Got Glycogen?: Development and Multispecies Validation of the Novel Preserve, Precipitate, Lyse, Precipitate, Purify (PPLPP) Workflow for Environmental DNA Extraction from Longmire’s Preserved Water Samples

Abstract

Unfiltered and filtered water samples can be used to collect environmental DNA (eDNA). We developed the novel “Preserve, Precipitate, Lyse, Precipitate, Purify” (PPLPP) workflow to efficiently extract eDNA from Longmire’s preserved unfiltered and filtered water samples (44–100% recovery). The PPLPP workflow includes initial glycogen-aided isopropanol precipitation, guanidium hypochlorite and Triton X-100–based lysis, terminal glycogen-aided polyethylene glycol precipitation, and inhibitor purification. Three novel eDNA assays that exclusively target species invasive to Australia were also developed: Tilapia_v2_16S concurrently targets Oreochromis mossambicus (Mozambique tilapia) and Tilapia mariae (spotted tilapia) while R.marina_16S and C.caroliniana_matK discretely target Rhinella marina (cane toad) and Cabomba caroliniana (fanwort), respectively. All 3 assays were validated in silico before in vitro and in situ validations using PPLPP workflow extracted samples. PPLPP workflow was concurrently validated in vitro and in situ using all 3 assays. In vitro validations demonstrated that 1) glycogen inclusion increased extracellular DNA recovery by ∼48-fold compared with glycogen exclusion, 2) swinging-bucket centrifugation for 90 min at 3270 g is equivalent to fixed-angle centrifugation for 5–20 min at 6750 g, and 3) Zymo OneStep Inhibitor Removal Kit, Qiagen DNeasy PowerClean Pro Cleanup Kit, and silica-Zymo double purification provide effective inhibitor removal. In situ validation demonstrated 95.8 ± 2.8% (mean ± SEM) detectability across all 3 target species in Longmire’s preserved unfiltered and filtered water samples extracted using the PPLPP workflow (without phenol:chloroform:isoamyl alcohol purification) after 39 d of incubation at room temperature and 50°C. PPLPP workflow is recommended for future temperate and tropical eDNA studies that use Longmire’s to preserve unfiltered or filtered water samples.

INTRODUCTION

Environmental DNA (eDNA), or DNA that is deposited into the environment by all living organisms through biologic processes (e.g., waste excretion, epidermal sloughing, decay), is an emerging, robust, and noninvasive genetic tool for target species and/or community surveys (reviewed by Goldberg et al.,¹ Deiner and Mächler,² Tsuji et al.,³ and Huerlimann et al.⁴). Unfiltered water samples are volume limited (e.g., 15–45 ml) and less commonly used for eDNA collection than filtration of larger water volumes (e.g., 250 ml to 100 L) through membranes of various types and pore sizes; however, given sample collection ease (i.e., nonexpert appropriate) and retention of extracellular and intracellular eDNA, unfiltered water samples provide a useful and effective alternative to filtration when eDNA is present in a relatively high concentration (reviewed by Goldberg et al.,¹ Tsuji et al.,³ and Huerlimann et al.⁴). More specifically, unfiltered water samples have been successfully used to detect the following: invasive wetlands bullfrog Rana catesbeiana^{5, 6}; commercially important Siberian sturgeon Acipenser baerii⁶; freshwater biodiversity of amphibians, fish, mammals, insects, and crustaceans⁷; elusive Burmese python Python bivittatus⁸; freshwater biodiversity of eubacteria and eukaryotes⁹; protozoan parasite Chilodonella hexasticha¹⁰; feral pig Sus scrofa^{11, 12}; invasive Asian carp Hypophthalmichthys molitrix¹³; endangered rainforest frogs Litoria lorica and L. nannotis¹⁴; and invasive freshwater plant Cabomba caroliniana.¹⁵

Recovery efficiency of eDNA from water samples can be influenced by many factors during extraction; thus, it is of escalating importance that eDNA studies consider the implications and limitations of chosen methodology.^{9, 16−23} Most notable is the lack of studies examining the addition of a coprecipitate [e.g., glycogen, linear polyacrylamide (LPA), yeast tRNA] when alcohol-salt precipitation(s) are undertaken despite known improvements to nucleic acid recovery.^{17, 24−27} Of note is that commercially available glycogen can contain trace quantities (picograms per microliter) of DNA from biological sources (e.g., mussels, oysters, bovine liver, or rabbit liver), whereas commercially available synthetic LPA typically does not²⁶; therefore, glycogen DNase pretreatment is recommended. To date, only 4 noninvasive genetics studies have utilized commercial glycogen as a coprecipitate during (e)DNA isolation from feathers,²⁸ sediment,^{24, 29} or filtered water samples³⁰; however, these studies did not test multiple glycogen concentrations for optimal recovery efficiency. Boström et al.²⁵ demonstrated a >6-fold increase in DNA recovery when tRNA (50 µg) was added as a coprecipitate and Lever et al.¹⁷ added LPA (20 µl/ml) to all precipitations without direct comparison of inclusion vs. exclusion on yield.

Both unfiltered and filtered water samples can be preserved against degradation by combining with ethanol (EtOH^{13, 31−33}) or nonalcohol buffers (e.g., Longmire’s and Queen’s; Table 1). Ethanol, which prevents hydrolysis of collected eDNA, has been the most broadly utilized for filtered eDNA samples (reviewed by Goldberg, et al.¹) and has recently been shown to preserve unfiltered eDNA samples for up to 6 d at 23.9°C.¹³ However, despite its effectiveness, EtOH-based preservation is often precluded from use in remote areas because of dangerous-goods shipping restrictions and total alcohol bans imposed by governing traditional landowners.⁴ Several nonalcohol alternatives exist for effective eDNA preservation (Table 1); however, of these, Longmire’s is the most broadly utilized to date for both unfiltered and filtered water samples. More specifically, Longmire’s is a lysis buffer³⁴⁻³⁶ that contains both ethylenediaminetetraacetate (EDTA) and sodium dodecyl sulfate (SDS), which collectively inhibit nuclease activity.³⁷ Like EtOH, Longmire’s has been demonstrated to provide effective eDNA preservative for both unfiltered and filtered water samples subjected to a range of temporal and thermal treatments (Table 1); however, unlike EtOH, Longmire’s lysis action requires that the volume used to preserve filtered water samples (i.e., intact or cut membranes) should not be discarded but rather precipitated to ensure retention of all lysed (i.e., dissolved) eDNA. For unfiltered water samples, Longmire’s has been demonstrated to provide efficient eDNA preservation when used at a 1:3 ratio (1 part Longmire’s to 3 parts environmental water^{12, 38}), whereas Queen’s, an alternative nonalcohol lysis buffer that also contains EDTA but uses N-lauroylsarcosine instead of SDS as the anionic surfactant^{36, 39} (Table 1), has yet to be tested for eDNA preservation effectiveness when used at a 1:3 ratio.

TABLE 1.

Summary of nonalcohol preservatives used for unfiltered and filtered water samples containing eDNA

Preservative	Chemical composition	Sample volume	Capture method	Preservation duration	Temperature	Reference
BAC	0.01% BAC	500 ml	Unfiltered	8 h	Ambient	79
		1 L	Unfiltered	8 h	Ambient	22
CTAB	100 mM Tris-HCl, 20 mM EDTA, 1.4 M NaCl, 2% (w/v) CTAB, 0.25 mM PVP	1.25 L	Filtered (polycarbonate track-etch)	0, 1, and 2 wk	20 and 45°C	76
		1 L	Filtered (cellulose nitrate)	Unspecified	−20°C	de Souza, et al., 201651
		250 ml	Filtered (polycarbonate)	Unspecified	−20°C	80
DMSOa	20% DMSO, 0.25 M EDTA, saturated with NaCl	250 ml	Filtered (polycarbonate track-etch)	Unspecified	Unspecified	76
GIT buffer	50 mM Tris-HCl pH 7.4, 25 mM EDTA pH 8, 5 M guanidine isothiocyanate, 0.8% 2-ME	1–2.5 ml	Unfiltered	5 d	Ambient	81
Longmire’s	100 mM Tris-HCl, 100 mM EDTA, 10 mM NaCl, 0.5% (w/v) SDS	1.25 L	Filtered (polycarbonate track-etch)	0, 1, and 2 wk	20 and 45°C	76
		2 L	Filtered (polycarbonate)	150 d	Ambient	82
		250 ml	Filtered (polycarbonate and cellulose nitrate)	8 d	4°C	83
		1 L	Filtered (n = 4 filter types)	5, 24 h, and 2 wk	Ambient	18
		15 ml	Unfiltered (1:3, 1:6, and 1:15 ratios)	56 d	7.3–34.2°C	38b
		5 L	Filtered (nitrocellulose)	Unspecified	Unspecified	116
		1 L	Filtered (cellulose acetate)	Unspecified	Ambient	84
		250 ml	Filtered (cellulose nitrate)	>1 mo (not specified)	4°C	85
		1 L	Filtered (cellulose acetate)	Unspecified	Unspecified	86
		1–2 L	Filtered (polycarbonate)	Unspecified	On ice	87
		45 ml	Unfiltered (1:3 ratio)	∼1 wk	Ambient	12
		250 ml	Filtered (cellulose nitrate)	Unspecified	Ambient	88
		250 ml	Filtered (glass microfiber)	3 wk	4°C	89
		300 ml	Filtered (various filter types)	Unspecified	Unspecified	90
		2.3–3.1 L	Filtered (nitrocellulose)	Unspecified	Ambient	91
		1.04 L	Filtered (Sterivex® cartridges)	2 wk	4°C	92
		15 ml	Unfiltered (1:3 ratio) and Filtered (nylon net)	39 d	Ambient and 50°C	This study
Preservation buffer	100 mM Tris-HCl, 100 mM EDTA, 10 mM NaCl, 1% (w/v) N-lauroylsarcosine, pH 7.5–8.0	45 L	Filtered (Envirocheck® HV capsule)	Unspecified	Ambient	93
		10 L	Filtered (Envirocheck® HV capsule)	Few days (not specified)	4°C	94
		20 and 60 L	Filtered (Envirocheck® HV capsule)	Unspecified	Unspecified	95
		60 L	Filtered (Envirocheck® HV capsule)	Unspecified	Ambient	96
		4 and 0.5 L	Filtered (nitrocellulose)	Unspecified	Ambient	97
		∼50 L	Filtered (Envirocheck® HV capsule)	≤1 mo	Ambient	98
Queen’s	10 mM Tris-HCl, 10 mM EDTA, 10 mM NaCl, 1% N-lauroylsarcosine, pH 7.5	15 ml	Unfiltered (1:3 ratio)	9 d	40°C	This study
RNAlatera	20 mM EDTA disodium, dihydrate (pH 8), 25 mM sodium citrate trisodium salt, 5.3 M ammonium sulfate, pH 5.2	250 ml	Filtered (polycarbonate track-etch)	Unspecified	Unspecified	76
		1 L	Filtered (n = 4 filter types)	5, 24 h, 2 wk	Ambient	18
		100–150 ml	Filtered (Sterivex® cartridge)	12 d	4°C	99
SLB	50 mM Tris-HCl, 20 mM EDTA, 200 mM NaCl, 750 mM sucrose, pH 9.0	1–2.5 ml	Unfiltered	5 d	Ambient	81
	50 mM Tris-HCl, 20 mM EDTA, 400 mM NaCl, 750 mM sucrose, pH 9.0	1 L	Filtered (Supor® 200)	<1 d	−80°C	100
		250–500 ml	Filtered (Supor® 200)	<1 d	−70°C	101
		12–24 L	Filtered (polysulfone)	<1 d	−20°C	102
		1 L	Unfiltered	8 h	Ambient	22

2-ME, 2-mercaptoethanol; BAC, benzalkonium chloride; CTAB, cetyltrimethyl ammonium bromide; EDTA, ethylenediaminetetraacetate; GIT buffer, guanidine isothiocyanate and Tris buffer; HV, high volume; NaCl, sodium chloride; PVP, polyvinylpyrrolidone; SDS, sodium dodecyl sulfate; SLB, sucrose lysis buffer; Tris-HCl, tris hydrochloride. ^aPilot study determined incompatibility with extraction protocol due to formation of qPCR inhibitive precipitate. ^bLongmire’s included 0.2% sodium azide.

