Biases from different DNA extraction methods in intestine microbiome research based on 16S rDNA sequencing: a case in the koi carp, Cyprinus carpio var. Koi

Abstract This study examined the technical bias associated with different DNA extraction methods used in microbiome research. Three methods were used to extract genomic DNA from the same intestinal microbiota sample that was taken from the koi carp Cyprinus carpio var. koi, after which their microbial diversity and community structure were investigated on the basis of a 16S rDNA high‐throughput sequencing analysis. Biased results were observed in relation to the number of reads, alpha diversity indexes and taxonomic composition among the three DNA extraction protocols. A total of 1,381 OTUs from the intestinal bacteria were obtained, with 852, 759, and 698 OTUs acquired, using the Lysozyme and Ultrasonic Lysis method, Zirmil‐beating Cell Disruption method, and a QIAamp Fast DNA Stool Mini Kit, respectively. Additionally, 336 OTUs were commonly acquired, using the three methods. The results showed that the alpha diversity indexes (Rarefaction, Shannon, and Chao1) of the community that were determined using the Lysozyme and Ultrasonic Lysis method were higher than those obtained with the Zirmil‐beating Cell Disruption method, while the Zirmil method results were higher than those measured, using the QIAamp Fast DNA Stool Mini Kit. Moreover, all the major phyla (ratio>1%) could be identified with all three DNA extraction methods, but the phyla present at a lower abundance (ratio <1%) could not. Similar findings were observed at the genus level. Taken together, these findings indicated that the bias observed in the results about the community structure occurred primarily in OTUs with a lower abundance. The results of this study demonstrate that possible bias exists in community analyses, and researchers should therefore be conservative when drawing conclusions about community structures based on the currently available DNA extraction methods.


| INTRODUC TI ON
Asian-origin koi carp (Cyprinus carpio var. Koi) are currently listed among the most important ornamental species because they can be reared in all the countries in the world. Their broad diversity of colors and color patterns are major factors contributing to their attractive market value (David et al., 2004). However, various infectious diseases in koi carp have emerged with the rapid development of industrial culture in recent years, resulting in great economic losses (Kumar et al., 2015;Pokorova et al., 2007).
Gut microbiota play important roles in fish health and physiology (Ganguly & Prasad, 2011). Studies have shown that gut microbiota are associated with many key functions of the host, such as resistance to infectious diseases and the decomposition of nutrients, and they provide the host with physiologically active materials including enzymes, amino acids and vitamins (Sugita, Kawasahi, & Deguchi, 1997). Accordingly, altered microbiota in the intestine can lead to changes in host immune functions as well as an increased risk of disease (Brown, DeCoffe, Molcan, & Gibson, 2012;Morgan et al., 2012). Over the last decade, an increasing number of studies have focused on the gut microbiota of fish (Narrowe et al., 2015;Xia et al., 2014). In early studies, conventional culture-dependent techniques were used; however, only a small percentage of the resulting bacterial flora was identified (Kathiravan et al., 2015). Recently, new technologies based on meta-genomics/high-throughput sequencing have been developed and successfully applied to analyzing the complex bacterial ecosystem of the gut. These new analytical approaches usually involve DNA extraction from stool samples or biopsies and the amplification of 16S ribosomal DNA (rDNA) followed by highthroughput sequencing. Increasing evidence shows that 16S rDNA sequencing approaches can be used to identify bacteria rapidly because they can overcome the limitations of culturebased bacterial detection methods.
The extraction of DNA from intestinal fecal samples is a key step in molecular biological analyses. Several protocols for extracting DNA from fish intestinal microflora have been described, including physical and chemical methods. Generally, common physical disruption methods have been employed, including freezing-thawing (Silva, Bernardi, Schaker, Menegotto, & Valente, 2012), sonication (Yang, Xiao, Zeng, Liu, & Deng, 2006) and bead beating (Carrigg, Rice, Kavanagh, Collins, & O'Flaherty, 2008). In addition, a variety of chemical lysis approaches has been used to obtain higher purity DNA samples, including cetyltrimethylammonium bromide (CTAB) (Chapela et al., 2007). However, different DNA extraction protocols can lead to biases with respect to the microbial diversity, community structure, proportions and number of reads and numbers of OTUs obtained based on the 16S rDNA high-throughput sequencing, subsequently influencing estimations of the microbial diversity and the taxonomic composition in the intestinal mucosa and intestinal content. Moreover, because there is no "gold standard" method for DNA extraction, it is difficult to determine the "true" diversity of the bacterial community. Some have suggested combining several extraction methods, if possible, to recover some of the loss in observable diversity that occurs when only one DNA extraction is used (Kashinskaya, Andree, Simonov, & Solovyev, 2016;Wen, He, Xue, Liang, & Dong, 2016).
This study examined the bias in results that were obtained using different extraction methods during microbiome research based on 16S rDNA high-throughput sequencing analyses. Three methods were used to extract the genomic DNA from the same sample of intestinal microbiota from the koi carp, Cyprinus carpio var. koi. Specifically, a protocol was modified from the lysozyme method developed by our laboratory and named the Combination of Lysozyme and Ultrasonic Lysis method (CLU); the Zirmilbeating Cell Disruption method (ZBC) referring to the research of Zoetendal et al. (2006) and a QIAamp Fast DNA Stool Mini Kit (QIA, Qiagen, Hilden, Germany), a common commercial kit, were also used.

