Depth-dependent differences between direct total DNA and an intracellular-DNA-enriched fraction in bulk-soil bacterial and fungal communities of a Lonicera japonica field

Abstract

Background

Direct total-DNA (tDNA) extraction from soil recovers both cell-associated and extracellular (relic) DNA, which can distort amplicon-based views of microbial communities; how strongly this biases bacteria versus fungi, and whether it varies with soil depth, remains poorly resolved for cultivated soils. Using a paired split-sample design, we compared bacterial 16S rRNA gene and fungal ITS profiles from directly extracted tDNA and an operationally defined intracellular-DNA-enriched (iDNA) fraction in bulk soil sampled at three depths from a continuously cultivated Lonicera japonica field.

Results

The two DNA pools gave related but clearly non-interchangeable pictures of the community, and the two kingdoms responded asymmetrically: bacteria were consistently the more sensitive, with higher richness recovered from tDNA and a marked separation between DNA pools throughout the profile, whereas the fungal contrast was smaller overall and, for alpha diversity and within-pair divergence, concentrated in the surface soil. DNA-pool choice acted mainly by selectively redistributing specific lineages rather than replacing the dominant community, while the measured soil physicochemical variables explained only a small share of community variation within either pool.

Conclusions

Direct total DNA and an iDNA-enriched fraction offer kingdom- and depth-dependent windows on the same soil, so the choice of DNA pool is a consequential design decision—especially for bacteria and for surface horizons—that should be matched to the ecological question and independently validated when a cell-associated fraction is the goal. More broadly, a single total-DNA snapshot cannot be assumed to represent the cell-associated community equally across taxa or depths; treating the DNA-extraction strategy as an ecological variable rather than a neutral technical step would strengthen the comparability and interpretation of bulk-soil bacterial and fungal community surveys.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12866-026-05436-3.

Background

Agricultural soil microbial communities regulate nutrient cycling, organic matter decomposition, and biogeochemical functions that underpin crop-soil systems [1]. In perennial cropping systems, sustained plant cover, repeated root turnover, progressive litter deposition, and comparatively limited physical disturbance collectively promote the accumulation of both actively cycling biological material and legacy organic matter throughout the bulk soil profile [1, 2]. The present study used a cultivated Lonicera japonica field as a model perennial cropping system. Lonicera japonica is an economically important medicinal perennial crop [3, 4]. Its growth and metabolite accumulation are influenced by environmental and edaphic conditions [5–7]. Recent evidence from the L. japonica soil–plant continuum indicates that bulk soil constitutes a major reservoir of plant-associated microbiota, positioning it as an ecologically meaningful microbial source rather than a passive background matrix [8]. For this reason, bulk soil from cultivated L. japonica fields provides a relevant agricultural matrix for testing whether DNA-pool choice changes soil microbial community profiles.

Most DNA-based soil microbiome studies rely on direct extraction of total soil DNA (tDNA) [9]; however, tDNA inevitably co-extracts DNA from both intact, metabolically active cells and from dead or membrane-compromised cells [10, 11]. Extracellular DNA released by cell lysis or active secretion can persist in the soil matrix for extended periods, in part because adsorption onto clay minerals and other soil colloids confers physical protection against nuclease degradation [10, 11]. As a consequence, tDNA-based amplicon profiles may partly reflect historical rather than extant microbial assemblages. The quantitative importance of this bias has been demonstrated empirically: using propidium monoazide–based viability PCR across a wide range of soil types, Carini et al. showed that relic DNA constituted, on average, 40% of both prokaryotic and fungal DNA, inflating apparent richness by up to 55% and causing significant misestimation of taxon relative abundances [12, 13]. More broadly, relic DNA has been shown to obscure spatial, temporal, and environment–community patterns in amplicon-based surveys, thereby complicating interpretation of how microbial communities respond to environmental gradients [14]. Consequently, the same bulk-soil sample may yield markedly different ecological interpretations depending on whether profiling targets total DNA or a more cell-associated DNA fraction.

These concerns may be particularly relevant in continuously cultivated perennial croplands, where vertical heterogeneity in organic matter input, root activity, and microbial biomass creates a spatially structured context for non-cell-associated DNA persistence. Soil depth is among the most consistent and powerful drivers of microbial diversity and community composition in agricultural soils, with both richness and biomass typically declining markedly below the surface horizon [15]. At the same time, perennial cultivation preferentially enriches microbial biomass and soil organic carbon in upper soil layers, where recent plant-derived inputs and residue turnover are most active [2, 15]. Under such conditions, the discrepancy between directly extracted tDNA and an operationally defined cell-associated DNA fraction, referred to here as the iDNA-enriched fraction, is expected to be both ecologically meaningful and depth dependent: the degree of overlap between contemporary biological activity and accumulated relic DNA inputs is likely greatest near the surface, where organic matter turnover and microbial necromass production are highest. This expectation finds support in recent work employing DNA-state-specific extraction approaches, which also reveals that the magnitude and direction of DNA-pool effects vary with extraction methodology and the physicochemical properties of the target matrix [2, 16]. The ecological consequences of DNA-pool choice therefore vary with extraction methodology and the physicochemical properties of the target matrix and require empirical evaluation within the target agroecosystem.

In this study, we examined bulk soil from a continuously cultivated L. japonica field using depth-resolved sampling at 0–3, 5–7, and 12–15 cm. Bacterial 16S rRNA gene and fungal ITS amplicon profiles from directly extracted tDNA were compared with those from an iDNA-enriched fraction generated by phosphate-buffer-assisted cell separation followed by membrane filtration. Selected soil physicochemical variables were integrated to evaluate whether DNA-pool choice altered alpha- and beta-diversity patterns, differential taxon recovery, and apparent environment–community associations. Specifically, we asked: (i) whether tDNA and the iDNA-enriched fraction yield discernibly different bacterial and fungal community profiles; (ii) whether any such differences vary with soil depth; and (iii) whether the response differs between microbial kingdoms. We hypothesized that direct total-DNA extraction would overrepresent historical microbial signals relative to the iDNA-enriched fraction, that this overrepresentation would be more pronounced for bacteria than for fungi, and that its magnitude would vary across the sampled depth range.

