ArticlePDF Available

Optimizing extraction and targeted capture of ancient environmental DNA for reconstructing past environments using the PalaeoChip Arctic-1.0 bait-set


Abstract and Figures

Sedimentary ancient DNA (sedaDNA) has been established as a viable biomolecular proxy for tracking taxon presence through time in a local environment, even in the total absence of surviving tissues. SedaDNA is thought to survive through mineral binding, facilitating long-term biomolecular preservation, but also challenging DNA isolation. Two common limitations in sedaDNA extraction are the carryover of other substances that inhibit enzymatic reactions, and the loss of authentic sedaDNA when attempting to reduce inhibitor co-elution. Here, we present a sedaDNA extraction procedure paired with targeted enrichment intended to maximize DNA recovery. Our procedure exhibits a 7.7–19.3x increase in on-target plant and animal sedaDNA compared to a commercial soil extraction kit, and a 1.2–59.9x increase compared to a metabarcoding approach. To illustrate the effectiveness of our cold spin extraction and PalaeoChip capture enrichment approach, we present results for the diachronic presence of plants and animals from Yukon permafrost samples dating to the Pleistocene-Holocene transition, and discuss new potential evidence for the late survival (~9700 years ago) of mammoth ( Mammuthus sp. ) and horse ( Equus sp. ) in the Klondike region of Yukon, Canada. This enrichment approach translates to a more taxonomically diverse dataset and improved on-target sequencing.
Content may be subject to copyright.
Optimizing extraction and targeted capture of ancient environmental
DNA for reconstructing past environments using the PalaeoChip
Arctic-1.0 bait-set
Tyler J. Murchiea,b* , Melanie Kucha,b, Ana T. Duggana,b, Marissa L. Ledgerc, Kévin Roched,e, Jennifer Klunka,f,
Emil Karpinskia,f, Dirk Hackenbergera,g, Tara Sadowaya,h, Ross MacPheei, Duane Froesej, Hendrik Poinara,b,g*
McMaster Ancient DNA Centre, McMaster University, Canada
Department of Anthropology, McMaster University, Canada
Department of Archaeology, University of Cambridge, United Kingdom
CNRS UMR 5607 Ausonius, University of Bordeaux Montaigne, France
CNRS UMR 6249 Chrono-Environment, University of Bourgogne Franche-Comté, France
Department of Biology, McMaster University, Canada
Department of Biochemistry, McMaster University, Canada
University Health Network, Toronto
Division of Vertebrate Zoology/Mammalogy, American Museum of Natural History
Department of Earth and Atmospheric Sciences, University of Alberta
*Corresponding authors: Tyler J. Murchie and Hendrik Poinar, Department of Anthropology, McMaster University, 1280 Main Street West, Hamilton,ON L8S
4L8, Canada. E-mail address: and
(RECEIVED August 13, 2019; ACCEPTED June 2, 2020)
Sedimentary ancient DNA (sedaDNA) has been established as a viable biomolecular proxy for tracking taxon presence
through time in a local environment, even in the total absence of surviving tissues. SedaDNA is thought to survive through
mineral binding, facilitating long-term biomolecular preservation, but also challenging DNA isolation. Two common limi-
tations in sedaDNA extraction are the carryoverof other substances that inhibit enzymatic reactions, and the loss of authentic
sedaDNA when attempting to reduce inhibitorco-elution. Here, we present a sedaDNA extraction procedure paired with tar-
geted enrichment intended to maximize DNA recovery. Our procedure exhibits a 7.719.3x increase in on-target plant and
animal sedaDNA compared to a commercial soil extraction kit, and a 1.259.9x increase compared to a metabarcoding
approach. To illustrate the effectiveness of our cold spin extraction and PalaeoChip capture enrichment approach, we present
results for the diachronic presence of plants and animals from Yukon permafrost samples dating to the Pleistocene-Holocene
transition, and discuss new potential evidence for the late survival (9700 years ago) of mammoth (Mammuthus sp.) and
horse (Equus sp.) in the Klondike region of Yukon, Canada. This enrichment approach translates to a more taxonomically
diverse dataset and improved on-target sequencing.
Keywords: Sedimentary ancient DNA; paleoenvironmental DNA; capture enrichment; environmental DNA; DNA
extraction; PowerSoil DNA extraction kit; late Quaternary extinctions; Pleistocene-Holocene transition; Yukon
paleoenvironment; PalaeoChip bait-set
Means of recovering and analyzing ecologically informative
environmental DNA (eDNA) have improved substantially
thanks to ongoing developments in high-throughput
sequencing (HTS) technologies (Taberlet et al., 2018). Sedi-
mentary ancient DNA molecules (sedaDNA, referring to a
subset of ancient eDNA sample types; see Rawlence et al.
[2014, p. 614]) have been successfully recovered from a
diverse range of depositional settings to evaluate what is
thought to be the local (Parducci et al., 2017, p. 930; Edwards
et al., 2018) diachronic presence of animals (Haile et al.,
2009; Giguet-Covex et al., 2014; Graham et al., 2016; Peder-
sen et al., 2016; Slon et al., 2017), plants (Anderson-
Carpenter et al., 2011; Willerslev et al., 2014; Epp et al.,
Cite this article: Murchie, T. J. et al 2020. Optimizing extraction and
targeted capture of ancient environmental DNA for reconstructing past
environments using the PalaeoChip Arctic-1.0 bait-set. Quaternary
Research 124.
Quaternary Research
Copyright © University of Washington. Published by Cambridge University Press, 2020.
2015; Alsos et al., 2016; Niemeyer et al., 2017), fungi (Bel-
lemain et al., 2013), microbiota (DCosta et al., 2011;Ahmed
et al., 2018), and eukaryotic parasites (Søe et al., 2018). It is
thought that much of sedaDNA survives in the absence of tis-
sues through the formation of organo-mineral complexes
(Greaves and Wilson, 1970; Lorenz and Wackernagel,
1987a;1987b; Ogram et al., 1988; Blum et al., 1997; Arnold
et al., 2011; Morrissey et al., 2015; Gardner and Gunsch,
2017) as extracellular genetic material binds to common con-
stituents of sediments such as humics (Crecchio and Stotzky,
1998), calcite (Cleaves et al., 2011), clays (Goring and Bar-
tholomew, 1952; Greaves and Wilson, 1969; Cai et al.,
2006), and other silicates (Lorenz and Wackernagel, 1987b;
Bezanilla et al., 1995). Soil minerals have been found to sta-
bilize a fraction of such complexes, allowing DNA molecules
to resist decomposition (Morrissey et al., 2015). However,
strong mineral binding can also result in poor sedaDNA min-
eral release when attempting to isolate these molecules during
DNA extraction (Alvarez et al., 1998; Saeki et al., 2010).
Mineral-bound sedaDNA is recovered in bulk in the form
of short, disseminated biomolecules from a diverse range of
organisms. This generally prohibits genomic reconstructions
of single individuals from eDNA samples, but it can allow for
identifying the presence (and, potentially, absence and rela-
tive abundance) of organisms through time. SedaDNA
methods show the most promise in reconstructing palaeoora
(Anderson-Carpenter et al., 2011; Willerslev et al., 2014;
Pedersen et al., 2016; Niemeyer et al., 2017; Sjögren et al.,
2017), or the variable presence of a particular taxon tempo-
rally (Graham et al., 2016). Comparative analyses using
eDNA, palynology, macroremains, and vegetation surveys
have argued that eDNA currently functions best as a comple-
ment to traditional palaeoecological methods (Jørgensen
et al., 2012a;2012b; Pedersen et al., 2013), as often each
proxy tends to identify a partially-overlapping (complemen-
tary) set of organisms (Parducci et al., 2017, p. 932). With
ongoing progress in ancient DNA (aDNA) methods, a single
DNA library (DNA fragments with articial adapter
sequences attached [ligated] to allow for high-throughput
sequencing [HTS]) can increasingly be used to identify a
diverse range of prokaryotic and eukaryotic organisms simul-
taneously from a single sample with shotgun sequencing
(sequencing a random subset of a DNA library to a select
depth, e.g., ten million molecules) or can be targeted to
amplify or enrich for specic taxa of interest. Using a targeted
approach signicantly improves the informative fraction of
sequenced DNA reads over shotgun sequencing by reducing
the proportion of off-target moleculesespecially microbial
DNA that otherwise tends to proportionally dominate shot-
gun data.
Fig. 1 - Colour online, B/W in print
Figure 1. (color online) Permafrost sampling sites in the Klondike region of Yukon, Canada (Table 1). Ice sheet data from Dyke (2004). Sea
level at last glacial maximum (LGM, 26.519 ka BP) (Clark, 2009) set to 126 meters below sea level (m bsl) based on midpoint between max-
imum and minimum eustatic sea level estimation models in Clark and Mix (2002). IFC = ice-free corridor. Base map data retrieved from Geo-
Yukon (, hosted by the Government of Yukon); contours elevation unit: meters above sea level.
2T.J. Murchie et al.
Table 1. Sample descriptions and read counts (SET-E).
Site, core, age Sample
DNA Targeting
Bait mapped &
of Total
Plant ref mapped
of Total
9685 cal yr BP
GB1 Metabarcoding DCosta
et al., 2011
109,233 311 0.3% 1,937 1.8%
SET256-MB Metabarcoding
(trnL only)
493,220 1,855 0.4%
SET257-MB 335,392 1,529 0.5%
SET258-MB 424,696 1,911 0.4%
SET256-En Enrichment 6,292,874 12,812 0.2% 63,562 1.0%
SET257-En 675,550 4,105 0.6% 18,541 2.7%
SET258-En 820,382 4,354 0.5% 17,695 2.2%
SET256-SG Shotgun 1,717,174 5 0.0% 48 0.0%
SET268-MB Metabarcoding
(trnL only) Cold Spin
373,049 1,959 0.5%
SET269-MB 442,926 2,417 0.5%
SET270-MB 330,692 1,942 0.6%
SET268-En Enrichment 1,704,149 122,282 7.2% 176,221 10.3%
SET269-En 1,782,291 104,760 5.9% 143,800 8.1%
SET270-En 1,269,901 102,542 8.1% 140,495 11.1%
SET268-SG Shotgun 2,202,687 133 0.0% 153 0.0%
Lucky Lady II
LLII 12-84-3
13,205 cal yr BP
LL3 Metabarcoding DCosta
et al., 2011
738,708 971 0.1% 5,615 0.8%
SET259-MB Metabarcoding
(trnL only)
452,855 2,115 0.5%
SET260-MB 272,395 1,399 0.5%
SET261-MB 386,793 1,655 0.4%
SET259-En Enrichment 515,685 2,563 0.5% 7,056 1.4%
SET260-En 692,434 3,470 0.5% 8,285 1.2%
SET261-En 1,060,105 2,368 0.2% 6,463 0.6%
SET259-SG Shotgun 1,517,583 17 0.0% 21 0.0%
SET271-MB Metabarcoding
(trnL only)
Cold Spin
565,986 2,449 0.4%
SET272-MB 437,902 2,269 0.5%
SET273-MB 483,637 2,583 0.5%
SET271-En Enrichment 1,837,939 119,031 6.5% 146,562 8.0%
SET272-En 1,343,928 110,609 8.2% 134,102 10.0%
SET273-En 1,267,881 112,855 8.9% 139,643 11.0%
SET272-SG Shotgun 2,122,805 145 0.0% 150 0.0%
Lucky Lady II
LLII 12-217-8
15,865 cal yr BP
LL1 Metabarcoding DCosta
et al., 2011
77,373 448 0.6% 2,359 3.0%
SET262-MB Metabarcoding
(trnL only)
459,492 2,140 0.5%
SET263-MB 295,806 1,675 0.6%
SET264-MB 552,797 2,082 0.4%
SET262-En Enrichment 534,677 452 0.1% 2,119 0.4%
SET263-En 873,741 312 0.0% 1,634 0.2%
SET264-En 418,730 648 0.2% 2,721 0.6%
SET264-SG Shotgun 1,497,940 6 0.0% 14 0.0%
SET274-MB Metabarcoding
(trnL only)
Cold Spin
358,119 1,320 0.4%
SET275-MB 448,934 1,720 0.4%
SET276-MB 458,110 1,579 0.3%
SET274-En Enrichment 867,275 6,600 0.8% 10,223 1.2%
SET275-En 996,802 5,614 0.6% 8,541 0.9%
SET276-En 352,658 2,802 0.8% 4,019 1.1%
SET274-SG Shotgun 1,992,150 40 0.0% 50 0.0%
Optimizing extraction and targeted capture of ancient environmental DNA 3
Despite rapid advances in aDNA techniques, two extrac-
tion related challenges persist that can limit the ability to
fully exploit sedimentary genetic archives: 1) the presence
of enzymatic inhibitors in sediment extracts, and 2) the loss
of ecologically informative sedaDNA due to overly vigorous
inhibitor removal. Inhibitors are substances that inhibit a
broad array of enzymatic reactions necessary to prepare
DNA for sequencing, such as adapter-ligation during library
preparation or polymerase chain reaction (PCR) amplication
during metabarcoding (described in the following subsec-
tion). Enzymatic inhibitors are often present in higher
concentrations in samples that are prepared with techniques
designed to maximize the recovery of aDNA (characteristi-
cally short, damaged fragments; Dabney et al., 2013a).
Commercial kits have been designed with reagents
that remove inhibitors from lysed samples, but these
reagents can also remove potentially informative DNA that
is already in low concentrations in ancient remains (Dong
et al., 2006).
To address these challenges, this study evaluates various
inhibitor removal treatments for their ability to reduce the car-
ryover of enzymatic inhibitors in sedaDNA extracts while
maximally retaining endogenous palaeoenvironmental DNA
that can successfully undergo library adapter-ligation and tar-
geted enrichment. Our aim is to minimize the need for exces-
sive PCR amplications on puried eluates by instead utilizing
a capture enrichment approach (Carpenter et al., 2013)thatto
date has typically been applied only to discrete, rather than dis-
seminated, materials (Rawlence et al., 2014, p. 614). We also
wanted to determine whether this approach would be as viable
for open-air sediments as the technique has been in cave
contexts (Slon et al., 2017). Capture enrichment would theoret-
ically aid in part to minimize the propagation of stochastic PCR
biases and allow for a far greater range of targetable genetic
loci that are more amenable to aDNA fragment sizes and dam-
age characteristics.
Four previously studied sediment core samples (DCosta
et al., 2011;Sadoway,2014; Mahony, 2015) recovered from
Yukon permafrost exposures (Fig. 1,Table 1) in the Klondike
region were chosen to experimentally optimize a new
sedaDNA extraction procedure. Further details of these itera-
tive extraction experiments are included in the supplementary
materials. We based these modications on our in-house lysis
solution and high-volume binding buffer, silica-spin column
extraction method as described by Dabney et al. (2013b).
Our sedaDNA modied Dabney et al. (2013b)extractionpro-
tocol is described here (Fig. 2) and referred to in-text as the cold
spin method. Experiments with alternative inhibition removal
techniques are detailed in the supplementary materials as sedi-
ment extraction/enrichment testing SET-A through SET-D.
The main-text experiment is referred to throughout as SET-E.
The cold spin optimized extraction protocol was used to eval-
uate taxonomic assignments between shotgun sequenced and
target enriched datasets from four Yukon permafrost samples.
These were enriched using a bait-set designed to capture mito-
genomes of extinct and extant northern animals (focused on
megafauna) as well as the chloroplast barcoding loci trnL,
rbcL, and matK from arctic and subarctic plants. This bait/
probe set is referred to here as the PalaeoChip Arctic-1.0 bait-
set. Results were also compared with shotgun and enriched
libraries extracted from the same samples using the DNeasy
PowerSoil DNA extraction kit (QIAGEN)one of the most
Table 1. Continued.