Extraction of eDNA from both unfiltered and filtered water samples typically involves initial precipitation to capture dissolved (i.e., extracellular) or suspended (i.e., intracellular) eDNA and/or terminal precipitation to purify extracted eDNA by mixing with absolute EtOH or absolute isopropanol and sodium acetate (NaAc) or sodium chloride (NaCl) (Table 2). Across eDNA studies, initial EtOH-NaAc precipitation is more commonly utilized than initial isopropanol-NaAc and isopropanol-NaCl precipitation; however, terminal isopropanol-NaAc and isopropanol-NaCl precipitation are more commonly utilized than terminal EtOH-NaCl precipitation (Table 2). Of note is that terminal precipitation with polyethylene glycol (PEG) and NaCl is the least commonly utilized approach during eDNA extraction from water samples to date (Table 2), despite historical use for low-copy DNA template recovery (e.g., human virus pathogens⁴⁰) and more recent demonstrations of effectiveness²⁶ and inhibitor removal.¹⁷

TABLE 2.

Overview of precipitation methods used during eDNA extraction

Name	Chemical composition	Incubation	Extraction	Centrifugation	References
Initial ethanol	2.2 volumes absolute EtOH + 0.1 volume 3 M NaAc	−20°C until extraction	Pelleted by centrifugation and extracted with QIAmp Tissue Extraction Kit (Qiagen)	5500 g for 35 min at 6°C	5
		−20°C until extraction	Pelleted by centrifugation and extracted with QIAmp Tissue Extraction Kit (Qiagen)	9400 g for 1 h at 6°C	6
		None (extracted immediately)	Pelleted by centrifugation and then either PowerWater DNA Isolation Kit (MoBio Laboratories) or QIAmp DNA Micro Kit (Qiagen)	3220 g for 45 min at 6°C	8
		−20°C until extraction	Pelleted by centrifugation and extracted with PCI, PowerWater DNA Isolation Kit (MoBio Laboratories), or DNeasy Blood and Tissue Kit (Qiagen)	5500 g for 35 min at 4°C	9
		10–120 min on ice, then −20°C until extraction	Pelleted by centrifugation and extracted with CTAB protocol	3220 g for 35 min at 6°C	32
		None (extracted immediately)	Pelleted by centrifugation and extracted with Bioline Isolate II Genomic DNA Kit	3200 g for 70 min at 6°C	64
		−20°C until extraction	Pelleted by centrifugation and extracted with DNeasy Blood and Tissue Kit (Qiagen)	10,000 g for 1 h at 4°C	103
		4°C until extraction	Pelleted by centrifugation and extracted with CTAB protocol	3200 g for 60 min at 6°C	11
		None (extracted immediately)	Pelleted by centrifugation, decanted, and air dried for 5 min (no EtOH pellet washes)	3220 g for 45 min at ?°Ca	12
		−20°C until extraction	Pelleted by centrifugation and extracted with DNeasy Blood and Tissue Kit (Qiagen)	8500 rpm for 30 min at ?°C	104
		−80°C until extraction	Pelleted by centrifugation and extracted with DNeasy mericon Food Kit (Qiagen)	9000 g for 15 min at ?°C	13
		Room temperature	Pelleted by centrifugation and extracted with DNeasy Blood and Tissue Kit (Qiagen)	15,000 g for 15 min at 6°C	98
		4 and 23.9°C for 0–6 d	Pelleted by centrifugation and extracted with IBI gMAX Mini Genomic DNA Kit for blood, tissue, and cultured cells (IBI Scientific, Peosta, IA, USA).	5000 g for 30 min at 20°C.	13
Terminal ethanol	2.5 volumes 99.6% EtOH, 0.1 volume 3 M NaAc, and 50 µg tRNA3	−20°C for 1 h	Pelleted by centrifugation, supernatant removed, one 70% EtOH wash, air dried, eluted	20,000 g for 20 min at 4°C	25
	2 volumes 100% EtOH and 40 µl 5 M NaCl4	−30°C overnight	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, air dried, eluted	10,000 rpm for 30 min at 4°C	9, 75
	2.5 volumes 100% EtOH, 0.2 volumes 5 M NaCl, and 20 µl/ml LPA5	Room temperature (2 h) or 4°C (overnight)	Pelleted by centrifugation, supernatant removed, one or two 70% EtOH washes, air dried, eluted	14,000 g for 30 min at 4°C	17
	2.2 volumes 100% EtOH and 0.1 volume 5 M NaCl	−20°C overnight	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, vacuufuge and air dried, eluted	15,000 g for 10 min at ?°C	76
	2 volumes 100% EtOH and 0.1 volume 5 M NaCl	−20°C overnight	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, air dried, eluted	20,000 g for 20 min at 4°C	19
	1 ml 100% EtOH, 50 µl 3 M NaAc, and 40 µg glycogen	−20°C overnight	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, air dried, eluted	13,000 rpm for 30 and 45 min at 4°C	30
Initial isopropanol	0.8 volumes 100% isopropanol and 0.1 volume 3 M NaAc	−20°C until extraction	Pelleted by centrifugation and extracted with DNeasy Blood and Tissue Kit (Qiagen)	10,000 g for 1 h at 4°C	103
	0.8 volumes 100% isopropanol, 0.2 volumes 5 M NaCl, and ≥4.4 µg/ml glycogen	4°C until extraction	Pelleted by centrifugation, resuspended in Lysis Buffer Ib, frozen, thawed, incubated at 50°C then precipitated again (see below)	3270 g for 90 min at 22°Cc or 6750 g for 10 min at 22°Cd	This study
Terminal isopropanol	0.6 volumes 100% isopropanol, 0.1 volume 3 M NaAc, and 50 µg tRNA	−20°C for 1 h	Pelleted by centrifugation, supernatant removed, one 70% EtOH wash, air dried, eluted	20,000 g for 20 min at 4°C	25
	1 volume 100% isopropanol and 0.5 volumes 5 M NaCl	−20°C overnight	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, air dried, eluted	16,873 g for 10 min at ?°C	16
	1.5 volumes 100% isopropanol, 0.1 volume 5 M NaCl, and 20 µl/ml LPA	−20°C for ≥2 h	Pelleted by centrifugation, supernatant removed, one or two 70% EtOH washes, air dried, eluted	14,000 g for 30 min at 4°C	17
	0.8 volume 100% isopropanol and 0.4 volumes 5 M NaCl	−20°C overnight	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, air dried, eluted	15,000 g for 10 min at ?°C	76
		−4°C overnight	Pelleted by centrifugation, supernatant removed, one 70% EtOH wash, air dried, eluted	13,000 rpm for 20 min at ?°C	20
	1 volume 100% isopropanol and 0.1 volume 3 M NaAc	−4°C overnight	Pelleted by centrifugation, supernatant removed, one 70% EtOH wash, air dried, eluted	13,000 rpm for 20 min at ?°C	89
		−20°C overnight	Pelleted by centrifugation, supernatant removed, one 70% EtOH wash, eluted	Unspecified	106
Terminal PEG	2 volumes 30% PEG6000 in 1.6 M NaCl, 30% PEG8000 in 0.4–2 M NaCl, and 20 µl/ml LPA	Room temperature (2 h) or 4°C (overnight)	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, eluted	14,000 g or 20,000 g for 30, 45, or 60 min at room temp	17
	2 volumes 30% PEG8000 in 1.6 M NaCl and ≥55.5 µg/ml glycogen	4°C overnight	Pelleted by centrifugation, supernatant removed, two 70% EtOH washes, eluted	20,000 g for 30 min at 20°C	This study

CTAB, cetyltrimethyl ammonium bromide; LPA, linearized polyacrylamide; NaAc, sodium acetate; NaCl, sodium chloride; PEG, polyethylene glycol; tRNA, transfer ribonucleic acid. ^aIndicates unspecified temperature. ^bSee Table 3. ^cSwinging-bucket rotor. ^dFixed-angle rotor.

Various lysis buffers have been used during extraction of DNA from noninvasively collected samples (Table 3). Following 37–65°C incubation for 15 min to overnight in a lysis buffer (with or without addition of Proteinase K and/or lysozyme), samples are most commonly subjected to 1 phenol:chloroform:isoamylalcohol (25:24:1; PCI) and/or 2 chloroform:isoamylalcohol (24:1; CI) phase separations followed by EtOH precipitation (Table 3). Given that PCI and CI phase separations are most commonly undertaken to remove proteins, lipids, and detergents from nucleic acids following enzymatic and/or mechanical lysis and before alcohol-salt precipitation (Table 3), we tested these phase separations for postlysis purification effectiveness as opposed to stand-alone extraction effectiveness. Regardless of definition, PCI and other non-PCI methods (e.g., Mu-DNA⁴¹) are less commonly used for processing eDNA samples than commercial kits (reviewed by Tsuji et al.³). Relative to commercial kits, PCI and non-PCI methods generally recover more eDNA (reviewed by Goldberg, et al.¹, Tsuji, et al.³); however, Boström et al.²⁵ demonstrated that inclusion of PCI purification reduced recovery of extracellular DNA and was not essential for in situ detection of target species. As such, processing unfiltered and 0.2 µM filtered water samples that potentially contain extracellular eDNA^{24, 43, 44} requires careful consideration regarding inclusion of PCI and/or CI phase separations for extraction or purification, and if utilized, initial and/or terminal precipitation(s) should include a coprecipitate to help maximize recovery of collected nucleic acids.^{17, 24−27}

TABLE 3.

