Illumina sequencing of eDNA libraries

[MPen92]

Hello everyone,

I am the current manager for the Marine Gene Probe Lab (MGPL) at Dalhousie University and we have been trying to sequence eDNA samples from freshwater lakes using the 12S MiFish-U primers from Miya et al. (2015). The lab has an in-house MiSeq for sequencing with v2 chemistry. Our protocol differs from the original paper and, while it gets some usable data, is often resulting in poor R2 quality and merge rates. We have also been having consistent issues with over-clustering, even when the libraries were diluted to 6pM prior to loading. I figure this is at least part of the reason for the poor R2 reads. We've also been getting higher amounts of adapter in our sequencing reads than we would like to see, with some instances taking up about 50% of the reads for a sample. I suspect those instances are due to poor template quality or insufficient template.

The actual preparation of the libraries is done by me (our Research Associate processes the data), and I'm wondering if anyone else on here is attempting similar experiments or has successfully generated good quality data from eDNA template. It's proving tricky for us. Thanks!

[ADoetsch]

I don't know the protocol and need to assume some things but it looks suspicious that you get overclustering with such low (perceived) library concentrations. What is the cluster density you experience with these libraries? With around 8 pM library concentration, we usually get around 600 - 800 clusters (I think the unit is per square mm). If you have much higher cluster density with lower concetration, maybe the concentration is underestimated.

When you talk about problems with adapters, do you mean adapter dimer sequences? If you get a lot of those, keep in mind that these tend to be much shorter, which affects the calculation of molarity a lot. How exactly do you calculate the molar concentration?

If primer/adapter dimers are a big problem, I recommend thorough size selection purification, e.g. with magnetic beads or gel purification.

[MPen92]

Hello there,

When I ran the same library at 8pM I got a density of 1021K/mm2. It's hard to say to what extent that hurt the reads since we also had a clog in the lines during the run, as well. It was a sad day.

The density I got at 6pM was 791K/mm2. Still a bit higher than we would have liked (we aim for 600-700). I'm starting to suspect the indices we use for quantification may be the problem. They were made in-house and are about three years old you, and subjected to multiple freeze-thaws. I'm going to test them against some new ones to see if the concentrations detected are similar, and I may try this with some diluted PhiX (known fragment size and concentration can act as controls).

Yes, the issue is adapter dimer sequence. Some of which appears to be in fragments that are the same length as our target library (400-450bp). I've had some back and forth about this with the lab, but I think the problem there is our initial annealing temperature (50). It's lower than anything I've seen in the literature, and results in large clouds of primer dimer when we run it with blanks. We calculate our library concentration using qPCR.

We currently do a gel purification (1% gel) and then a 1:1 ratio bead clean. I've tried this with a 0.6:1 ratio (bead:library) in the past and that seemed to help remove the smaller fragments. My ideal would be purifying the library with a two-sided size selection. We've also seen people bead-cleaning the first PCR to remove adapter dimers prior to indexing, which we're considering trying.

The hesitation I've been getting is with the cost associated with testing, which isn't cheap for us. Each MiSeq kit costs ~$3k for our lab including tax, shipping, and dollar conversion (we're in Canada). But I'm really not convinced that our protocol is optimized.

[ADoetsch]

I see. It sounds like you're already following many recommendations that you usually get in such cases.
I'd say that the cluster density ist not really very high for the MiSeq. We've had runs with higher density that produced good results.

One thing that comes to my mind is, if you have tried increasing the PhiX concentration already. How is sequence diversity in your target? We've had good experience with at least 20% PhiX in 16S-rRNA runs.

[MPen92]

I ran the previous run with a 20% PhiX Spike-in but the alignment was lower than that (~13.5%). Because our library fragments are smaller tha the PhiX by 50-100bp I suspect it's being out-competed for flow cell space. Small fragments may also be an issue if that is happening.

Diversity of our target is low (12S metabarcoding, but the lakes are species-poor for fish). Our R2 reads, in particular, are poor quality.