Site, core, age Sample
DNA Targeting
Bait mapped &
of Total
Plant ref mapped
of Total
Bear Creek
BC 4-2B
30,000 cal yr BP
BC Metabarcoding DCosta
et al., 2011
172,330 1348 0.8% 8,554 5.0%
SET265-MB Metabarcoding
(trnL only)
427,087 2,486 0.6%
SET266-MB 333,002 2,664 0.8%
SET267-MB 325,894 2,204 0.7%
SET265-En Enrichment 682,544 1,590 0.23% 4,340 0.6%
SET266-En 868,112 1,770 0.20% 4,392 0.5%
SET267-En 857,863 2,249 0.26% 5,533 0.6%
SET265-SG Shotgun 1,338,467 5 0.00% 0 0.0%
SET277-MB Metabarcoding
(trnL only)
Cold Spin
547,884 4,176 0.8%
SET278-MB 475,359 3,398 0.7%
SET279-MB 634,490 4,852 0.8%
SET277-En Enrichment 1,254,744 62,778 5.0% 73,276 5.8%
SET278-En 1,294,040 37,878 2.9% 45,323 3.5%
SET279-En 1,544,949 47,613 3.1% 57,484 3.7%
SET279-SG Shotgun 8,593,408 133 0.0% 131 0.0%
*Reads map-ltered to animal and plant baits, size ltered to 24 bp, de-duplicated, BLASTn aligned, and MEGAN LCA assigned. **Reads map-ltered to plant
references, with the same subsequent ltering parameters. Core section ages as per DCosta et al. (2011), Mahony (2015), and Sadoway (2014).
4T.J. Murchie et al.
commonly utilized sedaDNA extraction procedures. For each
of the four permafrost core sections, sediments were subsam-
pled and homogenized, then split into three 250 mg replicates
for both extraction methods and subsequent targeting strate-
gies in order to also assess intrasample variability. We com-
pared our results with previously sequenced PCR
metabarcoding data (Sadoway, 2014) derived also from
250 mg of the same core sections that had been extracted
with a similar (but different) approach (DCosta et al.,
2011). We then used metabarcoding to amplify the chloro-
plast intron trnL (Taberlet et al., 2007) on all of the same
PowerSoil and sedaDNA optimized extracts to evaluate the
variation in DNA recovery between our data and the previous
Sadoway (2014) metabarcoding approach.
Fig. 2 - Colour online, B/W in print
Figure 2. (color online) Subsampling to taxon assignment schematic comparing extraction, targeting, and bioinformatic ltering strategies.
See the Methods section for further details on extraction, double-stranded library preparation, capture enrichment, qPCR assays, and the bio-
informatic workow.
Optimizing extraction and targeted capture of ancient environmental DNA 5
Background: current approaches to analyzing
PCR metabarcoding
PCR metabarcoding is the most popular technique in eDNA
research (Taberlet et al., 2018, ch. 2). This approach targets
short, highly variable regions of the genome (regions with
enough variation to identify organisms to family, genus, or
species ranks) which also happen to be immediately anked
by deeply conserved regions (an identical genetic sequence
region across many organisms, such as all plants). Primers
are designed to bind to these two conserved anking regions,
facilitating a PCR reaction (Saiki et al., 1985,1988; Mullis
and Faloona, 1989) that amplies the hyper-variable internal
barcode at an exponential rate. A metabarcoding approach
exponentially amplies the selected metabarcode for all
DNA fragments of this target region from multiple organisms
simultaneously, overwhelming any other off-target eDNA
left in the sample extract. After these molecules are
sequenced and processed, the reads can then be identied
similarly to other approaches to determine a probable host
organism using algorithmic packages such as OBITools
(Boyer et al., 2016)orBLASTn (Altschul et al., 1990),
which compare the sequenced reads against a nucleotide ref-
erence database such as GenBank NCBI (National Center for
Biotechnology Information; Benson et al., 2013; Agarwala
et al., 2016) or otherwise curated references. To some degree,
PCR metabarcoding can mitigate the aforementioned chal-
lenges of inhibition through dilutions or additional purica-
tions (McKee et al., 2015), with the addition of reagents
such as bovine serum albumin (BSA) (Kreader, 1996; Gar-
land et al., 2010), with high cycle numbers to facilitate inhib-
itor denaturation, and with high polymerase concentrations.
The technique can also theoretically cope with substantial
DNA loss from inefcient DNA purications (in part, from
overly aggressive inhibitor removal techniques) by exponen-
tially amplifying low abundance molecules to sufcient con-
centrations for downstream manipulations.
PCR metabarcoding is widely used in eDNA applications
to great effect (e.g., Giguet-Covex et al., 2014; Willerslev
et al., 2014; Alsos et al., 2016; Epp et al., 2015; Sjögren
et al., 2017), but the technique has several aspects that theo-
retically limit the information that can be acquired from
ancient samples. Perhaps most signicantly, if the few
genetic regions amenable to metabarcoding fail to preserve
for particular organisms in the sediment samples being tested
(due to variable eDNA input as related to biomass turnover
and stochastic taphonomic processes), DNA from those
organisms will not be sequenced, resulting in abundant
false negatives, even with a large number of PCR replicates
per sample to control for stochastic amplications (Nichols
et al., 2018, p. 8). This problem is signicant for palaeoenvir-
onmental DNA applications where substantial DNA damage
and loss has occurred. Regions suitable for metabarcoding
often exist on the high end of the aDNA fragment spectrum
(typically >100 bp [base pairs]) whereas most aDNA
fragments have an average fragment length of approximately
50 bp or less. The chloroplast trnL intron P6 loop, used for
metabarcoding plants, is a relative exception with a amplicon
range of 49188 bp (with primer landing sites added [Taber-
let et al., 2007]). However even these short amplicons can
exceed the typical DNA fragment sizes expected with
degraded aDNA. Also, most sedaDNA does not come from
suitable metabarcoding regions, but rather from a random
metagenomic mixture of fragments in varying states of
decay across a range of nuclear and organelle genomes.
These non-barcoding regions are still useful in taxonomic
identications if they can be sequenced, either through
expensive and inefcient deep shotgun sequencing (sequenc-
ing a library to near exhaustion) or another targeted approach.
Metabarcoding is also vulnerable to differential amplication
rates due to variable molecular abundance per taxa, unequal
damage (by region and taxa), variability in metabarcode
amplication efciency, and variably optimal PCR condi-
tions depending on the eDNA mixture (Kanagawa, 2003;
Bellemain et al., 2010; Krehenwinkel et al., 2018; Nichols
et al., 2018; Sze and Schloss, 2019). These factors compound
downstream biases in taxonomic determinations, especially if
during inhibitor removal there was substantial loss of low-
abundance molecules from taxa with comparatively low bio-
mass turnover (Yoccoz et al., 2012, p. 3651). Metabarcoding
also does not allow one to evaluate DNA damage patterns
(such as cytosine deamination causing C to T base modica-
tions on the terminal ends of aDNA fragments) due to the reli-
ance on PCR amplications of discrete genetic loci with
intact primer landing sites. By contrast, shotgun sequencing
and enrichment techniques can utilize these characteristic
damage signals to build a case for ancient DNA authenticity
(Ginolhac et al., 2011; Jónsson et al., 2013).
Capture enrichment
As an alternative to PCR metabarcoding, capture enrichment
(also referred to as targeted capture) is a powerful means of
increasing the fraction of target molecules that can be recov-
ered from a DNA library (Mamanova et al., 2010; Schuene-
mann et al., 2011; Carpenter et al., 2013; Marciniak et al.,
2015). Capture enrichment involves designing a set of
RNA probes or baits that closely match DNA sequences
(including whole genomes) from organisms of interest,
which are then allowed to hybridize to the predetermined tar-
get molecules that may be present in the DNA library. After
hybridizing (either with in-solution biotinylated baits or on
a solid-state microarray with pre-attached baits, see Marciniak
et al., 2015, p. 29), the target molecules are sequestered while
the non-target fraction is washed away for discard or storage
for alternative use (Klunk et al., 2019). The technique allows
for whole genome capture of multiple target organisms
simultaneously, resulting in a huge increase in the proportion
and diversity of informative sequenced DNA without the
need for exponential PCR amplications and the compound-
ing biases thereinas well as substantially reduced costs
6T.J. Murchie et al.
compared to deep shotgun sequencing. Capture enrichment
has been primarily utilized with discrete materials such as
bone, having seen limited use to date with eDNA sample
types (Slon et al., 2016;2017). Targeted capture has its
own associated limitations such as incomplete off-target
exclusion, GC biases (due to stronger hybridization of
sequences rich in guanine or cytosine bases versus those
that are AT rich), and limitations in nding unexpected mol-
ecules due to reference-based bait design. However, capture
enrichment has the potential to allow for a far more diverse
recovery of ecologically informative eDNA across entire
genomes, especially in geographic regions with relatively
complete nucleotide reference databases of organisms, such
as the arctic (Sønstebø et al., 2010; Willerslev et al., 2014).
Being unrestricted by intact priming sites (compared to meta-
barcoding) can also allow for an assessment of molecular
damage on an organism-by-organism basis (Ginolhac et al.,
2011). Capture enrichment has the best potential when
using a DNA extraction procedure intended to maximize
DNA recovery. But, as discussed previously, this becomes
a signicant challenge with sedaDNA because of the high
co-elution of inhibitory substances that inhibit enzymes
used for adapter-ligation during library preparation.
Field sampling
The Yukon permafrost core samples (Fig. 1,Table 1)usedin
this analysis were previously collected and analyzed by
DCostaetal.(2011), Mahony (2015), and Sadoway (2014)
from near-vertical sediment exposures, then kept in cold stor-
age at the McMaster Ancient DNA Centre. All four samples
were analyzed by Sadoway (2014) using PCR metabarcoding.
Prior to core collection by all three original research teams, the
sampling area was cleared of eroded materials back to frozen
sediments to create a fresh coring surface for a 10 cm diam-
eter coring tube 30 cm in length. Horizontal core samples
were drilled with a small portable drill, recovered frozen, stored
individually in plastic bags, and transported frozen to the Uni-
versity of Alberta or McMaster University for subsampling.
Subsamples for aDNA were taken only from core interiors
(described below in Methods subsection 3) as exterior and
cut surfaces are more likely to contain contaminant DNA.
Bear Creek (sample BC 4-2B)
The Bear Creek site is located 11 km east of Dawson City,
Yukon, in the Klondike mining district. Mining activities
exposed a 10 m vertical sampling surface consisting of 3 m
of alluvial sediment overlain by 7 m of ice-rich loessal silt
(DCosta et al., 2011, p. Supp. 3). The Dawson tephra is prom-
inent at the site, dating to 25,300
CyrBP(30,000 cal yr BP)
(Froese et al., 2006), and is situated between 5.2 and 6 m from
the base of the exposure. Horizontal core sample BC 4-2B was
collected 50 cm below the tephra under a stratied lens of ice,
likely the remnant of a surface icing similar to other sites
associated with Dawson tephra in the area (Froese et al.,
2006). The core sample was collected from reddish-brown ice-
poor silts that extend below the tephra. These sediments are
interpreted as the palaeosurface and include the palaeoactive
layer that existed at the time of Dawson tephra deposition.
This unit was preserved due to the rapid deposition of the tephra
(80 cm thick at this site) that shifted the active layer upward.
Observations of the palaeoactive layer and preservation of the
ice body indicate that there was no thawing or water migration
in these relict permafrost sediments following deposition of the
Dawson tephra (DCosta et al., 2011,p.Supp.3).
Upper Goldbottom (sample MM12-118b)
This site is located 28 km south of Dawson along Goldbottom
Creek, a tributary of Hunker Creek and the Klondike River.
The 28.5 m exposure was divided by Mahony (2015,
pp. 6582) into ve units, dating roughly between 46,000
cal yr BP near the base, to 6000 cal yr BP near the surface.
The core sample MM12-118b used in this study comes from
Unit 4 near the top of the exposure. The sediments consist of
black and grey organic-rich silts with thin interbedded lenses
of gravels and sand, as well as components of green-grey silts
and interbedded humied brown organic silts. In situ grami-
noid and shrub macrofossils were also identied. Unit 4 is
estimated to have begun deposition ca. 10,600 cal yr BP
(9395 ± 25
C yr BP, UCIAMS-114910) (Mahony, 2015,
p. 77). The permafrost core used in this analysis
(MM12-118b) was dated to the early Holocene at ca. 9685
cal yr BP (Mahony, 2015, p. 189).
Lucky Lady II (samples LL2-12-84-3 and LL12-12-
The Lucky Lady II site is located 46 km south of Dawson in the
Sulphur Creek tributary of the Indian River. The site has an
11.5 m exposure that Mahony (2015,pp.8295) divided into
two units; ve vertical cores were taken at the site for high-
resolution isotopic and radiocarbon analyses. The two core
samples utilized in this study come from the lowermost unit,
03.5 m, which is estimated to have been deposited from ca.
16,500 to 13,140 cal yrs BP (13,680± 35
UCIAMS-51324 to 11,290 ± 160
C yr BP, UCIAMS-56390)
(Mahony, 2015,p.85).Thisunitconsistsofgreysiltwitha
thick black organic-rich horizon (palaeosol) at 2.7 m dating
from 13,410 to 13,140 cal yrs BP (11,580 ± 35
UCIAMS-143308; 11,290 ± 160
C yr BP, UCIAMS-56390)
with several thinner palaeosol horizons above. The unit
includes in situ graminoid macrofossils and multiple Spermo-
philus parryii (arctic ground squirrel) nests; several Equus sp.
(horse) and Bison sp. (bison) bones were also identied. The
unit is suggestive of a steppe-tundra landscape. Twocore sam-
ples were selected from this site for experimental testing based
on work by Sadoway (2014). Core sample LLII-12-84-3 was
dated to 13,205 cal yr BP, while core sample LLII-12-217-8
was dated to 15,865 cal yr BP (Sadoway, 2014,p.29).
Optimizing extraction and targeted capture of ancient environmental DNA 7
Lab setting
Laboratory work was conducted in clean rooms at the McMas-
ter Ancient DNA Centre. These rooms are subdivided into ded-
icated facilities for sample preparation, stock solution setup,
and DNA extraction through library preparation. Post-indexing
and enrichment clean rooms are in a physically isolated facility,
while high-copy PCR workspaces are in separate building with
a one-way workow progressing from low-copy to high-copy
facilities. Each dedicated workspace is physically separated
with air pressure gradients between rooms to reduce exogenous
airborne contamination. Prior to all phases of laboratory work,
dead air hoods and workspaces were cleaned using a 6% solu-
tion of sodium hypochlorite (commercial bleach) followed by a
wash with Nanopure puried water (Barnstead) and 30 minutes
of UV irradiation at >100 mJ/cm
Metal sampling tools were cleaned with commercial bleach,
rinsed with Nanopure water immediately thereafter, and
heated overnight in an oven at 130°C. Once the tools had
cooled, work surfaces were cleaned with bleach and Nano-
pure water and covered with sterile lab-grade tin foil. Sedi-
ment cores previously split into disks (DCosta et al., 2011,
p. SI. 45; Sadoway, 2014, ch. 1) and stored at 20°C had
the upper 1 mm of external sediment chiselled off to create
a fresh sampling area free of exogenous contaminants for a
hollow cylindrical drill bit. The drill bit (diameter 0.5 cm)
was immersed in liquid nitrogen prior to sampling and a
drill press was used to repeatedlysubsample the disk sections
(see DCosta et al., 2011, Fig. S3). Sediment was pushed out
of the drill bit using a sterile nail and the bottom 12mmof
sediment from the bit was removed before dislodging the
remaining sample. This exterior core portion was carefully
removed as it has a higher chance of containing sedaDNA
from other stratigraphic contexts due to coring and core split-
ting. A bulk set of subsampled sediment from the same core
disk was homogenized by stirring in a 50 mL falcon tube and
stored at 20°C for subsequent extractions. This process was
repeated individually for each core sample (Fig. 2).