Whether bacterial and fungal communities respond to DNA-pool choice symmetrically has received comparatively little direct attention, particularly across the shallow depth layers most relevant to perennial crop-soil systems. Most existing comparisons address a single microbial kingdom in isolation, leaving the relative sensitivity of bacteria and fungi to the same methodological contrast within identical soil samples poorly characterized. This study addresses that gap within an integrated design encompassing both kingdoms and three depth intervals in a continuously cultivated L. japonica field.

Materials and methods

Study site and soil sampling

Bulk soil was collected in October 2025 from a continuously cultivated Lonicera japonica field near Fengqiu County, Henan Province, China. Samples were obtained from three depth intervals: 0–3, 5–7, and 12–15 cm. At each depth, five biological replicates were collected independently, yielding 15 original soil samples in total. Each sample was homogenized and subdivided for direct total-DNA extraction, iDNA-enriched fraction preparation, and soil physicochemical analyses. For amplicon sequencing, each original soil sample generated two DNA pools — directly extracted total DNA (tDNA) and an iDNA-enriched fraction — resulting in 30 paired DNA samples, each of which was profiled with a bacterial 16S rRNA gene and a fungal ITS amplicon library (60 amplicon libraries in total); Fig. 1 summarizes the depth-resolved sampling, homogenization, and paired split-sample processing used to generate the two DNA-pool materials. The sampling was intended to represent bulk soil rather than targeted rhizosphere soil; exact distances from plant stems or roots were not recorded. Field-management metadata, including duration of cultivation, planting history, fertilization and irrigation regimes, tillage practice, and exact geographic coordinates, were unavailable at the time of sampling and are therefore not reported.

Fig. 1.

Paired iDNA-enriched and tDNA workflow. The schematic shows bulk-soil sampling from three depth intervals in a cultivated Lonicera japonica field, homogenization of 15 original soil samples, and split-sample processing into matched iDNA-enriched and direct total-DNA (tDNA) workflows

Separation of the iDNA-enriched fraction

The separation procedure was performed under aseptic conditions using sterile consumables and autoclaved buffers, with samples and reagents maintained on ice throughout. For each sample, 3 g of fresh soil was transferred to a sterile 50-mL centrifuge tube containing 6 mL of pre-chilled 0.1 M sodium phosphate buffer (pH 7.4; NaP buffer). The suspension was incubated on ice for 1 min, shaken horizontally at 175 rpm for 5 min with intermittent manual inversion, cooled on ice for 1 min, and then subjected to a second shaking round at 175 rpm for 5 min. The suspension was centrifuged at 500 × g for 10 min at 4 °C in a horizontal rotor to pellet coarse particles while minimizing cell disruption. The supernatant was decanted and kept on ice, and the pellet was re-extracted with 3 mL of pre-chilled NaP buffer following the same procedure. Five extraction rounds were performed per sample, and all recovered supernatants were pooled. The pooled supernatant was filtered through a sterile 0.22 μm polycarbonate membrane, followed by an additional wash with 10 mL of NaP buffer passed through the same membrane. The membrane was folded, transferred to a 2-mL microcentrifuge tube, immediately snap-frozen in liquid nitrogen, and stored at − 80 °C until DNA extraction. The DNA fraction recovered from the membrane-retained material is operationally defined as the iDNA-enriched fraction throughout this study. Because extracellular DNA removal and taxon-specific recovery were not independently quantified, this fraction should be interpreted as cell-associated DNA enriched by the phosphate-buffer extraction and membrane-filtration workflow, rather than as a validated measurement of exclusively intracellular DNA. This procedure was adapted from the cell-washing and membrane-filtration approach described by Medina Caro et al. [17] and applied to bulk soil.

DNA extraction, PCR amplification, and amplicon sequencing

Direct total DNA (tDNA) was extracted from soil using the MagBeads FastDNA Kit for Soil (cat. no. 116564384; MP Biomedicals, CA, USA) according to the manufacturer’s instructions. For the iDNA-enriched fraction, the 0.22 μm membrane was transferred directly to the extraction tube and processed using the same kit. DNA quantity and quality were assessed by NanoDrop NC2000 spectrophotometry (Thermo Fisher Scientific, Waltham, MA, USA) and 0.8% agarose gel electrophoresis. No extraction or sequencing negative control (blank) was included, and no computational decontamination step was applied. Pre-PCR DNA concentrations were not systematically recorded; because tDNA (from bulk soil) and the iDNA-enriched fraction (from filter membranes) originate from different sample matrices, a direct pre-PCR concentration comparison between the two pools would not be biologically meaningful.

For bacterial community analysis, the V3–V4 region of the 16S rRNA gene was amplified using primers 338 F (5′-ACTCCTACGGGAGGCAGCA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′). For fungal community analysis, the ITS1 region was amplified using primers ITS5 (5′-GGAAGTAAAAGTCGTAACAAGG-3′) and ITS2 (5′-GCTGCGTTCTTCATCGATGC-3′). Sample-specific barcodes were incorporated into the forward primers for multiplex sequencing. All PCR reactions were performed in 25 μL volumes containing Q5 High-Fidelity DNA Polymerase (M0491L; New England Biolabs, Ipswich, MA, USA), 5 μL of 5 × Q5 Reaction Buffer, 5 μL of 5 × Q5 High GC Buffer, 2 μL of 10 mM dNTPs, 1 μL each of 10 μM forward and reverse primer, and 2 μL of template DNA. For bacterial 16S amplification, thermal cycling consisted of an initial denaturation at 98 °C for 5 min, followed by 25 cycles of 98 °C for 30 s, 52 °C for 30 s, and 72 °C for 45 s, with a final extension at 72 °C for 5 min. For fungal ITS amplification, the same programme was used except that the annealing step was 55 °C for 45 s and 30 amplification cycles were performed. Amplicons were purified, quantified using the Quant-iT PicoGreen dsDNA Assay Kit (Invitrogen, Carlsbad, CA, USA), pooled in equimolar amounts, and subjected to library preparation and paired-end sequencing (2 × 250 bp) on an Illumina NovaSeq 6000 platform at Shanghai Personal Biotechnology Co., Ltd. (Shanghai, China).