Overview of buffers used during extraction of DNA from noninvasive samples

Name	Chemical composition	Incubation	Extraction	References
CTAB buffer	100 mM Tris-HCl pH 8, 1.4 M NaCl, 2% (w/v) CTAB, 20 mM EDTA, 0.25 mM PVP	65°C for 1 h	PCIa then EtOH + precipitation	75, 107
	100 mM Tris-HCl pH 8, 1.4 M NaCl, 2% (w/v) CTAB, 0.4% (v/v) 2-ME, 1% (w/v) PVP, pH 8.0	50°C for 15 min	CIb then isopropanol precipitation	108, 109
	100 mM Tris-HCl pH 8, 0.5 M NaCl, 0.5% (w/v) CTAB, 50 mM EDTA plus 2% SDS and 100 µg ProK	55°C for 3 h	CI then EtOH precipitation with 1 mg/ml glycogen	29
Digestion buffer	50 mM Tris-HCl pH 8.5, 1.0 mM EDTA, 0.5% Tween-20 plus 15.9U ProK	55°C for 15 min then 95°C for 7 min (ProK inactivation). Optional second incubation at 60°C for 10 min after addition of 150 µl digestion buffer and 7.95U ProK	DNA eluted from spin-column by centrifugation	110
Lysis buffer I	30 mM Tris-HCl pH 8, 30 mM EDTA pH 8, 800 mM guanidium hydrochloride, 0.5% Triton X-100, pH 10	50°C for 1–3 h	PCI purification (optional) then EtOH, isopropanol, or PEG precipitation with 20 µg/ml LPA	17
		50°C for ≥3 h	PCI purification (optional) then PEG precipitation with ≥55.5 µg/ml glycogen	This study
Lysis buffer II	100 mM Tris-HCl pH 8.0, 5 mM EDTA, 200 mM NaCl, 0.2% SDS, and 80 µg ProK	55°C overnight	PCI then EtOH precipitation	75
Lysis buffer III	10 mM Tris-HCl pH 8, 10 mM EDTA, 0.5% SDS, and 50 µg/ml ProK	56°C for 15 min	PCI then EtOH precipitation	33
Lysis buffer IV	50 mM Tris-HCl pH 8.0, 20 mM EDTA (pH 8.0), 150 mM NaCl, 68 mM N-lauroylsarcosine, 0.5% 2-ME, 0.33 mM DTT, and 170 µg ProK (twice)	37°C overnight (second 170 µg ProK added immediately prior to overnight incubation)	PCI then EtOH precipitation	111
Lysis solution	67 mM Tris-HCl (pH 8), 26.5 mM EDTA (pH 8), ∼100 mM guanidine thiocyanate, ∼228 mM trisodium phosphate dodecahydrate, and ∼34 mM NaCl, pH 9.0. For water samples, add ∼1.3% SDS (“Water Lysis Additive”)	None	Mu-DNA protocol	41
SDS buffer	10 mM Tris-HCl, 100 mM EDTA, 200 mM NaCl, and 1% SDS	1 mg/ml lysozyme added then 37°C for 30 min then 1 mg/ml ProK added then 55°C overnight	PCI then EtOH precipitation	19
SLB I	50 mM Tris-HCl, 40 mM EDTA, 400 mM NaCl, 750 mM sucrose, pH 9.0 plus lysozyme (1 mg/ml), 0.5% SDS, and 160 µg ProK	37°C for 20 min (lysozyme) then 37°C for 2 h (SDS + ProK)	PCI then EtOH precipitation	100, 112
SLB II	50 mM Tris-HCl, 40 mM EDTA, 400 mM NaCl, 750 mM sucrose, pH 9.0 plus lysozyme (1 mg/ml), 1% SDS, and 160 µg ProK	37°C for 20 min (lysozyme) then 37°C for 2 h (SDS + ProK) then 56°C for 15 min (phenol)	PCI then EtOH precipitation	113
SLB III	50 mM Tris-HCl, 20 mM EDTA, 400 mM NaCl, 750 mM sucrose, pH 9.0 plus 1% SDS and 100 µg ProK	37°C for 30 min (SDS + ProK) then 55°C for 10 min (SDS + ProK)	PCI then EtOH precipitation	102
SLB IV	50 mM Tris-HCl, 20 mM EDTA, 400 mM NaCl, 750 mM sucrose, pH 9.0 plus lysozyme (1 mg/ml), 1% SDS, and 100 µg ProK	37°C (lysozyme) then 55°C (SDS + ProK)	PCI purification (optional) then EtOH or isopropanol precipitation with 50 µg tRNA	25

CI, chloroform:isoamyl alcohol (24:1); CTAB, cetyltrimethyl ammonium bromide; DTT, dithiothreitol; EDTA, ethylenediaminetetraacetate; EtOH, ethanol; LPA, linearized polyacrylamide; NaCl, sodium chloride; PCI, phenol:chloroform:isoamyl alcohol (25:24:1); ProK, proteinase K; PVP, polyvinylpyrrolidone (mol wt 360,000); SDS, sodium dodecyl sulfate; SLB, sucrose lysis buffer Tris-HCl, tris hydrochloride; tRNA, transfer ribonucleic acid.

Environmental inhibitors (e.g., humic and fulvic acids^{45, 46}) can be coextracted with eDNA and, thus, generally require removal (reviewed by Goldberg, et al.¹, Huerlimann, et al.⁴). For example, OneStep PCR Inhibitor Removal Kit (hereafter referred to as Zymo; Zymo Research Inc., Irvine, CA, USA) has become one of the most commonly used commercial spin-column inhibitor purification methods^{20, 47−50}; however, direct comparisons between Zymo and other inhibitor purification methods (e.g., dilution or commercial kits) are infrequent (e.g., Minegishi et al.²¹ and McKee et al.⁵¹). Despite broad use of Qiagen DNeasy kits for eDNA extraction (e.g., DNeasy Blood and Tissue Kit, DNeasy mericon Food Kit; reviewed by Tsuji, et al.,³), Qiagen DNeasy PowerClean Pro Cleanup Kit (hereafter referred to as Qiagen; Qiagen Australia Pty Ltd, Chadstone, VIC, Australia) has only recently been used for inhibitor purification of in vitro⁵³ and in situ^{21, 54, 55} eDNA samples. Nonkit inhibitor purification using silica has also been utilized for plant and ancient DNA⁵⁶⁻⁵⁹; however, effective silica removal is required because of the potential inhibition of downstream enzymatic reactions if trace amounts are carried over into final elution.^{56, 57}

Here, we present a novel eDNA extraction workflow called “Preserve, Precipitate, Lyse, Precipitate, Purify” (PPLPP) that provides eDNA researchers with an optimized, efficient, cost-effective, and minimally toxic workflow for extraction of eDNA from Longmire’s preserved unfiltered and filtered water samples. We also present 3 concurrently developed eDNA assays that target 4 species invasive to Australia. All 3 assays were validated in silico before in vitro and in situ validations using PPLPP workflow extracted samples (Fig. 1). The PPLPP workflow was concurrently validated in vitro and in situ using all 3 species-specific assays. Protocols are provided for preparation of Longmire’s, Lysis Buffer I, and PEG8000-NaCl solutions (Supplemental Protocol 1); in-house DNase-treated glycogen (Supplemental Protocol 2); and silica inhibitor purification (Supplemental Protocol 3).

FIGURE 1.

Schematic overview of in silico, in vitro, and in situ validations conducted during PPLPP workflow development as well as PPLPP workflow overview (see Fig. 2 for detailed version of the PPLPP workflow). Arrows indicate the order in which in silico (light grey box), in vitro (medium grey boxes), and in situ (dark grey box) validations were conducted. In vitro validations 3–7 were used to derive the optimal PPLPP workflow (inset), which was then used for in vitro validations 8–11 and in situ validation (dashed lines). Target species and number of biologic replicates (n) are provided for each in vitro and in situ validation. Tilapia: O. mossambicus; cane toad: R. marina; Cabomba: C. caroliniana.

MATERIALS AND METHODS

PPLPP workflow overview

PPLPP workflow uses NaCl instead of NaAc for initial precipitation because NaCl is recommended when DNA is precipitated from solutions that contain SDS (Maniatis et al.¹²¹), such as Longmire’s (Supplemental Protocol 1 and Table 1). Centrifugation is undertaken at room temperature (except for pilot centrifugation-only test; see Preservatives and terminal precipitants) because centrifugation at warmer temperatures (i.e., 22°C vs. 4°C) assists with pellet formation and avoids coprecipitation of NaCl. Initial precipitation uses isopropanol (Table 2) because the 0.8 volumes required fits into 50 mL collection tubes that contain 15 mL unfiltered water samples preserved with 5 mL Longmire’s whereas the 2.2 volumes required for EtOH precipitation of Longmire’s preserved samples do not. Lastly, supernatants were decanted following initial isopropanol precipitation and pipetted following terminal precipitation with PEG8000 dissolved in NaCl (hereafter referred to as PEG8000-NaCl; Supplemental Protocol 1 and Table 2) given the risk of pellet loss if PEG-based precipitation supernatants are decanted.¹⁷

Target species

Given their long-standing and substantial impact on Australian ecosystems, we targeted both Oreochromis mossambicus (Mozambique tilapia) ^{60, 123} and Tilapia mariae (spotted tilapia) ^{113, 137} as well as Rhinella marina (cane toad) ^{121, 125, 131, 134, 136} and Cabomba caroliniana (fanwort) ^{62, 69, 71, 72} (see Supplemental Information).

Assay design

We developed a revised assay that targets and amplifies a mitochondrial 16S gene region equally for both O. mossambicus and T. mariae (Tilapia_v2_16S⁶⁰) as well as new assays that target R. marina 16S (R.marina_16S¹⁴) and C. caroliniana matK (C.caroliniana_matK¹⁵) (see Supplemental Information). Geneious analysis software (version R11)¹²² was used to obtain all available sequences for O. mossambicus and T. mariae 16S (n = 11), R. marina 16S (n = 27), and C. caroliniana matK (n = 13). Sequences for each species were aligned using ClustalW¹³² and assessed for conserved regions wherein target species exhibited ≥ 1 base pair mismatch with humans and non-target Australian fishes (n = 82), frogs (n = 18), turtles (n = 12), and aquatic plants (n = 36) (Supplemental Tables 2 and 3).

Sources of genomic, synthetic, and environmental DNA

Genomic DNA (gDNA) was extracted using cetyltrimethylammonium bromide (CTAB)¹⁰ in the dedicated DNA extraction room of the Molecular Ecology and Evolution Laboratory (MEEL; James Cook University, Townsville, Australia). Extracted gDNA was quantified before and after dilution using the QuantiFluor fluorometer and associated QuantiFluor ONE dsDNA System (Promega Co., Alexandria, NSW, Australia).

Synthetic DNA (sDNA) fragments based on the target region of O. mossambicus 16S, R. marina 16S, and C. caroliniana maturase K (matK) sequences (185, 317, and 385 bp respectively) were synthesized as gBlocks (Integrated DNA Technologies Australia Pty Ltd, Baulkham Hills, NSW, Australia). Dried sDNA pellets were resuspended in 100 µl 1x Tris-EDTA (TE) buffer, quantified in triplicate using QuantiFluor ONE dsDNA System (Promega Co., Alexandria, NSW, Australia), diluted 1:500 with DNA-free water, and used to generate species-specific sDNA standard curves.^{14, 15, 60}

Unfiltered water samples used for PPLPP workflow in vitro validations (Fig. 4 and Supplemental Figs. S1-S7 and S9) were collected from an ≈800-L tank that contained ≈40 adult O. mossambicus and underwent complete turnover with UV-sterilized water every ≈38 h (TropWATER Facility, James Cook University, Townsville, QLD, Australia: 19° 19′ 39″ S, 146° 45′ 39.24″ E). Unfiltered and filtered water samples used for PPLPP workflow in situ validation (Fig. 5 and Supplemental Fig. S8) were collected from Ross River under Nathan Street bridge (Townsville, QLD, Australia: 19° 18′ 21.96″ S, 146° 45′ 38.52″ E), wherein O. mossambicus, R. marina, and C. caroliniana are known to occur.⁶¹⁻⁶³

FIGURE 4.

