I have used the 1391f and EukBr primers (with mammalian blocking primer) for a human faecal microbiome project, where our goal was to detect microbial eukaryotes. I adpoted a different sequencing strategy from the one in the protocol - instead of ordering unique primers with barcodes in them, I ordered the primers with Illumina adapters attached for the Nextera XT dual-indexing strategy.
However, when analysing my data I found that these primers actually amplified the V9 region of bacterial 16S rRNA as well as the V9 region of 18S rRNA - 19-99% (mean 76%, median 86%) of the reads in each sample was simply “Eukaryota” classified no further. There were several ASVs with thousands of counts each, and when I BLASTed these sequences they all hit 16S rRNA. (I aligned against the SILVA 18S database only so I expect this is why they were still called "Eukaryota" - no bacteria in there).
I aligned these 18S rRNA primers to the E. coli 16S rRNA gene in silico and found that 1391f aligns to it perfectly with no mismatches, and EukBr aligns with 4 mismatches. In the EMP protocol, it says:
Relative to the 1389F primer of Amaral-Zettler et al. (2009), our 1391f primer’s degeneracy and its 3′ end were modified to minimize amplification of bacteria and archaea.
I'm struggling to see how this can be the case when they align to E. coli quite well. Additionally, the 1389f primer in Amaral-Zettler et al. is described as "universal", not "eukaryotic".
So I am wondering - has anyone else used these primers and found that they amplify bacterial 16S rRNA really well? Does anyone have any suggestions as to why this may have happened in my situation, but apparently has not been an issue for the EMP?
@rachaellappan - I have asked around on this end, I haven’t tracked anyone down who has used these primers. @Luke_Thompson, any experience? Maybe @William?
I have used the EMP 18S primers on seawater samples and tried classifying them in several different ways. I think having a chunk of sequences unclassified is fairly common, but your numbers do seem larger than mine. Here’s a quick overview of my preliminary findings - from ~65ish samples (all seawater), OTUs/ASVs assigned using deblur, and trimmed to 170 bp:
Using mothur’s classify.seqs with the latest SILVA database…
3.78% bacteria
Only 1 OTU archaea
6% “Unknown; unknown_unclassified”
So ~90% Eukaryota.
Another thing I tried was BLASTing against the SILVA database which gave me slightly higher percentages of bacteria/archaea, but those didn’t necessarily have high % identity or coverage so I’m guessing those would’ve been what assigned into the “Unknown;unknown_unclassified” category by classify.seqs. Still ended up with ~90% Eukaryota.
Finally, I tried out the PR2 database (which should be all Euks) and in this case 22% of the sequences came out as completely “Unclassified.” I’m assuming that these include, but are not limited to, the bacteria/archaeal sequences.
None of this is particularly helpful for your situation but I hope it’s at least useful as a point of comparison - I do imagine that your fecal sample had a higher bacteria:euk ratio than my seawater samples, so there’s that of course. One thing you didn’t mention is what length you trimmed the sequences to? I believe EMP trimmed to <100 bp but I kept it longer - what was yours? Maybe you’d want to try the entire SILVA database just to see what it classifies thing as if you do that? And finally, possibly trying PR2? I got slightly higher resolution for the eukaryotic sequences using that.
I would be interested in hearing whether you have gotten any other good suggestions or ideas - seems like the 18S-primer-using community is quite a bit smaller so I’m always eager to hear others’ experiences!
I'm sorry to hear you're getting so many prokaryotic sequences with the EMP 18S primers. I do not have much personal experience with these primers, and the vast majority of studies in the EMP have used the 16S primers only. I have asked colleagues who have more experience with the 18S primers to comment, and I see Maitreyi has provided some feedback already.
I spoke to Tony Walters, and his search of the literature revealed that the 1391f primer is in fact NOT modified. It is the same as published, and it's expected to target all three domains. We have modified the language on the EMP website accordingly:
The 18S protocol detailed here is designed to amplify eukaryotes broadly with a focus on microbial eukaryotic lineages. The primers target the 18S SSU rRNA and are based on those of Amaral-Zettler et al. (2009) and Stoek et al. (2010). The constructs are designed to be used with the Illumina platform. The forward primer, 1391f, is a universal primer, while the EukBr reverse primer favors eukaryotes, but can, with mismatch(s), bind and amplify bacteria and archaea.
Would you mind elaborating on your protocol for handling the ASV assignment with the EMP primers (1391/EukBr)? I am primarily interested in the trimming. Is this a trim to 170bp after merging reads, or before denoising? Additionally, could you share your trimmed SILVA database used for this classification? I am also having difficulties with unclassified ASVs, and am now wondering if it is an issue with the alignment between the amplicon and database.