Bioinformatics and quality control

Raw demultiplexed reads were processed in QIIME 2 (v2024.5) [18]. Adapter and primer sequences were trimmed with the cutadapt plugin [19]. Reads were then quality filtered, denoised, merged, and chimera checked using the DADA2 plugin [20] to generate amplicon sequence variants (ASVs). ASVs were inferred with the DADA2 denoising and chimera-removal steps as implemented in the plugin; no additional dataset-level abundance threshold was applied. ASVs classified as chloroplast or mitochondria were removed from the bacterial dataset (49 ASVs, 0.02% of bacterial reads). Representative bacterial sequences were aligned with MAFFT [21] and used to construct a phylogenetic tree with FastTree2 [22]. Taxonomy was assigned with the QIIME 2 classify-sklearn naïve-Bayes classifier [23] against SILVA release 138 (16S rRNA gene) [24] and UNITE release 9.0 (ITS) [25]. Non-fungal ITS reads, including plant host sequences, were removed by retaining only ASVs classified as Fungi.

Soil physicochemical analyses

All soil physicochemical analyses were conducted by Shanghai Personalbio Technology Co., Ltd. (Shanghai, China). Soil pH, electrical conductivity (EC), ammonium-N (NH₄⁺-N), and nitrate–N (NO₃⁻-N) were quantified for all 15 original soil samples (five biological replicates per depth interval) and used in subsequent inferential statistical analyses. Total nitrogen (TN), total phosphorus (TP), total potassium (TK), soil organic matter (SOM), and gravimetric moisture content were determined on depth-wise composite samples prepared by pooling equal masses from the five replicates within each depth layer (three composite samples in total); these variables are reported as descriptive edaphic background information and were not subjected to inferential testing.

Soil pH and EC were measured electrometrically in soil-to-water extracts [26]. Nitrate–N and ammonium-N were extracted with 1 mol/L KCl and quantified by dual-wavelength UV spectrophotometry [27] and the indophenol blue colorimetric method [26], respectively. Gravimetric moisture content was determined by oven-drying to constant mass [26]. TN was quantified by the Kjeldahl method after sulfuric acid–catalyst digestion [26]. TP and TK were determined after NaOH alkali fusion by molybdenum–antimony spectrophotometry [28] and flame photometry [26], respectively. SOM was quantified by the potassium dichromate oxidation–external heating method [29].

Statistical analyses

All statistical analyses were performed in R (v4.3.3) [30], with bacterial and fungal datasets analyzed independently. Because the between-pool richness contrast is sensitive to library size, each dataset was rarefied to 95% of its minimum library size (61,133 reads per sample for bacteria; 24,650 for fungi) to equalize sequencing effort (per-sample sequencing depths before and after rarefaction are given in Table S4); we note that variance-stabilizing and compositional alternatives exist. Alpha diversity was characterized by Observed ASVs and Shannon diversity index [31], both calculated from the rarefied ASV tables. Because each tDNA and iDNA library derived from the same homogenized soil sample, within-depth comparisons between DNA pools were treated as paired. For each depth, Observed richness and Shannon diversity were compared between pools using paired t-tests on the five matched tDNA–iDNA differences, reporting the mean paired difference (full per-metric paired results in Table S5), its 95% confidence interval and Cohen’s dz; the paired Wilcoxon signed-rank test was used as a distribution-free robustness check (with n = 5 its exact two-sided p cannot fall below 0.0625). p-values were adjusted for multiple testing across the three depths within each kingdom and metric using the Benjamini–Hochberg procedure [32].

Beta diversity was assessed using Bray–Curtis dissimilarities [33] calculated from rarefied ASV tables and visualized by principal coordinates analysis (PCoA). Community composition was tested by PERMANOVA on Bray–Curtis dissimilarities with depth, DNA pool and their interaction (9999 permutations), restricting permutations within paired soils (blocks = original soil sample) to respect the paired design [34]. Multivariate dispersion was assessed with PERMDISP as the average distance of individual samples to their group centroid and compared between DNA pools within each depth interval by permutation [35]. dbRDA models used replicate-level pH, EC, NH₄⁺-N and NO₃⁻-N, and were fitted separately within each DNA pool (n = 15 per model); adjusted R² is reported. Phylum- and genus-level relative-abundance tables were generated for compositional summaries. Genus-level comparisons are reported descriptively and are not significance tested: with five matched pairs per depth the exact two-sided paired signed-rank test cannot yield p < 0.0625, and an unpaired test would discard the paired design. For each depth we therefore report mean and median relative abundance, prevalence, log2 fold-change, the mean paired difference, the paired rank-biserial correlation as a paired effect size, and Cliff’s delta as an unpaired distributional effect size computed across all 25 iDNA-versus-tDNA sample comparisons per depth; all are oriented as the iDNA-enriched fraction relative to tDNA (Tables S7 and S8).

For community–environment analyses, only variables measured across all 15 original soil samples were included in inferential models: pH, EC, NH₄⁺-N, and NO₃⁻-N. Collinearity among environmental variables was screened using variance inflation factors (VIF), and variables with VIF > 10 were excluded prior to modelling. Community-level associations between microbial composition and soil variables were assessed separately for each kingdom using distance-based redundancy analysis (dbRDA) with Bray–Curtis dissimilarities [34]. Because each of the 15 original soil samples yielded one iDNA and one tDNA library but only a single set of soil measurements, a model fitted to all 30 libraries would assign 15 environmental observations to 30 rows; dbRDA was therefore fitted separately within each DNA pool (n = 15 per model). Overall model significance was tested by permutation, and each soil variable was tested marginally, i.e. given the other three (vegan anova.cca with by = "margin"; 9999 permutations; Table S9). Variables measured only on depth-wise composite samples (TN, TP, TK, SOM, and moisture) were used for descriptive characterization of the edaphic gradient and were excluded from all inferential tests. All tests were two-sided, and p < 0.05 was considered statistically significant.

Results

Soil physicochemical background and dataset overview

Across the three depth intervals, the soil profile showed a consistently alkaline and low-conductivity edaphic background. At the individual-replicate level, pH ranged from 8.02 to 8.35 and EC from 96.9 to 128.9 μS cm⁻¹ across individual samples (Additional file 2: Table S1; depth-interval means are summarised in Table 1). Ammonium-N and nitrate–N also varied within relatively narrow ranges across the 15 original soil samples, ranging from 1.11 to 1.79 mg kg⁻¹ and from 4.75 to 8.83 mg kg⁻¹, respectively. Full sample metadata are provided in Table S1, and the underlying soil physicochemical measurements are provided in Table S2A–S2B.

