¶ Membership of the UK IBD Genetics Consortium is provided in the Acknowledgments.
The authors have declared that no competing interests exist.
Conceived and designed the experiments: NK AW PL SD HF JP JS CL GH. Performed the experiments: NK AW SB SD FF JT. Analyzed the data: NK AW PL GH. Wrote the paper: NK GH.
Determining bacterial community structure in fecal samples through DNA sequencing is an important facet of intestinal health research. The impact of different commercially available DNA extraction kits upon bacterial community structures has received relatively little attention. The aim of this study was to analyze bacterial communities in volunteer and inflammatory bowel disease (IBD) patient fecal samples extracted using widely used DNA extraction kits in established gastrointestinal research laboratories.
Fecal samples from two healthy volunteers (H3 and H4) and two relapsing IBD patients (I1 and I2) were investigated. DNA extraction was undertaken using MoBio Powersoil and MP Biomedicals FastDNA SPIN Kit for Soil DNA extraction kits. PCR amplification for pyrosequencing of bacterial 16S rRNA genes was performed in both laboratories on all samples. Hierarchical clustering of sequencing data was done using the Yue and Clayton similarity coefficient.
DNA extracted using the FastDNA kit and the MoBio kit gave median DNA concentrations of 475 (interquartile range 228-561) and 22 (IQR 9-36) ng/µL respectively (p<0.0001). Hierarchical clustering of sequence data by Yue and Clayton coefficient revealed four clusters. Samples from individuals H3 and I2 clustered by patient; however, samples from patient I1 extracted with the MoBio kit clustered with samples from patient H4 rather than the other I1 samples. Linear modelling on relative abundance of common bacterial families revealed significant differences between kits; samples extracted with MoBio Powersoil showed significantly increased
This study demonstrates significant differences in DNA yield and bacterial DNA composition when comparing DNA extracted from the same fecal sample with different extraction kits. This highlights the importance of ensuring that samples in a study are prepared with the same method, and the need for caution when cross-comparing studies that use different methods.
The last decade has seen a marked rise in interest in the bacterial communities that coexist within humans, facilitated by the availability of modern molecular techniques. The Human Microbiome Project
Determining the bacterial community structure in fecal samples through amplification and sequence analysis of extracted DNA has revolutionized gastrointestinal microbiology research over recent years. These culture-independent techniques for assessing diversity have largely replaced traditional culture based approaches as they are considered to be less biased in terms of defining true diversity and considerably less labor-intensive
Previous studies have evaluated differences between DNA extraction methods from fecal samples, exploring detection with conventional PCR
The aim of this study was to analyze bacterial communities in healthy volunteer and IBD patient fecal samples extracted using the MoBio and FastDNA DNA extraction kits in two established gastrointestinal research laboratories. The MoBio Power Soil DNA extraction kit and the MP Biomedicals Fast DNA Spin Kit for Soil DNA extraction kit are two commonly used extraction procedures for fecal microbial diversity studies
Fecal samples were taken from two patients with IBD (I1 and I2) and from two healthy controls (H3 and H4) using the Fisher Fecal Commode Collection Kit. Fecal samples were kept at 4°C and processed within 4 hours of collection. This short period of storage is not expected to influence molecular estimation of microbial community composition
IMS: Institute of Medical Sciences, University of Aberdeen, Foresterhill, Aberdeen; RI: Rowett Institute of Nutrition and Health, University of Aberdeen, Bucksburn, Aberdeen.
Ethical approval was granted by North of Scotland Research Ethics Service (03/0137 and 12/NS/0061) on behalf of all participating centers and written informed consent was obtained from all subjects.