Physical disruption, chemical lysis, and extraction
Subsamples from the four cores were homogenized by core
and split into triplicates for each extraction method (24
extracts + 3 blanks), which were each used for shotgun,
enrichment, and trnL metabarcoding. DNeasy PowerSoil
DNA Extraction kit samples were extracted following manu-
facturer specications; puried DNA was eluted twice with
25 μL EBT buffer (10 mM Tris-Cl, 0.05% Tween-20). Sam-
ples extracted with our cold spin extraction method (a
sedaDNA modied version of the Dabney et al. [2013b]
extraction) were processed as follows (see Fig. 2):
Lysis (DNA release)
1) 500 μL of a digestion solution (see Table S1) initially
without proteinase K was added to PowerBead tubes
(already containing garnet beads and 750 μL 181 mM
NaPO4 and 121 mM guanidinium isothiocyanate from
the manufacturer).
2) 250 mg of homogenized sediment was added to each
PowerBead tube.
3) PowerBead tubes were vortexed at high speed for 15
minutes, then centrifuged briey to remove liquid from
the lids.
4) 15.63 μL of proteinase K (stock 20 mg/mL) was added to
each tube to reach a proteinase K concentration of 0.25
mg/mL in the digestion and PowerBead solution (a total
volume of 1.25 mL).
5) Tubes were nger vortexed to disrupt sediment and
beads that had pelleted in step 3.
6) PowerBead tubes were securely xed in a hybridization
oven set to 35°C and rotated overnight for 19 hours,
ensuring that the digestion solution, sediment, and
PowerBeads were moving with each oscillation.
7) PowerBead tubes were removed from the oven and cen-
trifuged at 10,000 x g for 5 minutes (the maximum speed
recommended for PowerBead tubes). Supernatant was
transferred to a MAXYMum Recovery 2 mL tube and
stored at 20°C.
Purication (DNA isolation)
8) The digestion supernatant was thawed, briey centri-
fuged, and added to 16.25 mL (13 volumes) of high-
volume Dabney binding buffer (see Table S2) in a 50
mL falcon tube and mixed by repeatedly inverting the
tube by hand.
9) Falcon tubes were spun at 4500 x g in a refrigerated cen-
trifuge set to 4°C for 20 hours overnight. (In subsequent
experiments, we have found that this speed can be
reduced to 2500 x g with no noticeable declines in inhib-
itor precipitation [data not shown].)
10) After centrifugation, falcon tubes were carefully
removed and the supernatant was decanted, taking
care to not disturb the darkly coloured pellet that had
formed at the base of the tube during the cold spin.
11) The binding buffer was passed through a high-volume
silica column (High Pure Extender Assembly, Roche
Diagnostics) and extraction proceeded as per Dabney
et al. (2013b).
12) Puried DNA was eluted off the silica columns with
two volumes of 25 μL EBT.
Prior to all subsequent experiments, both the cold spin and
PowerSoil extracts were centrifuged at 16,000 x g for 5 min-
utes to pellet remaining co-eluted inhibitors.
Library preparation
Double-stranded libraries were prepared for each extract as
described in Meyer and Kircher (2010) with modications
from Kircher et al. (2012) and a modied end-repair reaction
to account for the lack of uracil excision (Table S3). Samples
were puried after blunt-end repair with a QIAquick PCR
8T.J. Murchie et al.
Purication Kit (QIAGEN) to maximally retain small frag-
ments and after adapter ligation (Table S4) with a MinElute
PCR Purication Kit (QIAGEN), both using manufacturers
protocols. Library preparation master mix concentrations
can be found in Tables S3-S6.
qPCR: Inhibition spike tests, total quantication,
and indexing
Quantitative PCR (qPCR) is a technique that uses uorescent
dyes that intercalate with double-stranded DNA to monitor
PCR amplications in real-time, as opposed to at the end of
the reaction with standard PCR. When paired with DNA stan-
dards at known concentrations, qPCRcan be used to quantify
the starting concentrations of particular DNA molecules in a
sample (such as all DNA fragments derived from plants or
mammals) or to quantify the total number of molecules that
were successfully adapter-ligated during library preparation.
The technique can also be used to assess the inhibition load
of a sample as these inhibitory substances affect the rate of
PCR amplication (King et al., 2009). See supplementary
Appendix A (section SET-A. Inhibition Index and Fig. S6)
for details on the inhibition spike test. Total library DNA
quantications reported here and in the supplementary mate-
rials used the short amplication primer sites on the library
adapters to quantify total DNA copies per 1 μL with averaged
PCR duplicates prior to indexing PCR (Table S8).
For each of the fourcores and two extraction methodsthe sub-
sampled triplicate with the highest total DNA concentrations (as
based on the short amplication qPCR) was indexed for shotgun
sequencing (8 samples + 3 blanks). All subsampled triplicates
(24 samples + 3 blanks) were indexed separately thereafter for
targeted enrichment and sequencing. Metabarcoded samples
were processed identically, but with a trnLPCRamplication
prior to library preparation.
Targeted capture enrichment
The PalaeoChip Arctic-1.0 hybridization enrichment bait-set
was designed in collaboration with Arbor Biosciences to target
the mitogenomes of extinct and extant Quaternary animals
(focused primarily on megafauna; number of taxa 180),
and high-latitude plants based on curated reference databases
developed by Sønstebø et al. (2010), Soininen et al. (2015),
and Willerslev et al. (2014), initially targeting the trnL locus
(n 2100 taxa) in the chloroplast genome with the addition
of rbcLandmatK loci where available (see Appendix B in
the online supplementary materials for taxonomic list). Baits
were hybridized at 55°C for 24 hours. Further details on design
and wet-lab procedures can be found in the supplementary
online materials (Appendix A, section SET-E. Enrichment).
Post-indexing total quantication, pooling,
size-selection, and sequencing
Post-indexed libraries were quantied using the long-
amplication total library qPCR assay with averaged PCR
duplicates (Table S10). Thereafter, those libraries were
pooled to equimolar concentrations using the qPCR derived
molarity estimates. This is to equalize sequencing depth
between samples with wide ranges in molarity. Libraries
that were enriched, shotgun sequenced, or that had been
amplied with PCR metabarcoding were each pooled sepa-
rately. The three pools were size-selected with gel excision
following electrophoresis for molecules ranging from 150
bp to 600 bp. Gel plugs were puried using the QIAquick
Gel Extraction Kit (QIAGEN), according to manufacturers
protocols, then sequenced on an Illumina HiSeq 1500 with
a 2 x 90 bp paired-end protocol at the Farncombe Metage-
nomics Facility (McMaster University).
PCR metabarcoding
Sadoway (2014) previously worked with these and many
other Yukon permafrost core samples using a metabarcoding
approach. These libraries had been extracted in duplicate with
guanidinium protocols (Boom et al., 1990;DCosta et al.,
2011) from 250 mg of the same core sections, puried with
silica (Höss and Pääbo, 1993), and eluted twice (Handt
et al., 1996). Further details of Sadoways metabarcoding
approach can be found in the supplementary Appendix A,
section SET-E. Sadoway (2014) PCR metabarcoding.
As a follow-up to assess whether metabarcoding with
PowerSoil or our cold spin extractions would produce differ-
ent results than those from Sadoway, we used the same P6
loop trnL primers (Taberlet et al., 2007) to amplify all of
the extracts fromthe SET-E experiments (Table 1). Each sam-
ple extract was PCR amplied in triplicate (PCR conditions
and qPCR results are detailed in the supplementary Appendix
A, SET-E. PCR metabarcoding trnL), then puried with a
10K AcroPrep Pall plate using a vacuum manifold and pooled
into a single 50 μL sample extract in EBT. These trnL meta-
barcoded extracts were then library prepared, indexed, and
pooled at equimolar concentrations following the same afore-
mentioned procedures, but in a post-PCR facility. They were
sequenced at the Farncombe Metagenomics Facility on an
Illumina HiSeq 1500 with a 2 x 90 bp paired-end protocol
to an approximate depth of 500,000 reads.
Bioinformatic processing, MEGAN LCA, false
positives, and bubble-charts
Additional details on the bioinformatic workow can be
found in supplementary Appendix A (section SET-E. Bioin-
formatic workow). In brief, after trimming and merging
with leeHom (Renaud et al., 2014), the sequenced DNA
reads were mapped with network-aware-BWA (Li and Dur-
bin, 2009)(
to both the animal and plant baits as well as to the original
plant target references (targeting trnL, matK, rbcL) in order
to lter out off-target molecules, hereafter referred to as map-
ltering (see Fig. 2). Two map-ltering approaches were used
as we observed that while mapping to the baits is more
Optimizing extraction and targeted capture of ancient environmental DNA 9
conservative, it may unfairly bias against metabarcoding
amplicons as they may not map well to the articially tiled
(and highly curated) 80 bp probes (see section SET-E. Map-
ltered to plant reference sequences in supplementary
Appendix A for more details). Map-ltered reads were length
ltered to 24 bp (as smaller sequences generally have low
support when assessing taxonomy), string de-duplicated (to
remove molecules articially duplicated during PCR) with
the NGSXRemoveDuplicates module of NGSeXplore
(, and then
BLASTn (Altschul et al., 1990) aligned against a July 2018
local copy of the GenBank NCBI nucleotide database set to
return the top 100 alignments (taxonomic hits) per read.
Fasta and blast les (le types containing sequenced reads
[fasta] and the taxonomic alignments [blast]) were passed
to MEGAN (Huson et al., 2007;2016) where the BLAST
results were ltered through a lowest common ancestor
(LCA) algorithm (selected parameters detailed in supplemen-
tary Appendix A, section SET-E. Bioinformatic workow).
The MEGAN LCA algorithm determines the lowest taxo-
nomic rank at which a set of reads can be assigned based
on the assigned threshold of condence values. For example,
to call Bison bison as being present in this instance, at least 3
unique reads must align to a region of the B. bison mitoge-
nome with 95% identical similarity, with low e-values
and high bit scores (metrics used by BLASTn to assess the
likelihood of misalignments), and by matching or exceeding
other parameters used to dene condence in the taxon iden-
tication (the rational for select LCA parameters are dis-
cussed further in supplementary Appendix A SET-E.
Bioinformatic workow). Adjusting LCA parameters shifts
the trade-off ratio of false positive to false negative assign-
ments, although it would seem that optimal LCA parameters
may only exist on a project-by-project or even sample-by-
sample basis depending on the taxonomic molecular constit-
uents present, the degree of aDNA damage, and the research
question (see Huson et al., 2018). For example, if percent
identity is set to 100, only exact matches will be considered,
but then aDNA fragments with terminal base modications
(the majority of aDNA molecules) will be unassigned when
their taxonomic classication might otherwise be obvious.
Further, increasing the number of reads necessary to call a
taxon node as being present can articially bias the data to
organisms that release more eDNA irrespective of actual bio-
mass or had better organo-mineral complex preservation
characteristics. For example, woody plants at boreal sites
have been observed by Yoccoz et al. (2012, p. 3652) to
have proportionally less DNA presence relative to above-
ground biomass as compared with graminoids and forbs.
This would imply that DNA recovery cannot be easily corre-
lated with plant biomass across functional types without
some form of calibration, which would likely require exten-
sive region-specic experimentation. If minimum read counts
are set too high, we might expect that some rarer woody
plants would become undetectable while others would appear
to be in low abundance, and that forbs would bias towards an
Similarly, the LCA parameter top percent sets a percentage
threshold of hits to be used for taxonomic classication based
on the top bit-scores for a read as reported by BLAST. If this
value is set too high (> 50%), reads that might otherwise have
obviously best hits to a species or genus will instead be
assigned to higher (less taxonomically useful) ranks such as
family or order due to other BLAST hits for the same read
with lower bit-scores (less sequence similarity), thus inaccu-
rately inuencing read assignment. Setting this value too low
(< 5%), can likewise make taxonomic assignments inaccu-
rately specic by articially ignoring hits with marginally
lower bit-scores that are just as likely to represent the host
organism. These ambiguities in optimal LCA threshold values
are compounded by database incompleteness and the overrep-
resentation of certain organisms of economic or scientic
interest, the unknowns of palaeo-biogeography (Jackson,
2012), variably specic genetic loci (Kress and Erickson,
2007), erroneous GenBank references (Shen et al., 2013;Lu
and Salzberg, 2018), limitations of NCBI BLASTn and equiv-
alent algorithms (Shah et al., 2019), and other evolutionary
complexities that blur species boundaries such as introgressive
hybridization (Whitworth et al., 2007; Percy et al., 2014).
Alternative approaches, such as the phylogenetic intersec-
tion analysis (PIA) recently reported by Cribdon et al. (2020),
may be better suited to resolving some of these uncertainties
in determining optimal criteria for taxonomic identication.
These challenges are relatively typical of proxy measures of
palaeobiota (e.g. Jackson, 2012; Baker et al., 2013; Fiedel,
2018) and are in need of further research. In a rough sense
here, the more unique reads assigned to a taxon node (relative
to similar organisms), the higher the likelihood that those taxa
were indeed present. Further, relative shifts in read proportion
through time (such as the ratio of graminoids and forbs to
conifers and woody shrubs) can serve as a very rough esti-
mate of past relative abundance (although this must be inter-
preted carefully as many stochastic processes are involved in
eDNA release, degradation, preservation, and recovery). We
suggest that the LCA parameters utilized here balance the
challenges of false positive and false negative assignments
well for these particular samples, although more research is
needed to improve the criteria by which condence parame-
ters are selected and utilized algorithmically.
False positives
To alleviate observations of sporadic false positives in the
plant data for the genus level comparisons, all plant genera
were individually inspected and queried online to determine
if any extinct or extant species from those clades have been
observed in northern or alpine North America or northeastern
Eurasia. Clades assessed to be highly improbable (those with
known biogeographic extents limited to non-Holarctic
regions) were added to a running list of disabled taxa in
MEGAN that repeatedly were identied in this and other
10 T.J. Murchie et al.
parallel research (Table S18). We believe these false positives
are driven by an abundance of genetic research on specic
taxa within these unlikely clades and an absence of data for
whatever the real taxon is. Cribdon et al. (2020, pp. 24)
refers to these false positive hits as oasis taxa. In uneven data-
bases, sparsely populated genetic references within a clade
can succumb to oases where a set of well-studied organisms
drive an illusion of condence in genetic alignments because
those are the only goodalignments available for that read.
This makes the alignment seem condent and highly specic
because the wider range of organisms within those clades
have yet to be sequenced. This condence due to reference
oases makes it difcult to remove false positive hits via
stricter LCA parameters without also dropping a signicant
proportion of reads that have equally condent metrics, but
are instead driven by genetically well represented clades
that happen to make ecological sense. The MEGAN LCA
algorithm has been demonstrated to be robust to false posi-
tives (Huson et al., 2007;2018), but Cribdon et al. (2020)
argue that oasis taxa can remain a problem unless addressed
via manually removing so-called problematic taxa or using
an approach such as their PIA algorithm. We were unaware
of this software approach at the time of analysis, and instead
opted for a manual removal of suspected oasis taxa. A manual
exclusion approach may limit the possibility of identifying
biogeographically unknown and raretaxa that may other-
wise have evaded detection in Quaternary records to date,
but does allow for a more nuanced decision process that
might otherwise be obfuscated by a set of arbitrary cut-off
values. The oasis problem could also likely be alleviated
with a highly curated, non-redundant, and regionally specic
reference database, but this would further limit the ecological
reconstruction to only organisms one expects to nd. Oasis
reference taxa are likely to become less problematic as refer-
ence databases are improved over time, and as alignment
algorithms are better designed to cope with uneven database
coverage. Increasing the BLASTn top hits to 500 or more, as
suggested by Cribdon et al. (2020), is also something we have
found in parallel research aids in combating database uneven-
ness (data not reported here), but does often create very large
blast les.
