Long-term storage of feces at −80 °C versus −20 °C is negligible for 16S rRNA amplicon profiling of the equine bacterial microbiome

Introduction

Targeted marker gene sequencing allows for the characterization of biological communities at high-resolution (Cristescu, 2014). In particular, quantifying the resident bacteria of the gastrointestinal tract, or the gut microbiome, using deep sequencing of 16S rRNA amplicons has furthered our understanding of host physiology and disease etiology, and improved clinical diagnostics and treatment of gut dysbiosis (Kim et al., 2017; Vandeputte et al., 2017). As the cost of Next-Generation Sequencing (NGS) experiments continues to fall, marker gene surveys have become increasingly commonplace. However, the findings of different studies are often not readily comparable or generalizable because of a lack of standardized methodology (Vandeputte et al., 2017). For example, the use of different DNA extraction kits and amplification methodologies have been observed to introduce technical variation comparable to that of biological variation in metagenomic and amplicon-based experiments (Costea et al., 2017; Sinha et al., 2017). While these effects are substantial and well-documented, the spread of marker gene surveys has brought with it a renewed interest in optimizing and finding alternative sample storage solutions, which are also known as large sources of technical variation (Kim et al., 2017).

Short-term processing and storage of fecal samples for microbiome research has received considerable attention due to their importance in clinical applications (Table 1). Where immediate extraction is not possible, the conventionally accepted gold standard for short-term storage is rapid freezing at −80 °C without buffer (Vandeputte et al., 2017). While the process of freezing samples is still under contention (Bahl, Bergström & Licht, 2012; Fouhy et al., 2015; Metzler-Zebeli et al., 2016), it remains by far the most common approach for preventing rapid changes in the abundance of specific taxa under ambient conditions following sampling. This phenomenon, often referred to as microbial ‘blooms’, has been well documented. For example, a meta-analysis has shown that human rectal swabs stored at room temperature experience characteristic growths of Gammaproteobacteria (Amir et al., 2017). A similar increase in the abundance of proteobacteria was also seen in the microbiota of tadpoles (Nanorana parker) gut samples left thawing at ambient temperature for 12 h (Anslan et al., 2019), while equine feces collected 12 h post-defecation showed sharp increases of families from the Firmicutes phylum (Beckers, Schulz & Childers, 2017). Though the effect of sample storage at ambient temperature has been well-documented across study systems, it has remained unclear whether freezing at −80 °C is overly conservative. Notably, an increasing number of studies have claimed or demonstrated that storage at −20 °C is generally sufficient over the short-term (Song et al., 2016; Blekhman et al., 2016; Panek et al., 2018; Wang et al., 2018). Overall, the consensus emerging from these studies is that (any) differences in microbiome profiles caused by variation in freezing temperature are typically negligible when compared to biological patterns of interest (Lauber et al., 2010; Carroll et al., 2012; Dominianni et al., 2014; Blekhman et al., 2016; Bundgaard-Nielsen, Hagstrøm & Sørensen, 2018) and that storage at −80 °C may therefore be unnecessary for storage periods less than a few weeks or months.

While the impacts of short-term storage temperature on microbiome profiles are relatively well understood, less is known of the impacts of storage conditions over long periods of time. Delineating short-term from long-term storage is necessarily subjective, but a gap exists in the literature for storage periods greater than a few weeks or months (Table 1). A notable exception is the study of Kia et al. (2016) which showed that the microbiota from freeze-dried human fecal samples stored for 14 years at −20 °C were similar to those of samples stored at −80 °C in a cohort matched for age, sex, BMI and pre-existing health conditions (McDonald et al., 2018)⁠. However, a significant difference was nonetheless observed between groups when considering Bray-Curtis dissimilarities, and it is unclear if similar results would be obtained without the use of freeze-drying which is an uncommon practice in microbiome studies.

In the absence of relevant research, −80 °C has been widely adopted as the gold standard for long-term storage of biological samples intended for microbiome analyses including feces (Hale et al., 2015; Jenkins et al., 2018; Shaw et al., 2016), necessitating the use of ultra-low temperature freezers. Compared to regular −20 °C freezers, ultra-low temperature freezers incur higher capital and energy costs, have higher carbon footprints, and lower durability (Gumapas & Simons, 2013). In addition to constraints imposed by those with limited funding and resources (Vogtmann et al., 2017; Carruthers et al., 2019), immediate or prompt transfer to −80 °C freezers is also often logistically impractical, especially for studies conducted in remote areas (Song et al., 2016; Blekhman et al., 2016). Ultimately, meeting current gold standards often impose significant financial and logistical burdens on researchers, and failure to meet these standards creates uncertainty regarding sample integrity and the validity and comparability of results.

Given the need to validate storage protocols currently in use by our group and others for projects where access to −80 °C storage is limited or impossible, we evaluated the effect of long-term freezing storage temperature on fecal samples collected as part of a long-term study of feral horses on Sable Island, Nova Scotia, Canada (see Gold et al., 2019 and Regan et al., 2020 for details about this study). Specifically, we tested whether storing aliquots from the same fecal sample at either −20 °C and −80 °C for 4 years affected 16S rRNA amplicon characterization of the equine bacterial microbiome. This study is among the firsts to explicitly test the for impacts of long-term freezing temperature on 16S microbiome profiling, and will contribute to protocol development and evaluation in microbiome research.

