Bacteria, fungi, and nematodes occupy a large amount of biomass in the soil, playing pivotal ecosystem functions (Fierer et al., 2007; Bardgett and van der Putten, 2014; Nielsen et al., 2015; Wall et al., 2015; Bahram et al., 2018; Bar-On et al., 2018; Flemming and Wuertz, 2019; van den Hoogen et al., 2019). These soil organisms have also been recognized as indicators of the soil environment, targeted, for example, in assays of soil contamination by anthropogenic activities (Bongers, 1990; Bongers and Bongers, 1998; Bongers and Ferris, 1999). In uncovering and evaluating the diversity of those soil organisms, DNA metabarcoding (amplicon sequencing) based on high-throughput sequencing has been increasingly applied as a standard approach (Caporaso et al., 2012; Reuter et al., 2015). Through the recent enrichment of reference DNA databases of bacterial 16S rRNA, fungal internal transcribed spacer (ITS), and nematode 18S rRNA sequences, it has become possible to gain large datasets of soil biodiversity without longtime taxonomic expertise or experiences in microbial isolation (Quast et al., 2013; Nilsson et al., 2019; Kawanobe et al., 2021).

In expanding the use of DNA metabarcoding of soil biodiversity, it is of particular importance to develop standard protocols applicable to broad ranges of situations and organisms. For example, universal primers with high taxonomic coverage have been proposed for each of the bacterial 16S rRNA (Klindworth et al., 2013; Thijs et al., 2017), fungal ITS (Schoch et al., 2012; Lindahl et al., 2013; Op De Beeck et al., 2014), and nematode 18S rRNA regions (Sapkota and Nicolaisen, 2015; Ahmed et al., 2019; Waeyenberge et al., 2019; Sikder et al., 2020; Kenmotsu et al., 2021). For the bioinformatic processes, rapid and accurate algorithms for inferring operational taxonomic units (OTUs) or amplicon sequence variants (ASVs) have been implemented (Callahan et al., 2016). Likewise, theories on reference database search (Tanabe and Toju, 2013) and taxonomic assignment (Huson et al., 2007) have been proposed to allow automatic molecular identification. While all these processes are keys to reliability and reproducibility of DNA metabarcoding, results of molecular-based biodiversity inventories to depend greatly on DNA extraction methods (Thompson et al., 2017; Zielińska et al., 2017; Francioli et al., 2021). In particular, the amount of the soil subjected to DNA extraction possibly has major impacts on the α- and β-diversity estimates of metabarcoding samples.

Because soil properties such as pH and nitrogen concentrations vary within ecosystems, the taxonomic compositions of soil organisms often show high spatial heterogeneity (Franklin and Mills, 2003; Fierer, 2017; Flemming and Wuertz, 2019). Therefore, DNA extraction from small amounts of soil samples may lead to the underestimation of biodiversity or low reproducibility in the reconstruction of community structure. To date, a number of studies have been conducted to evaluate how soil amounts in DNA extraction can affect DNA metabarcoding results of bacteria and/or fungi (Martin-Laurent et al., 2001; Ellingsøe and Johnsen, 2002; Kang and Mills, 2006; Song et al., 2015; Penton et al., 2016). In the Earth Microbiome Project (Gilbert et al., 2014), for example, DNA extraction from 0.5 g of soil was described as a standard protocol.¹ For nematode, in contrast, studies based on real-time PCR and T-RFLP concluded that 100 g of soil could be required for inferring community structure (Wiesel et al., 2015). Because DNA extraction from 100 g of soil is impractical for high-throughput analyses with standard laboratory equipment, developing protocols of DNA extraction from smaller amount of soil is an important step for expanding the application of nematode DNA metabarcoding. Moreover, because soil ecosystem processes are driven by complex interactions between bacteria, fungi, and nematodes (Bonfante and Anca, 2009; Toju and Tanaka, 2019; Detrey et al., 2022), standardized DNA extraction protocols commonly applicable to the three organismal groups will provide a platform for reshaping our understanding of ecosystem functions and dynamics.

In this study, we examined how DNA metabarcoding data could be influenced by the amount of the soil subjected to DNA extraction. We collected soil in forest and farmland (soybean field) ecosystems and then performed DNA extraction at five levels of soil-amount settings (0.5, 2, 5, 10, and 20 g). Given the large body size and mobility of nematodes (Ferris et al., 1990; Ettema, 1998; Liu et al., 2019), DNA extraction protocols optimized for bacteria may lead to the failure of nematode diversity profiling (Wiesel et al., 2015). Thus, we tested the hypothesis that standard DNA extraction protocol optimized for bacteria (i.e., extraction from 0.5 g of soil) could result in potential biases in the estimation of nematode diversity. Specifically, based on the Illumina sequencing outputs, α- and β-diversity estimates were compared among the five soil-amount classes for each ecosystem type (forest or farmland) for each of the bacterial 16S rRNA, fungal ITS, and nematode 18S rRNA regions. We then explored the minimum soil amount enabling the simultaneous analysis of bacteria, fungi, and nematodes. The results shown in this study will provide a basis for designing DNA metabarcoding protocols commonly applicable to major taxa in soil ecosystems.