To address the probable false positive assignments here,
for all genus-level plant bubble-chart comparisons (Figs. 6
7, S23S27, as well as Fig. 4), we used a manual exclusion
approach using a list of improbable taxa (oasis candidates)
observed in these metagenomic reconstructions as well as
other permafrost and lake sediments processed in parallel
(list available in supplementary Appendix A, Table S18).
Other metagenomic comparisons reported here (Figs. 6,
S20S22, S28S33) do not include these improbable taxa
as disabled in MEGAN. Probable oasis references appear to
have very low LCA-assigned read counts even when summed
to high taxonomic ranks (e.g., Zingiberales) when compared
with taxonomic nodes that make palaeoecological sense (see
Figure 5 for an example). As such, these highly improbable
taxa do not appear to constitute a sizable fraction of the organ-
isms identied in this analysis.
Libraries were compared using bubble-charts with logarith-
mically scaled bubbles for visually proportional normaliza-
tions, but with absolute read counts retained. For the
site-by-site charts reporting both animal and plant reads
(Figs. 6, S20S22, S28S31), the plants were collapsed to
higher taxonomic ranks to allow for summarized compari-
sons. For the plant map-ltered charts (Figs. 67, S23
S27), all data were collapsed to the genus rankmeaning
all LCA assignments to species or subspecies were pushed
up to the rank genus and summed with other species within
that clade. This is to mitigate species-specic resolution
problems, driven in part by database incompleteness,
where only a subset of species in a genus may have been
sequenced for a particular locus to date. This can be seen
for example with Sitka willow (Salix sitchensis), which has
yet to be sequenced (and uploaded to GenBank) for the
trnL locus, as compared to Arctic willow (Salix arctica),
which is represented. This would theoretically increase the
likelihood of species misassignment within that genus if
the actual organism from which those DNA fragments is
derived has yet to be sequenced or if that locus is not
species-specic for the cladecompounded again with
ancient samples by taphonomy, diachronic biogeographic
shifts, and molecular evolution.
Methodological comparison
qPCRs on the DNA libraries show an up to 7.0x increase in
total adapted DNA over PowerSoil extractions among the
four core samples (average 3.6x increase) with our cold
spin extraction procedure, and an up to 5.6x increase in
endogenoustrnL library adapted chloroplast DNA (average
3.0x increase) (Figs. 3 and S16). Inhibition indices for our
cold spin extractions were lower than PowerSoil (average=
0.75 versus 0.95 for PowerSoil, see section SET-A. Inhibition
Index in supplementary Appendix A for a description of the
inhibition index), meaning that more inhibitors were retained
with our cold spin extractions compared with PowerSoil.
However, these inhibitors did not dramatically reduce enzy-
matic efciency during adapter ligation; in post-library prep-
aration qPCR assays (Figs. 3 and S16) these cold spin
samples quantify with signicantly more library adapted
DNA than PowerSoil extracts, despite some inhibitor reten-
tion. When these samples were extracted following standard
Dabney et al. (2013b) procedures without the cold spin,
they were completely inhibited (see supplementary Appendix
A, SET-A).
LCA-assigned reads from samples extracted using our
sedaDNA modied Dabney protocol paired with PalaeoChip
enrichments show a 7.719.3x increase in LCA assigned
reads over enriched PowerSoil extracts, and a 1.259.9x
increase compared with Sadoways(2014) plant and animal
PCR metabarcoding approach (Table S21). An equivalent
Optimizing extraction and targeted capture of ancient environmental DNA 11
increase is observed when comparing the trnL metabarcoded
PowerSoil and cold spin extracts with those that were
extracted with the cold spin and enriched (2.323.0x and
2.919.5x increases, respectively). Compared with shotgun,
the increase in map-ltered DNA for the cold-spin enriched
samples is consistently three orders of magnitude, where
shotgun sequencing recovered almost no ecologically infor-
mative DNA from our plant mapping references for any of
the core slices at a depth of 2 million reads. However, several
nuclear and broader chloroplast loci were identied in the
non-mapped shotgun data; this is discussed further in supple-
mentary Appendix A, SET-E. Additional data. LCA-assigned
read count is not necessarily the most informative metric
when comparing these methods, however, as many reads in
all variants were assigned to high taxonomic ranks with lim-
ited interpretive utility. More important is the breadth of
organisms identied by being able to target multiple loci
simultaneously and capture fragments too small for
metabarcoding. Targeted enrichments, when paired with the
cold spin extractions, recover the same predominant taxa as
the two metabarcoding approaches, but with a far greater
diversity of identied organisms (Figs. 47, S20S27).
Taxa with sufciently high LCA-assigned read counts from
the enriched libraries also show characteristic aDNA deami-
nation patterns and fragment length distributions with map-
Damage (Jónsson et al., 2013)(Fig. 8). We also observe
that extraction replicates show minimal variation in terms of
taxonomic assignments between homogenized subsamples,
suggesting that the trends observed are a result primarily of
differences in method and not intrasample variation.
While qPCR amplications indicate much higher starting
quantities of DNA for samples extracted with the cold spin
method compared to PowerSoil (Fig. S17), the downstream
metagenomic comparisons show no noticeable differences
in taxonomic proles between the two extraction methods
for trnL metabarcoded libraries (Figs. 67,S22S26). In
Fig. 3 - Colour online, B/W in print
Figure 3. (color online) Total DNA quantication of library-adapted molecules comparing both extraction methods by core sample (see Table
S8 for qPCR specications). The large range for modied Dabney extraction on core MM12-118B is driven by a single extraction replicate
with a lower copy number. Core LLII 12-217-8 consistently has low DNA recovery but also a low co-elution of DNA-independent inhibition.
Core LLII 12-127-8 likely contains predominantly highly degraded sedaDNA compared with the other three samples (discussed further in
supplementary Appendix A, subsection SET-D).2014; Mahony, 2015). Values indicate total reads assigned to that taxon node.
12 T.J. Murchie et al.
Fig. 4 - Colour online, B/W in print
Figure 4. (color online) Metagenomic summary comparison of all four permafrost core samples that were extracted with the sedaDNA
modied Dabney (cold spin) method, capture-enriched with the PalaeoChip baits, and map-ltered to the target bait sequences. Metabarcoding
and PowerSoil libraries are not depicted. Subsampled replicates merged in MEGAN. Only select organisms depicted.
Optimizing extraction and targeted capture of ancient environmental DNA 13
multiple instances, the cold spin metabarcoded samples
recovered a somewhat taxonomically broader set of eDNA
across subsampled replicates, but the effect is marginal at
best. We suspect that the high cycle numbers during PCR
amplication compensated for the relative DNA loss of the
PowerSoil extracts (observable in Fig. S17), amplifying the
lower concentration sedaDNA of PowerSoil to the same rel-
ative taxonomic proportions. Further modications to the
Fig. 5 - Colour online, B/W in print
Figure 5. (color online) Metagenomic comparison of the Bear Creek core sample (BC 4-2B). Reads map-ltered to the baits and compared
with absolute counts and logarithmically scaled bubbles. Sample dated to 30,000 cal yr BP (DCosta et al., 2011; Sadoway, 2014; Mahony,
2015). Values indicate total reads assigned to that taxon node for Animalia, and a clade summation of reads for Viridiplantae. Note: hits to
Arecales, Bromeliaceae, Restionaceae, Zingiberales, and Diosoreales are likely false positives driven by uneven reference coverages within
Commelinids (see Methods subsection 10).
14 T.J. Murchie et al.
cold spin extraction method may aid a metabarcoding
approach. But at this time, only samples intended for enrich-
ment or shotgun sequencing would be best served with the
cold spin.
We observed a diverse range of molecularly identied ani-
mals and plants from well-preserved sedaDNA in these per-
mafrost samples (Fig. 4). These data correspond well with
palaeoecological understandings of environmental change
around the Pleistocene-to-Holocene transition (Dyke, 2005;
Zazula et al., 2005,2006), but with some notable exceptions.
Our cold spin extraction method paired with targeted
enrichment (using the PalaeoChip Arctic-1.0 bait-set) recov-
ered potential sedaDNA evidence for the late survival of
mammoth (Mammuthus sp.) and horse (Equus sp.) in the
Klondike region of Yukon, Canada, as well as an early indi-
cation of low abundance pine (Pinus sp.)(Fig. 4). The signif-
icance and means by which to interpret these data is discussed
further under Palaeoecology below.
Our cold spin inhibitor removal procedure, paired with Dab-
ney et al. (2013b) aDNA purications and capture enrich-
ment, consistently recovered a broader taxonomic set of
on-target environmental molecules than the PowerSoil
Fig. 6 - Colour online, B/W in print
Figure 6. (color online) Metagenomic comparison of Upper Goldbottom core MM12-118b with reads map-ltered to the plant references, part
1 of 2. Compared with absolute counts and logarithmically scaled bubbles. Core slice dated to 9685 cal yr BP (Sadoway, 2014; Mahony, 2015).
Values indicate total reads assigned to that taxon node.
Optimizing extraction and targeted capture of ancient environmental DNA 15
extraction kit paired with targeted enrichment, shotgun
sequencing (with either extraction method), an older PCR
metabarcoding approach, or a plant-specictrnL metabar-
coding approach (with either extraction method). Over 70%
of the taxon classications identied with cold spin extrac-
tions paired with the PalaeoChip enrichments would not
have been identied with alternative approaches (Figs. 47,
S20S27). These results demonstrate the viability of targeted
enrichment for taxonomically diverse environmental samples
from permafrost exposures without the necessity of PCR
metabarcoding and the associated compounding biases
therein. These data also demonstrate the signicantly
improved breadth and sequencing efciency of eDNA recov-
ery when using capture enrichment compared with a shotgun
approach. While some of the taxa identied are biogeograph-
ically implausible, most make ecological sense. The breadth
of taxa identied with a capture enrichment approach (and
high DNA recovery extraction method) aid in expanding
the ecological scope of environmental changes that can be
observed when not restricted to DNA fragments amenable
to a metabarcoding approach. Deep shotgun sequencing to
library exhaustion would be ideal as it is the least biased
approach. However, until data storage, computational
power, database completeness, and sequencing costs are sig-
nicantly improved, deep sequencing strategies are largely
unachievable for most users except for those with immense
computational and sequencing resources.
Overcoming enzymatic inhibitors
Our ongoing experiments with a diverse range of sediments
suggest that extracts with inhibition indices over 0.3 are
still amenable to library preparation, but with reduced adapter
ligation efciency (see section SET-D in supplementary
Appendix A for a discussion of extract qPCR inhibition). Sol-
ution C3 (120 mM aluminum ammonium sulfate dodecahy-
drate) in the PowerSoil kit is effective at precipitating
humic substances, resulting in clear, inhibitor-free extracts.
Fig. 7 - Colour online, B/W in print
Figure 7. (color online) Metagenomic comparison of Upper Goldbottom permafrost core MM12-118b with reads map-ltered to the plant
references, part 2 of 2. Compared with absolute counts and logarithmically scaled bubbles. Core slice dated to 9685 cal yr BP (Sadoway,
2014; Mahony, 2015). Values indicate total reads assigned to that taxon node.
16 T.J. Murchie et al.
However, aluminum sulfate has also been demonstrated to
readily precipitate DNA along with those humic substances
(Dong et al., 2006). We have found that the PowerSoil kit
is effective at removing DNA inhibition, but that this
approach often resulted in less adapted DNA. The kit
would seem to be most effective with modern sediments
and soils, where degradative processes are comparatively lim-
ited, but this inhibitor removal approach might be too aggres-
sive at precipitating the tightly bound organo-mineral
complexes in which rare fragments of sedaDNA are pre-
served. There seems to be an important balance between
releasing enough DNA but not releasing too many inhibitors
during the lysis stage, and also removing enough inhibition
for enzymatic reactions, while not removing the majority of
the endogenous sedaDNA during purication (Fig. 9). We
suspect that with further optimization, the cold spin extraction
procedure could be improved to better remove DNA indepen-
dent inhibitors (such as humic substances) that impact
enzymatic activity (Allison, 2006) in other sediment contexts.
Samples with inhibitor types unaffected by the cold spin
may still be best extracted with a modied PowerSoil protocol
or alternative approaches. Of course, the inhibitor constituent
(as well as distribution of preserved aDNA) of a site or sample
likely varies considerably even at the molecular scale, making
it difcult to estimate the best approach at the outset of an
investigation without extensive site-specic experimentation.
Of interest for further research is how inhibitor precipita-
tion is affected by the interaction between the lysing detergent
sodium dodecyl sulfate (SDS) and the cold spin for (see sec-
tion SET-D. SDS and sarkosyl in supplementary Appendix
A). We suspect that the efciency of inhibitor precipitation
with the long cold spin method could be further optimized,
as our experiments suggest that the presence of SDS in the
lysis buffer signicantly contributes to the precipitation of
humics and other inhibitors, compared with extracts where
sarkosyl was used as the lysing detergent. For the permafrost
samples in this study, we have found that the cold spin tech-
nique is effective at removing enough inhibition for effective
library preparation while maximally retaining sedaDNA.
However, other samples (not in this study) with especially
high concentrations of humics and other unidentied inhibi-
tors retain much of their inhibition despite the cold spin (with
the cold spin producing thick pellets of inhibitor precipitates
[1 ccm], such as illustrated in Fig. 2), necessitating further
sample-specic modications. We have observed generally
(data not reported here) that lowering the sediment input
(<0.2 g), splitting the subsample into multiple low-input lys-
ing tubes and pooling after the cold spin over a Roche col-
umn, as well as increasing the duration of the cold spin,
helps further reduce the inhibitor load of especially challeng-
ing samples. Our cold spin technique is unlikely to be optimal
for all forms of sedaDNA inhibition however, as it has been
observed that identifying the specic inhibitory substances
involved is critical to mitigating the compound-specic
mechanisms that affect enzymatic reactions (Opel et al.,
2010). Further research is needed to identify the inhibitor
Fig. 8 - Colour online, B/W in print
Figure 8. (color online) Example mapDamage plots showing aDNA characteristic terminal deamination patterns and short fragment length
distributions (FLD) (length lter 24 bp, mapping quality lter 30). We suspect that the bimodal distributions in some of the plant FLDs is
due to non-specic mapping of closely related taxa. Taxa chosen for mapping are not necessarily accurate species of the ancient molecules,
such as in this case Equus caballus, but rather a closely related organism with an available NCBI-RefSeq organelle genome.
Optimizing extraction and targeted capture of ancient environmental DNA 17
constituents of sedaDNA target samples in order to improve
the inhibitor precipitation we observed while maximizing
sedaDNA retention.
