Taxonomic and predicted metabolic profiles of the ...

FEMS Microbiology Ecology Advance Access published August 23, 2016

Taxonomic and predicted metabolic profiles of the human gut microbiome in pre-Columbian mummies Tasha M. Santiago-Rodriguez1,2, Gino Fornaciari3,4, Stefania Luciani5, Scot E. Dowd6, Gary A. Toranzos7, Isolina Marota5 and Raul J. Cano1,2,* 1

Center for Applications in Biotechnology, California Polytechnic State University, San Luis

Obispo, CA, USA; 2 Department of Biology, California Polytechnic State University, San Luis

and Surgery, Division of Paleopathology, University of Pisa, Pisa, Italy; 4Center for Anthropological, Paleopathological and Historical Studies of the Sardinian and Mediterranean Populations, Department of Biomedical Sciences, University of Sassari, Sassari, Italy; 5

Laboratory of Molecular Archaeo-Anthropology/ancient DNA, School of Biosciences and

Veterinary Medicine, University of Camerino, Camerino, Italy; 6 Molecular Research LP (MR DNA), Shallowater, Texas, USA; 7Department of Biology, University of Puerto Rico, San Juan, PR Running title: Pre-Columbian mummies gut microbiomes Keywords: Ancient gut microbiomes; Chagas’ disease; Firmicutes; Leishmaniasis; mummies *Corresponding author: Raul J. Cano Mailing Address: Center for Applications in Biotechnology, California Polytechnic State University, San Luis Obispo, CA, USA Email: [email protected]

1

Downloaded from http://femsec.oxfordjournals.org/ by guest on August 25, 2016

Obispo, CA, USA; 3Department of Translational Research on New Technologies in Medicine

Abstract Characterization of naturally-mummified human gut remains could potentially provide insights into the preservation and evolution of commensal and pathogenic microorganisms, and metabolic profiles. We characterized the gut microbiome of two pre-Columbian Andean mummies dating to the 10-15th centuries using 16S rRNA gene high-throughput sequencing and metagenomics, and compared them to a previously characterized gut microbiome of an 11 th century A.D. preColumbian Andean mummy. Our previous study showed that the Clostridiales represented the

communities were also preserved during the process of natural mummification, as shown with the metagenomics analyses. The gut microbiome of the other two mummies were mainly comprised by Clostridiales or Bacillales, as demonstrated with 16S rRNA gene amplicon sequencing, many of which are facultative anaerobes, possibly consistent with the process of natural mummification requiring low-oxygen levels. Metagenome analyses showed the presence of other microbial groups that were positively or negatively correlated with specific metabolic profiles. The presence of sequences similar to both Trypanosoma cruzi and Leishmania donovani could suggest that these pathogens were prevalent in pre-Columbian individuals. Taxonomic and functional profiling of mummified human gut remains will aid in the understanding of the microbial ecology of the process of natural mummification.

2


majority of the bacterial communities in the mummified gut remains, but that other microbial

Introduction Ancient microbial DNA (amDNA) studies are augmenting our understanding of the evolution of commensal and pathogenic microorganisms in diverse human surfaces including dental calculus, bones and coprolites (Luciani et al., 2006, Adler et al., 2013, SantiagoRodriguez et al., 2013, Appelt et al., 2014a, Appelt et al., 2014b, Cano et al., 2014, Hunter, 2014, Warinner et al., 2015). There are several different mechanisms by which amDNA may be preserved in human samples including sporulation and resuscitation promoting factors (rpf)

Micrococcus luteus, which lacks the ability of forming spores, in amber (Greenblatt et al., 2004). While there are several mechanisms of DNA preservation, authentic amDNA usually exhibits a large degree of fragmentation, where fragment sizes may range from dozens to a couple hundred base pairs (bp). Unlike small amDNA fragments, which would probably require numerous independent PCR amplifications of short overlapping fragments, larger amDNA fragments may be more suitable for molecular-based methods such as PCR (Paabo et al., 2004, Knapp & Hofreiter, 2010). More recently, next-generation sequencing has been utilized to distinguish ancient indigenous cultures based on the effect of dietary habits to the gut microbiome (Cano et al., 2014), determine the evolution of pathogens and infectious diseases (Adler et al., 2013, Warinner et al., 2014), and characterize the microbial community structure of ancient human samples (Tito et al., 2012, Warinner et al., 2014),

