Skip to main content

Reproducibility, stability, and accuracy of microbial profiles by fecal sample collection method in three distinct populations.

Byrd DA, Chen J, Vogtmann E, Hullings A, Song SJ, Amir A, Kibriya MG, Ahsan H, Chen Y, Nelson H, Knight R, Shi J, Chia N, Sinha R

Publication: PLoS One (2019) • Vol 14 • Page e0224757PMID: 31738775View on PubMed

Purpose

This paper compared V4 16S rRNA sequencing results of the microbiome of case-matched fecal specimens that were stored for 0 days or 4 or 7 days at room temperature without preservation or among specimens that preserved on cards (Fecal occult blood test (FOBT) or Flinders Technology Associates (FTA)), in fecal immunochemical tests (FIT) tubes, in 70% ethanol, 95% ethanol, or in RNAlater. Specimens were collected from 132 healthy volunteers as part of three studies based in the United States (two studies were conducted at the Mayo Clinic, one study was conducted at the University of Colorado) and one study based in Bangladesh.

Conclusion

Based on outlier analysis, the authors excluded 0.25% (Mayo Clinic study 1) to 3.6% (University of Colorado study) of specimens. Interindividual differences were responsible for the majority of microbial variability (61-79% in unweighted UniFrac distance; 68-78% in generalized UniFrac distance; 63-78% in weighted UniFrac distance, and 77- 86% in Bray–Curtis distance), with <10% of the variability explained by collection/preservation method and <5% of the variability explained by storage duration at room temperature. Generally, technical reproducibility was high (interclass correlation coefficient, ICC >75%) for the abundance of four phyla, twelve genera, both alpha diversity metrics and all four beta diversity metrics, regardless of collection/preservation method, but there were exceptions for some phyla in specimens preserved in 70% or 95% ethanol. The ICC between Operational taxonomic units (OTU) abundances on days 0 and 4 (day 7 for the University of Colorado study) was high (>75%) for all metrics in specimens preserved on FOBT/FTA cards or in RNAlater, but lower ICCs were observed for one or more metric with all other preservatives. Similarly, ICCs between day 0 and day 4 were high for most genera for specimens preserved with FOBT/FTA cards, FIT tubes, 95% ethanol, or RNAlater, but were lower for specimens that weren’t preserved or specimens preserved in 70% ethanol. A significant log-fold change from day 0 to day 4 (day 7 for University of Colorado specimens) of selected (based on prevalence in ≥50% of specimens with mean of >10 reads) taxa (at the phyla, genera or class level) was found for 62% of taxa among specimens preserved in 70% ethanol, but only 7% of taxa among those preserved on FOBT/FTA cards. When compared to the “gold standard” that was frozen on day 0, the Spearman’s correlation coefficients for all preservation methods were strong (>0.70) for most alpha and beta diversity metrics but were lower for weighted UniFrac distance (0.485-0.799) and abundance of individual phyla. Compared to gold standard specimens, those preserved with 70% ethanol, RNAlater, FIT tubes, 95% ethanol, and FOBT/FTA cards demonstrated significant fold-changes in 31%, 28%, 24%, 24%, and 10%of selected taxa (based on prevalence in ≥50% of specimens with mean of >10 reads) , respectively. Finally, significant differences in genera distribution were observed between specimens from the three US-based studies and the study in Bangladesh (Healthy Effects of Arsenic Study, HEALS).

Studies (1)

Purpose

This study compared V4 16S rRNA sequencing results of the microbiome of case-matched fecal specimens that were stored for 0 days or 4 or 7 days at room temperature without preservation or among specimens that preserved on cards (Fecal occult blood test (FOBT) or Flinders Technology Associates (FTA)), in fecal immunochemical tests (FIT) tubes, in 70% ethanol, 95% ethanol, or in RNAlater. Fecal specimens were collected from 132 healthy volunteers as a part of four different studies: Mayo Clinic study 1 (20 volunteers), Mayo Clinic study 2 (52 volunteers), Bangladesh Healthy Effects of Arsenic Study (HEALS, 50 volunteers) and the University of Colorado study (10 volunteers). For the Mayo Clinic Study 1 specimens were mixed and aliquots were placed in tubes with no solution (3 replicates/timepoint), on FOBT/FTA cards (3 replicates/timepoint), in 70% ethanol (2 replicates/timepoint) and RNAlater (2 replicates/timepoint) and stored at room temperature for 0 or 4 days before freezing at -80°C. For the Mayo Clinic Study 2, specimens were mixed and aliquots placed in tubes with no solution (3 replicates day 0 only), in FIT tubes (2 replicates/timepoint), on FOBT/FTA cards (3 replicates/timepoint), in 95% ethanol (2 replicates/timepoint) and RNAlater (2 replicates/timepoint) and stored at room temperature for 0 or 4 days before freezing at -80°C. For the HEALS, specimens were mixed and aliquots were placed in tubes with no solution (2 replicates day 0 only), in FIT tubes (2 replicates/timepoint), on FOBT/FTA cards (3 replicates/timepoint), in 95% ethanol (2 replicates/timepoint) and RNAlater (2 replicates/timepoint) and stored at room temperature for 0 or 4 days before freezing at -80°C. For the University of Colorado study, specimens were mixed and aliquots placed in tubes with no solution (1 replicate, day 0 only), on FOBT/FTA cards (1 replicate/timepoint), in 70% ethanol (1 replicate/timepoint), in 95% ethanol (1 replicate/timepoint) and RNAlater (1 replicate/timepoint) and stored at room temperature for 0 days before freezing at -80°C or 7 days and used directly. DNA was extracted by the Knight Laboratory (Knight Laboratory and Mayo for Mayo study 1) using the MoBio Powersoil DNA Kit, and the V4 region of the 16S rRNA gene was PCR amplified and sequenced using Illumina’s MiSeq (250 bp Mayo 1 study) or HiSeq (all other studies) machine.  A biome table with each OTU was constructed from reads after they were demultiplexed and quality filtered (Phred quality score of >3) using QIIME 1.9, and read-errors were removed using deblur. The R Phyloseq package was used to calculate two alpha diversity metrics (the number of OTUs present and the Shannon Diversity Index). The R vegan package was used to calculate the four Beta diversity metrics (unweighted, generalized and weighted UniFrac distances, and the Bray-Curtis distance). Possible outliers were identified using the outlier index that was calculated based on the unweighted UniFrac distance. Specimens with an outlier index > 1.4 were further analyzed using principal coordinate analysis and were subsequently excluded if they appeared to cluster improperly. Intraclass correlation coefficients (ICCs) were used to investigate technical reproducibility. Spearman’s correlations (SCCs) were used to investigate accuracy.

