Deep Seq II: Cast – Dr Sara Rassner, Dr Andy Mitchell, PhD student André Soares. Many thanks to our miner minders Paul & Byron.
As mentioned in Science, we had a go at sequencing metagenomes in the subsurface last December. We extracted and sequenced DNA underground as a proof of principle for battery powered, standalone sequencing. As Elizabeth Pennisi, the writer of the Science article noted, we still had “a few kinks to iron out” in terms of read length and taxonomic classification on site. Since our main motivation is to go from sample to microbiological insight on-site, wherever that may be, we felt we ought to iron out these kinks. So, yesterday we headed back to the Big Pit coal museum, armed with our MinION sequencer.
What did we keep the same?
We sampled the same ochreous biofilm at the same site
We used the same rapid library preparation kit (RAD001, even though it is now outdated)
We used the same library loading protocol (SpotON, no beads) to the same version of flow cell (our last R9.4)
We used the same version of offline MinKNOW and local basecaller
(Note: We took an R9.5 and a RAD002 kit as a backup, but didn’t use them. Working in the field means that “two is one, one is none” when it comes to technical processes.)
What did we change?
Obviously: the time of year we sampled the biofilms. We have data suggesting ochreous biofilms at the surface of South Wales coalmines change seasonally in their bacterial community structures, but the scope for temporal changes in their subsurface equivalents is unknown. This factor, coupled with the n=1 nature of the study means temporal comparisons in metagenome composition should be approached very cautiously.
The DNA extraction protocol: PhD student André Soares has worked to optimize the extraction protocol to address two challenges:
- Compatibility with the battery centrifuge to avert, er, spoaradic yet catastrophic sample tube loss
- DNA yield and integrity
As we hoped, it appears that modification of the DNA extraction protocol made a very positive difference.
OK, so it’s the DNA extraction, stupid.
On the basis of experience with surface ochre, we were disappointed in the yield of DNA and the read length obtained in Deep Seq I: we used a standard PowerSoil protocol but with TerraLyzer bead-beating. Ultimately, using 10% of the recommended DNA input for nanopore sequencing we were pleasantly surprised to get any worthwhile reads at all. As biofilmed sediments such as the ochre sampled are characterized by a high ratio of mineral surfaces to biomass, they are particularly challenging. Exposed mineral surfaces can adsorb DNA during extraction, resulting in the loss of DNA before it can be purified.
Others have faced this challenge and developed improved DNA extraction protocols as a result. This protocol, developed by astrobiologists, uses a mixture of ethanol and phosphate buffer coupled with an additional heat treatment to help the dissociation of biomass from minerals, lysis, and crowding of mineral surface to reduce adsorption of DNA.
André extracting DNA
Faced with this situation, it isn’t uncommon for folk to simply extract from more starting material. But a 2 mL bead beating tube is a finite volume, and free movement of beads, buffer and sample is essential for effective bead beating, and incomplete suppression of inhibitors by downstream steps is likely. We felt it better to bead beat and remove inhibitors from standard volumes of sample, then combine samples onto the same spin column and elute in a standard 100 microlitre volume. With the use of the ethanol/phosphate buffer plus 2x30s bead beating and 20 minutes hot soak in André’s travel mug the yields of DNA from subsurface ochre improved considerably. But still not enough for the 200 ng DNA in 7.5 microlitres demanded for rapid library prep.
This left us with two options. Oxford Nanopore have recently released rapid low input kits which permit PCR amplification of transposase tagged DNA, permitting sequencing of 10 ng DNA (and with barcodes too). As attractive as this possibility is, we discounted it for several reasons
- Time. Due to the working patterns of Big Pit, our time on site is very limited. The PCR step took 1.5h on a lab based thermal cycler, plus Qubit quantification and AMPure clean up. It would take longer on a mini-PCR cycler
- Using (long) PCR would limit the read length to 5-10 Kbp. We’re not avid whale watchers, but I can admit we like seeing the occasional long read sneak through a pore.
- PCR in the field: one more thing to go wrong, one more source of contaminants, one more thing for the bias police.
So, in the end we decided to just go for a final AMPure bead based clean up of the PowerSoil eluate, to provide a quick additional cleaning step, and critically, up to a 20 fold concentration from 2×100 microlitre eluates to 10 microlitres which could be quantified and sequenced. Trials in the lab showed it did not affect DNA integrity adversely.
On the day itself though, we needn’t have worried: our duplicate extractions yielded 36-38 ng per microlitre when quantified by Qubit. Enough to exceed the 200 ng specification for the rapid kit. We occasionally take yes for an answer. So, in the interests of time we proceded without AMPure.
I think I’m starting to get the hang of this SpotON loading malarkey.
What did we find?
André inspects the taxonomic classification of metagenomic reads: from sample to insight, underground.
Overall, the read length distribution improved considerably. While our longest reads were just 59 kbp, using the older version of MinKNOW we couldn’t expect much longer.
The first 120 reads to land in the pass folder were hived off to a second laptop for Centrifuge based taxonomic classification, which provided identities for about 68% of those reads in 5 minutes. While visualization with Pavian didn’t pan out underground inspection of the csv file showed some interesting overlaps with DeepSeq I in terms of the proteobacteria detected, and the prevalence of Streptomycete reads among the Actinobacteria was again noteable. Melinabacteria and Pandoravirus were among the one read wonders: we’ll look into these data more carefully over the coming weeks.
What happened next?
Sara marching out the MinION.
As the run was still going at the end of the day underground, once more we moved the MinION out of the underground pumphouse, down a roadway and up the mineshaft to an office. The sequencing run continued unabated despite the shift of 100 metres elevation and 10 °C. As we had to return to Aberystwyth, we kept the MinION running in the back of the van on the way home, before putting the run out of its misery with a minimum of ceremony in a carpark in Builth Wells.
An uphill struggle?
On the way home I noticed a slight tendency for the number of pores actively generating sequence to change in relation to road conditions. Coming up from the Valleys means there’s a lot of uphills, downhills and roundabouts. Hardly sequencing at 17,000 kph but not the gentlest of rides either. Thanks to Matt “Maverick” Loose it’s clear MinIONs are resistant to inversion, so if this was a real effect, the mechanism is unclear. Who knows?
I think it’s time to update both our nanopore pre-prints and prepare for a long, cold summer.