Skip to main content

Propylene glycol-based antifreeze is an effective preservative for DNA metabarcoding of benthic arthropods


Preservation of DNA in bulk environmental samples is conventionally achieved using ethanol; however, transportation restrictions on ethanol, particularly from remote locations, are problematic, and ethanol requires a lengthy evaporation period to avoid polymerase chain reaction inhibition. We examined the efficacy of an easily accessible, non-toxic, propylene glycol-based antifreeze as an alternative to molecular-grade ethanol for preserving macroinvertebrate DNA from bulk-benthos DNA samples. We used 2 processing methods (no evaporation of preservative vs full evaporation) to test the differences in both cytochrome oxidase I (COI) exact sequence variants (ESVs) and COI taxonomic orders detected in both ethanol- and antifreeze-preserved samples. Our results suggest that antifreeze is a suitable alternative to ethanol for preservation of DNA in freshly collected samples (e.g., up to 3 d) because of the comparable ESV richness detected in antifreeze-preserved samples. We have demonstrated that by using antifreeze, it is possible to achieve sufficient taxonomic coverage and assess macroinvertebrate assemblages within bulk-benthos DNA samples. The application of this non-regulated preservative is particularly important for remote sampling (i.e., only air accessible) and sampling for community-based biomonitoring projects within Indigenous territories where alcohol is prohibited or not available.

Increasingly, DNA-based techniques are being used to detect and monitor biodiversity across a variety of ecosystems (Baird and Hajibabaei 2012, Taberlet et al. 2012, 2018, Thomsen and Willerslev 2015, Hajibabaei et al. 2016). DNA metabarcoding is a prominent approach that involves sequencing taxonomic marker genes, such as DNA barcodes, from bulk environmental samples, such as benthos (Taberlet et al. 2012,Goldberg et al. 2015, Pedersen et al. 2015, Thomsen and Willerslev 2015, Minamoto et al. 2016). Effective preservatives are required to prevent DNA degradation of environmental samples, and optimal DNA preservation methods vary depending on the type of sample and target taxa (Barnes and Turner 2016, Koziol et al. 2019, Sales et al. 2019). Maintaining the integrity of DNA from source to laboratory is particularly important for all DNA-based research and especially so for community-based monitoring programs (Steininger et al. 2015). Often, the biggest concern with community-based monitoring initiatives is the quality and consistency of data collection (Kelling et al. 2015, Kosmala et al. 2016), and this concern is no less relevant for DNA-based monitoring.

Environmental samples for DNA-based research are typically stored in absolute ethanol (Hajibabaei et al. 2012, Steininger et al. 2015, Minamoto et al. 2016, Koziol et al. 2019, Sales et al. 2019), but this preservative can create a variety of challenges. Ethanol is classified as a hazardous chemical (flammable and harmful), and elaborate and costly arrangements for storage and transit of environmental samples preserved in ethanol are required, particularly for air transportation (Williams 2007, Steininger et al. 2015). Air-transportation regulations create problems when attempting to collect environmental samples from remote locations within a country (i.e., air access only) and from locations overseas. Another issue with relying on ethanol as a preservative, apart from the cost of the chemical (Steininger et al. 2015), is the current bylaws regarding prohibition of alcohol in Indigenous territories across countries such as Canada, the United States, New Zealand, and Australia (Brady 2000, Clough and Bird 2015, Campbell 2016, Clifford and Shakeshaft 2017). Additionally, processing ethanol-preserved environmental samples requires a lengthy (~4–8 h) evaporation period prior to DNA extraction because of the polymerase chain reaction (PCR) inhibitory nature of ethanol in concentrations >1% (Demeke and Jenkins 2010, Schrader et al. 2012). This evaporation step substantially increases the laboratory processing time of ethanol-preserved environmental samples (Demeke and Jenkins 2010, Schrader et al. 2012). Ethanol regulations and availability coupled with processing time can result in limited locations for which biodiversity can be assessed using DNA metabarcoding, which ultimately results in geographic deficiencies in data (Pereira and Cooper 2006, McGee et al. 2019).