SedaDNA authenticity
DNA damage is often used as a proxy indicator for assessing
whether sets of molecules are ancient or the result of modern
contamination (Mitchell et al., 2005; Gilbert et al., 2007;
Ginolhac et al., 2011; Sawyer et al., 2012; Dabney et al.,
2013a; Jónsson et al., 2013). Damage proles for taxa with
sufciently high read counts (200 reads at minimum map
quality 30) show characteristic aDNA deamination patterns
and short fragment length distributions (Fig. 8). When map-
ping to the mitogenome, taxa with 200 reads typically show
an insufcient overlap of fragments to identify modications
of the terminal bases. This precludes condent assessment of
damage patterns and makes it difcult to authenticate rare
taxa with low read counts in this dataset. Comparisons of
the map-ltered with non-mapped libraries (in supplementary
Appendix A section SET-E. Additional data) suggests that
our quality ltering steps are sufciently conservative to
reduce the noise characteristic of metagenomic datasets (Lu
and Salzberg, 2018; Eisenhofer et al., 2019), but may also
strip out some potentially informative (but less condently
assigned) reads. Our pre-BLASTn map-ltering approach
allows for a more conservative and streamlined analysis.
Using a more regionally curated BLAST database might alle-
viate some of the potential for false positives due to ecolog-
ically nonsensical oasis taxa as discussed in Methods,
subsection 10. However, this will also potentially come at
the cost of alternative false positives and negatives due to
biases in database representation that result from limitations
in our proxy-driven understandings of palaeoecosystems
(Jackson, 2012). This theoretically could drive taxonomic
assignments towards organisms one is comfortable identify-
ing, because the actual unexpected organism is not included
as a possibility for taxonomic assignment, driving a false
sense of condence in the specicity of those reads. An
approach incorporating elements from the PIA algorithm
(Cribdon et al., 2020) with the interface and exibility of
the MEGAN software (Huson et al., 2007,2016; Huson and
Mitra, 2011) would likely be a good trajectory for moving
forward with an eDNA capture enrichment or shotgun
based approach.
Blank controls
Blank extraction, library, indexing, and metabarcoding reac-
tions (used to identify buffer and cross-contaminants) in the
shotgun and enriched libraries do not contain any map-
ltered reads (Table S23). The non-mapped LCA-assigned
reads for these blanks are predominantly adapter-
contaminated sequences (95%) (Fig. S32). The map-
ltered metabarcoding blanks only contain reads that could
be LCA assigned to high taxonomic ranks (Fig. S33). This
would suggest that patterns of library sequence composition
observed in the sample replicates are at minimum originating
from the samples themselves and are not the result of labora-
tory contamination.
This study is intended as a proof of concept to demonstrate
the viability of targeted enrichment for the recovery of eco-
logically complex, molecular taxonomic proxies from
open-air eDNA samples. Further research will utilize these
methods, and complementary palaeoecological techniques,
on Yukon lake sediments and permafrost cores from the
Klondike area to track ecological shifts during the
Pleistocene-Holocene transition. However, it is worth briey
contextualizing these broad taxonomic trends here for authen-
ticity purposes (as summarized in Fig. 4).
The Bear Creek (BC 4-2B, 30,000 cal yr BP [DCosta
et al., 2011]) and older Lucky Lady II samples (LLII
12-217-8, 15,865 cal yr BP [Mahony, 2015]) both date to a
period in which eastern Beringia is thought to have been
largely a herb tundra biome, dominated by exposed mineral
surfaces, prostrate willows, grasses, forbs (non-graminoid
herbs), and occupied by diverse and abundant megafauna
(Dyke, 2005; Zazula et al., 2005,2006). Our data reect
this environmental setting, particularly in the case of Bear
Creek (Figs. 45, S26S27) (DCosta et al., 2011). We iden-
tied a similar range of mammalian species as DCosta et al.
(2011) did using the same core sample, but with additional
taxa (e.g., caribou [Rangifer tarandus]). More specic taxo-
nomic assignments, especially if novel or possibly controver-
sial, need to be treated with caution. Thus, while DCosta
et al. (2011) identied Bos sp., we recovered a more specic
signal for yak (with hits to both B. grunniens and B. mutus
[domestic and wild yak, respectively]. Web-BLASTing
these yak-specic reads to the top 5000 hits (rather than top
Fig. 9 - Colour online, B/W in print
Figure 9. (color online) Conceptual representation of the balance
needed to overcome sedaDNA inhibition.
18 T.J. Murchie et al.
100) drops the LCA-assignment to Bos sp., which is consis-
tent with the results reported by DCosta et al. (2011). A
follow-up investigation with deeper sequencing is intended
to assess the possibility of B. mutus (wild yak) in eastern
Beringia during the Pleistocene.
Results from the younger Lucky Lady II sample (LLII
12-84-3, 13,205 cal yr BP [Sadoway, 2014]) indicate an
expansion of birch shrub tundra (Dyke, 2005), reected by
a decrease in grasses and a proportional increase in birch
(Betula sp.) and willow (Salix sp.)(Figs. 4, S23S24) relative
to the earlier two core samples. The youngest core sample
(MM12-118b, 9685 cal yr BP; Mahony, 2015) shows a pro-
portional increase in conifers, particularly spruce (Picea sp.)
but also potentially pine (Pinus sp.), that is consistent with
pollen records in southern Yukon for the expansion of forests
(Gajewski et al., 2014) and the establishment of the northern
boreal forest by 9000 cal yr BP (Dyke, 2005). Assignments
to Pinus sp. are unexpected as previous research has found
that lodgepole pine (Pinus contorta var. latifolia) had a north-
ern extent of 60°N (the Yukon-B.C. border) during the early
Holocene, and only reached its present-day northern limit
(63°N) by 1790 cal yr BP (Strong and Hills, 2013).
When the relevant set of pine positive samples is subjected
to more stringent bioinformatic testing (see supplementary
Appendix A, SET-E. Stringent LCA ltering for unexpected
taxa), these reads unambiguously retain their assignments
to the genus Pinus. The stratigraphic reliability of these
sedaDNA molecules in permafrost might be questionable,
however (see below), but if their age is accurate, it would sug-
gest a low-density northern expansion of pine beyond south-
ern Yukon 3500 years earlier than pollen and stomata
records have yet indicated (Schweger et al., 2011; Strong
and Hills, 2013; Edwards et al., 2015).
The mammalian sedaDNA constituents of these permafrost
samples also display a marked change, dwindling in relative
abundance and richness with time into the Holocene (Figs. 4
5, S2022), but perhaps less sharply than commonly thought.
For example, we recovered genetic evidence of both woolly
mammoth (Mammuthus primigenius) and horse (Equus sp.)
in the Upper Goldbottom core dated to 9700 cal yr BP
(Mahony, 2015). Previous radiocarbon dates on fossils indi-
cate that horses disappeared from high-latitude northwestern
North America relatively early, ca. 13,125 cal yr BP (last
appearance date11,500
C BP, based on AMNH
134BX36 from Upper Cleary Creek [Guthrie, 2003]). This
3400-year difference implies the existence of a substantial
ghost range (i.e., a spatiotemporal range extending beyond
the last appearance age, as indicated by directly dated fossils
or other associated remains; Haile et al., 2009). While this
nd cannot be corroborated by the macrofossil record for
Equus, it is consistent with previous sedaDNA results from
central Alaska (Haile et al., 2009).
In the absence of additional information, it is difcult to
assess whether this small signal (Fig. 4) may be considered
chronostratigraphically reliable or whether it has been
affected by factors such as leaching, cryo- or bioturbation,
or reworking (redeposition) (Arnold et al., 2011), altering
the relative positions of sedaDNA complexes. In the case
of the mammoth reads, after merging the sequenced data
from the three Upper Goldbottom core (MM12-118b) sub-
sampled replicates, coverage was insufcient (low read
counts mapping across the mitogenome) to reliably assess
characteristic aDNA damage patterns (Fig. S38). There is
arguably some indication of terminal damage with the
merged mapDamage and particularly short fragment length
distribution (FLD) proles, but greater sequencing depth is
needed to assess their authenticity. These LCA-assigned
mammoth (n = 41) and horse (n = 10) reads from the Upper
Goldbottom core (9700 cal yr BP, Fig. 4) were extracted,
concatenated, queried with the web-based BLASTn, and sub-
jected to stricter LCA parameters (see supplementary Appen-
dix A SET-E. Stringent LCA ltering for unexpected taxa).
Despite stricter ltering, three reads were LCA-assigned to
M. primigenius, 25 to Mammuthus sp., and 11 were identied
as Elephantidae. Equus sp. retained ve assigned reads. A
follow-up analysis with deeper sequencing of this sample
will further address the veracity of this signal.
Limitations of comparison
There are several caveats to keep in mind when assessing our
comparison of protocols and the potential of the PalaeoChip
Arctic-1.0 bait-set. First, the lysis stage of our PowerSoil and
of the sedaDNA modied Dabney protocols were not equiv-
alent in duration or reagents. We followed manufacturer spec-
ications for PowerSoil, but the lysis stage of extraction with
equivalent kits can be increased in duration and augmented
with additional reagents to theoretically increase DNA yield
(Niemeyer et al., 2017). Further, a recently released update
to the PowerSoil kit, the DNeasy PowerSoil Pro, claims to
have an 8-fold increase in DNA yield compared with compar-
ative commercial kits (it is unclear what the n-fold increase
over standard PowerSoil is with this updated kit). Our exper-
iments with the PowerSoil inhibitor removal solution C3
found consistently low DNA retention compared with our
longer duration 4°C spin as an inhibitor removal technique
(SET-B in supplementary Appendix A). The PowerSoil inhib-
itor removal solution is effective at rapidly precipitating enzy-
matic inhibitors, but this study suggests that it is often overly
aggressive and consistently precipitated viable sedaDNA in
the process (Fig. S7). We suspect that a longer lysis stage
with PowerSoil would increase overall yields, but would
not mitigate the substantial losses associated with overly
aggressive humic precipitation when utilizing solution C3
(at least at manufacturer recommended concentrations). We
found that the cold spin is sufciently effective at removing
enzymatic inhibition with these permafrost samples to allow
for successful adapter ligation, even if the extracts were not
as inhibitor free as PowerSoil (Fig. 3). However, we have
also found that samples from bogs or sites with high organic
loads remained highly inhibited despite the cold spin, likely
due to the high humic concentrations (among other forms
of inhibition). For difcult samples such as these, further ne-
tuning is needed to improve inhibitor removal.
Optimizing extraction and targeted capture of ancient environmental DNA 19
Second, metabarcoding is not directly equivalent to enrich-
ment when comparing taxonomic coverage and LCA-
assigned read counts. Mapping our data back to the baits
does strip out taxonomically informative hitspotentially
to a greater degree than with metabarcoding data that might
not map well to the curated bait sequences. To mitigate
this, we mapped the second set of plant specictrnL compar-
isons (Figs. 67, S2226) to the plant references rather than
the baits sequences to increase the metabarcoding reads avail-
able for BLAST. We observe that mapping to the curated baits
(which have low complexity and non-diagnostic regions
masked or removed) substantially reduces the number of
low condence (potential false positive) spurious hits but
does result in data loss (see Figs. S1819). Map-ltering to
the plant references alleviates this to some degree but in the
future this strategy might be better paired with a regionally
curated reference database or PIA approach (Cribdon et al.,
2020) as discussed earlier.
Finally, it should be emphasized that the PCR metabarcod-
ing data re-analyzed from Sadoway (2014) targeting multiple
plant and animal loci were from samples not puried with our
cold spin sedaDNA optimized extractions. The libraries that
were extracted with PowerSoil and our sedaDNA modied
Dabney procedure were only PCR amplied for trnL (rather
than the suite of loci initially assessed by Sadoway [2014]).
We observe that PowerSoil and cold spin extracted metabar-
coding samples generally outperform their counterparts pro-
cessed by Sadoway with the Boom et al. (1990) and
DCosta et al. (2011) extractions in terms of taxonomic
breadth, but that these three metabarcoding approaches do
generally identify the same predominant taxa. We suspect
that either of these newer extraction methods would have
resulted in a wider breadth of plant and animal taxa identied
for the Sadoway metabarcoded libraries. This limitation of
our comparison should moderate conclusions drawn from
this work.
The key observation in this study is that enrichment clearly
outperforms alternative targeting strategies in this dataset,
including the PowerSoil or cold spin extracted trnL metabar-
coded libraries. We suspect this is driven in large part by the
smaller fragment lengths available to an enrichment approach
compared with metabarcoding. For example, in the Upper
Goldbottom core (MM12-118b), LCA-assigned hits to
Betula sp. (Figs. 4 and 7) in the enriched libraries have a
FLD mode of 49 bp (n = 5397), whereas hits to Betula sp.
from the same extracts but with metabarcoded libraries have
a FLD mode of 98 bp (n = 500). In this case, the metabarcod-
ing libraries were restricted to targeting much longer frag-
ments that are comparatively rare with sedaDNA. This bias
towards large fragments with metabarcoding is likely one
of the main reasons why many rarer taxa are missed with
this approach. Low biomass organisms would have an
expectedly lower proportion of eDNA, with taphonomic pro-
cesses making long fragments (those needed for metabarcod-
ing) even more unlikely to survive. As an example where this
might be driving taxonomic false negatives, Lupinus sp.
(lupine) was identied in the Upper Goldbottom core with
a combined count of 353 unique reads in the cold spin
enriched libraries, but was absent from all of the metabarcod-
ing libraries. Lupinus sp. enriched libraries have a FLD mode
of 41 bp, with 79% of the reads being shorter than 69 bp. The
trnL metabarcode for L. arcticus (Arctic lupine) is 52 bp, but
with primer landing sites this fragment increases in length to
91 bp. With such a large fragment necessary for metabarcod-
ing detection, it is not surprising that this taxon and other low
biomass organisms were not detected with metabarcoding
when aDNA fragments of this length are exceedingly rare.
In the same way, 73% of Betula sp. reads in the enriched sam-
ple are shorter than 69 bp, and only 12% are 80110 bp,
whereas 83% of the metabarcoding reads for Betula sp. fall
in the larger fragment range. Figure S39 in supplementary
Appendix A depicts an FLD histogram of this example.
The goal of this analysis was to establish the viability of
enrichment for complex sedaDNA contexts, and to report
on a new inhibitor removal technique that may yet be further
optimized. Despite the limitations of the comparisons dis-
cussed above, the data clearly demonstrate the power of tar-
geted enrichment for eDNA, and we intend to further
expand on the PalaeoChip Arctic-1.0 bait-set with additional
target sequences for regionally specic vegetation, mammals,
insects, fungi, and microbiota. We also intend to optimize
PalaeoChip for other non-arctic/subarctic regions.
The experiments outlined in this report demonstrate the utility
of our cold spin inhibitor removal technique, paired with
Dabney et al. (2013b) purications, for overcoming enzy-
matic inhibitors in sedimentary materials. This technique uti-
lizes the high aDNA recovery potential of Dabney
purications, while also addressing the unique challenges
of sediments and soils where many inhibitory substances
tend to co-elute with target sedaDNA molecules. Other
extraction approaches, such as the PowerSoil kit, struggle
with signicant aDNA loss that ultimately limits the targeting
options for maximally exploiting the genetic archives pre-
served in sedimentary contexts. The improved DNA retention
of the cold spin inhibitor removal technique described here
facilities the signicantly expanded targeting scope of a cap-
ture enrichment approach for sedaDNA, in this case using the
PalaeoChip Arctic-1.0 bait-set. With this approach, we were
able to capture a highly complex set of plant and animal
DNA from permafrost sediments that outperformed an alter-
native extraction strategy, shotgun sequencing, and two ver-
sions of a PCR metabarcoding approach. Many of the
organisms identied here with our cold spin and enrichment
strategy were entirely missed with alternative methods,
including the potential late survival of woolly mammoth
(Mammuthus primigenius) and horse (Equus sp.) in the Klon-
dike of Yukon Canada, and the early Holocene appearance of
pine (Pinus sp.). Further work is needed to rene the potential
for false positives and negatives in metagenomic datasets due
to a variety of factors, but most notably database incomplete-
ness and unevenness. Recent work reported by Cribdon et al.