PPLPP workflow in vitro validation of glycogen and purification variables using R. marina and C. caroliniana low-copy tank-spike samples. Total O. mossambicus eDNA yield (nanograms ± SEM) is provided for multiple (A) and combined (A’) glycogen concentrations. Recovery of R. marina and C. caroliniana spiked-in gDNA (% ± SEM) is provided for multiple (B and C) and combined (B’ and C’) glycogen concentrations, respectively. Top and bottom letters indicate significant differences (Tukey’s HSD P < 0.05) among and within terminal inhibitor purification methods for each glycogen concentration tested (4.4, 22.2, and 44.4 µg/ml), respectively. Two-way ANOVAs for O. mossambicus (A), R. marina (B), and C. caroliniana (C) determined no glycogen effect within purification methods (bottom letters), whereas subsequent 1-way ANOVAs for O. mossambicus (A’), R. marina (B’), and C. caroliniana (C’) revealed significant species-specific differences among tested inhibitor purification methods. Unpurified: no terminal inhibitor purification; Zymo: OneStep PCR Inhibitor Removal Kit purification (Zymo Research Inc.); silica: silica purification (Supplemental Protocol 3); Qiagen: DNeasy PowerClean Pro Cleanup Kit purification (Qiagen Australia Pty Ltd).

FIGURE 5.

PPLPP workflow in situ validation of Longmire’s preservation effectiveness on multispecies eDNA collected in unfiltered (solid bars) and filtered (hashed bars) water samples from Ross River and subjected to 39 d of incubation at room temperature (RT; white bars) and 50°C (gray bars). Total eDNA yield (nanograms ± SEM; A) was determined for O. mossambicus, R. marina, and C. caroliniana in all treatments using Tilapia_v2_16S, R.marina_16S, and C.caroliniana_matK assays and species-specific gDNA standard curves (Table 4), respectively. Three-way ANOVA revealed no significant differences in eDNA yield among or within target species, capture methods, or incubation treatments (P > 0.3; see In situ validations). Detection rate (%; B) was subsequently determined for each target species (see Statistical analyses).

All unfiltered and filtered water samples were extracted in MEEL within either the low-copy DNA room or a dedicated eDNA laboratory located on a separate floor. Floors and benchtops in both laboratories were cleaned daily with 10% household bleach, water, and 70% ethanol to mitigate contamination risks. No field or extraction blanks exhibited target species amplification in any in vitro or in situ validations. All unfiltered and filtered water samples used for in vitro and in situ validations were collected and processed using 50 and 2 ml DNA LoBind tubes (0030122232 and 0030108078; Eppendorf South Pacific Pty Ltd, Macquarie Park, NSW, Australia), respectively, to ensure minimal loss of small DNA fragments.¹⁸

Quantitative polymerase chain reaction (qPCR)

In vitro validations 3–7 (Fig. 1; Supplemental Figures S1-S5) used initial tilapia assay⁶⁴ in 20 µl reactions with QuantiFast SYBR Green PCR Kit (Qiagen Australia Pty Ltd, Chadstone VIC, Australia), whereas all other in vitro and in situ validations (Figs. 3-5; Supplemental Figures S6-S8) used Tilapia_v2_16S, R.marina_16S, and/or C.caroliniana_matK assays (Table 4) in 10 µl reactions with PowerUP SYBR Green Master Mix (Thermo Fisher Scientific Australia Pty Ltd, Scoresby, VIC, Australia; see Supplemental Information). All in vitro and in situ validations were run with 4–6 technical replicates per biologic replicate in white 96-well plates (Thermo Fisher Scientific Australia Pty Ltd, Scoresby VIC, Australia) on a QuantStudio3 Real-Time PCR System (Thermo Fisher Scientific Australia Pty Ltd, Scoresby, VIC, Australia).

FIGURE 3.

PPLPP workflow in vitro validation of centrifugation, glycogen, and purification variables using R. marina and C. caroliniana low-copy clean-spike samples. Recovery efficiencies for R. marina gDNA (A–D) and C. caroliniana gDNA (E–H) were determined using R.marina_16S and C.caroliniana_matK assays under optimal conditions and species-specific gDNA standard curves for extrapolation (Table 4), respectively. The following PPLPP workflow variables were assessed: 1) initial precipitation centrifugation conducted in swinging-bucket rotor for 90 min (“SB_90 min”) or fixed-angle rotor for 5, 10, or 20 min (“FA_5 min,” “FA_10 min,” or “FA_20 min”; A, D, E, H); 2) low or high in-house glycogen concentration used in initial precipitation (44.4 µg/ml or 133.2 µg/ml; light or dark shading in A, B, E, F); and 3) exclusion or inclusion of terminal inhibitor purification using OneStep PCR Inhibitor Removal Kit (“Unpurified” or “Zymo”; solid or hashed bars in A and E), respectively. Different letter groupings in top and bottom letter rows (A and E) refer to significant differences (Tukey’s HSD P < 0.05) among and within centrifugation treatments, respectively, and different letters (B and F) refer to significant differences among treatments (Tukey’s HSD P < 0.05). Different letters (C and G) refer to a significant pairwise difference (2-tailed Welch’s t test P < 0.05). Treatment legends are provided (A, B, E, F). Bar plots represent mean ± SEM values.

TABLE 4.

Primer information and optimal conditions for Tilapia_v2_16S, R.marina_16S, and C.caroliniana_matK eDNA assays

Assay	Gene region	Annealing (°C)	[Primer] (nM)	Amplicon (bp)	gDNA			sDNA
					Efficiency	LOD (ng/µl)	Tm (°C ± 99.7% CI)	Efficiency	LOD (copies/µl)	Tm (°C ± 99.7% CI)	Oligonucleotide (5′–3′)
Tilapia_v2_16S_F	1273–1290	60	500	101	102.6%a 97.4%b	2.65E−06a 2.81E−06b	82.76 ± 0.69a 83.07 ± 0.52b	96.7%	3	83.02 ± 0.57	AATGTCTTTGGTTGGGGC
Tilapia_v2_16S_R	1377–1396										TTCTGTTGCTTGGAGTTGTA
R.marina_16S_F	30–49	65	250	290	93.3%	6.00E−05	80.79 ± 0.72	93.8%	3	80.05 ± 0.41	AGCCTGCCCAGTGACCATG
R.marina_16S_R	319–338										TGTTATGCTCCGTGGTCACC
C.caroliniana_matK_F	640–664	65	600	265	97.9%	6.86E−04	76.00 ± 0.47	83.9%	3	75.73 ± 0.60	GCTCCTTCTTTACATCTATTGCGAT
C.caroliniana_matK_R	884–903										GGTGCCACTACAAGACGCT

^aO. mossambicus. ^bT. mariae.

In silico validations

Following design of Tilapia_v2_16S, R.marina_16S, and C.caroliniana_matK assays (Table 4), each was validated in silico¹ using both targeted and nontargeted searches of the National Center for Biotechnology Information (NCBI) “nr” database via PrimerBLAST ⁶⁵ (see Supplemental Information).

In vitro validations

Assay efficiency

Tilapia_v2_16S, R.marina_16S, and C.caroliniana_matK amplification efficiencies and limits of detection (LODs) were determined using standard curves generated by serial dilution of both gDNA and sDNA templates. Amplicons produced from gDNA standards were considered positive for target species amplification and representative amplicons from each species-specific gDNA standard curve (n = 2–3) were Sanger sequenced bidirectionally for verification. Amplicons produced from sDNA standards were considered positive for target species amplification without Sanger sequencing confirmation given that sDNA templates were designed to be exact replicas of targeted 16S or matK regions.^{14, 15, 60}

Assay specificity

Assay specificity was validated by attempting to amplify gDNA of nontarget Australian fish, frog, and turtle species with Tilapia_v2_16S and R.marina_16S (Supplemental Table S2) or nontarget Australian aquatic plant species with C.caroliniana_matK (Supplemental Table S3) under optimal assay-specific conditions (Table 4). Any amplicon produced from nontarget species gDNA that exhibited a dissociation temperature (T_m) inside 99.7% confidence interval of target species gDNA standards (Table 4) was considered putatively positive for nontarget cross amplification and verified by bidirectional Sanger sequencing (∆T_m analysis⁶⁶).

Preservatives and terminal precipitants

Unfiltered water samples (30 ml; n = 16) were collected from the O. mossambicus tank (see Sources of genomic, synthetic, and environmental DNA) and preserved with Longmire’s or Queen’s (Table 1) before 9 d of incubation at 40°C and eDNA extraction using the centrifugation-only approach (i.e., no alcohol, salt, or coprecipitate⁶⁷) for initial precipitation and glycogen-aided PEG8000-NaCl or EtOH-NaCl for terminal precipitation (see Supplemental Fig. S1).

Longmire’s short-term preservation effectiveness

Unfiltered water samples (15 ml; n = 18) were collected from the O. mossambicus tank (see Sources of genomic, synthetic, and environmental DNA) and preserved with Longmire’s before constant 40°C incubation for 0, 3, 6, 9, 14, or 21 d and extraction using the initial PPLPP workflow (see Supplemental Fig. S2).

Lysis duration and ±PCI purification

Unfiltered water samples (15 ml; n = 4) were collected from the O. mossambicus tank (see Sources of genomic, synthetic, and environmental DNA) and preserved with Longmire’s before extraction with initial PPLPP using 2 lysis durations each with and without PCI purification (see Supplemental Fig. S3).

NaCl concentration, ±glycogen, and ±PCI purification (high-copy clean-spike)

DNA-free MilliQ (Thermo Fisher Scientific Australia Pty Ltd, Scoresby, VIC, Australia) water samples (15 ml; n = 8) were spiked with O. mossambicus gDNA and preserved with Longmire’s before extraction with the initial PPLPP workflow using 2 NaCl concentrations each with and without glycogen and PCI purification (see Supplemental Fig. S4).

Lysis duration and glycogen concentration test

Unfiltered water samples (15 ml; n = 8) were collected from the O. mossambicus tank (see Sources of genomic, synthetic, and environmental DNA) and preserved with Longmire’s before extraction with the initial PPLPP workflow using 2 glycogen concentrations during initial and terminal precipitations and 2 lysis durations (see Supplemental Fig. S5).