Biodiversity data deficiencies are particularly common in remote locations, often within Indigenous territories (Lele et al. 2010, World Wildlife Fund Canada [WWF Canada] 2017, Altman et al. 2018, Monsalve-Cuartas et al. 2019). DNA metabarcoding is a proven effective tool for generating large quantities of data on environmental health (Baird and Hajibabaei 2012, Fahner et al. 2016, Deiner et al. 2017, Porter and Hajibabaei 2018a) and can provide the answer for addressing data deficiencies if the issues with ethanol can be solved. Across Canada, a large proportion of watersheds are currently classified as data deficient (WWF Canada 2017). Out of 167 subwatersheds, 110 currently lack the data necessary to determine baseline health status (WWF Canada 2017). To combat this data deficiency, the Sequencing the Rivers for Environmental Assessment and Monitoring (STREAM) project was established. STREAM combines community-based monitoring with DNA metabarcoding and assesses the benthic macroinvertebrate assemblages in watersheds (particularly those with data deficient status) across Canada. STREAM’s collaborative approach brings together academic (University of Guelph) and non-academic partners (WWF Canada, Living Lakes Canada, and Environmental and Climate Change Canada) to facilitate community training and genomic analyses and create a baseline of watershed health. For that project, a number of Indigenous groups participate in sample collection in target watersheds and sub-watersheds that exist in remote locations. To achieve the desired community involvement and spatial coverage of samples, an effective, safe alternative to ethanol for preservation needs to be determined.

Previously, Steininger et al. (2015) compared the DNA-preserving ability of easily accessible glycol-based antifreeze with the preserving ability of ethanol for short-term preservation of invertebrate DNA. They determined that antifreeze was equally as effective at preserving DNA, even after 7 d under direct outdoor exposure (Steininger et al. 2015). That study, however, focused on the preservation of individual specimens as opposed to mixed-community or bulk environmental samples (Steininger et al. 2015). Similarly, other studies have investigated the relative suitability of a range of other preservatives on the DNA filters from solely water-derived environmental samples, including sodium acetate–ethanol solution (Ladell et al. 2019), Longmire’s solution (Williams et al. 2016), and cationic surfactant (Yamanaka et al. 2017). However, there have been no studies to test the efficacy of glycol-based antifreeze for preserving benthic macroinvertebrate DNA from environmental samples intended for DNA metabarcoding.

The primary aim of this study, therefore, was to compare the sequence diversity and derived taxonomic composition of benthic macroinvertebrate DNA from samples preserved in commonly used molecular-grade ethanol (>99%) vs samples preserved in commercially available propylene glycol-based antifreeze (Absolute Zëro™ RV Waterline Antifreeze; Recochem Inc., Montreal, Quebec, Canada). We hypothesized that median richness across sites from antifreeze-preserved samples would be > median richness across sites from ethanol-preserved samples. Second, we aimed to test the effect of 2 laboratory processing methods (no evaporation of preservative prior to DNA extraction vs full evaporation prior to DNA extraction) to determine if the evaporation step could be eliminated as a time-saving measure. We hypothesized that for antifreeze-preserved samples, there would be no detectable difference in benthic macroinvertebrate taxa richness between evaporated and unevaporated samples. This study will demonstrate how DNA-based biomonitoring initiatives can overcome obstacles to using ethanol-based preservatives.


Sample design and study sites

To explore the impacts of preservative type (ethanol vs antifreeze) on DNA preservation, we examined 2 types of sample: validation and test samples. We collected validation samples from riffle locations at 3 sites within tributaries of the Grand River (Waterloo, Ontario, Canada; Table S1, Fig. S1) to test the effect of ethanol vs antifreeze on the preservation of macroinvertebrate DNA in samples homogenized (in the laboratory) prior to the addition of either preservative (Fig. 1A). Test samples consisted of 2 benthos samples collected within close proximity to one another in riffle locations at 3 additional sites within tributaries of the Grand River (Figs 1B, S1). We collected test samples to simulate the effect of using ethanol or antifreeze to preserve bulk-benthos samples during shipping, including a brief (3 d) incubation period at room temperature. This incubation period was designed to mimic the approximate time samples would spend in postal transit from study site to laboratory. Test samples were preserved prior to homogenization (Fig. 1B). Natural variation between closely sampled ethanol- and antifreeze-preserved samples was expected. Validation samples provide the best insight into the effects of preservative on DNA preservation because we can directly compare homogenates. Test sampling provides a more realistic assessment because preservation typically occurs prior to homogenization.