One 500 mg fecal aliquot was used for MoBio PowerSoil DNA isolation kit extraction. 5 ml of MoBio lysis buffer was immediately added to the fresh fecal sample, which was then vortex mixed for 30–40 seconds. Fecal suspensions were then centrifuged (1,500 g×5 minutes) and 1 ml of the supernatant placed into the MoBio Garnet bead tubes containing 750 ul of MoBio buffer. These tubes were then heated at 65°C for 10 minutes, then at 95°C for 10 minutes. Samples were then stored at −80°C prior to processing in both laboratories following the manufacturer's instructions. DNA was eluted in 100 µL MoBio elution buffer.
For each fecal sample 2×500 mg aliquots were placed in FastDNA SPIN Kit lysing matrix E tubes and 978 µl of sodium phosphate buffer and 122 µl MT buffer were added to each tube and vortex mixed. One aliquot was then stored at −80°C and was defined as FastDNA method 1. The second aliquot was subjected to additional processing by heating at 65°C for 10 minutes, then at 95°C for 10 minutes followed by storage at −80°C. This was defined as FastDNA method 2. Both aliquots were then processed following manufacturer's (Qbiogene, MP Biomedicals, Illkirch, France) instructions. DNA was eluted in 100 µL FastDNA elution buffer.
Fecal DNA was quantified by Nanodrop mass spectrophotometry before dilution to 25 ng/µl. Initial PCR amplification was undertaken at each laboratory with Invitrogen AccuPrime Taq DNA Polymerase High Fidelity utilising a per-reaction mix of 2 µl of DNA template, 2 µl of Buffer II, 0.2 µl (2 µM) Fusion Primer A, 0.2 µl (2 µM) Fusion Primer B, 0.08 µl (1 U) Accuprime Taq and 15.52 µl sterile, deionized water to a final volume of 20 µl. Quadruplicate PCR reactions were set up per DNA sample. The 16S rRNA gene primers, spanning the V3-5 region of the 16S rRNA gene, were configured as follows: 338F,
Quantitative real-time PCR was performed as described previously
Bacterial family | Primer name | Primer sequence | Reference |
All bacteria | UniF | GTGSTGCAYGGYYGTCGTCA | |
UniR | ACGTCRTCCMCNCCTTCCTC | ||
Bac303F | |||
Bfr-Fmrev | CGCKACTTGGCTGGTTCAG | ||
Clep866mF | TTAACACAATAAGTWATCCACCTGG | ||
Clept1240mR | |||
Erec482F | |||
Erec870R | AGTTTYATTCTTGCGAACG | ||
EnterobactDmod2F | |||
Enter1432mod |
Analysis of sequence data was carried out using the Mothur software package
Comparisons in DNA yield were performed using Mann Whitney U testing. Linear modelling was used to assess the relative contribution of patient, DNA extraction method and extraction site to the measured proportions of different bacterial families. Log-transformed data was used to permit analysis of the fold change. A model was constructed for each bacterial family using the donor source, extraction method and extraction site as covariates. Bacterial families were reported where at least one sample had an abundance of 5% or more. For each family, samples were only included from participants with at least 0.5% abundance for that bacterial family in one or more of their samples. Modelling was also done in a similar manner using individual OTUs. When using linear modelling at the OTU level, Holm's method was used to correct for multiple testing. Correlation between pyrosequencing and qPCR data was done using Pearsons's correlation coefficient. Analysis was performed using R 2.15.2 (R Statistical Foundation, Vienna, Austria).