Centrifugation, glycogen, and inhibitor purification (low-copy clean-spike)

DNA-free MilliQ (Thermo Fisher Scientific Australia Pty Ltd, Scoresby, VIC, Australia) water samples (15 ml; n = 8) were spiked with 1 µl of R. marina gDNA (0.174 ng) and 1 µl of C. caroliniana gDNA (0.657 ng) and preserved with 5 ml Longmire’s (“low-copy clean-spike”). Samples were extracted following the optimal PPLPP workflow (without optional PCI purification; Fig. 2) using 44.4 µg/ml (n = 4) or 133.2 µg/ml (n = 4) in-house glycogen (Supplemental Protocol 2) in initial precipitations and 111.1 µg/ml in all terminal precipitations (n = 8). Initial isopropanol precipitations were conducted at room temperature in either a swinging-bucket rotor at 3270 g for 90 min (Allegra ×12R centrifuge with SX4750 rotor; Beckman Coulter Australia Pty Ltd, Lane Cove, NSW, Australia) or fixed-angle rotor at 6750 g for 5, 10, or 20 min (Heraeus Megafuge 8R with HighConicIII rotor; Thermo Fisher Scientific Australia Pty Ltd, Scoresby, VIC, Australia). The final 200 µl elution was divided equally with 100 µl left unpurified and 100 µl inhibitor purified using Zymo before gDNA recovery was determined using species-specific gDNA standard curves (Table 4). All amplicons with ∆T_m inside 99.7% confidence interval of species-specific gDNA standards were considered positive detections without Sanger sequencing confirmation given the known species specificity of spiked-in gDNA (see Sources of genomic, synthetic, and environmental DNA).

FIGURE 2.

Summary of the PPLPP workflow for unfiltered or filtered eDNA samples. Superscripts containing F_x, SF_x, or SP_x refer to corresponding Figure, Supplemental Figure, or Supplemental Protocol number, respectively. Optional PCI purification (step 13; grayed box) uses 1 phenol:chloroform:isoamylalcohol (25:24:1) and 2 chloroform:isoamylalcohol (24:1) phase separations. Hazardous substances are indicated by <!>. TE buffer, Tris-EDTA buffer. ^See Lysis Buffer (LB) I description (Supplemental Protocol 1 and Table 3). ^^See PEG8000-NaCl description (Supplemental Protocol 1 and Table 2). *For filters: vortex to dislodge material, transfer Longmire’s from both 2 ml LoBind tubes containing each filter half into one 50 ml LoBind tube, pulse spin to ensure complete transfer of Longmire’s, dilute combined volume up to 20 ml with DNA-free water, and then transfer both filter halves into one 2 ml LoBind tube (for step 9). **For filters: transfer 600 µl <!> LB < !> into 2 ml LoBind tube containing both filter halves (step 1c) to ensure complete lysis of filter-bound materials. ***For filters: transfer 600 µl <!> LB <!> into new 2 ml LoBind tube, pulse spin tube containing both filter halves, and transfer all residual <!> LB <!> into the same tube before terminal precipitation.

Glycogen and inhibitor purification (low-copy tank-spike)

Unfiltered water samples (15 ml; n = 9) were collected from the O. mossambicus tank (see Sources of genomic, synthetic, and environmental DNA), spiked with 1 µl R. marina gDNA (0.174 ng) and 1 µl C. caroliniana gDNA (0.657 ng), and preserved with 5 ml Longmire’s (“low-copy tank-spike”). Samples were extracted following the PPLPP workflow (without optional PCI purification; Fig. 2) using 4.4 µg/ml (n = 3), 22.2 µg/ml (n = 3), or 44.4 µg/ml (n = 3) in-house glycogen in initial precipitations and 111.1 µg/ml in-house glycogen in all terminal precipitations (n = 9). Initial isopropanol precipitations were conducted at 6750 g for 10 min at 20°C in fixed-angle rotor (Heraeus Megafuge 8R with HighConicIII rotor; Thermo Fisher Scientific Australia Pty Ltd, Scoresby, VIC, Australia). The final 200 µl elution was divided equally 4 ways with one 50-µl aliquot left unpurified and the remaining three 50-µl aliquots inhibitor purified using silica (Supplemental Protocol 3), Zymo, and Qiagen following the manufacturer’s instructions. Recovered eDNA and gDNA yields were determined using species-specific gDNA standard curves (Table 4). All amplicons with ∆T_m inside 99.7% confidence interval of species-specific gDNA standards were considered positive detections without Sanger sequencing confirmation given the known species-specificity of spiked-in gDNA and sole presence of O. mossambicus within TropWATER tank (see Sources of genomic, synthetic, and environmental DNA).

Double inhibitor purification (silica vs. silica-Zymo)

Unfiltered water samples (15 ml; n = 4) were collected from the O. mossambicus tank (see Sources of genomic, synthetic, and environmental DNA) and preserved with Longmire’s before extraction with the PPLPP workflow using silica-only and silica-Zymo double inhibitor purification (see Supplemental Fig. S6).

Longmire’s long-term preservation effectiveness

Unfiltered and filtered water samples (15 ml; n = 10 and 6 respectively) were collected from the O. mossambicus tank (see Sources of genomic, synthetic, and environmental DNA) and preserved with Longmire’s before constant 50°C incubation for 39 d and extraction using the PPLPP workflow (see Supplemental Fig. S7). Of note is that this in vitro validation was conducted concurrently with in situ validation (see In situ validations).

In situ validations

All 3 species-specific assays and the PPLPP workflow were concurrently validated using Longmire’s preserved 15 ml unfiltered (n = 6) and 15 ml filtered (n = 6) water samples, collected from Ross River (see Sources of genomic, synthetic, and environmental DNA) on March 22, 2018 (late wet season), and subjected to 39 d of incubation at room temperature and 50°C. Unfiltered samples were collected by decanting 15 ml from discrete water grabs into a new 50 ml LoBind tube that was preloaded with 5 ml Longmire’s followed by 5 inversions to ensure complete mixture. Filtered samples were collected by passing discrete 15 ml water grabs through presterilized 1.2 µM nylon net membrane filters (Merck Millipore Pty Ltd, Bayswater, VIC, Australia) using the Grover portable peristaltic pump (Grover Scientific Pty Ltd, Hermit Park, QLD, Australia). Each filter was immediately rolled, cut in half with ethanol-sterilized scissors, and preserved in a new 2 ml LoBind tube by submersion in ≈1 ml of preloaded Longmire’s (n = 12 filter halves). All unfiltered and filtered water samples were collected from knee-deep water ≈2 m from the river bank where C. caroliniana sprigs and R. marina tadpoles were observed while no O. mossambicus adults or juveniles were observed. Unfiltered and filter-half samples were incubated upright without agitation on the benchtop at room temperature (n = 3 and n = 6) or in a dry oven (BFD 53; Binder GmbH, Tuttlingen, Germany) set to 50°C (n = 3 and n = 6), respectively. Unfiltered samples were incubated in open-top racks and filtered samples were incubated in closed plastic freezer boxes. Unfiltered samples were submerged to the base of the lid in 10% bleach at incubation commencement and again after 39 d at 50°C to remove any potential DNA contamination on the tube exterior from field collection and circulated air within dry oven, respectively. Unfiltered samples were extracted following the optimal PPLPP workflow (without optional PCI purification; Fig. 2) using 4.4 and 55.5 µg/ml in-house glycogen in initial and terminal precipitations, respectively, with initial isopropanol precipitation conducted at 3270 g for 90 min in swinging-bucket rotor (Allegra ×12R centrifuge with SX4750 rotor; Beckman Coulter Australia Pty Ltd, Lane Cove, NSW, Australia) and Zymo inhibitor purification of the final 100 µl elution. For filtered samples, the Longmire’s used to preserve each filter-half (≈1 ml) was recombined in a new 50 ml LoBind tube, diluted up to 20 ml with DNA-free MilliQ water, and extracted as described above for unfiltered samples, except that following pellet resuspension 600 µl Lysis Buffer I was transferred into one 2 ml LoBind tube that contained both filter halves (see Fig. 2, steps 1c and 9). Recovered eDNA yields were determined using species-specific gDNA standard curves (Table 4). All assays that produced amplicons with ∆T_m inside 99.7% confidence interval of species-specific gDNA standards were considered putative positive detections, and representatives (n = 12 per species) were Sanger sequenced for confirmation.

Unfiltered and filtered Ross River water samples were also assessed for the presence of O. mossambicus using both initial and revised tilapia assays in order to validate that both assays equally detect O. mossambicus eDNA in situ (see Supplemental Fig. S8).

Statistical analyses

Technical qPCR replicates obtained for each biologic replicate were averaged using Microsoft Excel prior to statistical analyses.⁶⁸ All 1-, 2-, and 3-way ANOVAs; variance homogeneity monitoring (Brown-Forsythe); Dunnett’s post hocs; Tukey’s Honest Significant Difference HSD post hocs; Sidak post hocs; parametric 2-tailed Welch’s t tests; nonparametric 2-tailed Bonferroni-corrected Mann-Whitney U tests; linear regressions; and plots were completed in PRISM for Mac OSX (version 7.0d; GraphPad Inc., La Jolla, CA, USA). Alpha was set to 0.05 for all statistical analyses except when specified. Species-specific spike-in recovery rates (% ± SEM) were determined for each replicate of each treatment by dividing the extrapolated species-specific yield (nanograms) by the spiked-in species-specific gDNA quantity (nanograms) and multiplying by 100. Species-specific detection rates (%) were determined for each treatment by dividing the total number of positive detections (∆T_m analysis⁶⁶) by the total number of qPCR technical replicates run and multiplying by 100.

RESULTS

In silico assay validations

Targeted in silico validation for Tilapia_v2_16S and R.marina_16S assays yielded 15 and 5 species with 2–5 and 3–5 total mismatches (Supplemental Tables S4 and S5), respectively, whereas C.caroliniana_matK assay exhibited no species with mismatches (Supplemental Table S6).

Subsequent nontargeted in silico validation for Tilapia_v2_16S assay returned 28 species with no mismatches; however, only O. mossambicus and T. mariae are known to occur in Australia (Supplemental Table S7). Nontargeted in silico validation for the R.marina_16S assay returned 63 frog and toad species with ≤2 mismatches; however, none are native or known to occur within Australia (Supplemental Table S8). Nontargeted in silico validation of C.caroliniana_matK assay did not identify any species with ≤2 mismatches (Supplemental Table S9).

In vitro validations

Assay efficiency

Tilapia_v2_16S, R.marina_16S, and C.caroliniana_matK exhibited satisfactory gDNA and sDNA amplification efficiencies and LODs (Table 4).^{14, 15, 60} Sanger sequencing of O. mossambicus and T. mariae gDNA standards (n = 2 each) confirmed these to be positive detections (≥96.2% and ≥97.1% pairwise identity with GenBank accessions KU500883 and GQ168026, respectively). Sanger sequences of R. marina and C. caroliniana gDNA standards (n = 3 each) also confirmed these to be positive detections (≥98.4% and ≥99.2% pairwise identity with GenBank accessions KF665157 and KY392764, respectively).