Figure 1. 
Figure 1. 

Sampling design for both validation sampling (i.e., single sample collected/site and homogenized prior to preservation) (A) and test sampling (i.e., 2 samples collected/site and preserved prior to homogenization) (B).

To collect all samples, we used a standard Canadian Aquatic Biomonitoring Network kick net with a 400-µm-mesh net attached to a pole and net frame in a 10-m transect for 3 min/sample (ECCC 2012). After collecting samples, we drained the net, removed the benthic material, and placed the material in a sterile, white, 1-L polyethylene sample jar, which we filled no more than half full. We rinsed the net and inspected it to remove any remaining invertebrates. To avoid DNA carryover between sites, we used sterile nets to collect samples at each site, and field crew members wore clean nitrile gloves to collect and handle samples in the field and laboratory. We placed the samples (n = 9) on ice in a cooler for transport to the laboratory the same day.

Sample homogenization and DNA extraction

For the validation samples, we produced a crude homogenate by adding 300 mL of molecular-grade water and blending the sample in a standard blender (model BL2010BGC; Black and Decker™, Towson, Maryland). The blender had been previously decontaminated and sterilized with ELIMINase® (Decon Labs, King of Prussia, Pennsylvania) and then rinsed with deionized water and UV treatment for 30 min. After homogenization, we split the homogenate into 2 sample jars and added ethanol to one sample jar (to a final concentration of ~70%) and ~350 mL of antifreeze to the other jar (Fig. 1A). Validation samples remained on the laboratory bench at room temperature for 3 d. For the test samples, after the addition of either ethanol or antifreeze, samples were incubated on the laboratory bench at room temperature for 3 d before a crude homogenate was produced in the same manner as validation samples (Fig. 1B). For both validation and test samples, we then transferred subsamples of each homogenate to individual 50-mL Falcon® tubes (Corning, Glendale, Arizona) and centrifuged them at 2400 rpm for 2 min to pellet the tissue. We removed the supernatant and tested 2 methods of sample processing. For validation samples, we evaporated 3 antifreeze-preserved subsamples from each of the 3 sites (total of 9 subsamples) and 3 ethanol-preserved subsamples from each of the 3 sites (total of 9 subsamples) prior to DNA extraction (i.e., standard protocol). We also processed 3 antifreeze-preserved subsamples from each of the 3 sites (total of 9 subsamples) for DNA extraction without evaporating excess preservative from the pellet (Fig. 1A). For test samples, we evaporated 3 ethanol-preserved subsamples and 3 antifreeze-preserved subsamples/site (9 subsamples/preservation method) prior to DNA extraction. We processed 1 ethanol-preserved subsample and 3 antifreeze-preserved subsamples from each of the 3 sites (total of 9 subsamples) for DNA extraction without evaporating excess preservative from the pellet (Fig. 1B). The inhibiting effect of ethanol on DNA extraction is already known; therefore, the unevaporated ethanol-preserved subsamples were controls. We dried evaporated subsamples at 70°C until the preservative was fully evaporated (~4–8 h). Using a sterile spatula, we subsampled ~300 mg dry mass of homogenate into PowerBead tubes (Qiagen®, Hilden, Germany) containing garnet. We stored the remaining dry mass in the Falcon tubes at –20°C as a voucher.

We used a DNeasy® PowerSoil® kit (Qiagen) to extract DNA following manufacturer’s instructions. DNA was eluted in a final volume of 50uL of buffer C6. We used a NanoDrop™ spectrophotometer (Thermo Fisher Scientific™, Waltham, Massachusetts) to check purity and concentration of DNA for each site. We kept samples at –20°C for further PCR and sequencing. Each batch of DNA extractions included a negative control with no tissue. To verify the purity of the source preservatives, we used a DNeasy PowerWater® kit (Qiagen) to filter and extract 50 mL each of the source ethanol and antifreeze according to manufacturer’s protocols.

