Molecular biomass and MetaTaxogenomic assessment of soil microbial communities as influenced by soil DNA extraction procedure.

Three soil DNA extraction procedures (homemade protocols and commercial kit) varying in their practicability were applied to contrasting soils to evaluate their efficiency in recovering: (i) soil DNA and (ii) bacterial diversity estimated by 16S rDNA pyrosequencing. Significant differences in DNA yield were systematically observed between tested procedures. For certain soils, 10 times more DNA was recovered with one protocol than with the others. About 15,000 sequences of 16S rDNA were obtained for each sample which were clustered to draw rarefaction curves. These curves, as well as the PCA ordination of community composition based on OTU clustering, did not reveal any significant difference between procedures. Nevertheless, significant differences between procedures were highlighted by the taxonomic identification of sequences obtained at the phylum to genus levels. Depending on the soil, differences in the number of genera detected ranged from 1% to 26% between the most and least efficient procedures, mainly due to a poorer capacity to recover populations belonging to Actinobacteria, Firmicutes or Crenarchaeota. This study enabled us to rank the relative efficiencies of protocols for their recovery of soil molecular microbial biomass and bacterial diversity and to help choosing an appropriate soil DNA extraction procedure adapted to novel sequencing technologies.

Introduction

During the last two decades, novel molecular methods have been developed that are well‐suited to the characterization of soil microbial communities as they provide access to previously hidden genetic resources (for review see Torsvik and Øvreås, 2002). These methods are based essentially on characterizing soil DNA and significant efforts have been devoted to optimize the soil Dn class="Chemical">NA extraction procedure in order to obtain suitable representative extracts for quantitative and qualitative characterization of the microbial communities (Zhou ; Kuske ; Delmont ). These efforts led to the development of various homemade DNA extraction protocols, as well as commercial kits, which have been used in more than 1000 articles published yearly. Microbial abundance and community fingerprinting have already been assessed in various soils to compare the technical biases of these procedures (Martin‐Laurent ; Ranjard ). However, the recent development of high‐throughput sequencing technology (such as 454 or Illumina) allows the scientific community to assess ‘MetaTaxogenomic’ studies by getting hundreds of thousands of ribosomal rDNA gene sequences from a single metagenomic DNA (Roesch ; Will ; Maron ). The impressive power of these emerging tools calls for a more thorough examination of the relative efficiencies of different DNA extraction techniques to recover microbial diversity and discriminate soil bacterial composition as well as to detect the rare species.

Delmont and colleagues (2010) recently recommended the combination of several soil sampling and DNA extraction strategies to access the whole n class="Species">metagenome in terms of species richness. However, such a strategy is time‐ and cost‐consuming and inappropriate studying a large number of samples in wide extent studies. In this context, we aimed at comparing classical but different soil DNA extraction procedures for their recovery of soil DNA and ability to estimate bacterial diversity in seven different soils to deduce the most efficient and useful procedures for soil ‘MetaTaxogenomic’ studies. The seven soils were originated from the soil library of the French Soil Quality Monitoring Network (Dequiedt ) and were chosen for their contrasting physico‐chemical and land‐use characteristics (Table 1), known to strongly influence microbial abundance and diversity (Ranjard ). Three DNA extraction procedures were chosen for their recurrent use in the literature (Ultraclean soil DNA kit, MOBIO Laboratories) and for their strong optimization for soil bacterial diversity retrieval (GnS‐GII and Sy3). The GnS‐GII procedure was developed by the platform GenoSol to extract soil DNA in large‐scale soil surveys (Ranjard ). It is based on a combination of mechanical (FastPrep® bead‐beating, strongly optimized by V. Nowak and M. Lelievre, pers. comm.) and chemical (SDS) lyses of indigenous cells. The Sy3 procedure has been optimized for low bacterial density samples and gives a good DNA recovery of different bacterial phyla (like α‐, β‐, γ‐Proteobacteria, Firmicutes and Actinobacteria) (Maron ). It is based on a combination of mechanical (bead‐beating with Mikrodismembrator, BBraun Biotech), chemical (SDS), thermal (freeze‐thaw cycle) and enzymatic (proteinase K) lyses of indigenous cells. The Ultraclean soil DNA kit (MOBIO) is a commercial kit based on mechanical (bead‐beating) and chemical lyses (see ‘Experimental procedures’ in the Supporting Information file for a precise description of each procedure). For each protocol, DNA was extracted from one gram of soil (dry weight) according to the prerequisite for soil sampling size (Ranjard ). Regarding the detailed procedure of the lyses step, an order of practicability (including cost, fastness and technical levels) could be deduced as follows: MOBIO > GnS‐GII > Sy3.

