RNA seq and using of Poly(A) or non-Poly(A) based amplification of RNA

Question

I'm studying "Deep sequencing the circadian and diurnal transcriptome of Drosophila brain" Hughes et al., 2012. I've got some problems with the materials and methods.

Before RNAseq, the authors amplify RNA. They use two kits: Poly(A) based and non-poly(A) based. For the samples Poly(A) based amplified they do a sequencing to generate 100 bp paired-end and for the samples Non poly(A) based they do a sequencing to generate 75 bp single-ends.

I read in the strategy:

Total RNA was amplified, and ribosomal RNA was depleted, using a non-poly(A)-based amplification kit, which significantly diminishes the 3' bias in downstream libraries used for RNA-seq (see Methods).

1. I don't understand why they use both kits. Actually, they say that they want to asses the bias introduced by an alternative amplification methods BUT the following sequencing is not the same in both cases so there is more than one parameter which change between the two protocols so I don't understand how they can compare.

2. For the poly(A) based I understand that the rRNA are not polyadenylated so they remove rRNA from the amplified sample. However, don't they remove non-coding RNA too (long or short)?

3. For the non poly(A) based, I don't understand how they remove the rRNA but I understand that since they don't select the RNA according to their poly(A) tails they keep the rest of the RNAs which are not kept in the poly(A) based amplification.

4. When I look to the RUM statistics, I see that there are fewer total aligned reads in non-poly(A) based: 70-80% (non polyA) and 80-90% (polyA). So it seems like we lose some information with non-poly(A) based.

To sum up, why the authors use Poly(A) and non-poly(A)? Why do they show all the rest of the results from the non-poly(A) amplification since they have less aligned read? How do they compare since the following sequencing is not the same? How the rRNA are removed in both techniques and especially in non-poly(A)?

I didn't find the protocols kit with exactly the same name but here are the links to the protocols which are closer (I think).

http://www.mscience.com.au/upload/pages/nugen/nugen_ov_rna_seq-brochure.pdf

http://excilone.com/client/document/ug-arcturusE%E2%80%9E%C2%A2-riboamp%C3%82-hs-plus-amplification-kit-user-guide_38.pdf

Answer 1

1. Yes, it was odd that they said that. The second protocol with the polyA based method had another purpose though, which was "examine the reproducibility of period-regulated genes in an independent genetic background". I would suspect this was its main purpose, and not so much to look at fragment bias, which has already been done in other papers. They didnt even publish this data in the main paper, they put it in the supplement.

2. Yes, this would remove any transcript without a polyA tail.

3. One way of purifying is by using oligos that are antisense specifically to rRNA. The oligos will be attached to some solid phase, on a column or a bead or something, and the rRNA will stick to it and everything that doesnt is what will remain in your sample

4. It is not safe to say that you lose information with polyA based kits. The polyA based technique had reads that were single end and only 75bp long. This is compared to the other protocol which was 100bp paired. Reads that are 75bp will align much more easily than 200 bp (100 x 2 since its paired). However, there will not be as much certainty in the alignments, as there is a greater chance it will align to multiple locations in the genome, or align to the wrong location. So yes they have more reads aligned but I would be hesitant to make any conclusions from this. And if you look at total numbers of reads, the non-polyA method aligned more. Again, I dont think the main purpose was to show that one method is better than another.

Basically the method you choose depends on your experiment and your samples. Random hexamers will give you non-coding RNA which is important in some experiments, and that is its main advantage. It will also give more uniform coverage across the gene (no 3' bias). Poly(A) selection will give you only mRNA, however you will miss any transcripts that are degraded or partially degraded. If you have a high quality sample or a large amount of starting material this is less of a concern. As mentioned before you end up missing parts of the 5' end on some transcripts with this method. If you have a low abundance transcript that you are trying to find, polyA selection is probably best, as it is more likely to be captured and you will capture the whole thing, rather than just getting bits and pieces of it with random hexamers. They both have advantages and disadvantages. Hopefully this helps you or someone make a decision.