 

Summary of Findings

Based on the outlier analysis, the authors excluded 1 specimen (0.25%) sequenced by Knight Lab and 5 specimens (1.25%) sequenced by Mayo from the Mayo 1 study, 16 specimens (1.5%) from the Mayo 2 study, 15 specimens (1.5%) from HEALS, and 5 specimens (3.6%) from the University of Colorado study.

Interindividual differences were responsible for the majority of microbial variability (61-79% in unweighted UniFrac distance; 68-78% in generalized UniFrac distance; 63-78% in weighted UniFrac distance, and 77-86% in Bray–Curtis distance), with <10% of the variability explained by collection/preservation method and <5% of the variability explained by storage duration at room temperature.

Technical reproducibility for replicates was high (ICC >75%) for the abundance of four phyla, both alpha diversity metrics and all four beta diversity metrics, regardless of collection/preservation method. The ICC were also high (>75%) for all twelve genera investigated, with the exception of Blautia and Faecelibacterium in 70% ethanol specimens and Dorea in 95% ethanol specimens. The ICC between OTU abundances on days 0 and 4 (day 7 for the University of Colorado study) was high (>75%) for all metrics in specimens preserved on FOBT/FTA cards or in RNAlater, but lower ICCs were observed for Firmicutes in specimens without preservative, or specimens preserved in FIT tubes or with 70% or 95% ethanol; for Actinobacteria in specimens without preservative or specimens preserved with 70% ethanol; for Bacteroidetes in specimens without preservative or specimens preserved in FIT tubes, or with 70% or 95% ethanol; and for Proteobacteria in specimens without preservative or specimens preserved with 70% ethanol. Shannon Index and unweighted, generalized and weighted UniFrac distances in specimens without preservative and specimens preserved with 70% ethanol. Similarly, ICCs between day 0 and day 4 were high for most genera for specimens preserved with FOBT/FTA cards, FIT tubes, 95% ethanol, or RNAlater but were lower in specimens that weren’t preserved or specimens in 70% ethanol. A significant log-fold change from day 0 to day 4 (day 7 for Colorado specimens) of selected (based on prevalence in ≥50% of specimens with mean of >10 reads) taxa (at the phyla, genera or class level) was found for 62% of taxa for specimens preserved in 70% ethanol, but for only 7% of taxa when specimens were preserved on FOBT/FTA cards. When compared to the “gold standard that was frozen on day 0, Spearman’s correlation coefficients for all other preservation methods were strong (>0.70) for most alpha and beta diversity metrics but were lower for weighted UniFrac distance (0.485-0.799) and the abundance of individual phyla. Compared to gold standard specimens, those preserved with 70% ethanol, RNAlater, FIT tubes, 95% ethanol, and FOBT/FTA cards demonstrated significant fold-changes in 31%, 28%, 24%, 24%, and 10% of selected taxa (based on prevalence in ≥50% of specimens with mean of >10 reads), respectively. Finally, significant differences in genera distribution were observed between specimens from the three US-based studies and the HEALS study in Bangladesh.

  • Biospecimens

    FluidFeces
  • Preservation Types

    FrozenOther PreservativeEthanol
  • Diagnoses

    Normal
  • Platforms

    DNANext generation sequencing
  • Pre-analytical Factors

    ClassificationPre-Analytical FactorValue(s)
    Biospecimen AcquisitionLocale of biospecimen collection
    Bangladesh
    Mayo study 1
    Mayo study 2
    University of Colorado
    Biospecimen PreservationType of fixation/preservation
    Ethanol
    FOBT Card
    FTA Card
    Fecal immunochemical test (FIT)
    Frozen
    RNAlater
    StorageTime at room temperature
    0 days
    7 days (University of Colorado)