20 T.J. Murchie et al.
(2020) may serve as a viable next step towards further
improving molecular taxonomic identications to make full
use of the eDNA archives of palaeobiota being rapidly
unlocked by new methods.
An enrichment approach for eDNA avoids the myriad lim-
itations of a PCR metabarcoding strategy, and opens many
new possibilities for further study, such as whole genome
capture and assembly, as well as phylogenetic placement
without any surviving macroremains. By increasing the taxo-
nomic breadth of our environmental baits, and further opti-
mizing enrichment and sedaDNA extraction conditions, this
technique can continue to improve the sequenced fraction
of on-target molecules without deep shotgun sequencing, or
potentially biased PCR amplications. This technique
enables the recovery of a more holistic set of palaeoenviron-
mental DNA of widely varying molecular fragment lengths
from a diverse range of genetic loci. This expanded set of cap-
tured DNA targets allows for the simultaneous molecular
identication of organisms that might not have the biomass
to be readily detected with other palaeoecological methods,
and even in the complete absence of surviving tissues or
microfossils. PCR metabarcoding is likely to remain a viable
and important eDNA workhorse for the foreseeable future. As
the costs of metagenomic analysis continue to decrease, and
especially in situations where DNA preservation is favorable
and a wide set of targets are of interest, an enrichment
approach as shown here has the potential to recover a far
greater diversity of molecular taxonomic identiers to better
complement traditional palaeoenvironmental approaches.
Our thanks to Alison Devault at Arbor Biosciences for her invalu-
able assistance with designing the bait-set, as well as Brian Golding
and members of his bioinformatics research team at McMaster Uni-
versity for their computational resources and assistance, and all
members of the McMaster Ancient DNA Centre. We also wish to
thank the editors at Quaternary Research and peer-reviewers who
provided detailed and carefully considered critiques, which signi-
cantly improved the quality of this report. We wish to thank the Arc-
tic Institute of North America, the Gareld Weston Foundation, the
Natural Sciences and Engineering Research Council of Canada,
McMaster University and the Department of Anthropology, Polar
Knowledge Canada (POLAR), and the Social Sciences and Human-
ities Research Council of Canada for each funding various compo-
nents of this research.
The supplementary material for this article can be found at https://
Agarwala, R., Barrett, T., Beck, J., Benson, D.A., Bollin, C., Bolton,
E., Bourexis, D., et al., 2016. Database resources of the National
Center for Biotechnology Information. Nucleic Acids Research
44, D7D19.
Ahmed, E., Parducci, L., Unneberg, P., Ågren, R., Schenk, F., Rat-
tray, J.E., Han, L., et al., 2018. Archaeal community changes in
lateglacial lake sediments: evidence from ancient DNA. Quater-
nary Science Reviews 181, 1929.
Allison, S.D., 2006. Soil minerals and humic acids alter enzyme
stability: implications for ecosystem processes. Biogeochemistry
81, 361373.
Alsos, I.G., Sjögren, P., Edwards, M.E., Landvik, J.Y., Gielly, L., For-
wick, M., Coissac, E., et al., 2016. Sedimentary ancient DNA from
Lake Skartjørna, Svalbard: assessing the resilience of arctic ora to
Holocene climate change. The Holocene 26, 627642.
Altschul, S.F., Gish, W., Miller, W., Myers, E.W., Lipman, D.J.,
1990. Basic local alignment search tool. Journal of Molecular
Biology 215, 403410.
Alvarez, A.J., Khanna, M., Toranzos, G.A., Stotzky, G., 1998.
Amplication of DNA bound on clay minerals. Molecular Ecol-
ogy 7, 775778.
Anderson-Carpenter, L.L., McLachlan, J.S., Jackson, S.T., Kuch,
M., Lumibao, C.Y., Poinar, H.N., 2011. Ancient DNA from
lake sediments: bridging the gap between paleoecology and
genetics. BMC Evolutionary Biology 11, 115.
Arnold, L.J., Roberts, R.G., Macphee, R.D.E., Haile, J.S., Brock, F.,
Möller, P., Froese, D.G., et al., 2011. Paper II - Dirt, dates and
DNA: OSL and radiocarbon chronologies of perennially frozen
sediments in Siberia, and their implications for sedimentary
ancient DNA studies. Boreas 40, 417445.
Baker, A.G., Bhagwat, S.A., Willis, K.J., 2013. Do dung fungal
spores make a good proxy for past distribution of large herbi-
vores? Quaternary Science Reviews 62, 2131.
Bellemain, E., Carlsen, T., Brochmann, C., Coissac, E., Taberlet, P.,
Kauserud, H., 2010. ITS as an environmental DNA barcode for
fungi: an in silico approach reveals potential PCR biases. BMC
Microbiology 10, 189.
Bellemain, E., Davey, M.L., Kauserud, H., Epp, L.S., Boessenkool,
S., Coissac, E., Geml, J., et al., 2013. Fungal palaeodiversity
revealed using high-throughput metabarcoding of ancient DNA
from arctic permafrost. Environmental Microbiology 15, 117689.
Benson, D.A., Cavanaugh, M., Clark, K., Karsch-Mizrachi, I., Lip-
man, D.J., Ostell, J., Sayers, E.W., 2013. GenBank. Nucleic
Acids Research 41.
Bezanilla, M., Manne, S., Laney, D.E., Lyubchenko, Y.L., Hansma,
H.G., 1995. Adsorption of DNA to mica, silylated mica, and
minerals: characterization by atomic force microscopy. Langmuir
11, 655659.
Blum, S.A.E., Lorenz, M.G., Wackernagel, W., 1997. Mechanism of
retarded DNA degradation and prokaryotic origin of DNasesin non-
sterile soils. Systematic and Applied Microbiology 20, 513521.
Boom, R., Sol, C.J., Salimans, M.M., Jansen, C.L., Wertheim-Van
Dillen, P.M., van der Noordaa, J., 1990. Rapid and simple
method for purication of nucleic acids. Journal of Clinical.Mi-
crobiology 28, 495503.
Boyer, F., Mercier, C., Bonin, A., Bras, Y. Le, Taberlet, P., Coissac,
E., 2016. Obitools: a unix-inspired software package for DNA
metabarcoding. Molecular Ecology Resources 16, 176182.
Cai, P., Huang, Q., Zhang, X., Chen, H., 2006. Adsorption of DNA
on clay minerals and various colloidal particles from an alsol.
Soil Biology and Biochemistry 38, 471476.
Carpenter, M.L., Buenrostro, J.D., Valdiosera, C., Schroeder, H.,
Allentoft, M.E., Sikora, M., Rasmussen, et al., 2013. Pulling
out the 1%: whole-genome capture for the targeted enrichment
of ancient DNA sequencing libraries. American Journal of
Human Genetics 93, 852864.
Optimizing extraction and targeted capture of ancient environmental DNA 21
Clark, P.U., 2009. The last glacial maximum. Science 325, 710714.
Clark, P.U., Mix, A.. C., 2002. Ice sheets and sea level of the last
glacial maximum. Quaternary Science Reviews 21, 17.
Cleaves, H.J., Crapster-Pregont, E., Jonsson, C.M., Jonsson, C.L.,
Sverjensky, D.A., Hazen, R.A., 2011. The adsorption of short
single-stranded DNA oligomers to mineral surfaces. Chemo-
sphere 83, 15601567.
Crecchio, C., Stotzky, G., 1998. Binding of DNA on humic acids:
effect of transformation of bacillus subtilis and resistance to
DNase. Soil Biology and Biochemistry 30, 10611067.
Cribdon, B., Ware, R., Smith, O., Gaffney, V., Allaby, R.G., 2020.
PIA: more accurate taxonomic assignment of metagenomic data
demonstrated on sedaDNA from the NorthSea. Frontiers in Ecol-
ogy and Evolution 8, 112.
Dabney, J., Knapp, M., Glocke, I., Gansauge, M.-T., Weihmann, A.,
Nickel, B., Valdiosera, C., et al.., 2013b. Complete mitochon-
drial genome sequence of a middle Pleistocene cave bear recon-
structed from ultrashort DNA fragments. Proceedings of the
National Academy of Sciences 110, 1575815763.
Dabney, J., Meyer, M., Pääbo, S., 2013a. Ancient DNA damage.
Cold Spring Harbor Perspectives in Biology 5. doi:10.1101/
DCosta, V.M., King, C.E., Kalan, L., Morar, M., Sung, W.W.L.,
Schwarz, C., Froese, D., et al., 2011. Antibiotic resistance is
ancient. Nature 477, 457461.
Dong, D., Yan, A., Liu, H., Zhang, X., Xu, Y., 2006. Removal of
humic substances from soil DNA using aluminium sulfate. Jour-
nal of Microbiological Methods 66, 217222.
Dyke, A.S., 2004. An outline of the deglaciation of North America
with emphasis on central and northern Canada. Developments in
Quaternary Sciences 2, 373424.
Dyke, A.S., 2005. Late quaternary vegetation history of northern
North America based on pollen, macrofossil, and faunal remains.
Géographie physique et Quaternaire 59, 211.
Edwards, M.E., Alsos, I.G., Yoccoz, N., Coissac, E., Goslar, T.,
Gielly, L., Haile, J., et al., 2018. Metabarcoding of modern soil
DNA gives a highly local vegetation signal in Svalbard tundra.
Holocene 28, 20062016.
Edwards, M., Franklin-Smith, L., Clarke, C., Baker, J., Hill, S.,
Gallagher, K., 2015. The role of re in the mid-Holocene
arrival and expansion of lodgepole pine (Pinus contorta var.
latifolia Engelm. ex S. Watson) in Yukon, Canada. Holocene
25, 6478.
Eisenhofer, R., Minich, J.J., Marotz, C., Cooper, A., Knight, R.,
Weyrich, L.S., 2019. Contamination in low microbial biomass
microbiome studies: issues and recommendations. Trends in
Microbiology 27, 105117.
Epp, L.S., Gussarova, G., Boessenkool, S., Olsen, J., Haile, J.,
Schrøder-Nielsen, A., Ludikova, A., et al., 2015. Lake sediment
multi-taxon DNA from north Greenland records early post-glacial
appearance of vascular plants and accurately tracks environmental
changes. Quaternary Science Reviews 117, 152163.
Fiedel, S.J., 2018. The spore conundrum: does a dung fungus
decline signal humansarrival in the eastern United States? Qua-
ternary International 466, 247255.
Froese, D.G., Zazula, G.D., Reyes, A. V., 2006. Seasonality of the
late Pleistocene Dawson tephra and exceptional preservation of a
buried riparian surface in central Yukon Territory, Canada. Qua-
ternary Science Reviews 25, 15421551.
Gajewski, K., Bunbury, J., Vetter, M., Kroeker, N., Khan, A.H.,
2014. Paleoenvironmental studies in southwestern Yukon. Arctic
67, 5870.
Gardner, C.M., Gunsch, C.K., 2017. Adsorption capacity of multi-
ple DNA sources to clay minerals and environmental soil matrices
less than previously estimated. Chemosphere 175, 4551.
Garland, S., Baker, A., Phillott, A.D., Skerratt, L.F., 2010. BSA
reduces inhibition in a TaqMan® assay for the detection of Batra-
chochytrium dendrobatidis. Diseases of Aquatic Organisms 92,
Giguet-Covex, C., Pansu, J., Arnaud, F., Rey, P.-J., Griggo, C.,
Gielly, L., Domaizon, I., et al., 2014. Long livestock farming his-
tory and human landscape shaping revealed by lake sediment
DNA. Nature Communications 5, 3211.
Gilbert, M.T.P., Binladen, J., Miller, W., Wiuf, C., Willerslev, E.,
Poinar, H., Carlson, J.E., Leebens-Mack, J.H., Schuster, S.C.,
2007. Recharacterization of ancient DNA miscoding lesions:
insights in the era of sequencing-by-synthesis. Nucleic Acids
Research 35, 110.
Ginolhac, A., Rasmussen, M., Gilbert, M.T.P., Willerslev, E.,
Orlando, L., 2011. mapDamage: testing for damage patterns in
ancient DNA sequences. Bioinformatics 27, 21532155.
Goring, C.A.I., Bartholomew, W.V., 1952. Adsorption of mononu-
cleotides, nucleic acids, and nucleoproteins by clays. Soil Science,
74, 149164.
Graham, R.W., Belmecheri, S., Choy, K., Culleton, B.J., Davies,
L.J., Froese, D., Heintzman, P.D., et al.., 2016. Timing and
causes of mid-Holocene mammoth extinction on St. Paul Island,
Alaska. Proceedings of the National Academy of Sciences 113,
Greaves, M.P., Wilson, M.J., 1969. The adsorption of nucleic acids
by montmorillonite. Soil Biology and Biochemistry 1, 317323.
Greaves, M.P., Wilson, M.J., 1970. The degradation of nucleic
acids and montmorillonite-nucleic-acid complexes by soil micro-
organisms. Soil Biology and Biochemistry 2, 257268.
Guthrie, R.D., 2003. Rapid body size decline in Alaskan Pleisto-
cene horses before extinction. Nature 426, 169171.
Haile, J., Froese, D.G., MacPhee, R.D.E., Roberts, R.G., Arnold,
L.J., Reyes, A. V., Rasmussen, M., et al.., 2009. Ancient DNA
reveals late survival of mammoth and horse in interior Alaska.
Proceedings of the National Academy of Sciences of the United
States of America 106, 2235222357.
Handt, O., Krings, M., Ward, R.H., Pääbo, S., 1996. The retrieval of
ancient human DNA sequences. American Journal of Human
Genetics 59, 36876.
Höss, M., Pääbo, S., 1993. DNA extraction from Pleistocene bones
by purication method. Nucleic Acids Research 21, 39133914.
Huson, D.H., Albrecht, B., Bag
̆ci, C., Bessarab, I., Górska, A., Jolic,
D., Williams, R.B.H., 2018. MEGAN-LR: new algorithms allow
accurate binning and easy interactive exploration of metagenomic
long reads and contigs. Biology Direct 13, 117.
Huson, D.H., Auch, A.F., Qi, J., Schuster, S.C., 2007. MEGAN
analysis of metagenomic data. Genome Research 17, 377386.
Huson, D.H., Beier, S., Flade, I., Górska, A., El-Hadidi, M., Mitra,
S., Ruscheweyh, H.J., Tappu, R., 2016. MEGAN community edi-
tion - interactive exploration and analysis of large-scale micro-
biome sequencing data. PLoS Computational Biology 12,
Huson, D.H., Mitra, S., 2011. Comparative metagenome analysis
using MEGAN. In: F.J. de Bruijn (Ed.) Handbook of Molecular
Microbial Ecology I: Metagenomics and Complementary
Approaches. Wiley-Blackwell, Hoboken, New Jersey, pp. 341352.
Jackson, S.T., 2012. Representation of ora and vegetation in Qua-
ternary fossil assemblages: known and unknown knowns and
unknowns. Quaternary Science Reviews 49, 115.
22 T.J. Murchie et al.
Jónsson, H., Ginolhac, A., Schubert, M., Johnson, P.L.F., Orlando,
L., 2013. MapDamage2.0: Fast approximate Bayesian estimates
of ancient DNA damage parameters. Bioinformatics 29, 1682
Jørgensen, T., Haile, J., Möller, P.E.R., Andreev, A., Boessenkool,
S., Rasmussen, M., Kienast, F., et al., 2012a. A comparative
study of ancient sedimentary DNA, pollen and macrofossils
from permafrost sediments of northern Siberia reveals long-term
vegetational stability. Molecular Ecology 21, 19892003.
