Quantitative Amplicon Sequencing
This study represents a new method of meta-DNA relative abundance
analysis via sequencing. Given the dramatic decline in sequencing costs
as well as the sensitivity of current sequencing methods to detecting
SNPs, this method may represent a lower-cost alternative to
florescence-based qPCR. We also note that qAMPseq offers a
higher-accuracy alternative to relative read abundance that is not
subject to the issues that can arise when trying to quantify PCR
products from metabarcoding (Bell et al. 2019). That said, the protocol
here could be streamlined. One example is that it may be acceptable to
perform fewer bead cleaning steps to reduce cost and time at the bench.
Our protocol also necessitated four thermocyclers, running
simultaneously. However, similar to how gradient thermocyclers vary
annealing temperatures across reaction wells, we can envision a
modification of thermocycler heating blocks that might allow for
variation in the number of reaction cycles, and perhaps allow
qAMPseq reactions to be run on the same machine, also lowering costs
(e.g. Schicke and Hofmann, 2007).
One important difference between our study and studies that use markers
such as ITS2 and rcbL is that we had specific target species we
quantified in our samples. As such, we designed primers to amplify
regions where we knew there were SNPs that differentiated our target
species from each other, rather than relying on variation in ITS2 and
rcbL to distinguish Clarkia from each other. It is also important
to note that, because we were using closely related species within theClarkia genus, our application of the meta-barcoding approach was
likely not subject to many of quantitative biases identified by Bell et
al. (2019). These biases include copy number variation of the amplified
gene, differences in DNA isolation efficacy among samples, and variation
in primer amplification efficiency. If others are to use this method
with primers that target a broader range of possibly more divergent
species, these additional biases need to be carefully considered in
experimental design.
Finally, the difference between estimates of relative abundance from
qAMPseq versus relative read abundance depended on the tolerance with
which we filtered raw RRA values. In our study, the 10% cutoff of RRA
best matched the results from qAMPseq. A benefit to using qAMPseq,
rather than relative read abundance, is that it did not require an
arbitrary cutoff for the proportion of reads at the end of PCR that we
needed to filter out. However, qAMPseq still required that we define an
arbitrary number of reads as a threshold for amplification; this value
is analogous to the critical Ct value in the qPCR method of
quantitation. Importantly, it is likely that the similarity in relative
abundances between our qAMPseq approach and RRA was due to the fact that
our samples were diluted to a similar starting concentration of DNA (2
ng/uL). In studies wishing to use RRA in lieu of qAMPseq or qPCR, it
should be noted that RRA may not yield accurate estimates of relative
abundance if DNA concentrations are highly variable (Bell et al. 2019).
We also highlight that because all of our samples were diluted to the
same initial concentration, our analysis does not incorporate
information about the size of the sampled pollen ball, so we cannot draw
conclusions about the amount of pollen that different pollinators
transport in their pollen balls. Irrespective of approach, care must be
taken in determining the concentration of pre-amplification samples with
any particular primer/target combination, as well the interpretation of
the resulting data.