A survey of the sperm whale (Physeter catodon) commensal microbiome

Sample collection and DNA extraction

The present study sampled a living female sperm whale stranded near the bay area of Huizhou City in southern China on 12th March 2017, and we collected a blood sample of about 5 ml from the vein via syringe (Fig. S1). The whale died on 15th March, approximately 79 h after rescue. The dead animal was salvaged and post-mortem analysis was performed according to standard protocols (Kuiken, 1991). We collected fecal and muscle samples during the necropsy. We dissected the intestine with sterile scissors and forceps, then we used a swab to collect about 2 g of fecal matter from inside the incision to a sterile plastic tube. To collect a muscle sample, we cut the sperm whale cortex tissue and about 5g of muscle tissue was collected with scissors. Each sample was immediately stored in −20 °C freezers before transport to the laboratory within 12 h for further investigation.

The blood sample was combined with RNAlater to dissolve eukaryotic cells, and then centrifuged to obtain bacteria cells to extract DNA. Fecal and muscle sample DNA extractions were performed using the QIAamp DNA Stool Mini Kit (Qiagen, Valencia, CA, USA). To eliminate RNA contamination, DNase-free RNase was used to treat extracts. Finally, DNA quantity was determined using a Qubit 3.0 fluorometer, and DNA integrity was evaluated by gel electrophoresis (Pan et al., 2018). All lab work in this study was conducted in a sterile flow hood to prevent contamination. Blood and other samples were dealt in different laboratories to prevent same microbiota contamination. This study was approved by the Institutional Review Board of BGI (NO. BGI-R052-3 and NO.FT 17160) and the Institute of Deep-sea Science and Engineering Chinese Academy of Sciences (NO.SIDSSE-SYLL-MMMBL-01).

Library construction and sequencing

We totally constructed three metagenome pair-end libraries for the three samples and one 16S rDNA amplicon library for the fecal sample. To construct metagenome sequencing libraries, the extracted DNA was sheared into fragments between 500 bp and ∼800 bp in size. Fragments between 150 bp and 250 bp were selected using AMPure XP beads (Agencourt, Beverly, MA, USA) and then repaired using T4 DNA polymerase (ENZYMATICS, Beverly, MA, USA) to obtain blunt ends which were then 3-adenlyated to create sticky ends. These DNA fragments were ligated at both ends to T-tailed adapters and amplified for eight cycles. These amplification products were subjected to a single-strand circularization process using T4 DNA Ligase to generate a single-stranded circular DNA library. For constructing 16S V4 rDNA sequencing library, binding sites for sequencing primers of BGISEQ-500 were added at the 5′ end of V4 region common primers (515F: GTGCCAGCMGCCGCGGTAA, 806R: GGACTACHVGGGTWTCTAAT). The PCR products with target bands were mixed in equal amounts, and 2% agarose gel was used for electrophoresis and gel cutting, and then were subjected to generate a single-stranded circular DNA library.

All libraries were sequenced on the BGISEQ-500 platform in the paired-end model with 100 bp length reads for metagenome in BGI-Shenzhen and 150 bp for 16S rDNA library in BGI-Qingdao. Different sequencing machines were used for different libraries to prevent same microbiota contamination.

Metagenome assembly and taxonomic assignment

Adaptor contamination, low quality reads and duplication for metagenomes were filtered from the raw reads by SOAPnuke (1.5.6) (Chen, Chen & Shi, 2018), and the metagenome remaining reads were removed to eliminate host DNA on the basis of sperm whale reference (Fan et al., 2019) using SOAP2 (Version 2.21) (Gu, Fang & Xu, 2013) with the parameters “-v 8 -m 4”. To account for the possibility of contamination with human-associated microorganisms, clean data were mapped to the human genome, and samtools (Version 1.9) with the parameters “view -q 30 -F 4 -F 256” was used to account mapping reads. After removing the host data, the unmapped metagenome data of 16 Gb, 53 Gb and 15 Gb data for the blood, fecal and muscle samples, respectively, were subjected to further analysis. We then used metaSPAdes (Version 3.10.1) (Nurk, Bankevich & Antipov, 2013) to assemble each sample separately. MetaGeneMark (Version 3.38) (Zhu, Lomsadze & Borodovsky, 2010) was used to identify the coding sequences (CDSs) of the assembled sequences.