Library preparation and high-throughput sequencing

To compensate for known primer bias that results from relying on a single primer set (Hajibabaei et al. 2019), we amplified 3 fragments within the standard COI DNA barcode region with the following primer sets (B/ArR5 [~310 bp] called BR5, LCO1490/230_R [~230 bp] called F230R, and mICOIintF/jgHCO2198 [~313 bp] called ml-jg; Hajibabaei et al. 2012, Geller et al. 2013, Leray et al. 2013, Gibson et al. 2014) using a 2-step PCR amplification regime. These primer sets are specifically designed to target taxa from arthropods (BR5 and F230R) and to detect a broad range of metazoans (ml-jg). The 1st PCR used COI-specific primers, and the 2nd PCR involved Illumina®-tailed (Illumina, San Diego, California) primers. We assembled the PCR reactions in 25-μL volumes. Each reaction contained 2 μL of DNA template, 17.5 μL of molecular-biology-grade water, 2.5 μL of 10× reaction buffer (200 mM Tris–HCl, 500 mM KCl, pH 8.4), 1 μL of MgCl2 (50 mM), 0.5 μL of dNTPs mix (10 mM), 0.5 μL of forward primer (10 mM), 0.5 μL of reverse primer (10 mM), and 0.5 μL of Invitrogen™ Platinum™ Taq polymerase (model 5 U; Thermo Fisher Scientific). We initiated the PCR conditions with a heated lid at 95°C for 5 min, followed by a total of 35 cycles of 94°C for 40 s, 46°C for 1 min, 72°C for 30 s, a final extension at 72°C for 5 min, and hold at 4°C. We used the amplicons from the 1st PCR as templates in the 2nd PCR, and we used the same amplification condition from the 1st PCR in the 2nd, with the exception of using Illumina-tailed primers in a 35-cycle amplification regime. All PCRs were done using a Mastercycler® ep Gradient S thermal cycler (model MC Pro S; Eppendorf®, Hamburg, Germany). We included negative control reactions (no DNA template) for antifreeze, ethanol, DNA extraction kit, and the PCRs in all experiments.

We visualized PCR products on a 1.5% agarose gel to check the amplification success. We purified amplicons using a MinElute® PCR Purification Kit (Qiagen), eluting with 15 uL of molecular-grade water. We used a QuantIT™ PicoGreen™ dsDNA assay kit (Thermo Fisher Scientific) to quantify the purified amplicon samples from the 2nd PCR and normalized them to the same concentration based on those values. We pooled equimolar amounts of each amplicon from the same sample into the same tube prior to dual indexing. We then used a Nextera™ XT DNA library preparation kit #FC-131-1002 (Illumina) to dual index samples and pooled them into a single tube. We used AMpure magnetic beads (Beckman Coulter™, Brea, California) to purify the pooled library, quantified it using a QuantIT™ PicoGreen™ dsDNA assay (Thermo Fisher Sientific), and used the Agilent™ Bioanalyzer 2100 (model 2939A; Agilent Technologies, Santa Clara, California) to determine fragment length. The purified library was diluted based on the concentration and average fragment length and sequenced on a MiSeq™ (Thermo Fisher Scientific) using a V3 MiSeq sequencing kit #MS-102-2003 (300 × 2).

Bioinformatic processing

To assess the taxonomic composition in each sample, we processed the raw Illumina paired-end reads using the SCVUC v4 pipeline available from Briefly, we paired raw reads with SeqPrep (version 1.3.2; St John 2016) setting a minimum Phred score of 20 and minimum overlap of at least 25 bp. We trimmed primers with CUTADAPT (version 2.6; Martin 2011) ensuring a minimum trimmed fragment length of at least 150 bp, a minimum Phred score of 20 at the ends, and allowing a maximum of 3 Ns. We concatenated all primer-trimmed reads from individual sample files into a single file for a global exact sequence variant (ESV) analysis. We dereplicated reads with VSEARCH (version 2.14.1; Rognes et al. 2016) using the derep_fulllength command and the sizein and sizeout options. We performed denoising using the unoise3 algorithm implementation in VSEARCH with the cluster_unoise command (Edgar 2016). This method corrects sequences with potential errors and removes rare reads. Here we defined rare reads to be ESVs containing only 1 or 2 reads (singletons and doubletons; Callahan et al. 2017). We removed putative chimeric reads in a separate step using the uchime3 algorithm implementation in VSEARCH with the uchime3_denovo command. We identified and removed putative pseudogenes by translating denoised ESVs into every possible reading frame (plus strand) using ORFfinder (version 0.4.3; NCBIRC 2018), keeping the longest open reading frame, calculating cutoffs to identify open reading frames with outlier lengths (shorter than the 25th percentile – 1.5× interquartile length or longer than the 75th percentile + 1.5× interquartile length). We used the mitochondrial invertebrate genetic code for translation, and all subsequent analyses were limited to Arthropoda. We created an ESV × sample table with VSEARCH using the usearch_global command, mapping reads to ESVs with 100% identity. We taxonomically assigned ESVs by using the COI Classifier version 4, which includes a curated reference sequence set mined from BOLD and GenBank, and which uses the RDP classifier version 2.12 that uses a naive Bayesian algorithm (Wang et al. 2007).

