Plant-Bacteria transcriptome

[Alagurajvelu]

Our intention was to study simultaneous transcription of both Bacteria (Positive to plant growth) and Plant (Arabidopsis). We isolated plant root along with bacteria and started sequencing the total RNA after removing 16S rRNA and 18S rRNA. We sequenced 80 million reads through HiSeq. Now i find good transcription in plant (79.5 million reads mapped to Arabidopsis). I find very less reads (in thousands) on bacteria which have 4000 genes. I find lot of reads (30 million) on Chloroplast of Arabidopsis. I expected these many reads on bacteria. Unfortunately they map on chloroplast.

I am not a wet lab person. I want to understand what is wrong here: why there are no reads on bacteria - that i couldn't enumerate the expression level of bacteria ?. Could it be a error on experimental approach or bioinfo mapping error ?.

[SylvainL]

Hope you didn't prepare libraries using polyA enrichment...

[Alagurajvelu]

No PolyA enrichment. It is total RNA.

[kmcarr]

Are you certain that the RNA extraction protocol chosen was appropriate for both plant and bacteria?

[SNPsaurus]

Do you have any idea about the ration of bacteria to plant cells? Some estimates of the number of transcripts in E. coli put it in the 5000 per cell range, while eukaryotes may have several thousand transcripts per cell.

I've sequenced bacterial transcripts from fish guts, and the number of bacterial reads may be 1-2% of total. I wouldn't be surprised if root bacteria are even less well represented.

One issue may be that bacterial cells are hard to break open, so a protocol that works for plant cells may not work well for bacteria.

[cliffbeall]

We have seen something kind of similar with a bacteria - invertebrate infection system. The bacterial reads are a small percentage, even though they can kill off the hosts.

We're working on amplifying the bacterial signal but haven't solved it yet, would be really interested to hear of anything that works. One thing we are trying is polyA subtraction.

It doesn't seem to be breaking the cells because the method works fine on the same bacteria in culture.

[SNPsaurus]

In my experience it is hard to get rid of something efficiently enough to make a difference. Say you eliminate 50% of the stuff you don't want... you might see a rise in bacterial reads from 0.7% to 1.4%. Not very good. For the fish gut work we've been (as part of the META Center at the University of Oregon) making cheap capture baits by converting bacterial genomic DNA to biotinylated RNA, then doing capture hybridization to purify the bacterial transcripts. We get a majority of reads being bacterial. One caveat... you need to know what's there to make the probe.

[cliffbeall]

SNPsaurus, could you share that protocol? We do know the sequence of host and bacteria so it should be possible to do that in either direction.

[yzzhang]

we did a plant leaf RNA-seq which infected with pathogenic bacteria. Around 30 million hiseq pair end reads were generated from each sample, and around 5% of the reads mapped to the bacteria genome while around 70% of the total reads mapped to the plant genome (draft genome)

[SNPsaurus]

cliff, let me ask my student. Might take a moment to protocolize it.