| Sample preparation for DNA isolation
Koi carp were provided by the Gongwang koi fish-breeding center in Tianjin, China. The fish were transported to Tianjin Agricultural University, where they were maintained under optimal rearing conditions for 1 week in 20°C water. Aeration was provided to maintain optimal dissolved oxygen levels and the fish were fed commercial pellets twice daily. Genomic DNA was extracted from the intestinal contents and mucosa of adult koi carp that were 30-35 cm long and 380-410 g. In brief, the fish were euthanized with an overdose of MS-222 (Sigma-Aldrich, St Louis, MO, USA), after which their exteriors were wiped clean with 70% ethanol, their abdomens were opened at the ventral midline and the whole intestines were aseptically removed from the abdominal cavity. All the experimental procedures performed on these koi carp were approved by the Animal Care Committee of Tianjin Agricultural University, and the methods were performed in accordance with the approved guidelines and regulations.
The gut samples were used directly after their removal from the fish. The intestinal contents and mucosa of three fish were collected into a 50-ml centrifuge tube and homogenized in 15 ml of sterile phosphate-buffered solution (PBS, 0.01 mol/L, pH 7.2; Dingguo Changsheng, Beijing, China) by vortexing (IKA, Germany) three times at 158 g for 20 s each. The samples were then centrifuged at 110 g for 5 min at 4°C, after which the supernatant was dispensed into a new sterile 50-ml centrifuge tube. The supernatant was subsequently centrifuged at 2,739 g for 5 min. The bacterial precipitation was then resuspended in 3 ml of PBS. The bacterial suspension of mucosa and intestinal content from the three koi carp was divided into triplicate samples. One milliliter of bacterial suspension was used for the CLU method, one for the ZBC method, and another one for the QIA method.

| CLU method
The bacterial suspension (1 ml) was dispensed into a 2-ml microtube, and then it was disrupted using an Ultrasonic Cell Disruption System (Ningbo Scientz Biotechnology, Ningbo, China) 50 times for 2 s each with an interval of 5 s between each disruption. Next, the samples were centrifuged at 15,777 g for 5 min at 4°C, after which the upper aqueous layer was discarded. Each sample was then incubated for 30 min at 60°C in 750 μl of TE (10 mmol/L Tris-HCl, 1 mmol/L EDTA, pH 8.0) and 50 μl of lysozyme (20 mg/ml; Sangon Biotech, Shanghai, China). Subsequently, 10 μl of RNase A (20 μg/ml; Sangon Biotech) was added to the centrifuge tube, after which the suspension was incubated for 30 min at 30°C. The tube was then incubated for 60 min at 65°C with inversion every 20 min after adding 100 μl of 10% SDS (0.1 g/ml, pH 7.4; Sigma Aldrich) and 30 μl of Proteinase K (20 mg/ ml; Sangon Biotech). Thereafter, an equal volume of phenol: chloroform: isoamyl alcohol (25: 24: 1) was added and mixed by inversion. The samples were then centrifuged at 15,777 g for 2 min, after which the supernatant was collected in a new sterile 2-ml centrifuge tube. An equal volume of chloroform: isoamyl alcohol (24: 1) was then added to the tube, after which the suspension was mixed gently and centrifuged at 15,777 g for 2 min. The upper aqueous layer was subsequently transferred to another 2-ml sterile centrifuge tube, and the DNA was then precipitated using a 1/10 volume of NaAc (3 mol/L, pH 5.2) and 2 volumes of ice-cold (−20°C) 95% ethanol, followed by centrifugation at 15,777 g for 5 min at 4°C. Finally, the DNA pellet was washed twice with 1 ml of 70% ethanol before it was air-dried and finally resuspended in 100 μl of TE buffer that had been preheated to 50°C.