Table 1.

Soil physicochemical properties across three depth intervals in the Lonicera japonica field

Depth (cm)	pH	EC (μS cm−1)	NH4+-N (mg kg−1)	NO3−-N (mg kg−1)	TN† (g kg−1)	TP† (g kg−1)	TK† (g kg−1)	SOM† (g kg−1)	Moisture† (%)
0–3	8.16 ± 0.08	112.9 ± 5.3	1.41 ± 0.24	6.86 ± 1.70	2.20	2.05	11.55	33.94	25.74
5–7	8.12 ± 0.12	110.9 ± 7.7	1.31 ± 0.16	6.81 ± 1.40	1.79	1.89	11.32	31.30	22.42
12–15	8.20 ± 0.09	113.8 ± 12.2	1.26 ± 0.21	6.99 ± 1.33	1.39	1.79	11.21	21.17	20.18

Values for pH, electrical conductivity (EC), ammonium nitrogen (NH4+-N), and nitrate nitrogen (NO3−-N) are presented as mean ± standard deviation (n = 5 biological replicates per depth). †Total nitrogen (TN), total phosphorus (TP), total potassium (TK), soil organic matter (SOM), and moisture were measured on depth-wise composite samples and are presented as composite values without standard deviations. Composite variables were used descriptively and were not included in inferential community–environment analyses

Several composite resource-related variables declined with depth. Soil organic matter decreased from 33.94 g kg⁻¹ at 0–3 cm to 21.17 g kg⁻¹ at 12–15 cm, and total nitrogen decreased from 2.20 to 1.39 g kg⁻¹ over the same depth range. Soil moisture also declined from 25.74% to 20.18%, whereas total phosphorus and total potassium showed smaller decreases. Because TN, TP, TK, SOM, and moisture were measured on depth-wise composite samples, these variables were used descriptively and were not included in inferential community–environment analyses.

The bacterial and fungal amplicon datasets differed substantially in taxonomic breadth. Across the 30 sequencing libraries retained for downstream analyses, the bacterial dataset comprised 77,018 ASVs, whereas the fungal dataset comprised 1,276 ASVs. Bacterial communities were distributed across several major phyla, with Acidobacteriota (19.8%), Proteobacteria (18.8%), Gemmatimonadota (17.5%), Chloroflexi (8.9%), and Actinobacteriota (4.6%) as the dominant lineages. In contrast, fungal communities were more taxonomically concentrated and were dominated by Ascomycota (77.3%), followed by Basidiomycota (14.5%) and Mortierellomycota (6.9%).

DNA-pool choice differentially affected bacterial and fungal alpha diversity

Alpha diversity responded to DNA-pool choice more strongly and more consistently in bacteria than in fungi (per-sample values are provided in Table S3A–S3B). Bacterial Observed ASVs were higher in tDNA than in the iDNA-enriched fraction at all three depths (5/5 pairs at 5–7 cm and 4/5 pairs at 12–15 cm) and differed significantly at 5–7 cm (5149.6 vs 4400.0; mean paired difference 749.6 ASVs, 95% CI 432–1067; dz = 2.93; adjusted p = 0.008); the 0–3 cm and 12–15 cm differences were consistent in direction but not significant (adjusted p = 0.18 and 0.083). Bacterial Shannon diversity did not differ significantly between pools at any depth (adjusted p ≥ 0.12). Across both DNA pools, bacterial richness and Shannon diversity showed a numerical tendency to decline with increasing depth; no formal statistical tests for depth effects on alpha diversity were conducted within the present design.

Fungal differences were confined to the 0–3 cm layer, where tDNA showed higher Observed ASVs (90.6 vs 53.2; mean paired difference 37.4, 95% CI 15.5–59.3; dz = 2.12; adjusted p = 0.027) and Shannon diversity (2.75 vs 2.08; mean paired difference 0.67, 95% CI 0.21–1.14; dz = 1.80; adjusted p = 0.048); no fungal differences were detected at 5–7 or 12–15 cm (adjusted p ≥ 0.56). The restriction of the fungal effect to the 0–3 cm layer is consistent with its higher organic inputs: this layer had the highest SOM content (33.94 g kg⁻¹ vs 21.17 g kg⁻¹ at 12–15 cm; Table 1) and may plausibly receive the greatest input of non-cell-associated fungal DNA (e.g. through hyphal turnover, spore lysis, and root-associated mycelial senescence), although these processes were not measured here and this interpretation remains speculative. In deeper layers, where organic inputs and fungal biomass are lower, the difference between tDNA and the iDNA-enriched fraction may be too small to detect at the present sample size. Bacterial alpha diversity showed a consistent numerical trend toward higher values in tDNA across all depths, with a significant difference at 5–7 cm, whereas fungal alpha-diversity differences were significant only in the uppermost layer.

Community-composition differences between DNA pools were more distinct in bacteria than fungi

Community-composition analyses revealed a more distinct between-pool signal in bacteria than in fungi (Fig. 3A–D). In the bacterial PCoA, samples from the iDNA-enriched fraction and tDNA showed evident separation along PCoA1, which explained 28.1% of the total variation; PCoA2 explained an additional 18.1% (Fig. 3A). PERMANOVA with permutations restricted within paired soils (Table S6) showed that the DNA-pool workflow accounted for a substantial share of the measured bacterial community composition (R² = 0.273, p < 0.001), whereas the depth × DNA-pool interaction was small and not significant (R² = 0.073, p = 0.169). In fungi, the DNA-pool effect was smaller but likewise significant (R² = 0.076, p < 0.001) and the interaction was again non-significant (R² = 0.069, p = 0.151). Depth accounted for R² = 0.160 (bacteria) and R² = 0.097 (fungi) of the variation, but because depth is constant within each original-soil block it is not exchangeable under the blocked scheme; these values are reported descriptively and no permutation test of the depth main effect is claimed (Table S6). Because tDNA and the iDNA-enriched fraction represent two different extraction workflows applied to different physical inputs (bulk soil versus membrane-retained cells), this contrast reflects combined biological and methodological differences between DNA pools rather than a purely biological effect.

Fig. 3.