Table 1

Physico‐chemical characteristics and land‐use of studied soils.

Soils	Clay	Silt	Sand	Corg	C:N	pH	Calcium carbonate	Land use
399	258	394	348	1.8	9.1	8.2	804	Crop system in rotation with grassland
676	318	571	111	0.34	4.6	7.9	30.1	Crop system in rotation without grassland
2034	477	449	74	10.8	15.3	7.9	39.1	Deciduous forest
1850	174	128	698	1.7	14.4	7.8	2.2	Deciduous forest
1868	38	32	930	1.8	27.7	4	0	Coniferous forest
1901	707	275	18	4.9	10.7	5.4	0	Grassland
2116	322	405	273	0.6	7.2	8.3	134	Vineyards

Values are given in g kg−1 soil (dw), except for C:N and pH.

Results and discussion

The first step in the evaluation of the various DNA extraction procedures was based on the comparison of soil Dn class="Chemical">NA yields which were quantified as previously described (Bernard ). A significantly (P < 0.05) higher amount of soil DNA (in some cases, almost 10 times more) was recovered from all soils with GnS‐GII, as compared with the Sy3 and MOBIO procedures (Table 2). It is important to note that the improvement of DNA yield is not combined with an alteration of size integrity (as visually observed on agarose gel, data not shown). Based on our results the ranking of the protocol according to their efficiency for soil DNA recovery was: GnS‐GII >> Sy3 > MOBIO. This suggested that the most stringent cell lysis procedure in Sy3 (including mechanical, chemical, thermal and enzymatic steps) did not provide the highest DNA yield, maybe due to its inadequacy in the soil matrix and/or a possible alteration of DNA integrity. On the other hand, the highest efficiency of the GnS‐GII procedure might be explained by the mechanical lysis step, which was strongly optimized in terms of type and size of the glass‐beads as well as in terms of the strength and duration of grinding using FastPrep® ‐24 (V. Nowak and M. Lelièvre, pers. comm., see ‘Experimental procedures’ in the Supporting Information file). FastPrep® ‐24 grinding was carried out in three dimensions and improved cell lysis efficiency. Indeed, a significant increase in DNA recovery was obtained (data not shown) when the bead‐beating procedures in the MOBIO and Sy3 protocols were replaced by those of GnS‐GII, thereby confirming the importance of this step in the DNA extraction procedure.

Table 2

Numbers of filtered sequences produced by pyrosequencing the V4 region of the 16S rDNA gene; richness was estimated with numbers of taxonomic groups found at different taxonomic levels within soil samples, according to the DNA extraction procedure used; Evenness index was determined with taxonomy assigned at the genus level. Soil DNA recovery for each sample is also indicated.