| ZBC method
DNA was extracted from 1 ml of bacterial suspension according to the modified ZBC method (Zoetendal et al., 2006). In brief, the bacterial suspension was transferred to a 2-ml Lysing Matrix A tube (MP Biomedicals, Santa Ana, CA, USA), after which 150 μl of buffersaturated phenol was added to the tube. The sample was then oscillated at 4 m/s for 2 min, using a FastPrep ® -24 Instrument (MP Biomedicals), then cooled on ice for 30 s and purified with 150 μl chloroform: isoamyl alcohol (24: 1), and after that it was centrifuged at 15,777 g for 2 min at 4°C. At that stage, an equal volume of phenol: chloroform: isoamyl alcohol (25: 24: 1) was added and mixed in by inversion. Next, the sample was centrifuged at 15,777 g for 2 min and the supernatant was transferred to a new 2-ml sterile centrifuge tube. This step was repeated until the interface of the two layers was clean, after which an equal volume of chloroform: isoamyl alcohol (24: 1) was added to the tube. The sample was then mixed gently and centrifuged at 15,777 for 2 min, after which the supernatant was transferred into a new 2-ml centrifuge tube. Next, the DNA was precipitated with 1/10 volume of 3 mol/L NaAc (pH 5.2) and 2 volumes of cold 95% ethanol (−20°C) and stored at −20°C for 30 min.
The samples were then centrifuged at 15,777 g for 10 min, and the supernatant was discarded. The DNA was washed with 1 ml of cold (−20°C) 70% ethanol and centrifuged at 15,777 g for 5 min at 4°C.
Finally, the DNA pellet was dried by placing the tube upside down on tissue paper for 15 min, after which the dried DNA was rehydrated in 100 μl of TE buffer.

| QIA method
One-milliliter bacterial suspensions were centrifuged at 2,739 g for 5 min, after which the bacterial precipitation was resuspended with 220 μl of PBS. Next, DNA was extracted from 220 μl of bacterial suspension using a QIAamp Fast DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions.

| High-throughput 16S rDNA Illumina MiSeq sequencing
To analyze the microbial populations of the extracted DNA samples, the variable V3-V4 region of the 16S rDNA was amplified. To this end, a polymerase chain reaction (PCR) was conducted using the bacterial universal primers 341F (5′-CCCTACACGACGCTCTTCCG ATCTGCCTACGGGNGGCWGCAG-3′) and 805R (5′-GACTGGAGT TCCTTGGCACCCGAGAATTCCAGACTACHVGGGTATCTAATCC-3′) (Li et al., 2016). Barcodes unique to each sample were incorporated before the forward primers, which allowed for the identification of

| Data analysis
Following sequencing with the Illumina MiSeq, the sequencing reads were assigned to each sample according to their unique barcode.
Pairs of reads from the original DNA fragments were first merged, using FLASH (Magoč & Salzberg, 2011). A quality control procedure was used, including trimming the barcodes and primers and filtering low-quality reads by PRINSEQ (Schmieder & Edwards, 2011). The sequences that passed the above procedure were then denoised to correct for potential sequencing errors, and reads were discarded if they were identified by UCHIME as putative chimeras (Edgar, Haas, Clemente, Quince, & Knight, 2011). Finally, the filtered sequences were obtained. These sequences were classified into the same operational taxonomic units (OTUs) at an identity threshold of 97% similarity, using the Ribosomal Database Project (RDP) classifier (Wang, Garrity, Tiedje, & Cole, 2007). The Rarefaction, Shannon and Chao1 indexes were included in the alpha diversity analysis, using Mothur (Schloss et al., 2009). Weighted UniFrac metric distances were calculated to determine the beta diversity index, and the sample tree was used to examine the relationship of the community structures of the microbiota from different samples.

