Longitudinal gut metagenome stability in German adults

Background: Metagenomics has become a prominent approach for exploring the role of the gut microbiota in human health. However, the temporal variability of the healthy gut microbiome has not yet been studied in depth using metagenomics and little is known about the effects of different sampling and preservation approaches. We performed metagenomic analysis on fecal samples from seven subjects collected over a period of up to two years to investigate temporal variability and assess preservation-induced variation, specifically, fresh frozen compared to RNALater. We also monitored short-term disturbances caused by antibiotic treatment and bowel cleansing in one subject. Results: We find that the human gut microbiome is temporally stable and highly personalized at both taxonomic and functional levels. Over multiple time-points, samples from the same subject clustered together, even in the context of a large dataset of 888 European and American fecal metagenomes. One exception was observed in an antibiotic intervention case where, more than one year after the treatment, samples did not resemble the pre-treatment state. Clustering was not affected by preservation method. No species differed significantly in abundance, and only 0.36% of gene families were differentially abundant between preservation methods. Conclusions: Technical variability is small compared to the temporal variability of an unperturbed gut microbiome, which in turn is much smaller than the observed between-subject variability. Thus, short-term preservation of fecal samples in RNALater is an appropriate and cost-effective alternative to freezing of fecal samples for metagenomic studies.

Classified as human gut / stool shotgun DNA metagenome of adult by sonnet_x2 with confidence 0.65. Assay shotgun_dna.

• host run.scientific_name · “human gut metagenome”
• body site study.description · “fecal samples from seven subjects”
• life stage study.description · “gut microbiota in human health”

566 runs over 70 samples · mean 0.74 Gbp per run (median 0.34 Gbp, total 418.87 Gbp; 566 runs with a base count) · layout PAIRED 285, SINGLE 281 · platform ILLUMINA 566 · models Illumina HiSeq 2000 (566) · 55% of runs with a Sandpiper profile · first public 2015-03-16

Costea PI first · Coelho LP · Sunagawa S · Munch R · Huerta-Cepas J · Forslund K · Hildebrand F · Kushugulova A · Zeller G · Bork P last · Wang Z first · Lou H · Wang Y · Shamir R · Jiang R · Chen T last · Galkin F first · Mamoshina P · Aliper A · Lane E · Moskalev V · Gladyshev VN · Zhavoronkov A last · Olekhnovich EI first · Ivanov AB · Ulyantsev VI · Ilina EN last · Balaji A first · Sapoval N · Seto C · Leo Elworth RA · Fu Y · Nute MG · Savidge T · Segarra S · Treangen TJ last · Argentini C first · Lugli GA · Tarracchini C · Fontana F · Mancabelli L · Viappiani A · Anzalone R · Angelini L · Alessandri G · Longhi G · Bianchi MG · Taurino G · Bussolati O · Milani C · van Sinderen D · Turroni F · Ventura M last · Valguarnera E first · Tung J · Pak JJ · Russell SK · Hecht DW · Schwartz DJ · Wardenburg JB last

PRJEB8347.csv.gz (2.8 KB) · parquet (73.4 KB) — all 70 samples, every field with route and confidence · PRJEB8347_determinations.csv.gz (3.6 KB) — 566 evidence rows