Comparison of extraction methods for recovering ancient Comparison of extraction methods for recovering ancient microbial DNA from paleofeces microbial DNA from paleofeces

Objectives: Paleofeces are valuable to archeologists and evolutionary biologists for their potential to yield health, dietary, and host information. As a rich source of preserved biomolecules from host-associated microorganisms, they can also provide insights into the recent evolution and changing ecology of the gut microbiome. How-ever, there is currently no standard method for DNA extraction from paleofeces, which combine the dual challenges of complex biological composition and degraded DNA. Due to the scarcity and relatively poor preservation of paleofeces when compared with other archeological remains, it is important to use efficient methods that maximize ancient DNA (aDNA) recovery while also minimizing downstream taxonomic biases. Methods: In this study, we use shotgun metagenomics to systematically compare the performance of five DNA extraction methods on a set of well-preserved human and dog paleofeces from Mexico (~1,300 BP). Results: Our results show that all tested DNA extraction methods yield a consistent microbial taxonomic profile, but that methods optimized for ancient samples recover significantly more DNA. Conclusions: These results show promise for future studies that seek to explore the evolution of the human gut microbiome by comparing aDNA data with those generated in modern studies. amplification with Kapa HiFi Uracil+. The libraries were then purified by silica column (Qiagen MinElute PCR Purification kit) and quantified with a BioAnalyzer 2,100 using the High Sensitivity DNA reagents. MinElute centrifugation speeds were modified to 3,400 rcf, 9,400 rcf, and 17,900 rcf for binding, washing, and elution steps, respectively. All libraries were pooled in equi-molar amounts, size selected to 150 – 600 bp using a PippinPrep 2% aga-rose gel, and sequenced by Illumina HiSeq 2 × 100 paired-end sequencing.


| INTRODUCTION
The gut microbiome is a core component of human biology, contributing to a range of physiological functions from digestion to host immunity. To date, most studies of the human gut have focused on industrialized societies, but recent research on hunter-gatherer, horticulturalist, and pastoralist cultures has revealed previously unknown microbial diversity in populations living more traditional lifeways Schnorr et al., 2014;Yatsunenko et al., 2012). Such studies suggest that industrialized societies may have undergone recent changes in gut microbiome structure and function related to changing dietary and sanitation practices; however, testing this hypothesis requires the recovery of well-preserved gut microbiota from the archeological record. Paleofeces, alternatively known as coprolites, are the preserved remnants of feces from humans or other animals. Preserving only under extraordinary conditions, such as rapid desiccation or freezing, paleofeces are relatively uncommon in the archeological record; however, those specimens that do survive have the potential to shed light on the evolution of the gut microbiome (Schnorr, Sankaranarayanan, Lewis Jr., & Warinner, 2016;Warinner, Speller, Collins, & Lewis Jr., 2015).
The genetic comparison of paleofeces and fresh feces, however, presents two major challenges. The first is the degraded nature of ancient DNA (aDNA) itself, and the second is that reconstruction of a complex microbial community, such as the gut microbiome, can be influenced by DNA extraction methods (Wesolowska-Andersen et al., 2014). The aDNA is both low in abundance and highly fragmented, and as a result, aDNA extraction methods necessitate a minimum of handling steps and chemical reagents, as well as a specialized silica binding protocol, in order to mitigate DNA loss and efficiently recover short aDNA fragments (Dabney et al., 2013). On the other hand, modern microbiome reconstruction depends on the efficiency with which bacterial lysis occurs during DNA extraction, which is influenced by differences in cell wall structure, spore formation, and other factors. In order to reduce data variability in modern microbiome studies, many laboratories and large-scale projects have attempted to standardize DNA extraction methods. For example, both the Earth Microbiome Project (Marotz et al., 2017;Thompson et al., 2017) and the Human Microbiome Project (HMP) (Aagaard et al., 2013) have recommended using the Qiagen PowerSoil (formerly MoBio PowerSoil) DNA extraction kit, and consequently, this kit has become relatively standard in modern microbiome research.
To ensure efficient cell lysis, this kit uses extensive mechanical, chemical, and enzymatic lysis steps, followed by inhibitor removal using proprietary chemicals, and finally a DNA purification and concentration step using a silica spin column. In addition, some variants of this protocol have also advocated using an initial heat lysis step .
The fact that the standard methods for ancient and modern DNA extraction differ so greatly is a potential problem that could introduce bias when generating and comparing DNA sequence data obtained from paleofeces and fresh feces using two different protocols. However, on the other hand, using the same extraction method for both sample types is problematic because it would very likely yield suboptimal results for one sample type. For example, using an optimized aDNA protocol for modern microbiome DNA extraction is expected to result in inefficient cellular lysis and biased taxonomic recovery, and using an unmodified commercial DNA extraction kit for paleofeces would likely result in unacceptable aDNA losses.
To address this problem and determine the impact of extraction protocol on DNA recovery and reconstructed microbial profiles obtained from paleofeces, we systematically tested a panel of five DNA extraction protocols on well-preserved human and dog paleofeces from the Cueva de los Muertos Chiquitos archeological site. Located near the Rio Zape river in Durango, Mexico and dating to~1,300 BP, paleofeces from this site have been shown to contain well-preserved aDNA deriving from gut-associated bacteria (Tito et al., 2008;Tito et al., 2012). To assess protocol performance, we compare both total DNA yield and reconstructed microbial community structure. All genetic sequence data was generated using a shotgun metagenomics approach in order to avoid known aDNA amplification biases due to length polymorphisms in the 16S rRNA gene (Ziesemer et al., 2015). Overall, our findings show that protocols developed specifically for the recovery of aDNA result in significantly higher DNA yields compared with commercial DNA extraction kits. Importantly, however, all extraction protocols resulted in consistent taxonomic profiles from paleofeces, indicating that aggressive cell lysis and inhibitor removal steps are not necessary to efficiently recover DNA from paleofeces. Consequently, genetic sequences obtained from paleofeces using optimized aDNA extraction protocols can be compared with previously published gut microbiome data obtained using commercial kits without systematic taxonomic bias due to extraction protocol differences.