Mummified human gut remains are also known to be an important source of amDNA; yet, very few studies have examined the preservation of amDNA in this sample type. The process of natural mummification is highly unique, and results from the combination of low or 3


(Cano & Borucki, 1995, Greenblatt et al., 2004). Rpf have been associated with the survival of

elevated temperatures, low oxygen levels and dry conditions (Janaway et al., 2009). Natural mummification usually occurs in hyper-arid deserts such as those in Egypt, China, southwestern USA and the Andes. Unlike artificially-preserved mummies, environmental conditions aid in tissue desiccation, where the body shrivels to a dry leathery mass of skin and tendons that surround the bone. Natural mummification requires the water content to decrease below a critical threshold from 75-80 % to 20-27 %, resulting in the inhibition of liquefying putrefaction. Enzymatic action is then minimized or halted due to a drop in water content, as well as a

Previous studies have reported the preservation of bacterial DNA in mummified human gut remains. One study found sequences similar to Clostridium spp. in the gut of a naturallypreserved pre-Columbian Andean mummy from the 11th century (Ubaldi et al., 1998). This mummy was also included in the present study. Another study reported that the Firmicutes were among the most represented bacterial groups in mummified human gut remains (Tito et al., 2012). Our previous study characterizing the gut microbiome of an 11th century A.D. preColumbian Andean mummy using 16S rRNA gene high-throughput sequencing and metagenomics also demonstrated that the Firmicutes, particularly the Clostridiales, were among the most represented bacterial groups (Santiago-Rodriguez et al., 2015). That study also showed that sequences similar to other bacterial groups, archaea, fungi, viruses, including Human papillomavirus (HPVs), and pathogenic microbial eukaryotes, including Trypanosoma cruzi (Santiago-Rodriguez et al., 2015) were present in naturally-mummified human gut remains; however, little is still known about the microbial community composition, the preservation of pathogens, and preserved metabolic profiles of naturally-preserved human mummies. While in

4


reduction in temperature and changes in pH (Aufderheide, 2003).

our previous study we were able to draw some conclusions of the microbial community structure and function resulting from the process of natural mummification, the mummy included in our previous study exhibited a pathology associated with Chagas’ disease (Santiago-Rodriguez et al., 2015). Given that some disease states have been associated with perturbations to the gut microbiome (Kinross et al., 2011), it is of importance to characterize the gut microbiome of other naturally-preserved human mummies in order to understand the resulting microbial community structure and preserved functional profiles associated with the process of natural mummification.

process of natural mummification, metabolic profiles, disease states in ancient human populations, and the evolution of both commensal and pathogenic microorganisms in the human gut.

Therefore, by using 16S rRNA gene high-throughput sequencing and metagenomics, the present study aimed to: (i) characterize the gut microbiomes of two naturally-preserved preColumbian Andean mummies and compare them to a previously characterized pre-Columbian Andean mummy described above, (ii) identify preserved metabolic profiles and signatures associated with the process of natural mummification of the human gut and (iii) determine the evolution of genes associated with pathogens.

Results 16S rRNA gene analyses During DNA extraction of the mummified gut remains and prior to sequencing, DNA extraction controls or blanks were included both in the agarose gel and during DNA 5


Results will also open the opportunity to elucidate microbial preservation mechanisms during the

quantification to ensure amDNA authenticity, and to detect any potential contamination from reagents or the extraction process. DNA extraction controls did not show contamination. After quality filtering, a total of 79,752 (mummy FI3), 8,731 (mummy FI9) and 57,979 (mummy FI12) reads with an average length of 270 bp were processed. Data from mummy FI9 were previously generated and included in the present study (Santiago-Rodriguez et al., 2015). Prior to further analysis, we utilized SourceTracker to identify similarities with modern gut, skin, saliva and soil microbiomes. Mummy FI3 had the majority of the reads (84.1 %) matching modern gut

(Figure 1C) had the majority of the reads not matching any of the microbiomes included in the analysis. None of the reads matched the soil or skin microbiomes included in the analyses. Supplemental Figure 1 shows the alpha rarefactions curves using PD whole tree (Panel A), chao1 (Panel B) and observed Operational Taxonomic Units (OTUs) (Panel C) for mummies FI3, FI9 and FI12. We plotted alpha diversity measures present in the bacterial communities using PD whole tree (Figure 2A), chao1 (Figure 2B), and observed species (Figure 2C). Alpha diversity indices differed depending on the metric utilized and mummy studied.