Beta-diversity patterns across DNA pools and soil depths. A, B Principal coordinates analysis (PCoA) based on Bray–Curtis dissimilarities for bacterial (A) and fungal (B) communities. Ellipses indicate the distribution of samples grouped by DNA pool. C, D Within-group dispersion, quantified as distance to group centroid, for bacterial (C) and fungal (D) communities. Boxplot elements follow the same conventions as in Fig. 2. p values indicate within-depth comparisons of multivariate dispersion between the iDNA-enriched fraction and direct total DNA (tDNA), evaluated using permutation-based tests of homogeneity of multivariate dispersion. Asterisks denote p < 0.05. Colors distinguish DNA pool: blue, iDNA-enriched fraction; red, tDNA. E Within-pair Bray–Curtis distance between the iDNA and tDNA library of each matched soil, shown by depth for bacterial and fungal communities; the bracket and asterisk denote a Benjamini–Hochberg-significant pairwise difference between depths (adjusted p < 0.05)

In contrast, fungal community composition showed considerably more overlap between DNA pools. Fungal samples from the iDNA-enriched fraction and tDNA overlapped broadly in ordination space, and the first two PCoA axes explained lower proportions of variation than in the bacterial dataset (11.9% and 8.6%, respectively; Fig. 3B).

To assess whether DNA-pool choice also changed within-group compositional variability, we compared distances from individual samples to their group centroids. Multivariate dispersion between pools was higher in bacteria at 5–7 cm (PERMDISP p = 0.042, uncorrected for the six depth × kingdom dispersion tests) and marginal at 12–15 cm (p = 0.065), but did not differ at 0–3 cm (p = 0.60); fungal dispersion did not differ between pools at any depth (p ≥ 0.47). Because a difference in dispersion can itself contribute to a significant PERMANOVA result, we note that the clear centroid separation in ordination space (Fig. 3A) indicates that the bacterial DNA-pool effect was primarily a difference in community location, with an additional dispersion component at 5–7 cm. Together, the ordination and dispersion results indicate that DNA-pool choice produced a more distinct compositional signal in bacteria than in fungi under the sampled conditions. In a complementary paired analysis, the within-pair Bray–Curtis distance between the tDNA and iDNA library of the same soil varied with depth in fungi (Kruskal–Wallis p = 0.015; 0–3 vs 12–15 cm adjusted p = 0.037) but not in bacteria (p = 0.18), indicating that the magnitude of the tDNA–iDNA divergence was itself depth-dependent in fungi, being greatest in the surface layer and decreasing with depth (Fig. 3E).

Taxon-specific shifts between tDNA and the iDNA-enriched fraction

DNA-pool choice did not fundamentally replace the dominant taxonomic backbone of either kingdom. Instead, its main effect was expressed as changes in the relative recovery of selected lineages, with more pronounced taxon-level differences in bacteria than in fungi. In the bacterial dataset, the dominant phylum-level structure was broadly conserved across depths and DNA pools. Acidobacteriota, Proteobacteria, Gemmatimonadota, Chloroflexi, and Actinobacteriota remained the major bacterial phyla in both the iDNA-enriched fraction and tDNA (Fig. 4A, B), indicating that the shift between pools did not reflect wholesale community replacement. Class- and genus-level profiles, however, showed redistribution within selected lineages across depth intervals (Additional file 1: Fig. S1A, B and Fig. S2A, B). Among the top 10 bacterial genera by mean relative abundance (Fig. S2A, B), unclassified Vicinamibacterales (Acidobacteriota) and SBR1031 (Chloroflexi) were more abundant in the iDNA-enriched fraction than in tDNA at all three depth intervals, whereas Gemmatimonas and unclassified Gemmatimonadaceae (Gemmatimonadota), RB41 (Acidobacteriota) and candidate phylum NB1-j were consistently more abundant in tDNA. Because five matched pairs per depth cannot support genus-level significance testing, these comparisons are reported descriptively (Methods). Of the 43 bacterial genera examined, 28–33 per depth shifted in the same direction between DNA pools in all five soil pairs (paired rank-biserial r = ± 1), and 24 did so in the same direction at all three depths, indicating a directionally reproducible genus-level signal rather than a depth-restricted one. These results reflect taxon-selective redistribution between DNA pools rather than broadly supported differential abundance; full per-depth summaries and paired effect sizes are provided in Tables S7 and S8.

Fig. 4.

Phylum-level community composition. A Bacterial communities in the iDNA-enriched fraction. B Bacterial communities in direct total DNA (tDNA). C Fungal communities in the iDNA-enriched fraction. D Fungal communities in tDNA. Stacked bars show the mean relative abundance of dominant phyla at three soil depths (0–3, 5–7, and 12–15 cm). Low-abundance phyla are grouped as “Other”. Each bar represents the mean of five biological replicates

In the fungal dataset, Ascomycota dominated across all depths and DNA pools, with Basidiomycota and Mortierellomycota contributing smaller but variable proportions (Fig. 4C, D). Subordinate fungal phyla varied with depth and DNA pool, particularly in the deeper tDNA samples (Fig. 4D). Class- and genus-level profiles showed redistribution among selected fungal lineages (Fig. S1C, D; Fig. S2C, D). Among the top 10 fungal genera (Fig. S2C, D), Humicola and unclassified Sordariales were more represented in tDNA at 0–3 cm, whereas Cercophora was markedly more abundant in the iDNA-enriched fraction at that depth. At the deeper intervals, Aspergillus was more abundant in the iDNA-enriched fraction, while Linnemannia was more abundant in tDNA at every depth. The contrast with bacteria was marked: of the 52 fungal genera examined, only 19–21 per depth shifted in the same direction across all five soil pairs, and just 2 did so at all three depths (versus 24 of 43 bacterial genera), so the fungal genus-level signal was far less reproducible between DNA pools.

Overall, the taxonomic results suggest that DNA-pool choice mainly affected the relative recovery of selected lineages while leaving the dominant community structure largely intact. This pattern supports the broader conclusion that tDNA and the iDNA-enriched fraction provide related but non-identical views of the soil microbial community, with more pronounced taxon-level differences in bacteria.

