Limits...
Quantifying the biases in metagenome mining for realistic assessment of microbial ecology of naturally fermented foods

View Article: PubMed Central - PubMed

ABSTRACT

Cultivation-independent investigation of microbial ecology is biased by the DNA extraction methods used. We aimed to quantify those biases by comparative analysis of the metagenome mined from four diverse naturally fermented foods (bamboo shoot, milk, fish, soybean) using eight different DNA extraction methods with different cell lysis principles. Our findings revealed that the enzymatic lysis yielded higher eubacterial and yeast metagenomic DNA from the food matrices compared to the widely used chemical and mechanical lysis principles. Further analysis of the bacterial community structure by Illumina MiSeq amplicon sequencing revealed a high recovery of lactic acid bacteria by the enzymatic lysis in all food types. However, Bacillaceae, Acetobacteraceae, Clostridiaceae and Proteobacteria were more abundantly recovered when mechanical and chemical lysis principles were applied. The biases generated due to the differential recovery of operational taxonomic units (OTUs) by different DNA extraction methods including DNA and PCR amplicons mix from different methods have been quantitatively demonstrated here. The different methods shared only 29.9–52.0% of the total OTUs recovered. Although similar comparative research has been performed on other ecological niches, this is the first in-depth investigation of quantifying the biases in metagenome mining from naturally fermented foods.

No MeSH data available.


Related in: MedlinePlus

Variation in the recovery of eubacterial community structure as an effect of varied DNA extraction principles on different fermented food types.(a) Venn diagrams using three-fold rotationally symmetrical circles with unique and shared OTUs for different extraction methods are shown. Values in the parentheses indicate extraction methods and numbers inside each region indicate the number of unique or shared OTUs. (b) The canonical PCA was performed using log transformed (log xi + 1) values of species-level OTU relative abundance data. A significant difference in the microbial community structure between extraction methods is highlighted by ANOSIM, performed with 10,000 replicates using Bray-Curtis distances. Circles with solid colors denote data points for each extraction method. (c) Species-sample biplot of RDA shows the eubacterial species direction towards different extraction methods. Solid black arrows indicate species directions towards the method and their abundance. La: Lactobacillus acetotolerans, Bs: Bacillus subtilis, Gf: Gluconobacter frateurii, Tt: Thermus thermophilus, Ck: Clostridium kluyveri and Pv: Proteus vulgaris. (d) The significant difference in the mean relative abundance of major taxa between extraction methods as calculated by Student’s two-tailed paired t-test is shown. Other significant differences at various taxonomic levels are shown in Supplementary Fig. S1, Tables S6 and S7.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5037447&req=5

f3: Variation in the recovery of eubacterial community structure as an effect of varied DNA extraction principles on different fermented food types.(a) Venn diagrams using three-fold rotationally symmetrical circles with unique and shared OTUs for different extraction methods are shown. Values in the parentheses indicate extraction methods and numbers inside each region indicate the number of unique or shared OTUs. (b) The canonical PCA was performed using log transformed (log xi + 1) values of species-level OTU relative abundance data. A significant difference in the microbial community structure between extraction methods is highlighted by ANOSIM, performed with 10,000 replicates using Bray-Curtis distances. Circles with solid colors denote data points for each extraction method. (c) Species-sample biplot of RDA shows the eubacterial species direction towards different extraction methods. Solid black arrows indicate species directions towards the method and their abundance. La: Lactobacillus acetotolerans, Bs: Bacillus subtilis, Gf: Gluconobacter frateurii, Tt: Thermus thermophilus, Ck: Clostridium kluyveri and Pv: Proteus vulgaris. (d) The significant difference in the mean relative abundance of major taxa between extraction methods as calculated by Student’s two-tailed paired t-test is shown. Other significant differences at various taxonomic levels are shown in Supplementary Fig. S1, Tables S6 and S7.

