High number of read counts mapping to bacterial rRNA genes in RNA-Seq experiment

[buthercup_ch]

Hello to everyone.

In my group, we are using RNA-Seq to study the expression of bacteria under different conditions, isolated or as dual-transcriptome when in association with plants.

When checking the table of total counts for each gene, I can observe an important number of reads mapping against the rrna genes, mostly 16S and 23S. I also found a similar amount of read counts mapping against rrnas when analyzing a dual experiment where no bacteria was present, just plant material.

Illumina reads from HiSeq200 were treated by TopHat and Cufflinks. Library preparations followed a RiboZero Gold treatment + TruSeq Stranded Total RNA kit. Additionally, rrna depletion was verified by running Bioanalyzer.

Have any one an idea where this mapping might come?

Thanks in advance for your help.

[fanli]

Ribodepletion is never perfect. Did you use the appropriate RiboZero kit? What's your trace look like?

[buthercup_ch]

Hello fanli,

Thanks for your message.
To answer your question, yes the Ribozero kit used was appropriate for our samples. It was the Ribo-Zero Gold rRNA Removal Kit (Epidemiology). This kit removes cytoplasmic and mitochondrial rRNA from samples composed of human/mouse/rat, Gram-positive and Gram-negative bacterial RNA.

What do you mean by "your trace look like"?
If you mean how the bioanalyzer graph look like, they look perfect, with no peaks for 16S or 23S, but small peak for 5S rRNA.

I was thinking these sequences matching the ribosomal genes could actually come from an artifact during library preparation (PCR enrichment) or sequencing (over sequencing) causing a high number of overrepresented sequenced (high sequence duplication).

[fanli]

I don't know that I'd call it so much an artifact as a consequence of random amplification. You could look at the read duplication levels for your rRNA-mapped reads to see if it's a single PCR jackpot versus high levels of template rRNA. But the end result would be the same - you'll just have to live with the fact that you throw away some of your sequencing data.

By the way, what fraction was ribosomal anyways? I'd say anything less than 10% is fine..

[bastianwur]

Depending on the setup, 20% might also be acceptable.

What we've seen in some of our experiments is that, although the traces don't show any big peaks, we got a half ton of eukaryotic rRNA (bacterial was depleted) afterwards. We hypothesize that the smaller peaks in the surrounding are probably degraded rRNA, which can't be properly depleted (even with the right kit, as it seems). So it might not necessarily be visible during the QC.

I'd not have an idea where your bacterial data could come from in your non-bacterial example, unless it's a contamination from elsewhere.