| Samples and study design
Three paleofeces samples were selected for analysis from the Cueva de los Muertos Chiquitos archeological site, a cave situated approximately 15 m above the Rio Zape in Durango, Mexico. This site contains evidence of storage, agave feasting, waste disposal, and burial and is associated with the Loma San Gabriel Culture, a group of rural agriculturalists who occupied the area from approximately 1,200 to 1,400 years ago (Jimenez et al., 2012). Analysis of dental casts in quids from the site shows that at least 49 people were involved in feasting activity at the site (Hammerl, Baier, & Reinhard, 2015). Wooden materials associated with the paleofeces have been dated to 1,300 ± 100 BP (Brooks, Kaplan, Cutler, & Whitaker, 1962). The paleofeces were excavated from a midden above an adobe floor. The adobe floor sealed the remains of several partially mummified child burials, as well as offerings. This site was remarkably deep and the refuse deposits were separated from the entry of the cave by approximately 10 m.
This distance was composed of rock fall. Inside the rock fall was the midden overlying the adobe floor and extending a further 4-5 m deeper into the cave. This unusually isolated archeological context deep within a cave has contributed to their exceptional preservation, and paleofeces from this site have been previously shown to preserve DNA from gut microbiome-associated bacterial taxa (Tito et al., 2008;Tito et al., 2012). Paleofeces found within the midden are consistent with human and dog feces. The samples in this study are a subsample of 36 specimens analyzed for parasite remains (Jimenez et al., 2012).
One putative dog sample (Zape 2) and two human samples (Zape 5 and Zape 28) of paleofeces were selected for analysis. A canine origin for Zape 2 was suspected based on the presence of previously identified dog-associated parasites identified within the fecal material (Cleeland, Reichard, Tito, Reinhard, & Lewis Jr, 2013;Jimenez et al., 2012), and later confirmed in this study (see in the following). Because human and dog paleofeces make up the majority of paleofeces reported at archeological sites, we include paleofeces from both host species in this study.  >600 mAU/ml) to form a slurry for DNA extraction. Following extraction and purification, DNA was quantified using a Qubit 3.0 fluorometer with a double-stranded DNA High Sensitivity assay (Table S1).