Mentions: Illumina MiSeq barcoded amplicon sequencing of the SSU rRNA gene V4–V5 region was performed for in-depth analysis of the observed eubacterial community differences brought out by different methods. After applying quality-filtering protocols, a total of 2,778,997 high-quality sequence reads with an average of 17,929 ± 1,139 reads per sample and an average length of 354 ± 2 bp were obtained (see Supplementary Table S4). The differences in the eubacterial community composition were compared at different taxonomic levels. The shared species-level OTUs among the three methods were low (29.9% in milk, 30% in fish, 34.3% in bamboo and 43% in soybean) (Fig. 3a). Each method recovered a substantial number of additional OTUs that were not recovered by the other methods (see Supplementary Table S5). Our results clearly indicated that employing a single DNA extraction method for studying the microbial ecology of fermented foods would result in the underestimation of at least 25% of the total OTUs. Unsupervised PCA plotting of normalized species-level OTU abundance data grouped the samples based on methods (ANOSIM, p < 0.01) in the fermented bamboo shoot and milk products (Fig. 3b). Redundancy analysis (RDA) biplots also indicated the significant separation of method VII in fermented bamboo shoots (Monte Carlo permutation test, p = 0.002, F = 2.62) and method II in fermented milk products (p = 0.012, F = 2.11). Species directions towards different methods are shown in Fig. 3c. Similarity percentage analysis (SIMPER) analysis showed the major species level-OTUs that contributed towards method dissimilarity (50% of the total variability) (see Supplementary Table S8). We observed significant differences in the bacterial community composition recovered by the different extraction methods (Fig. 3d, Supplementary Fig. S1). In general, lactic acid bacteria (Lactobacillales) were more abundant when enzymatic lysis based extraction methods were employed (see Supplementary Tables S6 and S7). However, Bacillus (Bacillaceae), Gluconobacter (Acetobacteraceae), Clostridium (Clostridiaceae) and Proteobacteria were abundantly recovered when mechanical and chemical lysis principles were applied. For example, the preferential recovery of Bacillus subtilis (in bamboo shoot products), Clostridium bifermentans (in fish products) and Proteus vulgaris (in soybean products) by zirconia/silica bead beating method and recovery of Gluconobacter frateurii by guanidium thiocyanate treatment in milk products could be mentioned. Figure 4 shows the differential recovery of bacterial community by different DNA extraction methods from the four types of naturally fermented foods.


Quantifying the biases in metagenome mining for realistic assessment of microbial ecology of naturally fermented foods
Variation in the recovery of eubacterial community structure as an effect of varied DNA extraction principles on different fermented food types.(a) Venn diagrams using three-fold rotationally symmetrical circles with unique and shared OTUs for different extraction methods are shown. Values in the parentheses indicate extraction methods and numbers inside each region indicate the number of unique or shared OTUs. (b) The canonical PCA was performed using log transformed (log xi + 1) values of species-level OTU relative abundance data. A significant difference in the microbial community structure between extraction methods is highlighted by ANOSIM, performed with 10,000 replicates using Bray-Curtis distances. Circles with solid colors denote data points for each extraction method. (c) Species-sample biplot of RDA shows the eubacterial species direction towards different extraction methods. Solid black arrows indicate species directions towards the method and their abundance. La: Lactobacillus acetotolerans, Bs: Bacillus subtilis, Gf: Gluconobacter frateurii, Tt: Thermus thermophilus, Ck: Clostridium kluyveri and Pv: Proteus vulgaris. (d) The significant difference in the mean relative abundance of major taxa between extraction methods as calculated by Student’s two-tailed paired t-test is shown. Other significant differences at various taxonomic levels are shown in Supplementary Fig. S1, Tables S6 and S7.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5037447&req=5