Assay specificity

All 3 assays failed to amplify gDNA from empirically tested nontarget species^{14, 15, 60}. Of note is that Tilapia_v2_16S amplified an equal quantity of 16S copies per nanogram of O. mossambicus and T. mariae gDNA (2-tailed Welch’s t test; P = 0.3301)⁵⁹. Moreover, both tilapia assays amplified equal quantities of 16S copies per nanogram of O. mossambicus gDNA (2-tailed Welch’s t test; P = 0.2457) but significantly different quantities of 16S copies per nanogram of T. mariae gDNA (2-tailed Welch’s t test; P = 0.00048)⁵⁹.

Preservatives and terminal precipitants

Detection of O. mossambicus eDNA was 100% across all treatments (Supplemental Fig. S1). Given the lack of significant differences between preservation and precipitation treatments, all subsequent in vitro and in situ validations (Figs. 3, 4, 5 and Supplemental Figs. S2–S8) used Longmire’s instead of Queen’s for eDNA preservation (lower toxicity; see Discussion) and PEG8000-NaCl instead of EtOH-NaCl for terminal precipitation (enhanced inhibitor removal^{17, 45}).

Longmire’s short-term preservation effectiveness

Detection of O. mossambicus eDNA was 100% at days 0, 3, 6, 9, 14, and 21 (Supplemental Fig. S2).

Lysis duration and ±PCI purification

Detection of O. mossambicus eDNA was 100% for all treatments (Supplemental Fig. S3).

NaCl concentration, ±glycogen, and ±PCI purification (high-copy clean-spike)

Detection of O. mossambicus eDNA was 100% for all high-copy tank-spike treatments (Supplemental Fig. S4). Accordingly, all subsequent in vitro and in situ validations (Figs. 3–5 and Supplemental Figs. S5–S8) used 0.2 volumes 5 M NaCl (455 mM final) in initial precipitation; included 4.4 and 55.5 µg/ml glycogen (or higher) in initial and terminal precipitations, respectively; and excluded PCI purification. Of note is that PCI purification was retained as an optional step within the PPLPP workflow (see Fig. 2, step 13).

Lysis duration and glycogen concentration

Detection of O. mossambicus eDNA was 100% for all treatments (Supplemental Fig. S5). Accordingly, all subsequent PPLPP workflow extractions (Figs. 3–5 and Supplemental Figs. S6–S8) utilized glycogen at ≥4.4 and ≥55.5 µg/ml in initial and terminal precipitations, respectively.

Centrifugation, glycogen, and inhibitor purification (low-copy clean-spike)

Detection of R. marina gDNA was 100% for all treatments and gDNA recoveries (mean ± SEM) ranged from 45.5 ± 12.0% to 100.2 ± 7.2% (Fig. 3A). Three-way ANOVA revealed a significant centrifugation effect (F_3,16 = 4.859; P = 0.0137), purification effect (F_1,16 = 10.59; P = 0.0050), and glycogen × purification interaction (F_1,16 = 6.653; P = 0.0202). Tukey’s HSD comparisons within centrifugation treatments revealed that 1) 90 min swinging-bucket centrifugation yielded significantly lower gDNA recovery for unpurified 133.2 µg/ml samples than purified 133.2 µg/ml (P = 0.0029) and purified 44.4 µg/ml (P = 0.0183) samples, 2) 10 min fixed-angle centrifugation yielded significantly lower gDNA recovery for unpurified 133.2 µg/ml samples than unpurified 44.4 µg/ml (P = 0.0374) and purified 44.4 µg/ml (P = 0.0256) samples, and 3) 5 and 20 min fixed-angle centrifugations yielded no significant difference in gDNA recovery regardless of glycogen concentration or purification (P > 0.65). Tukey’s HSD comparisons among centrifugation treatments revealed that 1) purified 44.4 µg/ml samples extracted with 10 min fixed-angle centrifugation had significantly higher recovery than purified 44.4 µg/ml samples extracted with 5 min fixed-angle centrifugation (P = 0.0474) and that 2) unpurified 44.4 µg/ml, unpurified 133.2 µg/ml, and purified 133.2 µg/ml samples did not differ in gDNA recovery regardless of centrifugation condition (P > 0.07). Subsequent 2-way ANOVA (Fig. 3B) revealed a significant purification effect (F_1,28 = 6.483; P = 0.0167) with significant Tukey’s HSD comparisons between unpurified 133.3 µg/ml samples and purified 44.4 µg/ml samples (P = 0.0337) as well as purified 133.2 µg/ml samples (P = 0.0158). This was confirmed by a significant 2-tailed Welch’s t test comparison between merged purification method datasets (t = 2.374, df = 25.56; P = 0.0254; Fig. 3C); however, subsequent 1-way ANOVA of merged centrifugation datasets was nonsignificant (F_3,28 = 2.596; P = 0.0722; Fig. 3D).

Detection of C. caroliniana gDNA was 100% for all treatments, and gDNA recoveries (mean ± SEM) ranged from 14.5 ± 4.2% to 74.9 ± 4.5% (Fig. 3F). Three-way ANOVA revealed a significant centrifugation effect (F_3,16 = 5.67; P = 0.0077), glycogen effect (F_1,16 = 21.77; P = 0.0003), and glycogen × purification interaction (F_1,16 = 21.87; P = 0.0003). Tukey’s HSD comparisons within centrifugation treatments revealed that 1) 90 min swinging-bucket centrifugation yielded significantly lower gDNA recovery for unpurified 133.2 µg/ml samples than purified 133.2 µg/ml (P = 0.0044), unpurified 44.4 µg/ml (P = 0.0036), and purified 44.4 µg/ml (P = 0.0132) samples; 2) 10 min fixed-angle centrifugation yielded significantly higher recovery for unpurified 44.4 µg/ml samples than unpurified 133.2 µg/ml (P = 0.0124) and purified 133.2 µg/ml (P = 0.0430) samples; 3) 20 min fixed-angle centrifugation yielded significantly higher recovery from unpurified 44.4 µg/ml than unpurified 133.2 µg/ml samples (P = 0.0351); and 4) 5 min fixed-angle centrifugation yielded no significant difference in gDNA recovery regardless of glycogen concentration or purification (P > 0.11). Tukey’s HSD comparisons among centrifugation treatments revealed that 1) unpurified 133.2 µg/ml samples had significantly lower recovery when extracted with 90 min swinging-bucket centrifugation compared with 20 min fixed-angle centrifugation (P = 0.0206) and 2) unpurified 44.4 µg/ml, purified 44.4 µg/ml, and purified 133.2 µg/ml samples did not differ in gDNA recovery regardless of centrifugation condition (P > 0.06). Subsequent 2-way ANOVA (Fig. 3F) revealed a significant glycogen effect (F_1,28 = 13.94; P = 0.0009) and glycogen × purification interaction (F_1,28 = 14.01; P = 0.0008) with significant Tukey’s HSD comparisons between unpurified 44.4 µg/ml and unpurified 133.2 µg/ml samples (P < 0.0001), unpurified 133.2 µg/ml and purified 44.4 µg/ml samples (P = 0.0371), and unpurified 133.2 µg/ml and purified 133.2 µg/ml samples (P = 0.0366). This was confirmed by a significant 2-tailed Welch’s t test comparison between merged glycogen level datasets (t = 3.152, df = 29.85; P = 0.0037; Fig. 3G); however, subsequent 1-way ANOVA of merged centrifugation datasets was nonsignificant (F_3,28 = 2.252; P = 0.1042; Fig. 3H).

Given that initial precipitation containing 133.2 µg/ml glycogen required inhibitor purification for maximum yield, the upper glycogen concentration tested in subsequent in vitro low-copy tank-spike validation was 44.4 µg/ml [Fig. 4; see section Glycogen and inhibitor purification (low-copy tank-spike)]. Moreover, given that 10 min centrifugation at 6750 g in fixed-angle rotor provided acceptable R. marina and C. caroliniana gDNA recoveries (Fig. 3A and E respectively), this centrifugation condition was used for the subsequent in vitro validation [see section Glycogen and inhibitor purification (low-copy tank-spike)].

Glycogen and inhibitor purification (low-copy tank-spike)

Detection of O. mossambicus eDNA was 100% for all treatments with yields (mean ± SEM) that ranged from 0.32 ± 0.02 ng to 2.34 ± 1.03 ng (Fig. 4A). Two-way ANOVA demonstrated a significant purification effect on O. mossambicus eDNA yield (F_{3, 24} = 5.521; P = 0.0050; Fig. 4A) and subsequent 1-way ANOVA on merged glycogen treatments confirmed that the terminal inhibitor purification method significantly affected O. mossambicus eDNA yield (F_{3, 32} = 5.673; P = 0.0031; Fig. 4A’). More specifically, samples subjected to terminal inhibitor purification with silica (Supplemental Protocol 3) had significantly lower yield than samples subjected to terminal inhibitor purification with Zymo (P = 0.0014); however, unpurified samples did not differ from samples subjected to terminal inhibitor purification with Zymo, silica, or Qiagen (P > 0.1; Fig. 4A’).

Detection of R. marina gDNA was 100% for all treatments while gDNA recovery (mean ± SEM) ranged from 7.94 ± 0.77% to 55.58 ± 4.87% (Fig. 4B). Two-way ANOVA demonstrated a significant purification effect on gDNA recovery (F_{3, 24} = 87.91; P < 0.0001; Fig. 4B) and subsequent 1-way ANOVA on merged glycogen treatments confirmed that the terminal inhibitor purification method significantly affected gDNA recovery (F_{3, 32} = 78.96; P < 0.0001; Fig. 4B’). More specifically, unpurified samples had significantly lower gDNA recovery than samples subjected to terminal inhibitor purification with Zymo (P < 0.0001) and Qiagen (P = 0.0016) but significantly higher gDNA recovery than samples subjected to terminal inhibitor purification with silica (P < 0.0001; Fig. 4B’). Terminal inhibitor purification with Zymo yielded significantly higher gDNA recovery than terminal inhibitor purification with Qiagen (P = 0.0023; Fig. 4B’).

Detection of C. caroliniana gDNA was 100% for all treatments while gDNA recovery (mean ± SEM) ranged from 19.30 ± 2.98% to 46.34 ± 3.64% (Fig. 4C). Two-way ANOVA demonstrated a significant purification effect (F_{3, 24} = 27.86; P < 0.0001; Fig. 4C) and subsequent 1-way ANOVA on merged glycogen treatments confirmed that the terminal inhibitor purification method significantly affected gDNA recovery (F_{3, 32} = 26.72; P < 0.0001; Fig. 4C’). More specifically, unpurified samples had significantly lower gDNA recovery than samples subjected to terminal inhibitor purification with Zymo (P < 0.0001) and Qiagen (P = 0.0100) but did not differ from samples subjected to terminal inhibitor purification with silica (P = 0.5176; Fig. 4C’). Terminal inhibitor purification with Zymo yielded significantly higher gDNA recovery than terminal inhibitor purification with Qiagen (P = 0.0001; Fig. 4C’).

Double inhibitor purification (silica vs. silica-Zymo)

Detection of O. mossambicus eDNA was 100% for both treatments (Supplemental Fig. S6).