Jørgensen, T., Kjær, K.H., Haile, J., Rasmussen, M., Boessenkool,
S., Andersen, K., Coissac, E., et al., 2012b. Islands in the ice:
detecting past vegetation on Greenlandic nunataks using historical
records and sedimentary ancient DNA meta-barcoding. Molecu-
lar Ecology 21, 19801988.
Kanagawa, T., 2003. Bias and artifacts in multitemplate polymerase
chain reactions (PCR). Journal of Bioscience and Bioengineering
96, 317323.
King, C.E., Debruyne, R., Kuch, M., Schwarz, C., Poinar, H.N.,
2009. A quantitative approach to detect and overcome PCR inhi-
bition in ancient DNA extracts. BioTechniques 47, 941949.
Kircher, M., Sawyer, S., Meyer, M., 2012. Double indexing over-
comes inaccuracies in multiplex sequencing on the Illumina plat-
form. Nucleic Acids Research 40, 18.
Klunk, J., Duggan, A.T., Redfern, R., Gamble, J., Boldsen, J.L.,
Golding, G.B., Walter, B.S., et al., 2019. Genetic resiliency
and the Black Death: no apparent loss of mitogenomic diversity
due to the Black Death in medieval London and Denmark. Amer-
ican Journal of Physical Anthropology 169, 240252.
Kreader, C.A., 1996. Relief of amplication inhibition in PCR with
bovine serum albumin or T4 gene 32 protein. Applied and Envi-
ronmental Microbiology 62, 11021106.
Krehenwinkel, H., Fong, M., Kennedy, S., Huang, E.G., Noriyuki,
S., Cayetano, L., Gillespie, R., 2018. The effect of DNA degra-
dation bias in passive sampling devices on metabarcoding studies
of arthropod communities and their associated microbiota. PLoS
ONE 13, 114.
Kress, W.J., Erickson, D.L., 2007. A two-locus global DNA bar-
code for land plants: the coding rbcL gene complements the non-
coding trnH-psbA spacer region. PloS ONE 2, e508.
Li, H., Durbin, R., 2009. Fast and accurate short read alignment
with Burrows-Wheeler transform. Bioinformatics 25, 175460.
Lorenz, M.G., Wackernagel, W., 1987a. Adsorption of DNA to
sand and variable degradation of adsorbed DNA. Applied and
Environmental Microbiology 53, 29482952.
Lorenz, M.G., Wackernagel, W., 1987b. Adsorption of DNA to
sand and variable degradation rates of adsorbed DNA. Applied
and Environmental Microbiology 53, 29482952.
Lu, J., Salzberg, S.L., 2018. Removing contaminants from metage-
nomic databases. PLoS Computational Biology 261859.
Mahony, M.E., 2015. 50,000 years of paleoenvironmental change
recorded in meteoric waters and coeval paleoecological and cry-
ostratigraphic indicators from the Klondike goldelds, Yukon,
Canada. Masters thesis, University of Alberta, Edmonton,
Alberta. doi:10.1145/3132847.3132886
Mamanova, L., Coffey, A.J., Scott, C.E., Kozarewa, I., Turner, E.H.,
Kumar, A., Howard, E., Shendure, J., Turner, D.J., 2010.
Target-enrichment strategies for next-generation sequencing.
Nature Methods 7, 1118.
Marciniak, S., Klunk, J., Devault, A., Enk, J., Poinar, H.N., 2015.
Ancient human genomics: the methodology behind reconstructing
evolutionary pathways. Journal of Human Evolution 79, 2134.
McKee, A.M., Spear, S.F., Pierson, T.W., 2015. The effect of dilu-
tion and the use of a post-extraction nucleic acid purication col-
umn on the accuracy, precision, and inhibition of environmental
DNA samples. Biological Conservation 183, 7076.
Meyer, M., Kircher, M., 2010. Illumina sequencing library prepara-
tion for highly multiplexed target capture and sequencing. Cold
Spring Harbor Protocols 5. doi:10.1101/pdb.prot5448
Mitchell, D., Willerslev, E., Hansen, A., 2005. Damage and repair
of ancient DNA. Mutation Research 571, 26576.
Morrissey, E.M., McHugh, T.A., Preteska, L., Hayer, M., Dijkstra,
P., Hungate, B.A., Schwartz, E., 2015. Dynamics of extracellular
DNA decomposition and bacterial community composition in
soil. Soil Biology and Biochemistry 86, 4249.
Mullis, K.B., Faloona, F.A., 1989. Specic synthesis of DNA in
vitro via a polymerase-catalyzed chain reaction. In: R. Wu,
L. Grossman, K. Moldave (Eds.) Selected Methods in Enzymol-
ogy, Recombinant DNA Methodology. Academic Press, San
Diego, California, pp. 189204.
Nichols, R. V., Vollmers, C., Newsom, L.A., Wang, Y., Heintzman,
P.D., Leighton, M., Green, R.E., Shapiro, B., 2018. Minimizing
polymerase biases in metabarcoding. Molecular Ecology
Resources 18, 927939.
Niemeyer, B., Epp, L.S., Stoof-Leichsenring, K.R., Pestryakova,
L.A., Herzschuh, U., 2017. A comparison of sedimentary DNA
and pollen from lake sediments in recording vegetation composi-
tion at the Siberian treeline. Molecular Ecology Resources 17,
Ogram, A., Sayler, G., Gustin, D., Lewis, R., 1988. DNA adsorp-
tion to soils and sediments. Environmental Science and Technol-
ogy 22, 982984.
Opel, K.L., Chung, D., McCord, B.R., 2010. A study of PCR inhi-
bition mechanisms using real time PCR. Journal of Forensic Sci-
ences 55, 2533.
Parducci, L., Bennett, K.D., Ficetola, G.F., Alsos, I.G., Suyama, Y.,
Wood, J.R., Pedersen, M.W., 2017. Ancient plant DNA in lake
sediments. New Phytologist 214, 924942.
Pedersen, M.W., Ginolhac, A., Orlando, L., Olsen, J., Andersen, K.,
Holm, J., Funder, S., Willerslev, E., Kjær, K.H., 2013. A com-
parative study of ancient environmental DNA to pollen and mac-
rofossils from lake sediments reveals taxonomic overlap and
additional plant taxa. Quaternary Science Reviews 75, 161168.
Pedersen, M.W., Ruter, A., Schweger, C., Friebe, H., Staff, R.A.,
Kjeldsen, K.K., Mendoza, M.L.Z., et al., 2016. Postglacial via-
bility and colonization in North Americas ice-free corridor.
Nature 537, 4549.
Percy, D.M., Argus, G.W., Cronk, Q.C., Fazekas, A.J., Kesanakurti,
P.R., Burgess, K.S., Husband, B.C., Newmaster, S.G., Barrett,
S.C.H., Graham, S.W., 2014. Understanding the spectacular fail-
ure of DNA barcoding in willows (Salix): does this result from a
trans-specic selective sweep? Molecular Ecology 47374756.
Rawlence, N.J., Lowe, D.J., Wood, J.R., Young, J.M., Churchman,
G.J., Huang, Y.T., Cooper, A., 2014. Using palaeoenvironmental
DNA to reconstruct past environments: progress and prospects.
Journal of Quaternary Science 29, 610626.
Renaud, G., Stenzel, U., Kelso, J., 2014. LeeHom: adaptor trim-
ming and merging for Illumina sequencing reads. Nucleic Acids
Research 42, e141.
Sadoway, T.R., 2014. A Metagenomic Analysis of Ancient Sedimen-
tary DNA Across the Pleistocene-Holocene Transition. Masters
thesis, Hamilton, Ontario. McMaster University.
Saeki, K., Sakai, M., Wada, S.I., 2010. DNA adsorption on syn-
thetic and natural allophanes. Applied Clay Science 50, 493497.
Optimizing extraction and targeted capture of ancient environmental DNA 23
Saiki, R.K., Gelfand, D.H., Stoffel, S., Scharf, S.J., Higuchi, R.,
Horn, G.T., Mullis, K.B., Erlich, H.A., 1988. Primer-directed
enzymatic amplication of DNA with a thermostable DNA poly-
merase. Science 239, 487491.
Saiki, R.K., Scharf, S., Faloona, F., Mullis, K.B., Horn, G.T., Erlich,
H.A., Arnheim, N., 1985. Enzymatic amplication of p-globin
genomic sequences and restriction site analysis for diagnosis of
sickle cell anemia. Science 230, 13501354.
Sawyer, S., Krause, J., Guschanski, K., Savolainen, V., Pääbo, S.,
2012. Temporal patterns of nucleotide misincorporations and
DNA fragmentation in ancient DNA. PloS ONE 7, e34131.
Schuenemann, V.J., Bos, K., DeWitte, S., Schmedes, S., Jamieson, J.,
Mittnik, A., Forrest, S., et al.., 2011. Targeted enrichment of
ancient pathogens yielding the pPCP1 plasmid of Yersinia pestis
from victims of the Black Death. Proceedingsof the National Acad-
emy of Sciences 108, E746E752. doi:10.1073/pnas.1105107108
Schweger, C., Froese, D., White, J.M., Westgate, J.A., 2011. Pre-
glacial and interglacial pollen records over the last 3 Ma from
northwest Canada: why do Holocene forests differ from those
of previous interglaciations? Quaternary Science Reviews 30,
Shah, N., Nute, M.G., Warnow, T., Pop, M., 2019. Misunderstood
parameter of NCBI BLAST impacts the correctness of bioinfor-
matics workows. Bioinformatics 35, 16131614.
Shen, Y.Y., Chen, X., Murphy, R.W., 2013. Assessing DNA bar-
coding as a tool for species identication and data quality control.
PLoS ONE 8, 15.
Sjögren, P., Edwards, M.E., Gielly, L., Langdon, C.T., Croudace,
I.W., Merkel, M.K.F., Fonville, T., Alsos, I.G., 2017. Lake sedi-
mentary DNA accurately records 20th century introductions of
exotic conifers in Scotland. New Phytologist 213, 929941.
Slon, V., Glocke, I., Barkai, R., Gopher, A., Hershkovitz, I., Meyer,
M., 2016. Mammalian mitochondrial capture, a tool for rapid
screening of DNA preservation in faunal and undiagnostic remains,
and its application to middle Pleistocene specimens from Qesem
Cave (Israel). Quaternary International 398, 210218.
Slon, V., Hopfe, C., Weiß, C.L., Mafessoni, F., de sa Rasilla, M.,
Lalueza-Fox, C., Rosas, A., et al., 2017. Neandertal and Deniso-
van DNA from Pleistocene sediments. Science 356, 605608.
Søe, M.J., Nejsum, P., Seersholm, F.V., Fredensborg, B.L., Hab-
raken, R., Haase, K., Hald, M.M., et al., 2018. Ancient DNA
from latrines in northern Europe and the Middle East (500 BC
1700 AD) reveals past parasites and diet. Plos ONE 13, e0195481.
Soininen, E.M., Gauthier, G., Bilodeau, F., Berteaux, D., Gielly, L.,
Taberlet, P., Gussarova, G., et al., 2015. Highly overlapping win-
ter diet in two sympatric lemming species revealed by DNA meta-
barcoding. PLoS ONE 10, 118.
Sønstebø, J.H., Gielly, L., Brysting, A.K., Elven, R., Edwards, M.,
Haile, J., Willerslev, E., et al., 2010. Using next-generation
sequencing for molecular reconstruction of past arctic vegetation
and climate. Molecular Ecology Resources 10, 10091018.
Strong, W.L., Hills, L. V., 2013. Holocene migration of lodgepole
pine (Pinus contorta var. latifolia) in southern Yukon, Canada.
Holocene 23, 13401349.
Sze, M.A., Schloss, P.D., 2019. The impact of DNA polymerase
and number of rounds of amplication in PCR on 16S rRNA
gene sequence data. bioRxiv 565598. doi:10.1101/565598
Taberlet, P., Bonin, A., Zinger, L., Coissac, E., 2018. Environmen-
tal DNA: For Biodiversity Research and Monitoring. Oxford Uni-
versity Press, Oxford, U.K.
Taberlet, P., Coissac, E., Pompanon, F., Gielly, L., Miquel, C., Val-
entini, A., Vermat, T., Corthier, G., Brochmann, C., Willerslev,
E., 2007. Power and limitations of the chloroplast trnL (UAA)
intron for plant DNA barcoding. Nucleic Acids Research 35,
e14. doi:10.1093/nar/gkl938
Whitworth, T.L., Dawson, R.D., Magalon, H., Baudry, E., 2007.
DNA barcoding cannot reliably identify species of the blowy
genus Protocalliphora (Diptera: Calliphoridae). Proceedings of
the Royal Society B 274, 17311739.
Willerslev, E., Davison, J., Moora, M., Zobel, M., Coissac, E.,
Edwards, M.E., Lorenzen, E.D., et al., 2014. Fifty thousand
years of arctic vegetation and megafaunal diet. Nature 506, 4751.
Yoccoz, N.G., Bråthen, K.A., Gielly, L., Haile, J., Edwards, M.E.,
Goslar, T., Von Stedingk, et al., 2012. DNA from soil mirrors
plant taxonomic and growth form diversity. Molecular Ecology
21, 36473655.
Zazula, G.D., Froese, D.G., Westgate, J.A., la Farge, C., Mathewes,
R.W., 2005. Paleoecology of Beringian packratmiddens from
central Yukon Territory, Canada. Quaternary Research 63, 189198.
Zazula, G.D., Schweger, C.E., Beaudoin, A.B., McCourt, G.H.,
2006. Macrofossil and pollen evidence for full-glacial steppe
within an ecological mosaic along the Bluesh River, eastern
Beringia. Quaternary International 142143, 219.
24 T.J. Murchie et al.
... 4,5 However, even here, the macrofossil record is significantly lacking in small-bodied fauna (e.g., rodents and birds), whereas questions related to migration and extirpation, even among well-studied taxa, remain crudely resolved. The growing sophistication of ancient environmental DNA (eDNA) methods have allowed for the identification of species within terrestrial/aquatic ecosystems, [6][7][8][9][10][11][12] in paleodietary reconstructions, [13][14][15][16][17][18][19] and facilitated genomic reconstructions from cave contexts. 8,[20][21][22] Murchie et al. 6,23 used a capture enrichment approach to sequence a diverse range of faunal and floral DNA from permafrost silts deposited during the Pleistocene-Holocene transition. ...
... The growing sophistication of ancient environmental DNA (eDNA) methods have allowed for the identification of species within terrestrial/aquatic ecosystems, [6][7][8][9][10][11][12] in paleodietary reconstructions, [13][14][15][16][17][18][19] and facilitated genomic reconstructions from cave contexts. 8,[20][21][22] Murchie et al. 6,23 used a capture enrichment approach to sequence a diverse range of faunal and floral DNA from permafrost silts deposited during the Pleistocene-Holocene transition. 24 Here, we expand on their work with the mitogenomic assembly and phylogenetic placement of Equus caballus (caballine horse), Bison priscus (steppe bison), Mammuthus primigenius (woolly mammoth), and Lagopus lagopus (willow ptarmigan) eDNA from multiple permafrost cores spanning the last 40,000 years. ...
... These libraries contain distinctive terminal deamination patterns, 25 and the negative controls contain no reads that passed map-filtering and could be BLASTn identified or mapped to the reference genomes that make up the PalaeoChip Arctic v1.0 bait set. 6 Combined with the temporally distinctive palaeoecological reconstructions detailed in Murchie et al. 6,23 (where these cores were Bayesian age modeled), the consistent damage patterns and clear negative controls serve as strong support that these sequenced reads originate from the loessal permafrost silts themselves and are characteristically ancient. ...