f3: Variation in the recovery of eubacterial community structure as an effect of varied DNA extraction principles on different fermented food types.(a) Venn diagrams using three-fold rotationally symmetrical circles with unique and shared OTUs for different extraction methods are shown. Values in the parentheses indicate extraction methods and numbers inside each region indicate the number of unique or shared OTUs. (b) The canonical PCA was performed using log transformed (log xi + 1) values of species-level OTU relative abundance data. A significant difference in the microbial community structure between extraction methods is highlighted by ANOSIM, performed with 10,000 replicates using Bray-Curtis distances. Circles with solid colors denote data points for each extraction method. (c) Species-sample biplot of RDA shows the eubacterial species direction towards different extraction methods. Solid black arrows indicate species directions towards the method and their abundance. La: Lactobacillus acetotolerans, Bs: Bacillus subtilis, Gf: Gluconobacter frateurii, Tt: Thermus thermophilus, Ck: Clostridium kluyveri and Pv: Proteus vulgaris. (d) The significant difference in the mean relative abundance of major taxa between extraction methods as calculated by Student’s two-tailed paired t-test is shown. Other significant differences at various taxonomic levels are shown in Supplementary Fig. S1, Tables S6 and S7.
Mentions: Illumina MiSeq barcoded amplicon sequencing of the SSU rRNA gene V4–V5 region was performed for in-depth analysis of the observed eubacterial community differences brought out by different methods. After applying quality-filtering protocols, a total of 2,778,997 high-quality sequence reads with an average of 17,929 ± 1,139 reads per sample and an average length of 354 ± 2 bp were obtained (see Supplementary Table S4). The differences in the eubacterial community composition were compared at different taxonomic levels. The shared species-level OTUs among the three methods were low (29.9% in milk, 30% in fish, 34.3% in bamboo and 43% in soybean) (Fig. 3a). Each method recovered a substantial number of additional OTUs that were not recovered by the other methods (see Supplementary Table S5). Our results clearly indicated that employing a single DNA extraction method for studying the microbial ecology of fermented foods would result in the underestimation of at least 25% of the total OTUs. Unsupervised PCA plotting of normalized species-level OTU abundance data grouped the samples based on methods (ANOSIM, p < 0.01) in the fermented bamboo shoot and milk products (Fig. 3b). Redundancy analysis (RDA) biplots also indicated the significant separation of method VII in fermented bamboo shoots (Monte Carlo permutation test, p = 0.002, F = 2.62) and method II in fermented milk products (p = 0.012, F = 2.11). Species directions towards different methods are shown in Fig. 3c. Similarity percentage analysis (SIMPER) analysis showed the major species level-OTUs that contributed towards method dissimilarity (50% of the total variability) (see Supplementary Table S8). We observed significant differences in the bacterial community composition recovered by the different extraction methods (Fig. 3d, Supplementary Fig. S1). In general, lactic acid bacteria (Lactobacillales) were more abundant when enzymatic lysis based extraction methods were employed (see Supplementary Tables S6 and S7). However, Bacillus (Bacillaceae), Gluconobacter (Acetobacteraceae), Clostridium (Clostridiaceae) and Proteobacteria were abundantly recovered when mechanical and chemical lysis principles were applied. For example, the preferential recovery of Bacillus subtilis (in bamboo shoot products), Clostridium bifermentans (in fish products) and Proteus vulgaris (in soybean products) by zirconia/silica bead beating method and recovery of Gluconobacter frateurii by guanidium thiocyanate treatment in milk products could be mentioned. Figure 4 shows the differential recovery of bacterial community by different DNA extraction methods from the four types of naturally fermented foods.

View Article: PubMed Central - PubMed

ABSTRACT

Cultivation-independent investigation of microbial ecology is biased by the DNA extraction methods used. We aimed to quantify those biases by comparative analysis of the metagenome mined from four diverse naturally fermented foods (bamboo shoot, milk, fish, soybean) using eight different DNA extraction methods with different cell lysis principles. Our findings revealed that the enzymatic lysis yielded higher eubacterial and yeast metagenomic DNA from the food matrices compared to the widely used chemical and mechanical lysis principles. Further analysis of the bacterial community structure by Illumina MiSeq amplicon sequencing revealed a high recovery of lactic acid bacteria by the enzymatic lysis in all food types. However, Bacillaceae, Acetobacteraceae, Clostridiaceae and Proteobacteria were more abundantly recovered when mechanical and chemical lysis principles were applied. The biases generated due to the differential recovery of operational taxonomic units (OTUs) by different DNA extraction methods including DNA and PCR amplicons mix from different methods have been quantitatively demonstrated here. The different methods shared only 29.9&ndash;52.0% of the total OTUs recovered. Although similar comparative research has been performed on other ecological niches, this is the first in-depth investigation of quantifying the biases in metagenome mining from naturally fermented foods.

No MeSH data available.


Related in: MedlinePlus