Statistical analyses

To investigate similarities and differences between the assemblages detected in ethanol- and antifreeze-preserved samples, we conducted diversity analyses in Rstudio (version 1.1.456) using R (version 3.5.1; R Project for Statistical Computing, Vienna, Austria) with the vegan package (Oksanen et al. 2019). To assess sequencing depth for both preservative types, we plotted rarefaction curves using the rarecurve function. Unless otherwise stated, read depth/sample was normalized to the 15th percentile library size using the rrarefy function. Rarefaction adjusts for differences in library sizes across samples to aid comparisons of alpha diversity (Willis 2019). We assessed the recovery of ESVs from ethanol samples compared with antifreeze samples and assessed the proportion of all ESVs that could be taxonomically assigned with high confidence. Taxonomic assignments were deemed to have high confidence if they had the following bootstrap support cutoffs: species ≥0.70 (95% correct), genus ≥0.30 (99% correct), and family ≥0.20 (99% correct), as is recommended for 200-bp fragments when using the COI Classifier (Porter and Hajibabaei 2018b). Assignments to more inclusive ranks (e.g., order) do not require a bootstrap support cutoff to ensure that 99% of assignments are correct, with the assumption that the query sequence is present in the reference sequence database.

We calculated ESV richness for different groups of the data to compare assemblages among sites and preservation methods (ethanol- or antifreeze-preserved samples). To test for differences, we first checked the normality of the data using visual methods with the ggdensity and ggqqplot functions from the ggpubr package (Kassambara 2020) in R followed by testing for normality using the Shapiro–Wilk test (Shapiro and Wilk 1965). Because our data were not clearly normally distributed (W = 0.9631, p = 0.049), we used a conservative approach in subsequent analysis. We used a paired Wilcoxon test (Wilcoxon 1945) to test if median richness across sites from antifreeze-preserved samples was greater than median richness across sites from ethanol-preserved samples.

Because of the COI primer design specificity, we focused on just the Arthropoda to compare assemblages. Additionally, Arthropoda contains the Ephemeroptera, Plecoptera, and Trichoptera (EPT) orders, which are typically sensitive to changes in freshwater environmental conditions (Sandin et al. 2014). Also, EPT taxa have been detected using COI metabarcoding to survey freshwater macroinvertebrates (Hajibabaei et al. 2012, 2019, Elbrecht and Leese 2017, Emilson et al. 2017). To investigate arthropod assemblages identified in the ethanol- and antifreeze-preserved samples (from evaporated samples only), we used non-metric multi-dimensional scaling (NMDS) on Sorensen dissimilarities using the metaMDS function in the vegan package, and we calculated a Shephard’s plot and goodness of fit for the NMDS ordination using the stressplot and goodness functions. To assess differences among groups, we used the vegdist function to create a Sorensen dissimilarity matrix, the betadisper function to check for heterogeneous distribution of dissimilarities, and the adonis function to perform a permutational analysis of variance (PERMANOVA) to check for strong interactions between groups (preservative, site, processing method) in the dissimilarity matrix. To maintain a balanced design, we limited PERMANOVA tests to evaporated samples.

