Sampling challenges
@Zerina Rahic
Environment: pond vs lake vs river
Challenges
A universal methodology for eDNA analysis may not be appropriate across habitat types as water bodies vary considerably in their biological, physical, and chemical properties (Goldberg et al., 2016). These fundamental differences can affect eDNA behaviour, including origin, state, fate, and transport (Barnes & Turner, 2015), and may ultimately have repercussions for eDNA detection.
Sampling
The distribution and dispersion of eDNA in ponds complicates the design of sampling strategies. In contrast to lotic systems, eDNA has a patchy distribution in lentic systems due to an uneven distribution of organisms.
- Ponds: eDNA distribution and dispersion in ponds is limited both horizontally by the presence of barriers to water movement and vertically by chemical stratification of the water column due to minimal wind-mixing (Sayer et al., 2013). This large variation in eDNA on fine spatial scales has severe consequences for species detection.
- Lakes: A lake experiment with caged Northern pike Esox lucius (Linnaeus, 1758) also revealed a substantial reduction (* 80%) in eDNA detection probability as distance from cages increased (Dunker et al., 2016). More recent caging experiments of fish and amphibians in ponds revealed a strong decrease in eDNA detection probability with distance from the cage, with most species nearly undetectable after a few metres.
- River: In streams and rivers, eDNA detection probabilities of the black warrior waterdog Necturus alabamensis (Viosca, 1937) and the flattened musk turtle Sternotherus depressus (Tinkle & Webb, 1955) were also influenced by sampling season and were consistent with current knowledge on timing of foraging and reproduction in these species (de Souza et al., 2016). More recent research demonstrated eDNA concentration and subsequent detection probability of fish may be improved during the spawning season when gametes are released into the water
The patchy distribution of pond eDNA means one sample of surface water will not sufficiently represent true biodiversity. Representation can be achieved with a timely, thought-out sampling strategy that accounts for location, number and volume of samples, and method of collection.
Metabarcoding field sampling - room for improvement
@Michelle Anne Hughes
From this paper's analysis, there needs to be better standardization of sampling protocols to reduce subjective methods and become reproducible practices:
"methodologically sound field sampling is the foundation for subsequent analyses. We reviewed field sampling methods used for metabarcoding studies of both terrestrial and freshwater ecosystem biodiversity over a nearly three -year period (n = 75). We found that 95% (n = 71) of these studies used subjective sampling methods, inappropriate field methods, and/or failed to provide critical methodological information. It would be possible for researchers to replicate only 5% of the metabarcoding studies in our sample, a poorer level of reproducibility than for ecological studies in general. Our findings suggest greater attention to field sampling methods and reporting is necessary in eDNA -based studies of biodiversity to ensure robust outcomes and future reproducibility. Methods must be fully and accurately reported, and protocols developed that minimise subjectivity. Standardisation of sampling protocols would be one way to help to improve reproducibility, and have additional benefits in allowing compilation and comparison of data from across studies"
"valid inferences are critically dependent on valid sampling techniques (Crawley, 2015). Further, sample collection at a specific site and specific time can be done only once. This contrasts with later steps (e.g., DNA amplification, sequencing and bioinformatic analyses), which can be repeated or re-run from archived sample)"
"Major sources of field contamination include: pre-existing DNA on sampling equipment, DNA from the researcher and their personal microbiome, carry-over between samples, and, for some samples, unintentional movement of DNA from the surface into a sample"
Sampling Criterion
- "all sampling occurs within a "sampling universe". Defining the sampling universe requires specifying the area that samples are intended to be representative of (including, for example, geopolitical constraints, pragmatic limitations, and ecosystem types) and the criteria for excluding portions of that area from potential sampling (including safety and practicality constraints) ... [we need to] how these locations were chosen to be representative of any larger area."
- "the location of samples or plots within that universe must be determined, either using objective or subjective methods" - objective is the scientifically sounder option
- of objective protocols is that the precise location of a plot is specified as an exact and absolute location, based on either true random, gridbased, or more complicated sampling designs (e.g., Robertson et al., 2013)
- with subjective sampling, in contrast, the location of the plot is only loosely specified by the experimental design, with the exact location selected by the researcher, commonly in a way to be both "representative" of a site and to avoid what the researcher views as unusual or disturbed sites
- OR plots are located haphazardly within broad categories ... haphazard sampling includes a strong element of convenience, such as sampling along existing trails or roads (Anderson, 2001). Haphazard sampling makes replication by others nearly impossible
- Size of area sampled
- Subsample locations within the plots
- the location of individual subsamples within plots also varies substantially among studies. Subsamples may be explicitly located based on a regular pattern, truly random, haphazard, or subjectively distributed across the plot.
- haphazard and subjective methods provide no clear advantages except for convenience
- An advantage of subsampling is that it allows for characterisation of a given area which can be resampled in the future, whereas single point sample cannot be resampled.
- one study using T-RFLP found that eight subsamples were sufficient to distinguish bacterial communities among different land-uses in Australia (Osborne et al., 2011).
- Determining how many samples to take
- Power analysis can help find an optimal sampling level.
- Define substrate being sampled
Field Negative controls
Sample field equipment to determine contamination potential: "Most molecular ecologists routinely include negative controls in PCR reactions, recognising the power of PCR to detect very low levels of DNA. Typically, negative controls account only for contamination in the laboratory. Arguably, accounting for contamination in the field may be more important, albeit also more challenging. Field equipment can, and probably should, be sampled through swabbing in order to gain some insight into the potential for contamination. Sample storage media and containers can also be tested. No study in our review reported whether or how field negative controls were included. This stands in contrast to the almost universal use of negative controls in laboratory stages of analysis, an environment where maintaining sample integrity is relatively straightforward."
Avoiding Sample Contamination
Single use versus decontamination: "Some studies avoid contamination through single-use, pre-sterilised equipment, particularly for water sampling. Where sampling equipment is reused, soaking in a solution of sodium hypochlorite, such as commercial bleach, is an effective method of decontamination (Prince and Andrus, 1992; Kemp and Smith, 2005) provided the length of exposure and solution concentration are sufficient. Household bleach is a variable concentration solution of sodium hypochlorite (3 - 8%). For stringent decontamination, a solution as strong as 2 to 3% sodium hypochlorite may be needed (Kemp and Smith, 2005), but general cleaning should be effective with as low as 0.55% [10% v/v dilution of commercial bleach (Prince and Andrus, 1992)]. The effectiveness of bleach depends on the length of time of exposure, requiring at least a few minutes at typical concentrations... it may be most efficient to have multiple sets of sampling equipment, allowing multiple samples to be taken before having to decontaminate"
Sterilization is not the same as decontaminating: "Alcohol sterilizes by killing microbes, but does not remove DNA contamination. Indeed, ethanol is routinely used in DNA precipitation and processing."
Sample Storage and Transport
"Important decisions include the temperature samples are stored at, whether naturally anaerobic samples are maintained in anaerobic state, and the length of time samples are stored. Recommended times between collection and freezing are as low as 2 hours (von Wintzingerode et al., 1997). In some cases it may be possible to preserve samples chemically, both preventing microbial growth and loss of DNA (Seutin et al., 1991; Frantzen et al., 1998)."
References