Our facility is performing 16S rDNA next-gen amplicon sequencing on the Ion Torrent PGM platform of mock microbial community containing reference gDNA isolates of 30 bacterial strains. The mock community has been quantified so as to contain 1x10^4 copies of each bacterial organism. We use fusion primers targeting the V1-2, V4, and V7-8 hypervariable 16S gene regions. Out of the 30 reference targets only 12 are detected in the V1-2 region, 23 the V4 region, and 23 in the V7-8 regions. The rest fall out completely and are not even detected as slash or indiscriminate calls. The calls also have a tendency of fluctuating from run-to-run. Does anyone have any suggestions as to how we can make our system more robust and why our system is failing to capture all the reference targets? Do we need to increase our microbial input? Or perhaps our analysis parameters are too stringent.
Several things can go wrong in your setup, including your choice of targets organisms, choice of primers, and so on. It would be better for you to use known, reliable entities: for gDNA, use the standard mock communities from BEI (they're free), and for primers, use the 515F-806R which capture all the genomes in the BEI communities. Once you have robust success with that, you can start experimenting with new DNA or primers if you wish.
Actually 517F/806R is the V4 primer we are using. The V1-2 primer is 27F/388R, and the V7-8 is 1099F/1392R. The 30 reference mock samples were obtained from a combination of BEI and ATCC DNA isolated reference samples. We used ATCC for those strains that could not be obtained from BEI.
Zohar's question about target organisms is a good one. There are a number of lineages that, even at the genus level, are difficult to tell apart using 16S rRNA. For example, members of Escherichia, Shigella and sometimes Salmonella, show >99% homology across the full length of the 16S gene, so can be very hard to separate in amplicons. This is more severe if you're clustering your DNA into OTUs before classifying them.
If you have full length 16S sequences from them, you could create an in silico amplicon pool to test how recoverable they are bioinformatically. This would help to determine whether the lack of detection is a biological or technical problem.
All the above is fine but we tend to forget about basic wet lab practice. I'd try to adjust the PCR parameters, that's the main cause of poor taget amplification. Try changes in TM, adjust primer-sample ratio, include magnesium salts or even do a two steps PCR; amplify the ampilcon of the first PCR because the target will be enriched. One you get enough amplicon to be sequenced you'll identify it and will know if the problem is ACTUALLY your primers don't hit your target. You can already do it since you have available sequences. Find reference sequences for those targets you missed and perform an in silico hibridization with your primers to test their performance. Also, even if you optimize all the steps, it is normal to get a bias in the number of taxa identified using certain regions (Yarza et al. 2014, Nature - Uniting the classification of cultured and uncultured bacteria and archaea using 16S rRNA gene sequences).
You should use lower number of cycles on the PCR amplification, avoiding saturation. To this, you verify, by using of agarose gel, how many cycles do you need. In a unique survey you will remove the tubes time by time (i.e. 10, 15, 20, 25 and 30 cycles) and then, observe the intensity of the bands in the gel.
Thank you all for your suggestions. I wanted to post an update on how we resolved this issue. It ended up being a data analysis issue. Our analysis settings were too stringent, and as a result we were losing sensitivity (classic case of sensitivity vs. specificity). When we relaxed the constraints of our bioinformatic analysis pipeline we started capturing all our expectant targets.
- Badges
- Science method