Collection media and delayed freezing effects on microbial composition of human stool

Abstract

Background: Different bacteria in stool have markedly varied growth and survival when stored at ambient temperature. It is paramount to develop optimal biostabilization of stool samples during collection and assess long-term storage for clinical specimens and epidemiological microbiome studies. We evaluated the effect of collection media and delayed freezing up to 7 days on microbial composition. Ten participants collected triplicate stool samples each into no media as well as RNAlater® with and without kanamycin or ciprofloxacin. For each set of conditions, triplicate samples were frozen on dry ice immediately (time = 0) or frozen at -80 °C after 3-days and 7-days incubation at 25 °C. Microbiota metrics were estimated from Illumina MiSeq sequences of 16S rRNA gene fragments (V3-V4 region). Intraclass correlation coefficients (ICC) across triplicates, collection media, and incubation time were estimated for taxonomy and alpha and beta diversity metrics.

Results: RNAlater® alone yielded the highest ICCs for diversity metrics at time = 0 [ICC median 0.935 (range 0.89-0.97)], but ICCs varied greatly (range 0.44-1.0) for taxa with relative abundances <1%. The 3- and 7-day freezing delays were generally associated with stable beta diversity for all three media conditions. Freezing delay caused increased variance for Shannon index (median ICC 0.77) and especially for observed species abundance (median ICC 0.47). Variance in observed species abundance and in phylogenetic distance whole tree was similarly increased with a 7-day delay. Antibiotics did not mitigate variance. No media had inferior ICCs at time 0 and differed markedly from any media in microbiome composition (e.g., P =0.01 for relative abundance of Bacteroidetes).

Conclusion: Bacterial community composition was stable for 7 days at room temperature in RNAlater® alone. RNAlater® provides some stability for beta diversity analyses, but analyses of rare taxa will be inaccurate if specimens are not frozen immediately. RNAlater® could be used as collection media with minimal change in the microbiota composition.

Fig. 1

Comparison of major phyla in fecal samples stored in different media at baseline

Fig. 2

The relative abundance of more common taxon tend to be more reproducible under condition RNAlater at time 0. The figure is based on taxa with median relative abundance greater than 0.1 % across all samples. Each diamond represents a taxon. The line was fitted by linear regression using all data points

Fig. 3

Principal coordinate analysis of fecal microbiota to evaluate structure reproducibility under different sampling and storage conditions. Samples from each of the ten subjects are presented with a different color. a Time 0, weighted beta diversity with all media conditions. b Time 0, unweighted beta diversity at time 0 with all media conditions. c Weighted beta diversity across all media conditions and three time points. Variation explained was 54.1, 11.6, and 5.7 % for PCoA1, 2, and 3, respectively. d Unweighted beta diversity across all media conditions and three time points. Variation explained was 14.3, 11.1, and 7.6 % for PCoA1, 2, and 3, respectively

Fig. 4

Flow diagram for the analysis of 16S rRNA sequence data, classification of operational taxonomic units, and generation of diversity metrics