In total, 4,000 bacterial (220,607 reads), 1,622 fungal (148,058 reads), and 701 nematode (169,073 reads) ASVs were detected across the samples. The number of bacterial, fungal, and nematode genera were 234, 221, and 48, respectively. Likewise, the number of bacterial, fungal, and nematode families were 171, 158, and 38, respectively. The number of bacterial, fungal and nematode order were 151, 88, and 10, respectively. For all the three organismic groups, interpolation and extrapolation curve were saturated: sample coverage was higher than 0.9 in all the analyzed samples (Supplementary Figure 1). α-diversity of the three organismal groups are shown in Figure 1. The taxonomic compositions of three organismic groups in each soil amount setting were visualized as well (Figures 2–4).

For bacteria and fungi, soil amount in DNA extraction did not have significant effects on the α-diversity estimates at the ASV, genus, family, and order levels (Table 1; Supplementary Table 1 and Figure 1). Meanwhile, the interaction between ecosystem type (forest or soybean field) and soil amount had significant effects on the Shannon and Simpson diversity of bacteria at the genus, family and order levels (Table 1 and Supplementary Table 1). This result suggests that the effects of soil amount varied between ecosystem types, although combined effects of soil amount and ecosystem type were difficult to interpret.

In the nematode dataset, significant effects of soil amount on taxonomic richness, Shannon diversity, and Simpson diversity were observed at the ASV and family levels (Table 1; Supplementary Table 1). Likewise, soil amount had significant effects on taxonomic richness and Shannon diversity at the genus level (Supplementary Table 1). Meanwhile, such effects of soil amount on α-diversity index values were not observed at the order level (Supplementary Table 1). At the ASV, genus, and family levels Tukey’s HSD test indicated that α-diversity scores had higher values at 20 g than at 0.5 g soil DNA extraction settings in more than half of the combinations of ecosystem type and α-diversity indices (Figure 1).

The level of dissimilarity in bacterial community compositions among replicate samples significantly varied depending on the soil amount settings at the ASV, but not at the genus, family and order levels (Table 2; Supplementary Table 2 and Figure 5). In the Tukey’s HSD test, among-sample community dissimilarity of bacteria was lowest at 10 g among the soil amount settings in the soybean field data (Figure 5). For fungi, dependence of community dissimilarity level on the soil amount settings was observed at the ASV, genus, and order levels (Table 2 and Supplementary Table 2), although the pattern was inconsistent between the forest and soybean field ecosystems (Figure 5). Likewise, the level of among-replicate dissimilarity in nematode community compositions was observed at all the examined taxonomic levels (Table 2; Supplementary Table 2 and Figure 5) and there was significant difference among different soil amount in ANOVA test (Table 2). Moreover, Tukey’s HSD test indicate that the community composition dissimilarity had significantly lower values at 10 and 20 g soil amount setting (Figure 5).

The among-replicate community dissimilarity, which represented heterogeneity of community structure, varied depending also on organismal groups (Figure 5). Specifically, fungi showed the highest level of among-sample community heterogeneity at the genus, family, and order levels. At the ASV level, bacteria showed comparable and higher among-sample community heterogeneity in the forest and soybean field ecosystems, respectively (Figure 5). Such heterogeneity level of community structure was moderate for nematodes (Figure 5).

We here examined how the amount of soil subjected to DNA extraction affects the results of biodiversity surveys targeting three major groups of soil organisms, namely, bacteria, fungi, and nematodes. We then found that DNA metabarcoding protocols optimized for bacteria (and fungi) could result in potential biases in the estimation of α- and β-diversity of nematodes in the soil as detailed below.

For bacteria and fungi, there have been several studies evaluating potential effects of soil sample amount on DNA metabarcoding-based biodiversity inventories. Among those studies, some suggested that 10 g of soil were required to gain reproducible ASV richness data (Kang and Mills, 2006; Penton et al., 2016), while others concluded smaller amounts of soil were enough for surveys of bacteria and fungi (Ranjard et al., 2003; Song et al., 2015). In our study, the amount of soil subjected to DNA extraction did not directly affect the α-diversity estimates (i.e., taxonomic richness, Shannon diversity, and Simpson diversity) of bacteria and fungi (Table 1; Supplementary Table 1 and Figure 1). Therefore, we conclude that DNA extraction from 0.5 g of soil and that from larger amounts of soil can yield qualitatively similar metabarcoding results in the analyses targeting bacteria and fungi. Meanwhile, taking into account the previous studies reporting potential bias related to DNA extraction from small amounts of soil (Kang and Mills, 2006; Penton et al., 2016), careful management of experimental protocols would be required in DNA extraction from 0.5 g of soil.