DNA yields were significantly higher with either method of the FastDNA kit than with the MoBio kit, with median DNA concentrations of 476 ng/µL (interquartile range [IQR] 290–519) for FastDNA method 1, 453 ng/µL (IQR 228–689) for FastDNA method 2 and 22 ng/µL (IQR 9-36) for the MoBio method (p<0.001 for both comparisons,
Compositional analysis indicated a higher proportion of
Clustering of the microbiota composition derived from the sequence data for these samples was carried out using both the Jaccard and the Yue and Clayton calculators. The Jaccard calculator is used to describe overlap in community membership between different samples and ignores the proportional abundance of each OTU while, in contrast, the Yue and Clayton calculator takes the proportional abundance of each OTU into account when comparing community similarities. Jaccard-based calculations revealed a clear clustering of samples primarily by subject of origin (
Linear modelling of the family level data for the top nine families represented in the pyrosequencing data is shown in
Bacterial Family | Kit | Extraction Site | |||||
FastDNA 2 fold change | p | MoBio fold change | P | RINH fold change | p | Patients included | |
0.96 (0.74–1.25) | 0.775 | 0.63 (0.49–0.81) | 0.001 | 1.17 (0.95–1.44) | 0.160 | H3,H4,I1,I2 | |
1.13 (0.79–1.63) | 0.501 | 2.13 (1.49–3.05) | <0.001 | 1.09 (0.81–1.46) | 0.561 | H3,H4,I1,I2 | |
0.94 (0.79–1.13) | 0.524 | 1.32 (1.11–1.58) | 0.005 | 0.95 (0.82–1.10) | 0.516 | H3,H4,I1 | |
1.08 (0.74–1.57) | 0.695 | 0.61 (0.43–0.88) | 0.016 | 0.85 (0.63–1.15) | 0.311 | I1,I2 | |
0.77 (0.18–3.37) | 0.735 | 1.11 (0.26–4.69) | 0.892 | 3.84 (1.18–12.46) | 0.031 | H3,H4,I1,I2 | |
1.00 (0.77–1.30) | 0.976 | 0.46 (0.36–0.59) | <0.001 | 0.88 (0.71–1.08) | 0.243 | I1,I2 | |
1.46 (0.41–5.19) | 0.560 | 4.03 (1.16–14.01) | 0.035 | 0.70 (0.26–1.94) | 0.502 | H3,H4,I1,I2 | |
1.21 (0.81–1.81) | 0.361 | 0.32 (0.21–0.47) | <0.001 | 0.88 (0.64–1.22) | 0.445 | H3,H4,I1,I2 | |
0.35 (0.16–0.76) | 0.016 | 0.72 (0.33–1.56) | 0.418 | 0.65 (0.35–1.19) | 0.181 | H3,H4 |
RINH: Rowett Institute of Nutrition and Health.
Participants were excluded if all data points for that bacterial family were < 0.5%. Reference sample was from participant H3 using FastDNA method 1 and extracted at the Institute of Medical Sciences. Differences are shown as fold change with 95% confidence intervals.
At the OTU level, 18 OTUs were significantly different between the MoBio kit and the FastDNA kit after correction for multiple testing (
Genus | Family | Order | Class | Phylum | Fold change | p | Corrected p | Patients included |
Coriobacteriales | Actinobacteria | Actinobacteria | 0.00 (0.00–0.00) | 5.09×10−9 | 5.55×10−7 | I1 | ||
Clostridiales | Clostridia | Firmicutes | 0.00 (0.00–0.01) | 1.70×10−7 | 1.83×10−5 | H3,H4,I1 | ||
Clostridiales | Clostridia | Firmicutes | 0.01 (0.00–0.04) | 2.01×10−7 | 2.15×10−5 | H3,H4,I1 | ||
Bacteroidales | Bacteroidia | Bacteroidetes | 2.60 (1.92–3.52) | 3.26×10−7 | 3.46×10−5 | H3,H4,I1,I2 | ||
Clostridiales | Clostridia | Firmicutes | 0.36 (0.30–0.43) | 3.91×10−6 | 0.0004 | I1 | ||
Clostridiales | Clostridia | Firmicutes | 0.01 (0.00–0.05) | 1.97×10−5 | 0.0020 | H3,H4,I1 | ||
