So I had two thoughts when I read the article title “Indexed PCR Primers Induce Template-Specific Bias in Large-Scale DNA Sequencing Studies“…
#1: Ack! Yet another thing to worry about in the insanely complicated craziness that is 16S amplicon sequencing.
#2: I probably shouldn’t be the person to blog about this because I’ve honestly never quite understood the primer details
Having read the article now, I think I feel better about #1, but caveat emptor because of #2.
Basically the authors are concerned about template bias in a sequence-specific manner, meaning that samples with some indices will amplify better or worse than others. That would be bad, very bad… and in fact the authors show this to be the case with the primer design that they’re testing. And not just tiny changes, they demonstrate a change in abundance of >75% in some cases. Happily, they demonstrate that this bias can be removed through a double-PCR method. Here’s the full abstract below:
Massively parallel sequencing is rapidly emerging as an efficient way to quantify biodiversity at all levels, from genetic variation and expression to ecological community assemblage. However, the number of reads produced per sequencing run far exceeds the number required per sample for many applications, compelling researchers to sequence multiple samples per run in order to maximize efficiency. For studies that include a PCR step, this can be accomplished using primers that include an index sequence allowing sample origin to be determined after sequencing. The use of indexed primers assumes they behave no differently than standard primers; however, we found that indexed primers cause substantial template sequence-specific bias, resulting in radically different profiles of the same environmental sample. Likely the outcome of differential amplification efficiency due to primer-template mismatch, two indexed primer sets spuriously change the inferred sequence abundance from the same DNA extraction by up to 77.1%. We demonstrate that a double PCR approach alleviates these effects in applications where indexed primers are necessary.
However, I don’t think (hope) this is of concern for many of the amplicon studies that I’m aware of with 16S. For example, the primers used in our lab (modified from the Earth Microbiome protocol) have a quite significant primer pad and linker between the primer binding site and the index sequence. Therefore it’s hard to imagine that the index would interfere with primer binding as described in this paper… where the index sequence is directly adjacent to the primer binding site.
Or am I crazy?
I agree. I felt some anxiety as well upon reading the article’s title because we used to do two PCRs until about a year ago (and some in our group still do). However, it’s important to note that the primers used here were for mitochondrial 16S DNA from metazoans exclusively and not the standard 16S rRNA primers used to characterize bacteria.
Good point, but it seems the bigger issue is the location of the barcode itself (relative to the primer binding site)?
Wow, this definitely scared me when I first read the post title. So are Sonia’s primers in trouble of having this bias?
I just randomly found this post and thought I’d chime in (I led the study).
The one thing I hoped people would take away from this is that you can test your fears yourself: Just sequence replicate PCRs from the same DNA sample using different indexes.
In my field (macrobiologists), it seems all too common that researchers, journals, and funding agencies are so excited about being associated with a novel technique that they throw regard for basic principles of science out the window. In this case, it is a replicated design that allows one to test for an effect of lab procedures. I’m not as familiar with the microbiome literature, but my sense is that the problem exists to some degree there, too.
As far as I can tell, this issue should not be a problem in the single PCR approach *IF* all of the templates you’re interested in amplifying are invariant at the site where the index would bind. Again, I am not terribly familiar with microbial genomics or protocols, so I can’t venture a guess as to whether or not this is a problem for any specific protocol.
Happy to chat about this any time. You can drop me an email if you’d like (jodonnellbio at the email run by Google).
Cheers!