Sample	Soil DNA recovery µg g−1 soil (dry weight) (± standard deviation)	Sequences after equalization and filtering steps	Richness: number of taxonomic groups found (Evenness index)
			Phyla	Class	Order	Family	Genus
399 GnS‐GII	5.61 (± 0.98)	13 420	29	57	126	195	292 (0.99)
399 MoBio	0.43 (± 0.04)	13 750	24	56	122	183	275 (0.98)
399 Sy3	1.77 (± 0.43)	13 268	29	60	129	196	285 (1.00)
676 GnS‐GII	5.1 (± 1.16)	13 428	29	58	129	196	284 (0.93)
676 MoBIO	1.01 (± 0.28)	13 772	33	63	128	188	278 (0.97)
676 Sy3	1.11 (± 0.48)	13 475	29	65	138	211	302 (0.98)
2034 GnS‐GII	8.21 (± 1.21)	13 415	24	49	114	170	271 (0.97)
2034 MoBIO	0.85 (± 0.12)	13 232	22	49	105	162	252 (0.95)
2034 Sy3	1.42 (± 0.59)	13 299	27	52	110	168	258 (0.89)
1850 GnS‐GII	4.36 (± 0.45)	13 476	24	53	120	177	269 (0.95)
1850 MoBIO	0.44 (± 0.07)	13 740	21	55	115	167	243 (0.89)
1850 Sy3	1.35 (± 0.21)	7 806	25	51	101	153	212 (0.90)
1868 GnS‐GII	1.15 (± 0.11)	14 391	20	31	52	83	127 (0.70)
1868 MoBIO	0.26 (± 0.05)	14 472	18	28	48	68	106 (0.72)
1868 Sy3	0.61 (± 0.08)	10 347	20	34	57	86	125 (0.76)
1901 GnS‐GII	1.66 (± 0.13)	14 217	19	43	87	122	174 (0.74)
1901 MoBIO	0.35 (± 0.15)	14 265	15	33	66	94	129 (0.50)
1901 Sy3	0.99 (± 0.24)	14 149	21	48	91	130	182 (0.70)
2116 GnS‐GII	1.20 (± 0.39)	13 596	24	54	122	187	277 (0.96)
2116 MoBIO	0.05 (± 0.001)	0	NA	NA	NA	NA	NA
2116 Sy3	0.58 (± 0.03)	13 805	27	59	123	189	275 (0.96)

Numbers in italic indicate the significant differences between soil DNA recovery and soils for a given procedure. Bold numbers indicated the samples with the highest numbers of detected taxonomic groups for each studied soil between protocols for each taxonomic level.

NA, not applicable.

Numbers of filtered sequences produced by pyrosequencing the V4 region of the 16S rDn class="Chemical">NA gene; richness was estimated with numbers of taxonomic groups found at different taxonomic levels within soil samples, according to the DNA extraction procedure used; Evenness index was determined with taxonomy assigned at the genus level. Soil DNA recovery for each sample is also indicated.

Numbers in italic indicate the significant differences between soil Dn class="Chemical">NA recovery and soils for a given procedure. Bold numbers indicated the samples with the highest numbers of detected taxonomic groups for each studied soil between protocols for each taxonomic level.

Significant variations in DNA recovery between soils were observed with the GnS‐GII protocol. The greatest amount of soil Dn class="Chemical">NA (8.21 ± 1.21 µg DNA g−1 soil) was obtained from an alkaline fine‐textured soil with high organic carbon content and under deciduous forest (soil 2034), whereas the smallest amounts were detected in an acid sandy soil type with high C:N ratio and under coniferous forest (1.15 ± 0.11 µg DNA g−1 soil, soil 1868), and in an alkaline silty‐clay soil under vineyards (1.20 ± 0.39 µg DNA g−1 soil, soil 2116). As already demonstrated in a large‐scale study, soil DNA recovery is positively correlated with soil characteristics such as texture (silt content r = 0.56, P < 0.011), organic carbon content (r = 0.549, P < 0.010) and pH (r = 0.564, P < 0.008) (Marstorp ; Dequiedt ). In addition, our results confirm the significant impact of land use, the amount of soil DNA being much lower under coniferous forest and vineyards, independently of soil characteristics (Dequiedt ). Similar trends between soils were also observed for the MOBIO and Sy3 protocols but the differences were less significant. Consequently, in a context where the amount of soil DNA might be used as a robust indicator of soil microbial biomass (Marstorp ; Dequiedt ), our study underlined the need to choose a reliable DNA extraction procedure suitable for sensitive and early detection of quantitative changes due to soil characteristics and management.

Soil bacterial diversity was assessed by pyrosequencing the V4 region of 16S rDNA using the genome sequencer FLX system, with long‐read n class="Disease">GS FLX Titanium chemistry (Roche Applied Science, Indianapolis, Indiana) at the Genoscope center (Evry, France) (see ‘Experimental procedures’ in the Supporting Information). The raw sequence libraries were first equalized and filtered to remove reads originating from sequencing errors or putative chimeric sequences (see ‘Experimental procedures’ in the Supporting Information file for detailed bioinformatic procedure). As a result, the number of high quality sequences ranged from 13 232 to 14 472 (average read length: 256 bp), except for three soils which gave 10 000 or fewer raw sequences, or no PCR amplification (Table 2). The final step included a clustering of rare reads to abundant reads, with k = six differences, but not counting differences in homopolymer lengths, which are a major source of errors with the 454 technology. Rarefaction curves were drawn with the obtained clusters (at k = 6 differences, Fig. 1) and showed that: (i) saturation was reached for all soils and for each DNA extraction procedure, thereby indicating that about 15 000 sequences were sufficient to obtain a representative covering of soil bacterial diversity; and (ii) significant differences between the DNA extraction procedures were recorded in the slope and level of the plateau of the curve for three soils (1850, 1901 and 1868, Fig. 1). However, it was difficult to deduce a general ranking of the procedures according to their efficiency to recover OTUs. Indeed, GnS‐GII and MOBIO were more efficient for soil 1850, Sy3 and GnS‐GII for soil 1901 and Sy3 for soil 1868.

