INTRODUCTION
Environmental DNA (eDNA) can be defined as the mixture of complex, often degraded, DNA that organisms leave behind in their environment (i.e. soil, water, sediments, etc.). By studying short, taxonomically-informative DNA fragments obtained from eDNA samples, it is possible to identify the associated taxa and therefore to survey biodiversity. Coined as “eDNA metabarcoding”, this approach has revolutionized several branches of ecology and environmental sciences during the last decade, by providing relatively quick, non-invasive, and standardized assessments of present or past biodiversity of animals, plants and microorganisms (Taberlet, Bonin, Zinger, & Coissac, 2018). Metabarcoding is particularly valuable for monitoring biodiversity over large geographical or taxonomic scales (De Vargas et al., 2015; Delgado-Baquerizo et al., 2018; Zinger et al., 2019b). Furthermore, it gives access to biodiversity components that are elusive to conventional survey methods. For instance, it allows the rapid assessment of microbial soil biodiversity, which is extremely complex, time-consuming and imperfect when using direct observations, culturing techniques or microscopy (Giovannoni, Britschgi, Moyer, & Field, 1990; Ward, Weller, & Bateson, 1990).
Metabarcoding relies on a succession of several steps: 1) sampling; 2) preservation of the collected material until lab processing; 3) DNA extraction; 4) PCR amplification of a selected genomic region; 5) high-throughput sequencing of amplicons; and 6) analysis of sequences using bioinformatics and statistical tools (Zinger, Bonin, et al., 2019a). Each step is critical to obtain robust taxonomic inventories and diversity estimates, and an increasing number of studies have assessed how methodological choices across the different steps could influence the conclusions of a study (Calderón‐Sanou, Münkemüller, Boyer, Zinger, & Thuiller, 2020; Cantera et al., 2019; Chen & Ficetola, 2020; Nichols et al., 2018; Taberlet et al., 2018). So far, despite this growing body of literature, little attention has been accorded to the effect of different preservation conditions of the collected environmental material before lab processing (i.e. step 2). We thus know neither under which conditions the collected material should be stored, nor how long it can be stored to avoid biases in taxonomic inventories.
While more is known for water samples (see e.g. Kumar, Eble, & Gaither, 2020; Majaneva et al., 2018), in the case of soil biodiversity research, methodological analyses on the effects of sample preservation are largely dismissed probably because the majority of metabarcoding studies have so far been performed in temperate areas where access to lab facilities is often easy (Hoffmann, Schubert, & Calvignac-Spencer, 2016; Huerlimann et al., 2020). In such cases, sample preservation is sometimes not necessary at all, or at least not over long periods of time. However, one great promise of metabarcoding is its potential for providing biodiversity data for remote areas, where biodiversity monitoring is essential but difficult. When sampling in remote or inaccessible areas (e.g. tropical and arctic areas; mountain chains), samples are rarely collected nearby lab facilities and an immediatein situ DNA extraction is generally not possible due to logistic constraints (but see Zinger, Taberlet, et al., 2019b for a notable exception). More generally, with the ever-increasing number of samples analyzed during a typical metabarcoding study, sample preservation is more and more indispensable, and the time lag between sample collection and subsequent molecular processing makes it particularly relevant to understand the impact of sample preservation, and to identify preservation strategies that do not bias the conclusions of studies.
In an optimal metabarcoding study, communities recovered from preserved samples should ideally be identical to those retrieved if samples had been processed immediately after sampling. However, inappropriate preservation conditions can cause both DNA degradation and the proliferation of certain taxonomic groups, with respect to others, before DNA extraction (Cardona et al., 2012; Orchard, Standish, Nicol, Dickie, & Ryan, 2017). This can in turn affect taxa detection and also the relative contributions of different taxonomic groups to the overall biodiversity. A recent review suggested that the majority of eDNA metabarcoding studies do not provide accurate information about sample treatment before processing (Dickie et al., 2018). Almost half of the studies do not report how samples were stored and conserved, and 30% of them store samples at 0-4°C, and thus at a temperature where many bacteria and fungi continue to be active and potentially affecting the whole sample. About 15% of the studies stored samples in a range of 5-35°C, which is considered as a poor practice, and only 10 % stored them below 0°C (Dickie et al., 2018).
So far, the consequences of preservation practices and the resulting deviations from immediate processing and analyses have rarely been studied quantitatively. Yet, Lauber, Zhou, Gordon, Knight, & Fierer (2010) tested the effect of storing samples from soil, human gut and skin at different temperatures and did not detect any significant effect on bacterial communities, while Orchard et al. (2017) found that storage time and temperature can affect colonization by arbuscular mycorrhizal fungi, with subsequent impacts on the reconstruction of communities. Differences between these studies may be due to their different protocols. However, they also focused on different taxonomic groups, which may react differently to storage period and temperature. Other studies use desiccation for conserving plant and animal tissues for subsequent genomic studies (e.g. Chase & Hills, 1991), which has proven efficient and convenient. Although not widely used for metabarcoding samples, desiccation is another attractive option, and has a potential for being largely implemented in soil sample preservation. A clear understanding of the effect of different preservation methods, especially across various groups of taxa, is thus pivotal for a robust application of eDNA metabarcoding to biodiversity monitoring in general, and that of remote areas in particular.
Here, using eDNA metabarcoding of different taxonomic groups in soil systems, we tested: (i) how preservation methods influence overall richness estimates and what the role of rarely observed taxa is; (ii) how preservation methods influence identified community structure and its turn-over between different habitats; and (iii) what the best practices are under limited laboratory access. More specifically, we first selected three soil preservation methods (room temperature, 4°C, desiccation by addition of silica gel) because they are commonly used in the literature (room temperature and 4°C) or because they are easy to implement in the field (desiccation and room temperature). Then, we assessed the impact of these preservation methods applied to different durations in order to mimick logistic constraints, and compared the communities obtained with those observed in ideal conditions, i.e. when eDNA is extracted immediately after sampling (within less than one hour). We examined bacterial, fungal and eukaryotic communities to cover a broad taxonomic range, since different taxa can be differentially affected by sample preservation conditions (Cardona et al., 2012; Orchard et al., 2017).