Soil physicochemical variables explained little of the community variation

Across both kingdoms, the measured soil physicochemical variables were associated with community structure but accounted for only a small fraction of it. The correlation structure among physicochemical variables (Fig. S3) confirmed that TN, TP, TK, SOM, and moisture formed a coherent depth-related cluster, separate from the replicate-level variables used in inferential analyses. Among the four replicate-level variables available for community–environment analyses, correlations were generally moderate. EC was negatively associated with NO₃⁻-N, whereas NH₄⁺-N was positively associated with NO₃⁻-N. In contrast, TN, TP, TK, SOM, and moisture showed stronger positive associations with one another, consistent with a depth-related resource gradient. Because these latter variables were measured only on depth-wise composite samples, they were used descriptively and were not included in inferential dbRDA models.

Inferential community–environment analyses were restricted to pH, EC, NH₄⁺-N, and NO₃⁻-N. Because the two libraries of a soil sample share a single set of soil measurements, dbRDA was fitted separately within each DNA pool (n = 15 per model). All four models explained little community variation (adjusted R² = 0.053–0.186), indicating that most community variation reflects factors not captured by the measured soil chemistry (Table S9).

In bacteria, soil chemistry was associated with community composition in both DNA pools, and somewhat more strongly in tDNA (iDNA-enriched fraction: adjusted R² = 0.133, F = 1.536, p = 0.042, Fig. 5A; tDNA: adjusted R² = 0.186, F = 1.798, p = 0.017, Fig. 5B). The first two constrained axes explained 18.0% and 8.4% of the total variation in the iDNA-enriched fraction, and 23.3% and 8.3% in tDNA. Marginal permutation tests, in which each soil variable is tested given the other three, identified NH₄⁺-N as the only significant variable in the iDNA-enriched fraction (p = 0.010), whereas both NH₄⁺-N (p = 0.002) and NO₃⁻-N (p = 0.028) were significant in tDNA. In both pools, samples showed an apparent tendency toward depth-related positioning along dbRDA1; because depth was not a term in the constrained model, this pattern should be interpreted descriptively.

Fig. 5.

Community–environment relationships within each DNA pool. Distance-based redundancy analysis (dbRDA) of Bray–Curtis dissimilarities constrained by four replicate-level soil variables: pH, electrical conductivity (EC), ammonium nitrogen (NH₄⁺-N), and nitrate nitrogen (NO₃⁻-N). Each DNA pool was modelled separately (n = 15) because the two libraries of a soil sample share one set of soil measurements. A Bacterial iDNA-enriched fraction. B Bacterial tDNA. C Fungal iDNA-enriched fraction. D Fungal tDNA. Point colour denotes DNA pool (blue, iDNA-enriched fraction; red, tDNA); point shape denotes depth interval (circles, 0–3 cm; triangles, 5–7 cm; squares, 12–15 cm). Solid black arrows mark soil variables with a significant marginal effect, i.e. tested given the other three variables (p < 0.05); dashed grey arrows mark non-significant variables. Ellipses outline the 68% normal-probability region of each panel’s samples. Adjusted R² and the global model p value are given above each panel. Full model statistics are in Table S9

In fungi, the constrained models were weaker, and the two DNA pools diverged: the model for the iDNA-enriched fraction did not reach significance (adjusted R² = 0.067, F = 1.249, p = 0.068, Fig. 5C), whereas the tDNA model did (adjusted R² = 0.053, F = 1.195, p = 0.041, Fig. 5D). The first two constrained axes explained 10.8% and 9.3% of the total variation in the iDNA-enriched fraction, and 10.0% and 9.5% in tDNA. In tDNA, pH (p = 0.039) and NH₄⁺-N (p = 0.012) had significant marginal effects; no variable reached significance in the iDNA-enriched fraction (smallest p = 0.052 for NO₃⁻-N). The measured soil chemistry therefore constrained fungal communities more weakly and less consistently than bacterial ones.

Overall, the measured replicate-level soil variables were associated with bacterial and fungal community composition but explained only a small proportion of it in every DNA pool (more than 80% of the variation remained unexplained). Explanatory power was higher for bacterial tDNA than for the bacterial iDNA-enriched fraction (adjusted R² = 0.186 versus 0.133); in fungi the adjusted R² was slightly lower for tDNA (0.053 versus 0.067), although only the fungal tDNA global model reached p < 0.05. Soil physicochemical factors therefore contributed to the observed community patterns, but these analyses provide no evidence that DNA-pool choice fundamentally alters community–environment relationships; rather, it modestly changes how strongly such relationships are detected.

Discussion

DNA-pool choice affected bacterial and fungal community profiles differently

Bacterial community profiles showed a more consistent response to DNA-pool choice than fungal community profiles. In bacteria, tDNA showed higher richness than the iDNA-enriched fraction at all three depths, with a significant paired difference in Observed ASVs at 5–7 cm; bacterial Shannon diversity followed the same direction, though no within-depth comparison reached significance (Fig. 2A, B). Bacterial communities also showed more distinct between-pool separation in Bray–Curtis ordination space, and DNA-pool choice explained a substantial fraction of bacterial community composition (27%) but only a small fraction in fungi (8%). Dispersion between pools was also greatest at 5–7 cm in bacteria (uncorrected PERMDISP p = 0.042; Fig. 3A, C). By contrast, fungal alpha-diversity differences were confined to the shallowest layer (Fig. 2C, D), fungal ordination showed broad overlap between pools, and dispersion did not differ at any depth (Fig. 3B, D). These results indicate that the observed DNA-pool signal was more consistently expressed in bacteria than in fungi.

Fig. 2.

Alpha-diversity responses to DNA-pool choice. A Bacterial Observed ASVs. B Bacterial Shannon diversity index. C Fungal Observed ASVs. D Fungal Shannon diversity index. Boxplots summarize five biological replicates per depth interval and DNA pool; horizontal lines indicate medians, boxes span the interquartile range, whiskers extend to 1.5 × the interquartile range, and points beyond the whiskers represent outliers. p values correspond to within-depth paired comparisons between the iDNA-enriched fraction and direct total DNA (tDNA), evaluated using paired t-tests with Benjamini–Hochberg adjustment across depths within each kingdom and metric. Asterisks denote p < 0.05. Colors distinguish DNA pool: blue, iDNA-enriched fraction; red, tDNA

