10 min Read

circRNAs vary greatly by genesis and size from their linear RNA counterparts and thus cannot be purified from other RNAs by typical means of size or sequence. Instead they can be enriched by a series of treatments which can be applied in various orders. These include rRNA depletion, polyadenylation and/or RNase R treatment (21). Recent literature suggests that downstream circRNA analysis by RNA-seq can vary greatly depending on the enrichment method used. Factors in jeopardy include: number of different circRNAs identified, precision and sensitivity. These variations are due to the overall complexity of the methods, the degree to which they achieve high purity circRNAs and eliminate contaminating linear RNAs which can interfere with downstream analysis. Currently there are several common circRNA enrichment methods each with their own advantages and disadvantages.

1. Ribosomal RNA Depletion (rRNA-)

This method involves the removal of rRNA which is the most abundant RNA species making up 80-90% of the total RNA population. Since circRNAs only constitute < 0.1% of total RNA, it is necessary to eliminate the bulk rRNAs in order to achieve better sensitivity. There are two main rRNA depletion strategies: subtractive hybridization and RNaseH degradation. Briefly, subtractive hybridization involves the hybridization of biotinylated ssDNA probes to rRNAs, whereby streptavidin magnetic beads are then applied to eliminate the rRNAs. The RNaseH degradation method involves ssDNA probes which hybridize specifically to rRNAs, subsequently the duplexed DNAs are degraded by the addition of the RNaseH enzyme. While highly effective at removing rRNAs especially in preparation for NGS, rRNA depletion is the most inefficient at removing all other linear RNAs and thus by itself has the poorest sensitivity for circRNA enrichment.

2. Polyadenylation (polyA) + RNase R

Studies have revealed that many linear RNA species are resistant to RNase R degradation due to certain RNAs containing highly stable G-quadruplex structures as well as other 3’ structured ends. To amend this, a poly(A) step is introduced which adds an unstructured poly(A) tail to the 3’ end of all linear RNA species. This long unstructured end provides adequate binding for RNase R which exhibits 3’ to 5’ exonuclease activity resulting in improved purity of circRNAs. While this strategy does not result in the highest number of circRNAs, it does thoroughly remove linear RNAs, possesses good sensitivity and the highest precision.

3. rRNA- + poly(A) + RNase R

This method utilizes a combination of the above techniques starting with rRNA depletion. In terms of linear RNA removal, rRNA- + poly(A) + RNase R demonstrates improvement over poly(A) + RNase R, but the number of circRNAs obtained, sensitivity and precision are quite similar between the two strategies.

4. poly(A) + RNase R + rRNA-

This strategy utilizes a combination of the above techniques however it starts with a polyadenylation step. Interestingly, shifting the rRNA depletion step last significantly improved removal of linear RNAs and results in the highest sensitivity and number of circRNAs obtained. In contrast, the poly(A) + RNase R + rRNA- strategy also demonstrates the poorest precision of all discussed circRNA enrichment methods.

Table 2: Summary of circRNA enrichment methods performance

abm's Rnase R and Poly(A) Polymerase, E. coli can be utilized in different circRNA Enrichment Methods

1. RT-PCR using Divergent Primers

Individual circRNAs can be detected, quantified and validated by RT-PCR using divergent primers (22). On the linear RNA, these primers are designed in opposing directionality and span the circRNA back splice junction. Divergent primers work to specifically amplify circRNA and not the counterpart linear RNA. This is because the divergent primers will only become convergent primers if the RNA is successfully circularized.

2. RNA-seq

Following circRNA enrichment, the sample next undergoes RNA library construction and the Next Generation Sequencing technique RNA-seq, which provides the required raw reads and input datasets. Numerous circRNA identification software and tools exist which can be divided into three categories (23).

2.1 Back splice Junction-Based circRNA identification tools

The Back splice junction (BSJ) read is representative of a molecular signature which can be used to identify a circRNA. Many algorithms that are embedded into these tools are either based on splitting reads (segmented-read-based) or are based on pre-defined BSJ and flanking circRNA sequence. The tools then map the read to a reference database allowing for circRNA discovery. Example BSJ-based tools include CIRI, CIRCexplorer, KNIFE, UROBORUS and circ_Finder.

2.2 Integrated circRNA identification tools

These are representative of ensemble tools as they are able to integrate multiple tools and merge the results together. One issue with identifying new circRNAs is the false positive rate. This can be greatly reduced by using integrated tools as the identification of circRNAs must be supported by multiple tools across multiple data sets using different enrichment treatments. Usage of this type of tool increases the ease and reliability of identifying new circRNAs. Example integrated-based tools include CirComPara which integrates CIRCexplorer, CIRI, find_circ and segemehl.

2.3 Machine learning-based circRNA identification tools

Unlike BSJ and integrated-based identification tools, machine learning does not require RNA-seq input datasets and instead uses circRNA specific features and knowledge to train a classification model. Specific features would include Alu repeats, sequence and structural motifs which are unique to circRNAs. Example machine learning-based tools include PredcircRNA, WebCircRNA and DeepCirCode.

If you’re ready to start your circRNA expression project, browse abm’s circRNA Disease Expression Vector Library or let us clone your custom circRNA vector!