Longmire’s long-term preservation effectiveness

Detection of O. mossambicus eDNA was 100% for all treatments (Supplemental Fig. S7). Of note is that neither unfiltered nor filtered samples differed significantly from baseline (T = 0) following 39 d of room temperature or 50°C incubation.

In situ validations

Species-specific eDNA yields ranged across unfiltered and filtered Ross River samples from least to most as follows (Fig. 5A): O. mossambicus (9.5 × 10⁻⁴ to 1.15 × 10⁻² ng) < R. marina (1.66 × 10⁻³ to 2.41 × 10⁻¹ ng) < C. caroliniana (1.98 × 10⁻² to 20.79 ng). Three-way ANOVA revealed that these differences in eDNA yield were, however, nonsignificant for species (F_2,24 = 1.133; P = 0.3385), collection method (F_1,24 = 0.9138; P = 0.3486), temperature (F_1,24 = 1.039; P = 0.3183), species × collection method (F_2,24 = 0.9199; P = 0.4122), species × temperature (F_2,24 = 1.108; P = 0.3486), collection method × temperature (F_1,24 = 0.9577; P = 0.3375), or species × collection method × temperature (F_2,24 = 0.9463; P = 0.4022). Of note is that 1 filter contained ≈310-times-more C. caroliniana eDNA than the mean (±SEM) of the other 5 filters (20.79 vs. 0.067 ± 0.019 ng). Unfiltered room temperature, unfiltered 50°C, filtered room temperature, and filtered 50°C samples exhibited slight differences in detection rate for O. mossambicus (91.67%, 100%, 100%, and 66.67%) and R. marina (100%, 100%, 100%, and 91.67%) whereas C. caroliniana exhibited 100% detection across all treatments (Fig. 5B). Sanger-sequenced putative positive amplicons aligned with O. mossambicus 16S (GenBank accession KU500883), R. marina 16S (GenBank accession KF665157), and C. caroliniana matK (GenBank accession KY392764) nucleotide sequences with ≥98.5%, ≥98.4%, and ≥99.0% pairwise identity, respectively. Lastly, initial and revised tilapia assays did not differ in O. mossambicus eDNA detectability among in situ samples (P > 0.8; Supplemental Fig. S8).

DISCUSSION

Based on in vitro validations (Figs. 3 and 4 and Supplemental Figs. S1–S7) the optimal PPLPP workflow (Fig. 2) was determined to include 1) preservation in Longmire’s solution at a 1:3 ratio for unfiltered samples or undiluted for filtered samples; 2) initial overnight precipitation at 4°C with 0.8 volumes absolute isopropanol, 0.2 volumes 5 M NaCl, and ≥4.4 µg/ml glycogen (commercial or in-house); 3) centrifugation in a swinging-bucket rotor at 3270 g for 90 min at room temperature or fixed-angle rotor at 6750 g for 10 min at room temperature; 4) pellet resuspension in 600 µl Lysis Buffer I (Table 3); 5) freezing at ≤−20°C for ≥30 min; 6) thawing at room temperature for ≥30 min in the dark; 7) intensely vortexing at 2600 rpm for 30 s (bead beating alternative); 8) lysing at 50°C for ≥3 h under gentle agitation; 9) terminal precipitation overnight at 4°C in 2 volumes PEG8000-NaCl (Table 2) containing ≥55.5 µg/ml glycogen (commercial or in-house); 10) centrifugation in a fixed-angle rotor at 20,000 g for 30 min at 22°C; and 11) inhibitor purification (commercial kit or silica; see Supplemental Protocol 3). Of note is that the PPLPP workflow also applies to unfiltered water samples preserved with EtOH-NaAc,^{5, 13} except that the first 2 steps are combined (i.e., add coprecipitate before direct precipitation of preserved eDNA). Lastly, the PPLPP workflow requires only the following standard molecular laboratory equipment: 1) nonrefrigerated swinging-bucket centrifuge with capability of ≥3270 g or fixed-angle centrifuge(s) with capabilities of ≥6750 g and 20,000 g, 2) rotors with 50 and 2ml capacities, 3) dry oven (≥50°C) with rocker, 4) vortex (≥2600 rpm), 5) refrigerator (≤4°C), 6) freezer (≤−20°C), 7) standard micropipettes (1–1000 µl), and 8) standard chemical preparation (e.g., pH meter, stir plate, balance, and fume arm).

The observed species-specific eDNA yields from in situ validation (Fig. 5A), although not significantly different, were consistent with expectations in that O. mossambicus were not observed during collection but are known to have invaded Ross River (i.e., predicted to be least abundant) whereas R. marina tadpoles were observed near the collection location in the shallows immediately off the river bank (i.e., predicted to be intermediately abundant) and C. caroliniana sprigs were observed in abundance at the collection location ≈2 m from the river bank (i.e., predicted to be most abundant; see In situ validations). The <100% detection rate for O. mossambicus and R. marina (Fig. 5B) was more likely due to incomplete mixing prior to template loading than loss of extracellular eDNA due to 1.2 µM filtration (see Turner et al.⁴³) given the nonsignificant difference in eDNA yield among treatments for both species (Fig. 5A). Degradation of captured eDNA was also an unlikely cause of the observed <100% detection rate for O. mossambicus and R. marina given the demonstrations of Longmire’s short- and long-term preservation effectiveness presented here (Supplemental Figs. S2 and S7, respectively) and elsewhere (Table 1). The ≈310-times-more C. caroliniana eDNA observed for 1 filtered sample (Fig. 5A) was most likely due to the inadvertent collection of tissue fragment(s) within this sample given the abundant presence of sprigs, which are known to reproduce via fragmentation⁶⁹⁻⁷³; however, we did not undertake histologic investigation to determine cell types collected in unfiltered and filtered water samples. Most importantly, Longmire’s preserved Ross River samples extracted using the PPLPP workflow (without optional PCI purification; Fig. 2) confirmed the specificity of Tilapia_v2_16S, R.marina_16S, and C.caroliniana_matK assays for in situ detection of O. mossambicus, R. marina, and C. caroliniana eDNA, respectively. Tilapia_v2_16S was also verified for in situ detection of T. mariae using filtered tropical river water samples that were rescreened using revised tilapia assay following no detections with initial tilapia assay.⁷⁴

In addition to concurrent in situ validation of the PPLPP workflow and all 3 species-specific assays, Queen’s buffer was also demonstrated for the first time to be a viable alternative to Longmire’s for eDNA preservation of unfiltered water samples for up to 9 d at constant 40°C incubation (Supplemental Fig. S1 and Table 1) when used at a 1:3 ratio. Use of Queen’s and Longmire’s at a 1:3 ratio appears to provide equally efficient eDNA preservation in that 100% O. mossambicus detection was observed for both preservatives despite extraction using the uncommon centrifugation-only approach (see Preservatives and terminal precipitants; Woldt et al.⁶⁷). Preservation buffer, which contains 10 times more Tris-HCl and EDTA than Queen’s, has been successfully used for the preservation of eDNA captured on various filter membrane types (Table 1) but has yet to be tested for eDNA preservation effectiveness when mixed directly with unfiltered water samples. However, given the lesser toxicity due to use of SDS instead of N-lauroylsarcosine as anionic surfactant, Longmire’s (Table 1) is recommended for both short-term (up to 21 d at ≤40°C; Supplemental Fig. S2) and longer-term (up to 39 d at ≤ 50°C; Fig. 5 and Supplemental Fig. S7) preservation of unfiltered and filtered water samples containing eDNA. These findings corroborate previous findings regarding Longmire’s eDNA preservation effectiveness (Table 1); however, this study extends the upper limit of Longmire’s preservation effectiveness for unfiltered (1:3 ratio) and filtered (undiluted) samples and, in so doing, suggests suitability for preservation of eDNA samples collected, transported, and/or stored under extreme thermal conditions (e.g., remote tropical study sites; reviewed by Huerlimann et al.⁴).

The PPLPP workflow exhibited in vitro spike-in recovery rates (44–100%; Figs. 3 and 4 and Supplemental Fig. S4) and in situ detection rates (95.8 ± 2.8%, mean ± SEM; Fig. 5) across all 3 target species. In light of these empirical demonstrations of efficiency and effectiveness, as well as previous demonstrations that precipitation-based eDNA extraction methods generally produce higher yields than commercial DNA extraction kits,^{1, 3, 17} the PPLPP workflow was not compared with a commercial DNA extraction kit (e.g., Qiagen DNeasy Blood and Tissue Kit, Qiagen Dneasy mericon Food Kit, QIAamp DNA Micro Kit, MagMAX-96 AI/ND Viral RNA Extraction Kit, PowerBiofilm DNA Isolation Kit, or MolBio PowerWater DNA Isolation Kit). Instead, 3 inhibitor purification methods (silica, Zymo, and Qiagen) were directly compared with unpurified samples in vitro for all 3 target species (Fig. 4), given the influence that inhibitor purification method can have on final eDNA recovery⁵² and the general lack of comparisons between inhibitor removal methods to date.⁷³ Moreover, given the significant increase in eDNA yield that was recently demonstrated for frozen filters extracted with the Mu-DNA protocol compared with Qiagen DNeasy Blood and Tissue Kit extractions,⁴¹ future studies should consider comparing PPLPP and Mu-DNA workflows for target species and/or community detectability using Longmire’s preserved filter membranes collected from various environmental water sources (e.g., freshwater vs. marine, lentic vs. lotic, clear vs. turbid, temperate vs. tropical).

A meaningful comparison can, however, be made between observed O. mossambicus eDNA yields from 3 PPLPP in vitro validations (Fig. 4 unpurified only, Supplemental Figs. S5 and S7 unpurified T = 0 only) and O. mossambicus eDNA yields obtained from similar in vitro validations using a benchmark extraction method (Robson et al.⁶⁴; Table 2). Note that Supplemental Fig. S3 (-PCI only) data were excluded from the meta-analysis because O. mossambicus eDNA yields from this in vitro validation were significantly higher (Tukey’s HSD P < 0.0001) than the 3 other relevant PPLPP in vitro validations (ANOVA F_3,21 = 99.31; P < 0.0001). Note also that this comparison confirms the nonsignificant (Tukey’s HSD P > 0.11) influence of qPCR master mix (Qiagen QuantiFast vs. Thermo Fisher PowerUp) on O. mossambicus eDNA yield (Supplemental Fig. S5 vs. Fig. 4 and Supplemental Fig. S7), respectively. All 3 relevant PPLPP in vitro validations collected 15 ml unfiltered water samples in late winter and late summer from an outdoor, shaded, and recirculating tank that contained ≈1 O. mossambicus per 20 L (see Sources of genomic, synthetic, and environmental DNA). Similarly, the in vitro validation by Robson et al.⁶⁴ collected 15 ml unfiltered water samples from 23, 29, and 35°C non-recirculating 60 L tanks that contained 3 O. mossambicus each following 5 d of eDNA accumulation. All PPLPP and Robson et al.⁶⁴ samples were extracted without PCI purification in the same laboratory (MEEL) using swinging-bucket or equivalent fixed-angle centrifugation (Fig. 3 and Table 2). Robson et al.⁶⁴ utilized the benchmark EtOH-NaAc precipitation followed by commercial spin-columns (Isolate II Genomic DNA; Bioline Inc., Taunton, MA, USA; Table 2). PPLPP samples quantified total O. mossambicus eDNA yield without inhibitor purification using either initial or revised tilapia assays (40 cycles, 60°C, 500 nM each primer, efficiency range 93–103%) while Robson et al.⁶⁴ similarly quantified total O. mossambicus eDNA yield without inhibitor purification using initial tilapia assay (40 cycles, 60°C, 260 nM each primer, efficiency range 94–121%). Comparison of total O. mossambicus eDNA yield revealed that the minimum PPLPP yield (0.48 ng) was greater than the maximum yield reported by Robson et al.⁶⁴ (Supplemental Fig. S9A). Subsequent comparisons to Robson et al.⁶⁴ 35°C treatment yields revealed that PPLPP workflow extracted samples contained ≈3- to 6-fold more O. mossambicus eDNA (Supplemental Fig. S9B). This comparison was made against the Robson, et al.⁶³ 35°C treatment only to ensure conservativeness in that the tank sampled for PPLPP in vitro validations was not temperature controlled (i.e., most conservative to assume 35°C). As such, the improved O. mossambicus eDNA yield observed for PPLPP workflow extracted samples was more likely due to improved extraction efficiency of the PPLPP workflow (e.g., inclusion of coprecipitate) than due to the collection of more O. mossambicus eDNA (i.e., PPLPP samples collected from recirculating tank whereas Robson, et al.⁶³ samples collected from non-recirculating tanks; see above) or other methodological differences (e.g., centrifugation or qPCR conditions; see above.