One possible explanation is that direct tDNA extraction captured a broader mixture of cell-associated and non-cell-associated bacterial DNA, increasing apparent bacterial richness relative to the iDNA-enriched fraction [12, 14]. Because extracellular DNA was not quantified directly, this explanation remains a mechanistic interpretation rather than a direct measurement of the non-cell-associated DNA pool; this caveat applies to all mechanistic interpretations below. The direction of the bacterial alpha-diversity shift and the more distinct between-pool separation in ordination space are consistent with this possibility (Figs. 2A, B and 3A). A related pattern has been reported in forest soil, where RNA-based and total-DNA-based communities differed substantially and showed marked spatial stratification [36]; although the iDNA-enriched fraction does not target transcriptionally active cells, this comparison supports the broader expectation that DNA-pool choice can shift community profiles. The confinement of significant fungal alpha-diversity differences to the 0–3 cm layer (Observed ASVs: 90.6 vs 53.2, adjusted p = 0.027; Shannon: 2.75 vs 2.08, adjusted p = 0.048) may reflect a surface horizon with higher organic matter input and greater fungal biomass turnover, but this mechanism was not measured directly. Alternative explanations—including differences in fungal cell wall architecture, spore persistence, and the kinetics of DNA release and persistence—cannot be excluded [10, 11, 37]. The fungal response was therefore less consistently expressed than the bacterial response, and its mechanism remains an open question.

The expression of DNA-pool effects was not uniform along the soil profile

The effect of DNA-pool choice was not expressed uniformly across the three sampled layers. In bacteria, the largest observed alpha-diversity difference occurred at 5–7 cm, and the largest within-depth dispersion difference between DNA pools (uncorrected p = 0.042) was also observed at this depth (Figs. 2A and 3C). In fungi, by contrast, significant alpha-diversity differences between tDNA and the iDNA-enriched fraction were detected only at 0–3 cm (Fig. 2C, D). The between-pool contrast thus varied with depth rather than representing a uniform methodological offset. The depth-dependence of the DNA-pool effect was expressed in the alpha-diversity pattern — the depth at which pools differed significantly was kingdom-specific (bacteria at 5–7 cm, fungi at 0–3 cm) — rather than as a significant depth × DNA-pool interaction in community composition, which we did not detect under the paired test. This fits the well-documented depth-dependent structuring of soil bacterial communities, where diversity and biomass decline with depth alongside gradients in carbon availability [38, 39].

This interpretation is consistent with, but not proven by, the measured edaphic background. Although pH and EC varied only within relatively narrow ranges, SOM, TN, and moisture declined from the shallowest to the deepest layer, indicating a depth-related reduction in resource-related conditions (Table 1). Such vertical heterogeneity may influence the relative contribution of currently cell-associated DNA, recently released DNA, and more persistent non-cell-associated DNA [10, 11]. Such gradients in carbon availability and biomass alter both lineage-level composition and community similarity across depths [38, 39], suggesting that similar gradients modulate accumulation and persistence of non-cell-associated DNA. Direct support for this interpretation comes from a study of soil bacterial communities across a climate and depth gradient in Chile, where depth was found to influence the iDNA community significantly but not the eDNA community, with the authors attributing this asymmetry to depth-related differences in metabolic activity and cell-associated DNA abundance [40]. This matches the pattern observed here, where the bacterial DNA-pool contrast was strongest at 5–7 cm and not statistically significant at 12–15 cm, though still directionally consistent (Fig. 2A), although the mechanism in this alkaline, cultivated soil may differ from that in a climate-gradient setting. The available edaphic variables do not directly quantify the non-cell-associated DNA pool, so the link between SOM and TN gradients and depth-dependent pool effects remains inferential. The magnitude of the DNA-pool effect thus differed across depths—most directly shown by the depth-dependent within-pair divergence in fungi (Fig. 3E)—whereas the formal depth × DNA-pool interaction in the beta-diversity model was not significant in either kingdom.

DNA-pool choice changed selected taxonomic and soil-association patterns

DNA-pool choice did not broadly replace the dominant community backbone; rather, it altered the relative recovery of selected lineages. At the phylum level, both bacterial and fungal communities retained broadly similar dominant structures between tDNA and the iDNA-enriched fraction (Fig. 4A–D), but class- and genus-level profiles showed more visible redistribution within selected lineages, particularly in the bacterial dataset (Figs. S1A, B and S2A, B). This is ecologically relevant because community-level interpretation in soil microbiome studies often depends not only on whole-community diversity metrics but also on which taxa are emphasized at lower taxonomic ranks [12, 14].

The community–environment results reinforce this picture. Replicate-level soil variables, including pH, EC, NH₄⁺-N, and NO₃⁻-N, were associated with low-explanatory-power dbRDA models in both DNA pools of both kingdoms, although the model for the fungal iDNA-enriched fraction did not reach significance (Fig. 5A–D). The fitted pH vectors in the bacterial and fungal ordinations are consistent with the well-established role of soil pH as a primary determinant of bacterial community composition across broad spatial scales [41, 42]; even within the narrow pH range of the present soil profile, pH-correlated variation in lineage-level community structure is plausible. Inorganic nitrogen availability similarly modulates soil microbial community composition and diversity [43], providing a basis for the NH₄⁺-N and NO₃⁻-N vectors in the constrained ordinations. Formally, however, pH reached a significant marginal effect only in the fungal tDNA model, and NO₃⁻-N only in the bacterial tDNA model. NH₄⁺-N was the only variable with a significant marginal effect in an iDNA-enriched model (bacteria), and it was significant in both tDNA models, making it the most consistent single correlate of community structure across pools (Table S9). The low adjusted R² values, however, indicate that the measured variables accounted for only a small proportion of total community variation, and the constrained ordinations recovered only weak environmental structure within either DNA pool, particularly in fungi (Fig. 5C, D). DNA-pool choice thus affected which diversity and taxonomic patterns are emphasised, particularly in bacteria, but did not fundamentally alter community–environment relationships across the sampled conditions.

tDNA and iDNA-enriched profiles thus offer complementary but non-interchangeable perspectives on the same community. Direct tDNA captures both cell-associated and non-cell-associated DNA, whereas the iDNA-enriched fraction approximates the cell-associated pool without being equivalent to a measurement of viable or metabolically active cells [12, 17]. This distinction is most relevant when studies compare soil depths, microbial kingdoms, or low-biomass matrices in which non-cell-associated DNA persistence or cell-recovery bias could change apparent richness, beta-diversity, and selected lineage abundances.