A total of 422 (mummy FI3), 222 (mummy FI9) and 477 (mummy FI12) OTUs were identified in each mummy, and only a few of these OTUs were shared (Figure 3). Table 1 describes the shared OTUs between the mummies. Taxonomic assignments at the order level for all three mummies using 16S rRNA gene analyses are shown in Table 2. Mummy FI3 had 96.834 % of the sequences belonging to the Bacillales. Other common inhabitants of modern human guts with small relative abundances and that were present in mummy FI3 included the Clostridiales (0.034 %), Enterobacteriales (0.003 %) and Turicibacterales (0.005 %). Mummy 6


microbiomes (Figure 1A), while mummies FI9 (96.8 %) (Figure 1B) and FI12 (99.0 %)

FI9 had 97.739 % of the sequences belonging to the Clostridiales. Other bacterial groups known to be part of modern human gut microbiomes and that were found in mummy FI9 included the Bacillales (0.352 %), Neisseriales (0.020 %), Pseudomonadales (0.125 %), Sphingomonadales (0.186 %) and Turicibacterales (0.494 %). Mummy FI12 had the majority of the sequences belonging to the Clostridiales (99.622 %). Other bacterial groups present in mummy FI12 included the Bacillales (0.084 %) and Enterobacteriales (0.009 %).

When reads were retrieved from CLC Genomics Workbench as a SAM file and processed using mapDamage for further ancient DNA authentication as described previously (Ginolhac et al., 2011), results did not show the peculiar damage patterns usually observed when working with other types of ancient DNA (Supplemental Figure 2). Typical sequence patterns resulting from aDNA damage include short sequence length, an excess of cytosine to thymine (C-to-T) misincorporations at 5-ends of sequences, and complementary guanine to adenine (Gto-A) misincorporations at the 3-termini (Ginolhac et al., 2011).

Microbiome community composition A total of 16,805,260 (mummy FI3); 146,081,692 (mummy FI9) and 16,537,474 (mummy FI12) reads, with an average length of 100 bp, were generated from shotgun metagenomic sequencing. A total of 16,301,102 (mummyFI3), 130,621,282 (mummy FI9) and 13,503,041 (mummy FI12) were assembled, generating a total of 71,660 (mummy FI3), 53,340 (mummy FI9) and 13,552 (mummy FI12) contigs, with an average length of 2,103 bp (mummy 7


Metagenome analyses

FI3), 733 bp (mummy FI9) and 1,578 bp (mummy FI12). Metagenome binning results showed that an average of 0.14 % of the sequences had no match when using the “existing taxonomy” training set or user-selected taxonomy, whereas an average of approximately 99.86 % of the signatures had a match to the “existing taxonomy” training set or user-selected taxa. Binning results also showed the presence of sequences sharing similarity to other bacterial groups in all three mummies (e.g. Acidobacteriales, Alteromonadales, Bifidobacteriales, Chlamydiales, Legionellales, Leptospiralles, Spirochaetales and Verrucomicrobiales), fungi

varying abundances, depending on the mummy (Figure 4). For instance, mummy FI3 had the Chlamydiales (9.686 %), Clostridiales (13.004 %) and Fusobacteriales (3.596 %) in higher relative abundances. Acidobacteriales (1.969 %), Actinomycetales (6.877 %), Bifidobacteriales (1.747 %), Burkholderiales (3.532 %), Caulobacteriales (3.440 %), Pseudomonadales (3.881 %) and Sphingomonadales (7.010 %) were more represented in mummy FI9. Mummy FI12 had a higher representation of Bacteroidales (7.991 %), Flavobacteriales (2.826 %), Methanobacteriales (6.560 %) and Sphingobacteriales (13.445 %) (Figure 4).