Clostridiales | Clostridia | Firmicutes | 0.17 (0.12–0.24) | 3.13×10−5 | 0.0032 | H3 | ||
Clostridiales | Clostridia | Firmicutes | 0.00 (0.00–0.01) | 3.49×10−5 | 0.0036 | H4,I1 | ||
Clostridiales | Clostridia | Firmicutes | 0.10 (0.04–0.23) | 5.23×10−5 | 0.0053 | H3,H4 | ||
Enterobacteriales | Gamma-proteobacteria | Proteobacteria | 0.41 (0.28–0.59) | 1.36×10−4 | 0.0136 | I1,I2 | ||
Clostridiales | Clostridia | Firmicutes | 5.05 (2.55–9.97) | 2.24×10−4 | 0.0222 | H3,I1 | ||
Clostridiales | Clostridia | Firmicutes | 0.48 (0.34–0.68) | 0.0002 | 0.0232 | H3,H4,I1 | ||
Clostridiales | Clostridia | Firmicutes | 2.50 (1.92–3.26) | 0.0003 | 0.0250 | H3 | ||
Clostridiales | Clostridia | Firmicutes | 3.19 (2.26–4.50) | 0.0003 | 0.0298 | H3 | ||
Bacteroidaceae | Bacteroidales | Bacteroidia | Bacteroidetes | 2.03 (1.42–2.89) | 0.0004 | 0.0342 | H3,H4,I1,I2 | |
Lachnospiraceae | Clostridiales | Clostridia | Firmicutes | 0.00 (0.00–0.06) | 0.0004 | 0.0343 | H3,H4,I1 | |
Lachnospiraceae | Clostridiales | Clostridia | Firmicutes | 0.00 (0.00–0.00) | 0.0004 | 0.0358 | I1 | |
Lachnospiraceae | Clostridiales | Clostridia | Firmicutes | 0.06 (0.01–0.25) | 0.0004 | 0.0381 | H3,H4,I1,I2 |
Samples were excluded if all data points for that bacterial family were <0.5%. Reference sample was from patient H3 using either FastDNA method and extracted at the Institute of Medical Sciences. Differences are shown as fold change with 95% confidence intervals.
Correlation between pyrosequencing and qPCR data was generally good (
Bacterial Family | Kit | Extraction Site | |||||
FastDNA 2 fold change | p | MoBio fold change | P | RINH fold change | p | Patients included | |
0.75 (0.48–1.17) | 0.209 | 0.74 (0.48–1.16) | 0.199 | 1.35 (0.94–1.94) | 0.107 | H3,H4,I1,I2 | |
1.19 (0.94–1.51) | 0.147 | 1.25 (0.99–1.57) | 0.066 | 0.98 (0.81–1.18) | 0.822 | H3,H4,I1,I2 | |
0.69 (0.47–1.01) | 0.070 | 2.32 (1.58–3.39) | <0.001 | 1.26 (0.92–1.71) | 0.157 | H3,H4,I1 | |
1.28 (0.97–1.69) | 0.102 | 0.65 (0.48–0.87) | 0.011 | 0.91 (0.72–1.15) | 0.436 | I1,I2 |
RINH: Rowett Institute of Nutrition and Health.
Participants were excluded if all data points for that bacterial family were <0.5%. Reference sample was from participant H3 using FastDNA method 1 and extracted at the Institute of Medical Sciences. Differences are shown as fold change with 95% confidence intervals.
With the recognition that cultured bacteria cover only a small proportion of gut microbial diversity
This study highlights important differences in the performance of two commercially available kits for DNA extraction from fecal samples. Significantly lower DNA yields were seen with the MoBio kit than the FastDNA kit. This is consistent with results published previously
The importance of the observed differences will depend on the analysis techniques used. However, whenever a relative quantification technique is used, the results for even a single organism will be influenced by the effects of the extraction technique on the total number of bacteria isolated relative to that specific species. The methods for both kits used here involved physical disruption by bead-beating.