Materials and Methods

Results

Community composition

Sequencing generated a total of 1,615,043 reads, with a mean of 100,940 reads per sample (87,049–124,423). After sequence processing including quality filtering, denoising, merging and chimera removal, an average of 52,600 reads remained per sample (45,827–64,386). A total of 4427 ASVs were identified, of which 126 were singletons and a further 33 were identified as contaminants. After removing singletons and contaminants, an average of 71.0%, 88.0% and 97.7% of sequences within a sample were assigned to the genus, family and phylum levels with at least 80% confidence, respectively. Bacterial communities were dominated by the Firmicutes (average proportion 40.8%; 32.1–53.2%) and Bacteroidota phyla (average proportion 35.5%; 22.3–41.4%, Fig. S1). A total of 89 unique families were observed, and 65 of these were present at proportional abundances of 1% or less within a sample (Fig. 1). At the genus level, 138 unique genera were identified of which 114 were present at proportional abundances of 1% or less within a sample (Fig. S2).

Alpha diversity comparisons

The number of unique ASVs per sample ranged from 1,565 to 2,095 with a mean (±1 standard deviation) of 1,797 ± 172 (Fig. 2). Chao1 estimates ranged from 1,677 to 2,305 with a mean of 1,935 ± 200, and Shannon values ranged from 5.79 to 6.44 with a mean of 6.00 ± 0.17 (Fig. 2). Observed ASV richness (p = 0.39), Chao1 indices (p = 0.64) and Shannon diversity (p = 0.11) did not differ significantly between samples stored at different temperatures.

Beta diversity comparisons

Aliquots from the same sample generally yielded consistent results, as apparent in NMDS ordinations (Fig. 3). The proportion of multivariate variance explained by sample ID was large (79–88%) and statistically significant (p < 0.001) for all distance measures considered (Table S1). In contrast, storage temperature explained a comparatively small (2–3%) and non-significant amount of multivariate variation across distances. Similarly, no difference in beta dispersion was observed between temperature storage groups indicating homogeneity of variance in the community structure between treatment groups (Table S1).

Discussion

Immediate extraction of DNA for amplicon sequencing experiments is preferable but many study systems and clinical applications require the storage of samples by way of freezing. We aimed to determine if equine fecal samples stored at −20 °C for multiple years could recapitulate the bacterial community structure inferred from samples stored at −80 °C. Samples stored at different temperatures did not differ in ASVs richness, Chao1 indices, or evenness (Shannon). Furthermore, variation in community composition among samples was largely driven by sample (individual) ID with no detectable impact of storage temperature. In short, our study demonstrates that archiving equine feces at either −20 °C or −80 °C for multiple years has negligible impacts of 16S rRNA microbiome profiling.

Our findings mirror those obtained in similar studies that have looked at the impact of storage freezing temperature on microbiome profiling (Table 1). For example, Jenkins et al. (2018) noted that murine fecal samples stored at −20 °C and −80 °C for 33 days did not differ in terms of richness or the Chao1 estimator. Microbiome profiles from spider monkey feces stored at −20 °C for 8 weeks closely resembled those stored at −80 °C in composition and diversity (Hale et al., 2015). Using pyrosequencing reads, Hang et al. (2014) did not observe differences between oropharyngeal swabs at these temperatures using weighted UniFrac distances.

Though we did not compare frozen samples to freshly extracted samples, previous studies have shown that long-term storage at −20 °C generally gives rise to microbiota profiles similar to those that have been immediately extracted (Hale et al., 2015; Kia et al., 2016). In addition, limited differences were observed between fresh fecal samples and those stored at −80 °C in RNAlater for 5 years (Tap et al., 2019). Together with our findings, these suggest that archiving fecal samples for long periods of time at −20 °C likely yields uncompromised microbiome profiles, even when compared to DNA extracted from fresh samples. Studies comparing results from fresh fecal samples with those of samples archived at −20 °C for multiple aliquots would be valuable.

Based on this study and other accumulating evidence, we suggest the use of −20 °C as a cost-effective storage temperature to maintain microbiome integrity over multiple years comparable to that of samples stored at −80 °C, at least for targeted sequencing of short amplicons. However, it remains to be seen if this is generalizable to longer time periods such as decades, other sample types or host organisms, or non-amplicon based sequencing methods such as shotgun metagenomics and third generation sequencing. It is also possible that larger sample sizes would lead to the detection of subtle variation, if present. We also note that our study used manual defrost freezers, and that long-term archival of samples in self-defrosting freezers experiencing frequent temperature fluctuations may yield different outcomes.

In addition to general relevance, our findings are especially pertinent to clinicians and researchers conducting field or community-based studies where access to −80 °C storage is limited or absent. Wider adoption and acceptance of storage at −20 °C for amplicon-based microbiome research should greatly increase total storage space, encourage the re-use of samples, and promote longitudinal sampling (Björk et al., 2019). Increased reliance on −20 °C freezers, at least for storage up to a few years, may also help reduce drop-out rates in microbiome studies by alleviating the logistical burden of rapidly transferring samples to −80 °C (Vandeputte et al., 2017). In terms of standardizing study design, DNA extraction and amplification protocols may be of greater importance for resource allocation, which are likely larger sources of technical variation (Costea et al., 2017; Sinha et al., 2017) compared to long-term storage temperature.

Conclusion

We demonstrated that technical replicates of equine fecal samples stored at −20 °C and −80 °C for multiple years yield similar bacterial microbiome results in terms of alpha and beta diversity. Our findings add to mounting evidence indicating that standard domestic freezers are both economical and effective for microbiome research.

Supplemental Information