To investigate arthropod orders shared between preservative types, sites, and processing methods, we used the draw.pairwise.venn and draw.triple.venn functions in the VennDiagram package (Chen and Boutros 2011) in R. Focusing on EPT ESVs, we used the psych package (Revelle 2020) in R to calculate pairwise Pearson’s correlations between paired ethanol and antifreeze samples. We set the alpha level to 0.05 and used the Holm adjustment (Holm 1979) to correct for multiple tests. We plotted the correlations using the corrplot (version 0.84; Wei and Simko 2017) package in R. We plotted confidently identified arthropod genera from EPT orders represented by at least 2 ESVs in both ethanol-preserved and antifreeze-preserved samples to examine whether any genera were more associated with 1 type of preservative. Finally, to assess if any arthropod families, in particular from bioindicator orders, were represented more in 1 type of preservative compared with the other, we visualized the frequency of ESVs detected from arthropod families using a heatmap generated from the geom_tile function in the ggplot package (Wickham 2016) in R. We limited this analysis to evaporated samples for a balanced comparison.


Raw taxonomic data

Antifreeze samples produced clear bands in the PCR gel for both unevaporated and evaporated samples (Fig. S2). The BR5 primer set failed to amplify all 3 unevaporated ethanol samples. Ethanol samples generated bands similar to antifreeze samples for all sites. A total of 11,185,532 × 2 (for forward and reverse reads) Illumina paired-end reads were sequenced (Table S2). After bioinformatic processing, we retained 19,298 ESVs (9,908,826 reads). After taxonomic assignment, a total of 5198 arthropod ESVs (2,418,452 reads) were retained for data analysis (Table S3).

Bioinformatic processing

Out of all ESVs, 34.8% were assigned to Arthropoda, accounting for 60.5% of reads in all ESVs (Fig. S3). Rarefaction curves that reach a plateau show that our sequencing depth was sufficient to capture the ESV diversity in our PCRs (Fig. S4). In terms of classification, we confidently identified most arthropod genera, families, and species (Fig. S5). Negative controls (antifreeze, ethanol, extraction kit, and PCR) produced a small number of sequences from a total of 13 orders (Table S4), however the low number of ESVs returned (between 3–80) would not have influenced the sequencing results observed in the field samples dataset (Hornung et al. 2019).

ESV richness

ESV richness comparisons showed little difference between experimental groups. For the validation samples, mean overall ESV richness displayed similar values for antifreeze (229) and ethanol (222) samples (after normalization; Fig. 2A). For test samples, diversity was again similar between antifreeze- (251) and ethanol-preserved samples (245; Fig. 2B). For the 2 different processing methods for the test samples, there was a marginally higher ESV richness for unevaporated antifreeze-preserved samples (273 vs 271), and, as expected, evaporated ethanol-preserved samples produced greater ESV richness compared with unevaporated ethanol samples (220 vs 196; Fig. 2B). Overall, median arthropod ESV richness in antifreeze samples was not greater than median arthropod ESV richness in ethanol samples (p = 0.82), in contrast with our hypothesis.

Figure 2. 
Figure 2. 

Arthropod exact sequence variant (ESV) richness varies across unevaporated and evaporated antifreeze- and ethanol-preserved samples. Results are shown from 2 experiments based on validation sampling (A) and test sampling (B). Evaporated and unevaporated samples refer to samples that have either been evaporated fully before DNA extraction or have been extracted directly after removal of supernatant. Richness is based on rarefied data.

Arthropod diversity

To address whether using antifreeze as a substitute for ethanol as a preservative would affect biodiversity analyses, we compared alpha and beta diversity of the identified arthropod assemblages among experiment and treatment groups. Arthropod assemblages were relatively similar between evaporated ethanol- and antifreeze-preserved samples, with replicates clustering close together for most sites on the NMDS plot (Fig. 3A–C). PERMANOVA (evaporated samples) showed that differences between validation and test experiments explained 19% of the variation in assemblage composition (F = 8.2, p = 0.001; Table 1). PERMANOVA also showed that differences between preservative used explained 4% of the variation (F = 1.7, p = 0.001; Table 1). The interaction between experiment (validation vs test samples) and treatment (ethanol vs antifreeze) explained another 4% of the variation (F = 1.6, p = 0.001; Table 1).

Figure 3. 
Figure 3. 

Arthropod samples cluster mainly by site and sampling method. Non-metric multi-dimensional scaling (NMDS) plots show overlap between the binary Bray–Curtis (Sorensen’s) dissimilarities across ethanol- and antifreeze-stored samples (A), clustering by sampling site (B), overlap between evaporated and unevaporated processing methods (C), and clustering based on sample design (D) (stress = 0.124, R2 = 0.93). The same ordination is shown in each panel. Ordination is based on rarefied data.