PCI purification of eDNA samples after lysis and before terminal precipitation was included as an optional step in the PPLPP workflow (see Fig. 2, step 13) given the common use of PCI purification at this extraction step.^{20, 66, 75−77} However, given the increase in time, cost, and toxicity associated with PCI purification³ and the potential for extracellular DNA loss (Supplemental Fig. S4; ²⁵), the PPLPP workflow was tested with (Supplemental Figs. S1–S4) and without (Figs. 3–5 and Supplemental Figs. S5–S8) PCI purification after lysis with Lysis Buffer I (Table 3) and before terminal PEG8000-NaCl precipitation (Table 2). Most notably, the positive in situ detections obtained for all 3 target species when PCI purification was excluded demonstrated that additional purification was not essential for PPLPP workflow effectiveness (Fig. 5); however, future studies should empirically determine whether PPLPP workflow effectiveness differs when unfiltered and/or filtered in situ water samples are extracted with and without additional PCI purification given the recent demonstration that inclusion of phenol during eDNA extraction improved both target species and metagenomic datasets.²⁰

PPLPP workflow centrifugation can be performed using either room temperature swinging-bucket (90 min at 3270 g) or fixed-angle (5–20 min at 6750 g or 5 min at 12,104 g) centrifugation. Of note is that 5 min at 12,104 g is not recommended because 50 ml LoBind tubes exhibited signs of physical stress (e.g., etching) when subjected to this centrifugation condition. More specifically, the nonsignificant difference between 90 min of swinging-bucket centrifugation and 5, 10, or 20 min of fixed-angle centrifugations for both R. marina and C. caroliniana gDNA recoveries (Figs. 2H and 3D), respectively, confirmed that the utilized 10 min fixed-angle centrifugation (Fig. 4 and Supplemental Fig. S6) was equivalent to 90 min swinging-bucket centrifugation (Fig. 5 and Supplemental Figs. S1–S5, S7 and S8). PPLPP workflow extractions undertaken by future eDNA studies should utilize one of these spike-in recovery validated centrifugation conditions, depending on available equipment, given that this aspect of eDNA extraction tends to be unstandardized across studies (Table 2).

The linear relationship between microliters of 20 mg/ml commercial glycogen added and O. mossambicus eDNA yield from tank water (see Sources of genomic, synthetic, and environmental DNA) demonstrated that 4.4 and 55.5 µg/ml glycogen are the minimum concentrations required for efficient initial and terminal precipitations, respectively (Supplemental Fig. S4C). Of note is that addition of ≤44.4 µg/ml glycogen to initial precipitation does not appear to directly inhibit qPCR (i.e., inhibitor purification not essential), whereas addition of >44.4 µg/ml glycogen to initial precipitation does appear to require inhibitor purification for optimal efficiency (Fig. 3) despite this concentration being ≈90-fold lower than previously reported glycogen inhibition of enzymatic reactions (>4 mg/ml⁷⁸). Higher glycogen concentrations added to initial and terminal precipitations (e.g., 22.2–133.2 and 111.1 µg/ml; Figs. 3 and 4) did not significantly increase O. mossambicus eDNA yield or R. marina and C. caroliniana gDNA recovery; however, we recommend that future eDNA studies that target other species and/or aquatic environments undertake in vitro and in situ validations of glycogen concentrations within or above these ranges (4.4–133.2 and 55.5–111.1 µg/ml in initial and terminal precipitations, respectively) to ensure optimal study-specific extraction efficiency. Lastly, given the cost of commercial DNase-treated glycogen when used at minimum recommended initial and terminal precipitation concentrations of 4.4 and 55.5 µg/ml ($3.27 per sample; 10901393001; Sigma-Aldrich Pty Ltd, Macquarie Park, NSW, Australia; Fig. 5 and Supplemental Figs. S2–S8), respectively, we recommend use of in-house DNase-treated glycogen (Supplemental Protocol 2) to decrease cost at these minimal concentrations ($0.02 per sample) and, in so doing, financially permit use of potentially advantageous higher concentrations (e.g., $0.09 per sample at 22.2 and 111.1 µg/ml). Future eDNA studies are encouraged to include DNase-treated glycogen (commercial or in-house), LPA,²⁶ or tRNA²⁵ as a coprecipitate whenever Longmire’s preserved unfiltered or filtered water samples are precipitated given the lysis action of this preservative (i.e., minimize loss of dissolved extracellular DNA^{34, 36}; see Introduction). Equal amounts of DNA-free glycogen and LPA (20 µg each) have been demonstrated to provide equal spike-in recovery efficiency for precipitated DNA²⁶; however, DNA recovery and cost-effectiveness comparisons between various glycogen, LPA, and tRNA concentrations have yet to be undertaken for unfiltered or filtered eDNA samples collected in situ.

Zymo inhibitor purification increased and decreased gDNA recovery efficiency from low-copy clean-spike samples for R. marina and C. caroliniana by 1.9% and 19.45% compared with unpurified samples (Figs. 2E and 3A), respectively. However, reduction in R. marina and C. caroliniana recovery between clean-spike and tank-spike samples was notably less for Zymo purified samples (≈45% and ≈2%) than unpurified samples (≈72% and ≈47%), respectively. This suggests that Zymo inhibitor purification effectively but incompletely removes environmental inhibitors in that R. marina and C. caroliniana still exhibited 100% detection despite ≈1 and ≈0.1 higher threshold cycles for tank-spike compared with clean-spike samples, respectively. Despite the common use of Zymo inhibitor purification, future studies that collect unfiltered or filtered freshwater in situ samples should consider undertaking direct comparisons of Zymo and Qiagen inhibitor purification kits given the recent demonstration that Zymo was not as effective as Qiagen for inhibitor removal from filtered marine in situ samples²¹. Alternatively, future eDNA studies that use Zymo inhibitor purification should consider adjusting qPCR cycling conditions to include ≥2 additional cycles to help avoid false-negative detections due to incomplete removal of coextracted environmental inhibitors.

Silica inhibitor purification was developed as a cost-effective (≤$1 per sample) eDNA-specific hybrid of 2 previous protocols (Li et al.⁵⁷ and Huanca-Mamani et al.⁵⁹; see Supplemental Protocol 3) to provide an alternative to more expensive commercial inhibitor purification kits (e.g., ≈$4 and ≈$6 per sample for Zymo and Qiagen inhibitor purifications, respectively). Despite diligence during protocol development, aliquots that were inhibitor purified with silica purification had significantly lower eDNA yield or gDNA spike-in recovery compared with aliquots that were inhibitor purified with Zymo (Fig. 4). These observed eDNA yield reductions could be due to downstream qPCR inhibition from silica carryover,^{26, 56} which is supported by the significant increase in O. mossambicus eDNA yield observed following silica-Zymo double inhibitor purification compared with silica-only inhibitor purification [Supplemental Fig. S6; see Double inhibitor purification (silica vs. silica-Zymo)]. Therefore, eDNA was not lost during silica inhibitor purification (detection rate still 100%), but rather samples were likely contaminated with silica carryover that inhibited downstream qPCR amplification by ≈1 cycle or ≈50%. As such, future use of silica inhibitor purification should consider subsequent purification with Zymo if targeting low-copy target species in which ≈1 cycle shift could result in a false-negative detection, as the combined cost and time of silica-Zymo is lower than Qiagen inhibitor purification. However, it remains to be tested whether silica-Zymo, silica-Qiagen, or Zymo-Qiagen double purification can provide more effective removal of abundant or resilient (i.e., coextracted) inhibitors from challenging (e.g., turbid) water bodies than silica, Zymo, or Qiagen can achieve independently. Adjusting qPCR cycling conditions to include ≥2 additional cycles could provide a cost-effective alternative for presence/absence or other nonquantitative studies to help avoid false-negative detections due to retained or imposed inhibition.

Although not directly comparable across the clean-spike and tank-spike samples, Qiagen inhibitor purification also increased recovery efficiency significantly compared with unpurified in vitro samples, suggesting this inhibitor purification method is effective. However, the significantly lower yield obtained from unfiltered in vitro samples inhibitor purified with Qiagen compared with Zymo appears to be more likely driven by elevated eDNA loss than less efficient inhibitor removal given the superior inhibitor removal performance observed following both Qiagen and Zymo inhibitor purification of filtered marine in situ samples.²¹ Loss of eDNA during Qiagen inhibitor purification, like silica inhibitor purification (Fig. 4), could be a consequence of the lengthier and more complex protocol (15 steps including pelleting, membrane-binding, washing, and eluting) than Zymo inhibitor purification (1 step: load and spin). We recommend that future eDNA studies targeting freshwater ecosystems undertake subsequent in situ comparisons of Qiagen and Zymo inhibitor purifications to verify which performs better for unfiltered and/or filtered water samples given the potential for these to contain different inhibitor types and loads than present in filtered marine in situ samples.²¹

Lastly, direct comparisons between silica, Zymo, Qiagen, Mu-DNA,⁴¹ and other less commonly utilized commercial inhibitor purification kits (e.g., Norgen Biotek CleanAll RNA/DNA Clean-Up and Concentration Kit and MoBio Laboratories PowerClean DNA Clean-Up Kit; Lever et al.¹⁷) should be incorporated into future eDNA studies given that the most effective inhibitor removal approach across capture methods (unfiltered vs. filtered) used to sample environmental water bodies (e.g., freshwater vs. marine) that contain unique inhibitor types and loads^{11, 45, 46} remains to be comprehensively elucidated.

Supplementary Material

This article includes supplemental data. Please visit http://jbt.abrf.org/ to obtain this information.