Figure 1

Rarefaction curves determined by pyrosequencing of V4 region of the 16S rDNA gene obtained for each soil with different soil DNA extraction procedures, based on randomly selected and filtered sequences. Rarefaction curves were determined using clustering at k = 6 differences between sequences (see Supporting Information for more information). Black line: GnS‐GII; light grey line: MOBIO; dark grey line: Sy3.

Rarefaction curves determined by pyrosequencing of V4 region of the 16S rDNA gene obtained for each soil with different soil Dn class="Chemical">NA extraction procedures, based on randomly selected and filtered sequences. Rarefaction curves were determined using clustering at k = 6 differences between sequences (see Supporting Information for more information). Black line: GnS‐GII; light grey line: MOBIO; dark grey line: Sy3.

Bacterial OTU matrices were subjected to principal component analysis (PCA) to obtain an ordination on a factorial map and compare the soil DNA procedures based on OTU composition (Fig. 2). The discrimination between soils was similar irrespective of the soil Dn class="Chemical">NA extraction protocol used, with soil 1868 being discriminated from soil 1901 and from the other soils. This result suggests that although various soil DNA extraction procedures could lead to different soil DNA yields for a given soil, the procedures led to a similar rough discrimination between soils, based on the global composition of the bacterial community. This observation is in accordance with the results of previous studies involving DNA fingerprinting approach which concluded that all procedures were biased and provided a robust global discrimination of bacterial community structure (Martin‐Laurent ). Although only a limited number of soils were studied, it is interesting to note that the soil bacterial community discrimination revealed a strong influence of soil pH since soils 1868 and 1901 were significantly discriminated and more acidic than the other soils (Table 1). Therefore, our results corroborate the high structuring effect of soil pH on soil bacterial community composition (Fierer and Jackson, 2006; Griffiths ).

Figure 2

Principal component (PC1xPC2) plots generated from the clustering of 16S rDNA OTU sequences at k = 6 obtained from the different soils using the various soil DNA extraction procedures (black square: GnS‐GII, grey circle: Sy3, white triangle: MOBIO).

Principal component (PC1xPC2) plots generated from the clustering of 16S rDn class="Chemical">NA OTU sequences at k = 6 obtained from the different soils using the various soil DNA extraction procedures (black square: GnS‐GII, grey circle: Sy3, white triangle: MOBIO).

The obtained reads were taxonomically identified at phyla, class, order, family and genus levels using a previously published procedure (Stoeck ; Behnke ), that had been slightly modified to incorporate the Silva 104 reference sequences (see Supporting Information for the detailed bioinformatic pipeline). Interestingly, application of the described method allowed the affiliation of as much as about 98% of the obtained sequences to known bacterial phyla (Fig. S1). Furthermore, significant differences in bacterial richness were observed between DNA extraction procedures for all soils (Table 2). At the phylum level, the GnS‐GII and Sy3 procedures always detected a higher number of taxa than MOBIO, excepted for soil 676. At the genus level, the GnS‐GII protocol was the most efficient in recovering the highest number of taxa for all soils, except for two soils (676 and 1901) where Sy3 and GnS‐GII gave similar results. Differences between the most and the least efficient procedures for recovering richness were soil dependent and ranged from 1% (soil 2116) to more than 26% (soil 1901). Thus, the deduced ranking of bacterial richness recovery efficiency was GnS‐GII > Sy3 > MOBIO that might partly explain the higher Dn class="Chemical">NA recovery obtained with the GnS‐GII procedure. On the other hand, when evenness was calculated at the genus level, the results implied that the variations between procedures were less significant than for richness (Table 2). This could suggest that the global pattern of population frequency was only slightly influenced by the DNA extraction procedure as discussed above for PCA analysis (Fig. S1).