Table 1. 

Statistical outputs from permutational multivariate analysis of variance tests. The formula = sora ~ experimentb × treatmentc (strata = site).

GroupF ModelR2p-value
Residuals 0.74 

a sor = Sorenson’s sample by exact sequence variant matrix.

b Experiment = validation samples or test samples.

c Treatment = ethanol or antifreeze.

View Table Image

To assess broad-scale differences in the relative detection of taxa from ethanol- or antifreeze-preserved samples, we assessed order-level differences in Venn diagrams (Fig. 4A–F). At the arthropod order level, there were a total of 22 orders (mix of terrestrial–riparian and aquatic) shared between the 2 preservative types for validation samples (Fig. 4A) and 26 shared in the test samples (Fig. 4D). Within the validation samples, 2 orders were unique to ethanol and 4 orders were unique to antifreeze. In contrast, the test samples indicated 4 orders unique to ethanol and 5 unique to antifreeze. Fourteen (validation samples; Fig. 4B) and 13 (test samples; Fig. 4D) arthropod orders were shared between all sites, with the greatest number of unique arthropod orders found at sites Clair 12 (4) and Laurel 10 (3). When looking at processing methods for antifreeze, 23 (validation samples; Fig 4C) and 27 (test samples; Fig. 4F) arthropod orders were shared.

Figure 4. 
Figure 4. 

High degree of taxonomic overlap between treatment groups. Venn diagrams of unique and overlapping arthropod orders for preservative types (A, D), sites (B, E), and evaporation processing method (C, F). Based on rarefied evaporated data.

Focusing in on finer-level differences in taxon recovery in important bioindicator orders, we plotted the number of unique ESVs detected from arthropod EPT genera (Fig. 5). A similar number of unique ESVs were detected from EPT genera from ethanol- and antifreeze-preserved samples. Some of the confidently identified arthropod genera represented by >2 ESVs, identified from both ethanol- and antifreeze-preserved samples, included: Caenis (Ephemeroptera), Polypedilum (Chironomidae), and Tipula (Chironomidae) (Fig. 5). Although differences were observed in the relative number of ESVs detected from EPT genera, overall, the presence of EPT ESVs detected from ethanol- and antifreeze-preserved samples were positively correlated (Fig. S6).

Figure 5. 
Figure 5. 

The number of arthropod exact sequence variants (ESVs) from genera detected from samples preserved in antifreeze or ethanol are similar. Each point represents a genus identified with high confidence. The points are color coded for the Ephemeroptera, Plecoptera, and Trichoptera arthropod orders detected in this study. Dotted line is a 1:1 correspondence line. Both axes are transformed by log10 to improve the spread of points with small values. Based on rarefied data. Only evaporated samples are shown for a balanced comparison.

At an even finer level, we assessed the relative richness detected from EPT families for both the validation and test samples (Fig. S7A, B). Ethanol- and antifreeze-preserved samples showed similar results for the detected EPT families across replicates at each site. Subtle differences in detection between preservative types was evident in the validation samples. In some cases, antifreeze-preserved samples produced a greater number of reads for detected families (e.g., Uenoidae; Fig. S7B), and in other cases, ethanol-preserved samples produced a greater number of reads for detected families (e.g., Leptophlebiidae; Fig. S7A). A large majority of the families detected were present in both ethanol- and antifreeze-preserved samples at the same read abundance across all sites.


To monitor biodiversity through metabarcoding studies, it is vital to use DNA preservatives that are effective at preventing DNA degradation to allow the detection of taxa present within an environmental sample (Rees et al. 2014, Strickler et al. 2015, Barnes and Turner 2016). Commonly, 95 to 100% ethanol is the standard preservative used for maintaining the integrity of both tissue and environmental samples containing DNA (Baird and Hajibabaei 2012, Hajibabaei et al. 2012, Rees et al. 2014). However, this one-size-fits-all approach is not feasible for optimal DNA preservation across all environment types (Barnes and Turner 2016) and for all communities participating in biomonitoring efforts. We assessed the ability to detect arthropod assemblages and species richness from environmental samples preserved with commercially available antifreeze vs samples preserved with absolute ethanol. Our study shows that detection of arthropod ESV richness was comparable between the 2 preservatives and that our ability to identify sampled arthropod assemblages, especially EPT taxa, was similar for both types of preservative.

