| Title | Claps | Level | Year | L/Y |
|---|---|---|---|---|
|
Repository of the Max Delbrück Center for Molecular Medicine (MDC) in the Helmholtz Association
67 auth. P. Costea, G. Zeller, S. Sunagawa, É. Pelletier, A. Alberti, 3. Florence, Levenez, Melanie Tramontano, Marja Driessen, R. Hercog, Ferris Jung, 4. JensRoat, Kultima, M. Hayward, Luis Pedro Coelho, ...
18 Metagenomic analysis of fecal samples suffers from challenges in comparability 19 and reproducibility that need to be addressed in order to better establish 20 microbiota contributions to human health. To test and improve current 21 protocols, we…
18 Metagenomic analysis of fecal samples suffers from challenges in comparability 19 and reproducibility that need to be addressed in order to better establish 20 microbiota contributions to human health. To test and improve current 21 protocols, we quantified the effect of DNA extraction on the observed microbial 22 composition, by comparing 21 representative protocols. Furthermore, we 23 estimated the effect of sequencing, sample storage and biological variability on 24 observed composition, and show that the DNA extraction process is the 25 strongest technical factor to impact the results. We characterized the biases of 26 different methods, introduced a quality scoring scheme and quantified 27 transferability of the best methods across labs. Finally, we propose a 28 standardized DNA extraction methodology for human fecal samples, and 29 confirm its accuracy using a mock community in which the relative abundances 30 are known. Use of this methodology will greatly improve the comparability and 31 consistency of different human gut microbiome studies and facilitate future 32 meta-analyses. 33 Over 3000 publications in the past five years have used DNAor RNAbased profiling methods to 34 interrogate microbial communities in locations ranging from ice columns in the remote arctic to the 35 human body, resulting in more than 160,000 published metagenomes (both shotgun and 16S rRNA 36 gene). To date, one of the most studied ecosystems is the human gastrointestinal tract. The gut 37 microbiome is of particular interest due to its large volume, high diversity and potential relevance to 38 human health and disease. Numerous studies have found specific microbial fingerprints that may be 39 useful in distinguishing disease states, for example diabetes, inflammatory bowel disease or 40 colorectal cancer. Others have linked the human gut microbial composition to various factors, such 41 as mode of birth, age, diet and medication. Such studies have almost exclusively used their own 42 specific, demographically distinct cohort and methodology. Given the many reports of batch effects 43 and known differences when analyzing data generated using different protocols, comparisons or 44 meta-analyses are limited in their interpretability. For example, healthy Americans from the HMP 45 study showed lower taxonomic diversity in their stool than patients with inflammatory bowel disease 46 (IBD) from a European study, although it is established that IBD patients worldwide have reduced 47 taxonomic diversity. It is thus currently very difficult to disentangle biological from technical 48 variation when comparing across multiple studies. 49 In metagenomic studies, the calculation of compositional profiles and ecological indices is preceded 50 by a complex data generation process, consisting of multiple steps (Figure 1), each of which is subject 51 to technical variability. Usually, a small sample is collected by an individual shortly after passing 52 stool and stored in a domestic freezer, prior to shipment to a laboratory. The location within the 53 specimen that the sample is taken from has been shown to impact the measured composition, 54 which is why in some studies larger quantities were homogenized prior to storage in order to 55 generate multiple, identical aliquots. Furthermore, different fixation methods can be used to 56 preserve the sample for shipping and long-term storage. Freezing at below -20°C is the standard, 57 though more practical alternatives exist. Eventually, the sample is subjected to DNA extraction, 58 library preparation, sequencing and downstream bioinformatics analysis (Figure 1). 59 Here we examined the extent to which DNA extraction influences the quantification of microbial 60 composition, and compared it to other sources of technical and biological variation. The majority of 61 the protocol comparison studies to date have used a 16S rRNA gene amplification approach, which 62 suffers from additional issues. Specifically, the choice of primer, PCR bias and even the choice of 63 polymerase can affect the results, which may lead to different conclusions when performing the 64 same DNA extraction comparison in a different setup – issues that are minimized using metagenomic 65 sequencing. We compared a wide range of extraction methods, using metagenomic shotgun 66 sequencing, in respect to both taxonomic and functional variability, while keeping all other steps 67 standardized. We investigated the most commonly used extraction kits with varying modifications 68 and additional protocols which do not make use of commercially available kits (see Supplementary 69 Table 1 and Supplementary Information). While other studies have previously investigated the 70 differences between extraction methods in a given setting, we here systematically tested for 71 reproducibility within and across laboratories on three continents, by applying strict and consistent 72 quality criteria. We further assessed the accuracy of the best performing extraction methods by using 73 a mock community of ten bacterial species whose exact relative abundance was known. This 74 community included both gram-positive and gram-negative bacteria and their relative abundance 75 spanned three orders of magnitude. Based on these analyses we recommend a standardized 76 protocol for DNA extraction from human stool samples, which, if accepted by the research 77 community, will greatly enhance comparability among metagenomic studies. 78
|
4
|
6 | 2018 |
Social Media Posts