A smaller effect was observed of the extraction site on relative abundance, with only
The qPCR data in general correlated well with that from pyrosequencing with the exception of
Ariefdjohan et al. previously assessed the effect of DNA extraction method on the measured bacterial composition of stool using denaturing gradient gel electrophoresis (DGGE)
This study is somewhat limited by its relatively small sample size, with fecal samples obtained from only four individuals. There were a small number of outliers; samples H4F2AA and H4F2AR had much higher relative abundance of
This study demonstrates important differences in the yield and relative abundance of key bacterial families for kits used to isolate bacterial DNA from stool. This highlights the importance of ensuring that all samples to be analyzed together are prepared with the same DNA extraction method, and the need for caution when comparing studies that have used different methods.
Barcodes used for each sample included in the study, and the respective ENA-deposited dataset that they can be recovered from.
(XLSX)
Relative abundances of the top nine bacterial families measured for each individual.
(DOCX)
Relative abundances of the bacterial families measured by quantitative PCR for each individual.
(DOCX)
UK IBD Genetics Consortium individual authors and affiliations: Lead = Parkes M: Inflammatory Bowel Disease Research Group, Addenbrooke's Hospital, University of Cambridge, Cambridge CB2 0QQ, UK.
Other members not named within the manuscript author list (alphabetical by surname):
Ahmad T: Peninsula College of Medicine and Dentistry University of Exeter, EX1 2LU, UK.
Anderson CA: The Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SA, UK.
Barrett JC: The Wellcome Trust Sanger Institute, Wellcome Trust Genome Campus, Hinxton, Cambridge, CB10 1SA, UK.
Drummond H: Gastrointestinal Unit, Centre for Genomic and Experimental Medicine, Institute of Genetics and Molecular Medicine, University of Edinburgh, Edinburgh EH4 2XU, UK.
Edwards C: Department of Gastroenterology, Torbay Hospital, Torbay TQ2 7AA, UK.
Hart A: Inflammatory Bowel Disease Unit, St Mark's Hospital, Watford Road, Harrow, Middlesex HA1 3UJ, UK.
Hawkey C: Nottingham Digestive Diseases Centre, Queen's Medical Centre, Nottingham NG7 1AW, UK.
Henderson P: Royal Hospital for Sick Children, Paediatric Gastroenterology and Nutrition, Glasgow G3 8SJ, UK.
Khan M: Department of Medical Genetics, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 0JH, UK.
Lamb CA: Department of Gastroenterology & Hepatology, University of Newcastle upon Tyne, Royal Victoria Infirmary, Newcastle upon Tyne NE1 4LP, UK.
Lee JC: Inflammatory Bowel Disease Research Group, Addenbrooke's Hospital, University of Cambridge, Cambridge CB2 0QQ, UK.
Mansfield JC: Department of Gastroenterology & Hepatology, University of Newcastle upon Tyne, Royal Victoria Infirmary, Newcastle upon Tyne NE1 4LP, UK.
Mathew CG: Department of Medical and Molecular Genetics, King's College London School of Medicine, 8th Floor Guy's Tower, Guy's Hospital, London, SE1 9RT, UK.
Mowat C: Department of General Internal Medicine, Ninewells Hospital and Medical School, Ninewells Avenue, Dundee DD1 9SY, UK.
Newman WG: Department of Medical Genetics, Manchester Academic Health Science Centre (MAHSC), University of Manchester, Manchester M13 0JH, UK.
Prescott NJ: Department of Medical and Molecular Genetics, King's College London School of Medicine, 8th Floor Guy's Tower, Guy's Hospital, London, SE1 9RT, UK.
Simmons A: Translational Gastroenterology Unit, Experimental Medicine Division, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
Simpson P: Translational Gastroenterology Unit, Experimental Medicine Division, John Radcliffe Hospital, Headington, Oxford OX3 9DU, UK.
Taylor K: Department of Gastroenterology, Guy's & St Thomas' NHS Foundation Trust, St Thomas' Hospital, London SE1 7EH, UK.
Wilson DC: Child Life and Health, University of Edinburgh, Edinburgh EH9 1UW, UK.