| Extraction Method A: Human Microbiome Project protocol with PowerSoil kit
This method follows the HMP protocol for the MoBio PowerSoil DNA extraction kit and was used as a test of the efficacy of a standard microbial extraction protocol on DNA recovery from paleofeces.
In brief, the paleofeces suspension was added to a PowerBead tube containing 750 μL guanidine thiocyanate and garnet bead solution, and samples were rotated for 2 hr at RT. The sample was then incubated at 65 C for 10 min with 60 μL of solution C1, followed by bead beating for 10 min. Following centrifugation for 1 min at 17900 rcf, the supernatant was then processed according to the PowerSoil kit manufacturer's instructions. In total, this protocol involves mechanical lysis, chemical lysis, heat lysis, chemical removal of inhibitors, and silica purification steps. Final DNA elution was performed with 60 μL of Solution C6 after a 5 min RT incubation of the elution buffer on the column.

| Extraction Method B: Modified MinElute protocol
This method was selected as a test of a highly efficient aDNA extraction protocol developed for bone (Dabney et al., 2013) on paleofeces as a source material. While this protocol has been shown to be highly effective for the recovery of aDNA from mineralized tissues, including bone, dentine, and dental calculus (Mann et al., 2018), it has yet to be empirically tested on paleofeces. In brief, the paleofeces suspension was added to a PowerBead tube containing 750 μL guanidine thiocyanate and garnet bead solution, and samples were rotated for 4 hr at RT, followed by bead beating for 10 min. Samples were then spun down at 3400 rcf for 5 min, and the supernatant was then added to 14 mL of Qiagen PB buffer. This was then centrifuged in a MinElute column (Qiagen) attached to a Zymo-Spin V column (Zymo Research) for 4 min at 1500 rcf, rotated 90 , and then centrifuged for an additional 2 min. The column was then dry spun for 1 min at 3400 rcf and washed twice with 700 μL Qiagen PE buffer at 9400 rcf. DNA was eluted from the column after a 5 min RT incubation in two rounds of 30 μL of Qiagen EB buffer at 17900 rcf for a total volume of 60 μL.  (Table S3), and the insert lengths of all mapped reads were computed.

| Extraction Method
The median insert length was calculated for each species in each sample, and their distributions are shown in Figure S1.
In order to determine the effect of extraction method on the observed microbial community of the paleofeces, the resulting BIOM tables were merged into a single BIOM table using the script merge_otu_tables.py in QIIME (Data S1), and beta diversity was calculated using the weighted UniFrac metric (Lozupone, Lladser, Knights, Stombaugh, & Knight, 2011) via the script beta_diversity_through_plots.py. Principal coordinates were extracted from the output file and visualized using R.

| Assessment of paleofeces microbial preservation
To determine whether the reconstructed microbial communities observed were endogenous to the paleofeces, Bayesian source tracking was performed using SourceTracker v2.1.0 . Reference fecal metagenomes obtained from industrialized human populations , non-industrialized human populations , and domesticated dogs (Li, Lauber, Czarnecki-Maulden, Pan, & Hannah, 2017), as well as soil metagenomes (Johnston et al., 2016) were prepared in QIIME following the same procedures used for paleofeces (see above), and the BIOM files were merged with those of the current study for use with SourceTracker2. Then, to determine whether the paleofeces DNA exhibits appropriate molecular behavior for aDNA, such as DNA damage, the merged reads from each sample were mapped to the genome of Prevotella copri (DSM 18205), and mapDamage 2.0 (Jonsson, Ginolhac, Schubert, Johnson, & Orlando, 2013) was used to measure the frequency of nucleotide deaminations that are characteristic of aDNA. P. copri was selected because it has been reported in both human and dog gut microbiota (Li et al., 2017;Tett et al., 2019). to each was log10 transformed and compared under the assumption that host DNA would be higher than other eukaryotic, non-host DNA for a given paleofeces sample. Next, the combined BIOM file containing OTUs from the paleofeces and reference fecal metagenomes was processed in QIIME. Beta diversity was again computed using the weighted UniFrac metric and visualized using principal coordinates analysis under the assumption that the microbial communities present in Zape 5 and Zape 28 would cluster with modern human feces, while

| Host determination of paleofeces
Zape 2 would cluster with dog feces.