Traditionally, paleontologists have relied on the morphological features of bones and teeth to reconstruct the evolutionary relationships of extinct animals.¹ In recent decades, the analysis of ancient DNA recovered from macrofossils has provided a powerful means to evaluate these hypotheses and develop novel phylogenetic models.² Although a great deal of life history data can be extracted from bones, their scarcity and associated biases limit their information potential. The paleontological record of Beringia³—the unglaciated areas and former land bridge between northeast Eurasia and northwest North America—is relatively robust thanks to its perennially frozen ground favoring fossil preservation.⁴,⁵ However, even here, the macrofossil record is significantly lacking in small-bodied fauna (e.g., rodents and birds), whereas questions related to migration and extirpation, even among well-studied taxa, remain crudely resolved. The growing sophistication of ancient environmental DNA (eDNA) methods have allowed for the identification of species within terrestrial/aquatic ecosystems,6, 7, 8, 9, 10, 11, 12 in paleodietary reconstructions,13, 14, 15, 16, 17, 18, 19 and facilitated genomic reconstructions from cave contexts.⁸,20, 21, 22 Murchie et al.⁶,²³ used a capture enrichment approach to sequence a diverse range of faunal and floral DNA from permafrost silts deposited during the Pleistocene-Holocene transition.²⁴ Here, we expand on their work with the mitogenomic assembly and phylogenetic placement of Equus caballus (caballine horse), Bison priscus (steppe bison), Mammuthus primigenius (woolly mammoth), and Lagopus lagopus (willow ptarmigan) eDNA from multiple permafrost cores spanning the last 30,000 years. We identify a diverse metagenomic spectra of Pleistocene fauna and identify the eDNA co-occurrence of distinct Eurasian and American mitogenomic lineages.
... Here, we present hybridization capture enriched sedimentary ancient DNA (sedaDNA) data derived from loessal silts preserved in permafrost (Fig. 1, Table 1) and recovered from four sites in the Klondike goldfields-an unglaciated region of west-central Yukon Territory, Canada 35 -dating to ca. 30,000-4000 calibrated (calendar) years before present (cal BP). This work builds on the methodological results reported in Murchie et al. 36 in which North American horse (Equus caballus) and woolly mammoth (Mammuthus primigenius) DNA was unexpectedly identified in a permafrost sample dating to~9700 cal BP. This post-dates the last macrofossil evidence (such as bones, teeth, and soft-tissues) of these animals in Alaska by some 3300 years. ...
... When their populations began to decline (from anthropogenic or environmental pressures), it triggered numerous knock-on effects including the disappearance of mammoth-steppe vegetation and its replacement by shrublands and wet tundra. Our second objective is to replicate and test the extent of the sedaDNA ghost ranges for Equus sp. and Mammuthus as identified in Murchie et al. 36 . ...
... Ancient environmental DNA (eDNA) 76 is a powerful method for directly assessing the local 77 presence of animals, plants, fungi, and microbiota through time 33,36,37,41,[78][79][80][81][82] . Most eDNA is quickly metabolized by bacteria or otherwise degraded through a variety of chemical and physical processes. ...
Full-text available
The temporal and spatial coarseness of megafaunal fossil records complicates attempts to to disentangle the relative impacts of climate change, ecosystem restructuring, and human activities associated with the Late Quaternary extinctions. Advances in the extraction and identification of ancient DNA that was shed into the environment and preserved for millennia in sediment now provides a way to augment discontinuous palaeontological assemblages. Here, we present a 30,000-year sedimentary ancient DNA (sedaDNA) record derived from loessal permafrost silts in the Klondike region of Yukon, Canada. We observe a substantial turnover in ecosystem composition between 13,500 and 10,000 calendar years ago with the rise of woody shrubs and the disappearance of the mammoth-steppe (steppe-tundra) ecosystem. We also identify a lingering signal of Equus sp. (North American horse) and Mammuthus primigenius (woolly mammoth) at multiple sites persisting thousands of years after their supposed extinction from the fossil record.
... Metabarcoding involves polymerase chain reaction (PCR) amplification of a short, but highly variable, gene region that is flanked by conserved regions. The design of PCR amplification primers target these conserved regions, which allows amplification of the variable region across all plant taxa present in an environmental sample (Murchie et al., 2020). The variable region most often used to recover plant DNA in environmental samples is the P6 loop of the chloroplast trnL (UAA) intron (Taberlet et al., 2006). ...
... This approach reduces overall sequencing costs and increases the amount of plant genetic information that can be recovered (Foster et al., 2020). Targeted capture approaches have been shown to recover more taxa from environmental samples compared to PCR-based metabarcoding for a range of different organisms (Dowle et al., 2016;Shokralla et al., 2016;Murchie et al., 2020), and given the potential to recover a large amount of sequence data from multiple gene regions, this approach is also likely to improve taxonomic assignment of sequences from various plant communities. ...
... By incorporating multiple gene regions, we aimed to accurately reconstruct plant communities in environmental samples to species level identification, removing the reliance on a single, short barcode. The bait set utilised in this study has not previously been applied to environmental samples and it is rare that studies are undertaken targeting multiple gene regions in a single assay from complex DNA mixtures (only; Murchie et al., 2020;Lentz et al., 2021) and never with this many gene regions. We conducted sensitivity and discriminatory power tests on artificial DNA mixtures to assess the threshold of detection and the level of taxonomic resolution offered by the approach before analysing soil samples from a study site where we could verify results using observations from the area. ...
Full-text available
Metabarcoding of plant DNA recovered from environmental samples, termed environmental DNA (eDNA), has been used to detect invasive species, track biodiversity changes, and reconstruct past ecosystems. The P6 loop of the trnL intron is the most widely utilised gene region for metabarcoding plants due to the short fragment length and subsequent ease of recovery from degraded DNA, which is characteristic of environmental samples. However, the taxonomic resolution for this gene region is limited, often precluding species level identification. Additionally, targeting gene regions using universal primers can bias results as some taxa will amplify more effectively than others. To increase the ability of DNA metabarcoding to better resolve flowering plant species (angiosperms) within environmental samples, and reduce bias in amplification, we developed a multi-gene targeted capture method that simultaneously targets 20 chloroplast gene regions in a single assay across all flowering plant species. Using this approach, we effectively recovered multiple chloroplast gene regions for three species within artificial DNA mixtures down to 0.001 ng/μL of DNA. We tested the detection level of this approach, successfully recovering target genes for 10 flowering plant species. Finally, we applied this approach to sediment samples containing unknown compositions of eDNA and confidently detected plant species that were later verified with observation data. Targeting multiple chloroplast gene regions in environmental samples, enabled species-level information to be recovered from complex DNA mixtures. Thus, the method developed here, confers an improved level of data on community composition, which can be used to better understand flowering plant assemblages in environmental samples.
... The development of high-throughput sequencing methods also enabled the emergence of robust studies of ancient pathogens [12] and their importance for human prehistory [13], microbiomes [14] as well as the high-resolution reconstruction of past ecological communities from sedimentary aDNA (e.g. [15][16][17]). The first publications of complete prehistoric human and Neanderthal genomes [18,19] opened the floodgates for studies using ancient genomes ( palaeogenomics), especially to trace human gene flow across continents [20]. ...
... [15,36,[43][44][45][46]), and target enrichment, in which sedaDNA molecules of interest are selectively enriched prior to sequencing (e.g. barcode or mitochondrial loci [16,[47][48][49]). Detailed descriptions of these methods applied to sedaDNA have been recently reviewed elsewhere [50,51]. ...
... [49]) do not find evidence of leaching. Furthermore, molecular biology protocols can be impacted by variations in substrate composition and the co-extraction of inhibitors, although new specialist methods, such as the cold-spin DNA extraction method [16], are beginning to mitigate these issues. ...
Full-text available
Ancient DNA (aDNA) has played a major role in our understanding of the past. Important advances in the sequencing and analysis of aDNA from a range of organisms have enabled a detailed understanding of processes such as past demography, introgression, domestication, adaptation and speciation. However, to date and with the notable exception of microbiomes and sediments, most aDNA studies have focused on single taxa or taxonomic groups, making the study of changes at the community level challenging. This is rather surprising because current sequencing and analytical approaches allow us to obtain and analyse aDNA from multiple source materials. When combined, these data can enable the simultaneous study of multiple taxa through space and time, and could thus provide a more comprehensive understanding of ecosystem-wide changes. It is therefore timely to develop an integrative approach to aDNA studies by combining data from multiple taxa and substrates. In this review, we discuss the various applications, associated challenges and future prospects of such an approach.
... DNA hybridization capture (a.k.a. target enrichment) has already been successfully applied to enrich low-concentration DNA-fragments of interest from aDNA samples including sediments (Slon et al., 2017;Murchie et al., 2020;Vernot et al., 2021). It relies on the ability of the DNA to hybridize with complementary nucleic acids, the so-called baits. ...
Full-text available
Eastern Africa has been a prime target for scientific drilling because it is rich in key paleoanthropological sites as well as in paleolakes, containing valuable paleoclimatic information on evolutionary time scales. The Hominin Sites and Paleolakes Drilling Project (HSPDP) explores these paleolakes with the aim of reconstructing environmental conditions around critical episodes of hominin evolution. Identification of biological taxa based on their sedimentary ancient DNA (sedaDNA) traces can contribute to understand past ecological and climatological conditions of the living environment of our ancestors. However, sedaDNA recovery from tropical environments is challenging because high temperatures, UV irradiation, and desiccation result in highly degraded DNA. Consequently, most of the DNA fragments in tropical sediments are too short for PCR amplification. We analyzed sedaDNA in the upper 70 m of the composite sediment core of the HSPDP drill site at Chew Bahir for eukaryotic remnants. We first tested shotgun high throughput sequencing which leads to metagenomes dominated by bacterial DNA of the deep biosphere, while only a small fraction was derived from eukaryotic, and thus probably ancient, DNA. Subsequently, we performed cross-species hybridization capture of sedaDNA to enrich ancient DNA (aDNA) from eukaryotic remnants for paleoenvironmental analysis, using established barcoding genes (cox1 and rbcL for animals and plants, respectively) from 199 species that may have had relatives in the past biosphere at Chew Bahir. Metagenomes yielded after hybridization capture are richer in reads with similarity to cox1 and rbcL in comparison to metagenomes without prior hybridization capture. Taxonomic assignments of the reads from these hybridization capture metagenomes also yielded larger fractions of the eukaryotic domain. For reads assigned to cox1, inferred wet periods were associated with high inferred relative abundances of putative limnic organisms (gastropods, green algae), while inferred dry periods showed increased relative abundances for insects. These findings indicate that cross-species hybridization capture can be an effective approach to enhance the information content of sedaDNA in order to explore biosphere changes associated with past environmental conditions, enabling such analyses even under tropical conditions.
... Although costly, deeper shotgun sequencing (whereby more sequencing resources are provided to each sample) could also be used to obtain even more eukaryotic sedaDNA sequences. Third, to better focus the sequencing effort, the application of RNA baits designed to target and enrich certain taxonomic groups (e.g., phytoplankton and other microalgae (Armbrecht et al., 2021;or plants and animals;Murchie et al., 2020), using a multi-loci approach (Foster et al., 2020;Horn, 2012) could be applied, acknowledging that this method usually requires ∼100 ng of input DNA, which was not available for this data set. Finally, through de novo assembly of short DNA fragments that are compiled into metagenomic assembled genomes, it is possible that more unassigned sequences can be identified. ...
Full-text available
Studies incorporating sedimentary ancient DNA (sedaDNA) analyses to investigate paleo‐environments have increased considerably over the last few years, and the possibility of utilizing archived sediment cores from previous field campaigns could unlock an immense resource of sampling material for such paleo‐investigations. However, sedaDNA research is at a high risk of contamination by modern environmental DNA, as sub‐optimal sediment storage conditions may allow for contaminants (e.g., fungi) to grow and become dominant over preserved sedaDNA in the sample. Here, we test the feasibility of sedaDNA analysis applied to archive sediment material from five sites in the North Atlantic, collected between 1994 and 2013. We analyzed two samples (one younger and one older) per site using a metagenomic shotgun approach and were able to recover eukaryotic sedaDNA from all samples. We characterized the authenticity of each sample through sedaDNA fragment size and damage analyses, which allowed us to disentangle sedaDNA and contaminant DNA. Although we determined that contaminant sequences originated mainly from Ascomycota (fungi), most samples were dominated by Emiliania huxleyi, a haptophyte species that commonly blooms in the study region. We attribute the presence of contaminants to non‐ideal sampling and sample storage conditions of the investigated samples. Therefore, while we demonstrate that sedaDNA analysis of archival North Atlantic seafloor sediment samples are generally achievable, we stress the importance of best‐practice ship‐board sampling techniques and storage conditions to minimize contamination. We highly recommend the application of robust bioinformatic tools that help distinguish ancient genetic signals from modern contaminants, especially when working with archive material.
... DNA hybridization capture (a.k.a. target enrichment) has already been successfully applied to enrich low-concentration DNA-fragments of interest from aDNA samples including sediments (Slon et al., 2017;Murchie et al., 2020;Vernot et al., 2021). It relies on the ability of the DNA to hybridize with complementary nucleic acids, the so-called baits. ...
Full-text available
Eastern Africa has been a prime target for scientific drilling because it is rich in key paleoanthropological sites as well as in paleolakes, containing valuable paleoclimatic information on evolutionary time scales. The Hominin Sites and Paleolakes Drilling Project (HSPDP) explores these paleolakes with the aim of reconstructing environmental conditions around critical episodes of hominin evolution. Identification of biological taxa based on their sedimentary ancient DNA (sedaDNA) traces can contribute to understand past ecological and climatological conditions of the living environment of our ancestors. However, sedaDNA recovery from tropical environments is challenging because high temperatures, UV irradiation, and desiccation result in highly degraded DNA. Consequently, most of the DNA fragments in tropical sediments are too short for PCR amplification. We analyzed sedaDNA in the upper 70 m of the composite sediment core of the HSPDP drill site at Chew Bahir for eukaryotic remnants. We first tested shotgun high throughput sequencing which leads to metagenomes dominated by bacterial DNA of the deep biosphere, while only a small fraction was derived from eukaryotic, and thus probably ancient, DNA. Subsequently, we performed cross-species hybridization capture of sedaDNA to enrich ancient DNA (aDNA) from eukaryotic remnants for paleoenvironmental analysis, using established barcoding genes ( cox1 and rbcL for animals and plants, respectively) from 199 species that may have had relatives in the past biosphere at Chew Bahir. Metagenomes yielded after hybridization capture are richer in reads with similarity to cox1 and rbcL in comparison to metagenomes without prior hybridization capture. Taxonomic assignments of the reads from these hybridization capture metagenomes also yielded larger fractions of the eukaryotic domain. For reads assigned to cox1 , inferred wet periods were associated with high inferred relative abundances of putative limnic organisms (gastropods, green algae), while inferred dry periods showed increased relative abundances for insects. These findings indicate that cross-species hybridization capture can be an effective approach to enhance the information content of sedaDNA in order to explore biosphere changes associated with past environmental conditions, enabling such analyses even under tropical conditions.
... Meanwhile, cellular content was observed in several specimens (Roche et al., 2019). This poor state of preservation not only decreased the chances of recovering exploitable ancient nucleic acids but may also have jeopardized their stability under stringent extraction chemicals, as was suggested for the PowerMax Soil C3 solution (Murchie et al., 2020). ...