Pairwise comparison of the relative abundances of identified taxa between procedures suggested that the higher bacterial richness obtained with GnS‐GII could mainly be explained by its greater efficiency to recover particular phyla. Figure 3A–C clearly indicated that MOBIO underestimated the dominant Proteobacteria and Acidobacteria groups and that Sy3 underestimated Actinobacteria. Furthermore, the GnS‐GII procedure was better to detect some phyla, such as Firmicutes and Crenarchaeota (Archaeabacteria) (Fig. 3D–F). This observation was highly significant for several classes of Firmicutes (Bacilli and Clostridia) (Fig. S1 and Table S1). Both Crenarchaeota and Firmicutes are known to be recalcitrant to detection due to their particular habitat, physiology or their spore forming ability (Fierer ; Bates ). Altogether, these data confirm that the lysis stringency developed for the GnS‐GII protocol led to a better detection of these taxa and allowed the access to a more genetically diverse proportion of the soil metagenome. From an ecological point of view Crenarchaeota, Firmicutes and Acidobacteria are known to be abundant and ubiquitous in soil environments and to be of major interest in soil functioning (Jones ; Bates ). Therefore, the ranking of procedures (GnS‐GII > Sy3 > MOBIO) deduced from the efficiency of bacterial richness recovery could facilitate the choice of an adequate procedure which, in turn, might have ecological relevance in a context where the major challenge in soil microbial ecology is to better understand community assembly and to link microbial diversity to ecosystem goods and services (Maron ).

Figure 3

Scatter plot comparison of relative abundance of each detected phylogenetic group in the studied soils (399: ◆, 676:●, 2034:▾, 1850:▴, 1868: , 1901: ◂ and 2116: ▸) with each procedure (GnS‐GII, MOBIO and Sy3). Only the 14 most abundant phyla are plotted for the comparison. (A), (B) and (C) scatter plots show the dominant phyla detected; and (D), (E) and (F) scatter plots the minor phyla. (A) and (D): GnS‐GII versus MOBIO; (B) and (E): GnS‐GII versus Sy3; (C) and (F): MOBIO versus Sy3. Dotted lines are unity lines.

Significant differences were also observed when a given procedure was used to compare bacterial diversity between different soils (Table 2). For example, the acidic soils 1868 and 1901 harboured the lowest bacterial richness, thereby corroborating the positive correlation between pH and bacterial richness previously depicted by Fierer and Jackson (2006). In addition, these soils exhibited a very different pattern of detected phylogenetic groups (Fig. S1 and Table S1) with several particular dominant phyla such as Verrucomicrobia for soil 1901. This phylum is known to be one of the most common and diverse phyla in soil habitats (Kielak ) and is commonly found in acidic grassland soils (Kuramae ). Furthermore, the lowest bacterial diversity observed in soil 1868 might also result from the coarse texture, low organic carbon content, high C:n class="Chemical">N ratio and a coniferous land cover (Table 1). All these parameters are known to be deleterious for soil microbial diversity (Lejon ; Dequiedt ).

Altogether, these results indicate that our biological vision of soil microbial abundance and biodiversity might be limited by the procedures used to recover soil metagenomic DNA. This statement has become critical with the recent expansion of 16S rDn class="Chemical">NA pyrosequencing technologies, which allow a more profound investigation of microbial diversity. Technically, our study confirms the pivotal importance of cell lysis in the soil DNA extraction procedure. It also emphasizes the need for incessant technical surveys to increase species richness per sequencing effort, without increasing the cost and practicality of the procedure, to be useful for wide extent studies. Consequently, we need to revisit our choice of extraction protocol to ensure that the soil DNA recovered is not only of good quality but also sufficiently representative in terms of richness and evenness of the bacterial populations. However, fungal communities constitute a significant portion of soil biodiversity and are also a key component in soil processes. In order to draw meaningful conclusions about the representativeness of the extracted soil DNA, our study could therefore be completed by an evaluation of fungal diversity recovery. This statement is of particular relevance in the current context where soil microbial ecology studies are carried out to better understand community assembly in space and time and its link with soil functioning and more widely with ecosystem services.