To determine any true effect of preservative type on the detection of macroinvertebrate DNA, using validation samples (i.e., samples that are homogenized prior to preservation) is important. Previous studies concerning the application of new preservatives for DNA-based research have not taken this approach (e.g., Stein et al. 2013). Natural biological variation can be high even within the same site, as demonstrated by our results (Figs 2A, B, S8), and we expected to observe higher dissimilarity between preservation methods in the test samples compared with the validation samples because of the addition of preservative prior to homogenization. However, overall ESV richness remained similar, especially for ESV richness in EPT groups. Preservatives used to maintain the integrity of DNA for biomonitoring are required to be consistent, in terms of preserving the existing taxa across different sites, to enable an accurate assessment of existing arthropod assemblages. Our study serves as a model for robustly testing new preservatives on environmental samples through the use of validation samples, and it also illustrates the ability of antifreeze to consistently preserve samples at a similar rate as ethanol.

Another strength of antifreeze for preserving benthic samples is that, unlike ethanol, it does not appear to require an evaporation period to avoid inhibiting the DNA extraction or PCR in the processing of samples. Optimizing DNA extraction protocols and generating results quickly without sacrificing quality or reliability are critical for minimizing consumable and labor costs of biomonitoring research (Bonada et al. 2006, Baird and Hajibabaei 2012, Vasselon et al. 2017). DNA metabarcoding research can be limited by PCR inhibition, and often time-consuming steps are taken to fully remove the ethanol preservative and purify DNA prior to sequencing (Demeke and Jenkins 2010, Schrader et al. 2012). Using propylene glycol-based antifreeze as a preservative for environmental samples allows the lengthy evaporation step to be bypassed, and DNA pellets formed from centrifuged samples can be extracted directly and instantly.

A primary goal of many freshwater biomonitoring programs is to investigate and compare macroinvertebrate assemblages across sites of varying geographic and quality status to build a picture of freshwater ecosystem health at larger-than-site scales. Results from this study have shown that despite the use of 2 different preservatives using 2 processing methods, the main source of variation in the data was among sites, followed by experiment (validation or test). Moreover, most arthropod orders were shared across ethanol- and antifreeze-preserved samples, and we have shown that the traditional bioindicator orders were also shared across preservative types.

Overall, antifreeze-preserved samples were comparable with ethanol-preserved samples for arthropod DNA detection at multiple taxonomic levels. Comparable recovery of arthropod ESVs and matched detection of key bioindicator orders and families in antifreeze samples highlights the suitability of propylene glycol-based antifreeze as an alternative to conventional absolute ethanol. Use of inexpensive, easily accessible antifreeze would facilitate widespread DNA-based research, particularly community-based monitoring initiatives, by improving feasibility for sampling remote sites, providing consistent and reliable sample preservation, reducing laboratory processing time, and simplifying the sample shipping process. Despite our study’s limited duration (3-d incubation period), there is evidence from previous research to suggest that propylene glycol-based antifreeze preserves DNA over a longer time period (e.g., ≥7 d; Steininger et al. 2015). Additional studies of the preservation ability of antifreeze over a longer time period and with additional taxa is needed to draw reliable conclusions of the suitability of this preservative for exclusive use for other DNA-based biomonitoring studies (e.g., Malaise traps; Lynggaard et al. 2019). Additional research into the applicability of antifreeze for additional environmental sample types, such as water and soil, would be beneficial to further investigate the potential for propylene glycol-based antifreeze to become a widely used DNA preservative.

Author contributions: CVR, MTW, and MH designed the study. MTW conducted field work and laboratory analyses. TMP and CVR conducted statistical analyses. CVR wrote the manuscript with assistance from all of the authors.

We would like to thank Christina Myrdal, Bianca Marcellino (Dougan & Associates), and Genevieve Johnson for help with collecting samples and Carley Maitland for assisting with sample collection and processing. This study is funded by the government of Canada through Genome Canada and Ontario Genomics.

Data accessibility: Raw sequences are available from NCBI SRA BioProject ID: PRJNA673628. The bioinformatic pipeline SCVUC v4 is available from GitHub at The COI classifier v4 we used is available from The code used to generate figures, including infiles, are available from GitHub at

Literature Cited