We also compared the taxonomic profiles of the mummified human gut remains to those of extant healthy Japanese and Amazonian subjects. The relative abundances of some microbial groups differed between the Japanese and Inca mummies, as shown with the ANOVA results, while others were virtually similar (Figure 4). For instance, the Alteromonadales (p=0.0189), Chlamydiales (p=0.0012), Lactobacillales (p=3.3019 e-7) Neisseriales (p=0.0072) and Vibrionales (p=0.0002) were among the bacterial groups that were significantly more represented in the mummified guts. Sequences similar to Saccharomycetales (p=0.0010) were 8


(Saccharomycetales), and Kinetoplastida, including both Trypanosoma and Leishmania, with

also more represented in the mummified guts. Kinetoplastida-associated sequences, including both Leishmania spp. and Trypanosoma spp. were also more represented in the mummies (Figure 4). No statistically significant differences were noted between the Clostridiales and Bacillales or other microbial groups in the Inca mummies and Japanese metagenomes. Differences were also noted between the Inca mummies and the modern Amazonians. For instance Enterobacteriales (p=0.0003) and Pseudomonadales (p=0.0224) were significantly more represented in the modern Amazonians, while the Lactobacillales (p=0.0026) are

Given that the Clostridiales were among the most abundant bacteria in the mummified gut remains, as shown with both the 16S rRNA gene and shotgun metagenomics data, we mapped the reads to the genomes of 40 Clostridium spp. to identify those species that could potentially be more represented in the mummified gut remains. Mummies FI3, FI9 and FI12 had 4.6 % (780,018 reads); 4.6 % (6,793,116 reads); and 0.9 % (142,649 reads) of the reads mapping to all the Clostridium spp. genomes included in the analyses, respectively. The Japanese gut metagenomes had 3.5 % (2,827reads) (MG-RAST ID 4524574.3); 4.8 % (4,255 reads) (MGRAST ID 4525093.3); 3.1 % (2,400 reads) (MG-RAST ID 4525311.3); 4.4 % (3,571 reads) (MG-RAST ID 4525312.3); and 0.5 % (429 reads) (MG-RAST ID 4525314.3) of the reads mapping to the included Clostridium spp. genomes, respectively. The modern Amazonian gut metagenomes had 2.0 % (2,221 reads) (MG-RAST ID 4461123.3); 4.1 % (5,978 reads) (MGRAST ID 4461138.3) and 3.1 % (8,147 reads) (MG-RAST ID 4461130.3) reads mapping to the Clostridium spp. genomes included, respectively. Clostridium botulinum (NC_017299) (Supplemental Figure 3A), Clostridium difficile (CP013196) (Supplemental Figure 3B) and 9


significantly more represented in the Inca mummies (Figure 4).

Clostridium perfringens (CP010994) (Supplemental Figure 3C) had the highest number of mapped reads in the mummified gut remains. Mummies FI3, FI9 and FI12 had 0.4 % (65,367 reads); 0.09 % (124,344 reads) and 0.05 % (8,491 reads) of the reads mapping to C. botulinum. Mummies FI3, FI9 and FI12 had 0.3 % (45,043 reads); 0.06 % (82,564 reads); and 0.05 % (8,197 reads) of the reads mapping to C. difficile. Mummies FI3, FI9 and FI12 had 0.2 % (27,102 reads); 0.04 % (52,830 reads) and 0.04 % (6,986 reads) of the reads mapping to C. perfringens. There was not any clear pattern in the most predominant Clostridium spp. in the stool

% (214 reads) and 0.5 % (437 reads) of the reads mapping to C. difficile, respectively. Subject 4525311.3 had 0.4 % (288 reads) of the reads mapping to Clostridium saccharolyticum. Subject 4525314.3 had 0.2 % (165 reads) of the reads mapping to Clostridium perfringens. Subject 4525312.3 had 0.3 % (284 reads) of the reads mapping to Clostridium symbosium. From the data it appears as if specific Clostridium species are preserved in the mummified gut remains, while the modern gut might be inhabited by a larger number of different Clostridium species.