| OTUs and alpha diversity analysis
After applying quality control measures and filtering the chimera, the numbers of filtered reads (Filtered-num) were 29,618, 41,379, and 48,586 for the DNA samples extracted, using the CLU, ZBC, and QIA methods, respectively (Table 1). These reads, which had a mean length of 415.7 bp, were assigned to 1,381 OTUs of intestinal bacteria based on a 97% similarity cut-off ( Figure 1). The numbers of OTUs were 852, 759, and 698 for samples extracted, using the CLU, ZBC, and QIA methods, respectively (Table 1). Moreover, samples extracted using the three methods shared 336 OTUs, accounting for 39.44%, 44.27%, and 48.14% of the respective OTU numbers for the CLU, ZBC, and QIA methods.
The Rarefaction, Shannon's and Chao1 alpha diversity indexes were also calculated ( Figure 2, Table 1). The ZBC and QIA methods showed similar trends in the rarefaction curves. However, the CLU method had a higher slope for the rarefaction curve than the other two methods (Figure 2a). The Shannon indexes were 3.41, 2.61, and 2.57 for the CLU, ZBC, and QIA methods, respectively (Figure 2b), while the Chao1 index values were 1,609, 1,458, and 1,320 for these methods (Figure 2c).

| Bacterial community
The 10 OTUs with the highest abundance at the phylum level that were obtained, using the three extraction methods were identified ( Figure 3a)

| Beta-diversity analysis
The relationship among the community structures of the microbiota from different samples was examined using a sample tree based on the weighted UniFrac distance matrixes. The microbial community structures and species richness were more similar between samples extracted, using the ZBC and QIA methods than those obtained, using the CLU and ZBC methods or the CLU and QIA methods ( Figure 4).  TA B L E 2 Ten OTUs from microorganisms with the highest abundance at the phylum level were identified on the basis of DNA samples extracted using all three methods  (Kormas, Meziti, Mente, & Frentzos, 2014). Previous studies also revealed a core microbiome from the intestines of fish dominated by Proteobacteria, Firmicutes, Bacteroidetes, Actinobacteria, and Fusobacteria (Smriga, Sandin, & Azam, 2010;Sullam et al., 2012;Ye, Amberg, Chapman, Gaikowski, & Liu, 2014). In a study conducted by Kashinskaya et al. (2016), the intestinal microbiota of the  Sullam et al., 2012;Uchii et al., 2006;Wong & Rawls, 2012;Ye et al., 2014). In the present study, the use of only one library construction method and a single primer set probably limited the ability to identify the observable diversity. In addition, the first centrifugation of 110 g used for the DNA isolation sample preparation aimed to discard the solid residue from the mucosa and intestinal contents.

| D ISCUSS I ON
This step would unavoidably result in some loss from the bacterial sample and consequently some loss in the potential biodiversity. To determine the "true" diversity of the bacterial community, the development of a universal methodology that is applicable to more samples is needed. The results of the present study suggested that bias is present among DNA extraction methods; therefore, researchers should be conservative in drawing conclusions about community structures.

ACK N OWLED G M ENTS
This study was partially supported by the Scientific Program of Tianjin City (15JCZDJC34000; 16JCZDJC33500), the Innovation Team of Tianjin Fisheries Research System (ITTFRS2017009), and the Technical Innovation Key Program of Higher Education of Guangdong Province (CXZD1114).

CO N FLI C T O F I NTE R E S T
None declared.

O RCI D
Jingfeng Sun http://orcid.org/0000-0002-8070-5304 F I G U R E 4 Distance-based weighted UniFrac analysis associated with the three genomic DNA extraction methods. The microbial community structures and species richness of DNA samples extracted using the ZBC and QIA methods were more similar than those obtained using the CLU and ZBC methods or the CLU and QIA methods