While there have been a number of methodological studies on bacterial and fungal DNA metabarcoding, few studies have been conducted to examine the effects of soil sample amount on nematode diversity surveys. In research on soil nematode communities, the traditional approach is to isolate nematodes from the soil matrix by means of the Baermann method, elutriator sucrose centrifugation, and filter separation, followed by microscopic morphological observations (Neher et al., 1995; Yeates et al., 1999; Ferris et al., 2012). This approach is usually taken based on sampling of several hundred grams of soil. Meanwhile, in recent years, DNA-based techniques have been recognized as an alternative approach and nematode community inventories from smaller amount of soil have started to be examined (Wiesel et al., 2015; Treonis et al., 2018; Quist et al., 2019; Zhang et al., 2021).

Our results indicated that soil amount in DNA extraction could considerably affect DNA metabarcoding results of nematodes α-diversity and community composition (Table 1; Supplementary Table 1 and Figure 1). All the α-diversity estimates increased with increasing soil amount, reaching maximum values at 10 g at the ASV, genus, and family levels (Figure 1). These results indicate that biodiversity surveys of nematodes require higher amounts of soil than those of bacteria and fungi. This finding is as expected because nematodes [typically, from 0.1 to 2.5 mm long (Hodda, 2022)] are much larger than bacterial cells and fungal hyphae: thus, their population density in the soil may be much lower than microbes. Even if “environmental DNA” derived from feces or remains of nematodes are detectable simultaneously with DNA from living nematode bodies, the scale of soil samples required for diversity inventories would be larger in nematodes than in bacteria and fungi (Peham et al., 2017). Given our data, we would suggest mixing as much soil as possible and then introducing subsampled 20 g of soil to lysis solutions for DNA extraction. Ultimately, the use of more soil in the lysis step would be ideal. However, 20 g may be the upper limit of molecular biological experiments given the size of centrifugeable tubes (≦ 50 ml). Consequently, including replicate sampling points within each study site is of particular importance in nematode DNA barcoding even if DNA extraction is performed at the 20 g soil scale.

The results of the DNA metabarcoding also indicated that the level of dissimilarity (β-diversity) among replicate samples could vary depending on soil amount settings in DNA extraction, especially in nematodes (Table 2; Supplementary Table 2 and Figure 5). Basically, DNA extraction from small amount of soil is expected to result in higher dissimilarity among replicate samples because inferred community structure is more likely to be influenced by stochastic factors. The expected decrease in β-diversity with increasing soil amount was observed in at the ASV level in the soybean filed ecosystem for bacteria and nematode community (Figure 5). Thus, DNA extraction from 20 g of soil is recommend for analyses including nematodes as target organisms with regard to both α- and β-diversity surveys.

Although the above results are informative, it should be acknowledged that the optimal amount of soil for DNA metabarcoding potentially depend on soil and ecosystem types. For example, the level of among-sample heterogeneity in fungal community structure can vary between grassland ecosystems and farmlands with continuous anthropogenic perturbations (e.g., hyphal disconnection by tillage). In addition, it has been reported that soil nematode density varies across continents depending on soil edaphic factors (van den Hoogen et al., 2019). Thus, for processing soil samples with highly heterogeneous community structure or low organismal densities, mechanical approaches for systematically mixing large amounts of soil before DNA extraction need to be developed in future studies as direct DNA extraction from soil.

Overall, we found that the standard DNA metabarcoding protocols developed for bacteria are not directly applicable to multicellular organisms in terms of the soil amount subjected to DNA extraction. Given the results on α- and β-diversity estimates, we conclude that DNA extraction from 20 g of soil is necessary for simultaneously investigating the biodiversity of bacteria, fungi, and nematodes in below-ground ecosystems. Because DNA-based inventories of biodiversity have become major tools in ecology, a next important step may be integration and comparison of diversity patterns among different groups of organisms. Toward the thorough understanding of metacommunity-scale dynamics and macroecological patterns of soil biodiversity, further attempts of technical standardization are necessary to target whole biomes including not only bacteria, fungi, and nematodes but also other invertebrates and protists.

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: https://www.ddbj.nig.ac.jp/, DRA014170.

TK and HT designed the work and wrote the manuscript. TK carried out the experiments and analyzed the data. Both authors contributed to the article and approved the submitted version.