (see Methods). The si-circANKRD12 exhibits an extended longevity period in SKOV3 cells, approaching nine doublings (18 days). The reduction in the level of cyclin D1 is on par with the observed silencing efficiency and longevity of siRNA (Fig. 7c). After the 10th doubling, the effect on cell proliferation induced by silencing of circANKRD12 becomes insignificant (Fig. 7d), suggesting that the knockdown efficiency is lost due to multiple passages of cells. As cyclin D1 is involved in cell cycle progression [28], cell cycle analysis of circANKRD12 knockdowns was conducted. Cell cycle analysis using FACS shows there is a significant G0/G1 cell cycle arrest in SKOV3 cells silenced with circANKRD12 compared to scrambled control (Fig. 8a, b).

As AMPK signaling and other metabolic pathways are affected in circANKRD12 knockdown (Additional file 4: Table S4). In order to determine whether the altered gene expression of these pathways translates to changes in basic metabolism, we analyzed metabolic phenotypes of circANKRD12 silenced MDA-MB-231 and SKOV3 cells using Seahorse extracellular flux analyzer. The oxygen consumption ratio (OCR) and ATP linked OCR analysis shows that knockdown of circANKRD12 decreases oxidative phosphorylation (OXPHOS) of MDAMB231 cells and SKOV3 cells (Fig. 8c-e). Previous reports have suggested that high invasive potential of cancer cells is negatively correlated to high energetic cancer phenotype [29, 30]. These results thus indicate that circANKRD12 silencing can induce phenotypic switching between highly proliferative cells to highly invasive cells through cyclin D1 deregulation and shifting the oncobioenergetics to a low energy phenotype to facilitate invasion.