We also identified genes that could be potentially important, particularly for Clostridium virulence. While no reads mapped to important C. botulinum virulence factors, we found sequences mapping to phage structural genes including a Clostridium phage tail gene (Figure 5A). The nucleotide sequence (genome region 2,702,140 to 2,702,340) was retrieved and BLASTx to further confirm the mapping results. The best hit was indeed a Clostridium phage tail protein, and exhibited an identity of 82 % (e-value=4 e-32). Reads also mapped to C. difficile VirE, or virulence-associated protein E. The nucleotide sequence (genome region 3,941,999 to 3,942,733) was retrieved and BLASTx, confirming that the region belonged to VirE and 10


metagenomes of Japanese individuals. For instance, subjects 4524574.3 and 4525093.3 had 0.3

exhibited an identity of 100 % (e-value =1e-172) (Figure 5B). Reads mapping to a C. perfringens hemolysin were retrieved (region 2,664,336 to 2,664,447) (Figure 5C) and BLASTx, confirming that the best hit was a hemolysin, and exhibited an identity of 86 % (e-value 1e-11).

Functional profiles associated with disease and metabolism Functional categories were determined using KO annotations from the metagenomic data

profiles in all three mummies included, but were not limited to lipid, carbohydrate and amino acid metabolism. Metabolic profiles of the three mummies were compared to those of extant (Japanese and Amazonian) gut microbiomes, and their relative abundances showed some resemblance and similar relative abundances (Figure 6A). There were, however, some differences as demonstrated with the ANOVA results. For instance, there were statistical significant differences between the mummified and extant Japanese gut metagenomes relative abundances associated with amino acid metabolism (p=0.0013), glycan biosynthesis and metabolism (p=0.0103), metabolism of cofactors and vitamins (p=0.0048), metabolism of other amino acids (0.0292), metabolism of terpenoids and polyketides (p=0.0376), as well as nucleotide metabolism (p=0.0346). Interestingly, no significant differences were noted in the metabolic profiles between the mummified gut remains and modern Amazonians (Figure 6A).

We also visualized the relative abundances of specific KOs associated with metabolism in each mummy and compared them to modern gut Japanese and Amazonian metagenomes to identify potential differences that could be attributed to the process of natural mummification. 11


(Figure 6). Metabolic profiles were also predicted using metagenomic data. Selected metabolic

Heatmap shows that the mummified gut remains clustered separately from the modern Japanese (red) and Amazonian (purple) guts, and that mummy FI9 clustered separately from mummies FI3 and FI12 (Figure 6B). ANOVA results show that there were statistically significant differences associated with K000280 (p=0.0013), K00290 (p=0.0158), K00400 (p=0.0490), K00540 (p=0.0179), K00550 (p=0.0482), K00740 (p=0.0489), K00780 (p=0.0187) and K00920 (p=0.0465) between the mummified gut remains and Japanese guts. K000280 is associated with valine, leucine and isoleucine degradation, and was more represented in the mummified gut

more represented in the Japanese guts. K00400 (phenylalanine, tyrosine and tryptophan biosynthesis), K00540 (lipopolysaccharide biosynthesis), K00550 (peptidoglycan biosynthesis), K00740 (riboflavin metabolism), K00790 (biotin metabolism) and K00920 (sulfur metabolism) were more represented in the Japanese guts. K00280 (p=0.0421) and K00760 (nicotinate and nicotinamide metabolism) (p=0.0476) were significantly more represented in the mummified gut remains when compared to modern Amazonians. K00190 (oxidative phosphorylation) (p=0.0126) and K00670 (one carbon pool by folate)(p=0.0284) were significantly more represented in the modern Amazonians (Figure 6B). Differences are highlighted in bold.

Pairwise comparisons between taxonomy and metabolic profiles were performed to identify potential signatures that could be associated with the process of natural mummification of the human gut. Overall, the correlation profiles of the mummified human gut remains (Figure 7A), and modern Amazonian (Figure 7B) and Japanese guts (Figure 7C) showed a noticeable number of differences. Negative, positive and no correlations are shown in black, red and white,

12


remains, while K000290 is associated with valine, leucine and isoleucine biosynthesis, and was

respectively. Significant differences are shown with asterisks (*). p-values ≤0.05, ≤0.01, ≤0.001 and ≤0.0001 are shown with *, **, *** and ****, respectively.

Evolution of pathogens associated with Chagas’ disease and leishmaniasis Given the presence of sequences associated with Kinetoplastida when performing the binning analysis in all three mummies, as well as the paleopathology of mummy FI9 suggesting

available strains of T. cruzi and four available strains of Leishmania donovani to identify sequences sharing similarity to extant sequences. Mummies FI3 and FI9 had sequences sharing similarity (