| DNA recovery by extraction method
Overall DNA recovery differed by paleofeces sample, with Zape 2 yielding the highest average amount of DNA (median, 6.7 ng/mg; range, 0.8-11.7 ng/mg), followed by Zape 5 (median, 2.9 ng/mg; range, 0.02-6.7 ng/mg), and Zape 28 (median, 2.5 ng/mg; range, 0.03-7.2 ng/mg) (Table S1). DNA recovery also differed significantly by extraction method (Figure 2), and the lowest DNA yields were observed for extraction methods using the PowerSoil kit F I G U R E 2 Choice of DNA extraction method affects DNA recovery from paleofeces. Log-fold change in the normalized yield of DNA (ng DNA per mg paleofeces) is shown compared with the median DNA yield using method A for each sample. Statistical comparisons were made using a pairwise Wilcoxon test with Benjamini-Hochberg correction for multiple tests. Methods b-e all resulted in significantly higher median DNA yield than method A (**, p < .01); other significant pairwise differences are indicated by *, p < .05 F I G U R E 3 Sample origin, not DNA extraction method, is primary determinant of microbial community structure in paleofeces. Principal Coordinates Analysis (PCoA) analysis of weighted UniFrac betadiversity in paleofeces indicates that DNA extraction method has little influence on reconstructed microbial structure. This contrasts with studies of modern feces, where extraction method has been shown to introduce systematic taxonomic biases To further investigate the high degree of DNA loss observed in Method A, we compared the DNA fragment lengths recovered using the five protocols ( Figure S1). Although protocols B-E were expected to retain shorter DNA fragments due the MinElute silica column, we did not observe significant differences between any of the methods, but rather high inter-sample variation. This supports our conclusion that the cell lysis and/or inhibitor removal steps of the PowerSoil kit protocol likely had the greatest impact on DNA recovery.
Methods B, C, and D all performed well, and differences between these methods were non-significant. However, using two silica columns to expand the total DNA binding capacity and reduce clogging (Method D) resulted in the highest median DNA yields, and we found that this method was more straightforward to implement than Method B, which frequently required cleaning steps to dislodge clogged material. Method C incorporated a phenol: chloroform purification step that was intended to remove unwanted material prior to silica column filtration, but we found that it did not reduce clogging of the columns. Because Method C uses hazardous chemicals and did not outperform the other methods, we do not recommend its use. Overall, we recommend Method D on the basis of DNA recovery and ease of use.

| Effect of DNA extraction protocol on microbial community reconstruction
The microbial community structure reconstructed from modern feces has been shown to be highly influenced by choice of DNA extraction method (Wesolowska-Andersen et al., 2014). By contrast, here we find that reconstructed microbial profiles from paleofeces are highly similar regardless of DNA extraction protocol and that differences in sample beta diversity are primarily driven by the paleofeces from which the sample derives ( Figure 3). This finding suggests that cellular degradation over time has sufficiently weakened the cellular structure of the ancient microbial cells such that even highly simplified DNA extraction protocols, such as those used in Methods B and D, recover a microbial community similar to complex and highly aggressive protocols, such as Method A.

| Gut microbiome preservation
To assess gut microbiome preservation in the paleofeces, Bayesian source tracking of the microbial taxa (OTUs) present in each sample was performed using published reference metagenomes of human feces from industrialized and non-industrialized populations, dog feces, and soil ( Figure 4a). We found that the majority of the OTUs present are consistent with microbes found within human or dog feces rather than soil, indicating wellpreserved paleofeces. Additionally, all three paleofeces exhibited DNA damage patterns consistent with authentic aDNA (Figures 4b   and S2). The high degree of endogenous gut microbiome preservation we observe is consistent with previous studies that have reported good molecular preservation of paleofeces at this site (Tito et al., 2008;Tito et al., 2012).

| CONCLUSIONS
Paleofeces are a valuable resource for investigating the evolution of the gut microbiome and the diet and health of past peoples, but until now no systematic studies had been conducted to assess the performance and potential biases of existing DNA extraction protocols on the recovery and reconstruction of microbiome profiles. In this study, we compared five DNA extraction methods on a panel of wellpreserved human and dog paleofeces and evaluated methodological performance on the basis of DNA recovery and taxonomic composition. We found that DNA extraction methods that have become fieldstandard in modern microbiome studies, such as the HMP PowerSoil protocol, recover significantly less DNA from paleofeces than methods that have been developed and optimized for ancient skeletal material. Additionally, we found that aDNA optimized-methods do not negatively impact the structure of the reconstructed microbial communities when compared with the HMP PowerSoil protocol, and in fact, all DNA extractions tested in this study yielded highly similar microbial communities. This finding supports future research that seeks to compare metagenomic data generated using optimized protocols for both ancient and modern fecal samples. For paleofeces, we Christina G. Warinner https://orcid.org/0000-0002-4528-5877