Our operationally defined iDNA-enriched fraction differs from viability-based approaches such as propidium monoazide (PMA) treatment or DNase digestion, which chemically or enzymatically remove extracellular or non-viable-cell DNA before amplification. Because our phosphate-buffer/membrane protocol was not benchmarked against these methods or against spike-in recovery controls, the fraction should be compared with, rather than equated to, PMA- or DNase-based relic-DNA removal; direct methodological comparison is a useful direction for future work [12, 13]. The most consistent finding was a kingdom-level asymmetry. Bacteria showed higher Observed ASVs in tDNA at every depth, with a significant difference at 5–7 cm, and more distinct between-pool separation and dispersion differences in ordination space. The fungal response was more localised: alpha-diversity differences were significant only at 0–3 cm, ordination showed broad pool overlap, and dispersion did not differ at any depth. For community profiling in cultivated L. japonica soils, this means DNA-pool choice matters across the full profile for bacteria, but mainly in the surface layer for fungi.

Limitations

Several limitations should define the scope of interpretation without invalidating the central comparison. First, the membrane-retained fraction was operationally defined as an iDNA-enriched, cell-associated fraction, and extracellular DNA removal, cell recovery efficiency, and taxon-specific recovery bias were not independently quantified. This means that differences between tDNA and the iDNA-enriched fraction should be interpreted as differences between two operational DNA-pool workflows—which may partly reflect extraction efficiency and cell-recovery bias rather than the relic-versus-intracellular contrast alone—and not as a direct measurement of viable or metabolically active microorganisms. Second, because each tDNA and iDNA library derived from the same homogenized soil sample, we analyzed the data as the paired design it is; paired tests increase power relative to unpaired comparisons, although with five pairs per depth the resolution of the nonparametric paired test is limited, so we emphasize effect sizes and the consistency of direction alongside adjusted p-values. Third, TN, TP, TK, SOM, and moisture were measured only on depth-wise composite samples and were used descriptively rather than inferentially; consequently, the discussion of these variables is limited to depth-related background rather than replicate-level driver testing. Fourth, field-management metadata and exact geographic coordinates were unavailable, and the study was restricted to a single Lonicera japonica field sampled once. Seasonal plant growth, biomass turnover, fertilization, irrigation, or manure inputs could alter the relative contribution of cell-associated and non-cell-associated DNA in agricultural soils, so the results should not be generalized beyond similar cultivated-field settings without further validation. Because no extraction or sequencing blank was included and no computational decontamination was applied, low-level reagent or laboratory contamination cannot be assessed; this is of particular concern for the low-biomass iDNA-enriched fraction, and low-abundance taxa should therefore be interpreted with caution. Finally, we compared only taxonomic amplicon profiles; whether these compositional differences translate into functional differences was not tested and cannot be inferred from taxonomy alone.

Notwithstanding these constraints, the combined alpha-diversity, beta-diversity, taxonomic-composition, and dbRDA results consistently show that DNA-pool choice changes amplicon-based profiles of bulk-soil microbial communities in this L. japonica field. Direct tDNA extraction remains informative for broad bulk-soil profiling, but when the objective is to enrich for the cell-associated fraction of the soil microbiome, tDNA and iDNA-enriched profiles should not be treated as interchangeable without validation in the target soil matrix and depth range.

Conclusions

DNA-pool choice influenced amplicon-based diversity and composition profiles of bulk-soil microbial communities in the Lonicera japonica field, but the effect was not uniform across microbial kingdoms or soil depths. Relative to fungi, bacterial communities showed more consistent responses to DNA-pool choice, with higher richness in tDNA across the soil profile, a significant paired richness difference at 5–7 cm, and more distinct separation between DNA pools in Bray–Curtis ordination space. In contrast, fungal paired alpha-diversity differences between DNA pools were significant only in the 0–3 cm layer (Observed ASVs: 90.6 vs 53.2, adjusted p = 0.027; Shannon: 2.75 vs 2.08, adjusted p = 0.048), with no significant differences at greater depths, and fungal communities showed substantial between-pool overlap across the profile. The localisation of the fungal DNA-pool effect at the surface layer may be related to higher organic matter input and biomass turnover in this horizon, although this mechanism was not directly tested.

At the taxonomic level, DNA-pool effects were expressed mainly as selective redistribution within specific lineages rather than as broad replacement of the dominant community backbone. Community–environment analyses showed that replicate-level soil variables explained a small fraction of community variation, with no clear evidence that DNA-pool choice fundamentally altered these relationships.

Direct total DNA and an iDNA-enriched fraction thus provide related but non-interchangeable, kingdom- and depth-dependent views of bulk-soil microbial communities in this cultivated perennial cropping system, with the divergence between DNA pools greatest for bacteria and in the surface soil. Direct total-DNA extraction remains informative for broad profiling, but when the goal is to enrich for the cell-associated fraction, the DNA-pool choice should be validated for the target soil matrix, kingdom, and depth range rather than treated as methodologically neutral. Our results suggest that DNA-pool choice may influence ecological interpretation in cultivated Lonicera japonica soils, and should be considered in the design of agricultural soil microbiome studies — particularly when comparing depths, microbial kingdoms, or matrices expected to differ in microbial biomass and non-cell-associated DNA persistence.

Abbreviations

ASV

Amplicon sequence variant

dbRDA

Distance-based redundancy analysis

EC

Electrical conductivity

iDNA

Intracellular-DNA-enriched fraction

ITS

Internal transcribed spacer

PCoA

Principal coordinates analysis

PERMANOVA

Permutational multivariate analysis of variance

QIIME 2

Quantitative Insights Into Microbial Ecology 2

SOM

Soil organic matter

tDNA

Direct total soil DNA

TK

Total potassium

TN

Total nitrogen

TP

Total phosphorus

Authors’ contributions

CM performed the experiments, analyzed the data, prepared the figures, and drafted the manuscript. XW conceived and supervised the study, revised the manuscript, and approved the final version. Both authors read and approved the final manuscript.

Funding

This work was supported by the Student Research Development Program (SRDP) of Ocean University of China [grant number 202510423295X].

Data availability

The raw 16S rRNA gene and ITS amplicon sequencing data generated in this study have been deposited in the NCBI BioProject database under accession PRJNA1404232. Associated SRA records will be publicly available upon release by NCBI.

Declarations

Ethics approval and consent to participate

Not applicable. This study did not involve human participants, human data, human tissue, animals, or clinical material.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.