Taxonomic assignment of the predicted genes was carried out using BLAST+ (Version 2.2.26, best match according to BlastN, evalue <1e−5, coverage >50%) by aligning against the prokaryotic representative genomes (Version 201608) and non-redundant nucleotide (NT) database (Version 20170616) (Zhao et al., 2018). The reads were then aligned to the predicted genes using Bowtie2 (Version 2.2.5) (Langmead & Salzberg, 2012), thus the sequence-based gene abundance profiling was calculated using PathoScope (Version 2.0.6) (Francis et al., 2013). The relative abundances of species were calculated from the relative abundance of their respective genes, and the microbes with a relative abundance below 1e−8 were removed to reduce contamination (Guo et al., 2018).

The blood, muscle and fecal clean reads were aligned to the VFDB (Virulence Factors of Bacterial Pathogens) database using Bowtie2 (Version 2.2.5) and the blood assembly contigs were aligned using blat (Version 3.2.1). The Virulence Factors identified in the two results that were annotated to Clostridium novyi and Clostridium botulinum were selected. The core bacterial virulent genes were aligned by MUSCLE (Version 3.8.31) with the same bacteria genome from NCBI and the trees were built by FastTree (Price, Dehal & Arkin, 2009).

16S rDNA community sequence analysis

In order to compare the gut microbiota with that of other marine mammals, we first collected the 16S V4 rDNA tags of the fecal samples of six species from the NCBI and EMBI database (see more data details in Table S1). Then the tags were clustered to the OTUs (Operational Taxonomic Units) using USEARCH (Version 7.0.1090) (Rognes et al., 2016) with a 97% threshold. Taxonomic assignment of OTU sequences were classified using Ribosomal Database Project (RDP) Classifier (v.2.2) trained on the RDP database (version trainset14_032015) (Cole et al., 2014), using 0.8 confidence values as a cutoff. The data of sperm whale fecal sample 16S rDNA amplicon was discarded to eliminate the low quality and adapter pollution, then paired-end reads with overlap were merged to tags using FLASH (v1.2.11). Finally, all tags were mapped into the OTU representative sequences using USEARCH GLOBAL to obtain the OTUs and species abundance profiles. Bray-Curtis distance was used to estimate β-diversity (between-sample diversity).

Community diversity and structure of the samples’ microbiome

To obtain a reference metagenomic sequences of the three tissues, we performed a strict data quality control and obtained ∼16 Gb, ∼53 Gb and ∼15 Gb high-quality data of blood, fecal and muscle samples, respectively (Table S2). Subsequently, we assembled the metagenomic reads of the three tissues, with an assembly size of ∼37 Mb, ∼470 Mb and ∼151 Mb for blood sample, fecal sample and muscle sample, respectively (Table 1). A total of 1,715 microorganisms were found in the three tissues, among which the microbial species in the fecal samples were the most abundant. There were notable differences in species richness and abundance in the different samples (Fig. 1A).

In all three tissues, only 113 microbial species were detectable, and they were those belonging to seven phyla Firmicutes, Bacteroidetes, Deferribacteres, Spirochaetes, Fusobacteria, Proteobacteria and Tenericutes (Table S3). A total of 21 phyla and 74 genera were identified in at least two tissues (Fig. 1B). Each sample had a dominance of different taxa; for instance, the blood sample was dominated by Firmicutes. In comparison with blood and muscle samples, the fecal sample was more abundant in Bacteroidetes and Firmicutes. The muscle sample shared fewer genera when compared to the other two samples.

Detection of potentially pathogenic microorganisms

To detect potentially pathogenic microbes of the sperm whale, we screened the pathogenic bacterial species in the metagenome sequence data of three tissues. We found pathogenic microbes including Enterococcus faecium, Fusobacterium nucleatum, Pseudomonas aeruginosa, Streptococcus anginosus, Streptococcus pneumoniae, Streptococcus suis, and five toxigenic Clostridium in the bloodstream and intestine, and six of them appeared in muscle tissue (Table 2) (Chevalier, Bouffartigues & Bodilis, 2017; Engholm, Kilian & Goodsell, 2017; Fortier, 2017; Gao, Howden & Stinear, 2018; Han, 2015; Okwumabua et al., 2017). Then we aligned the blood, muscle and fecal assembly results and sequencing data against the VFDB databases to identify bacterial virulence genes (Chen et al., 2016). Twelve virulence genes were found in Clostridium novyi and Clostridium botulinum. By aligning the pathogenic genes to all strain sequences of C. novyi and C. botulinum, we found the nearest bacteria to them were C. novyi A str. 4540 (GCA_000724445) and C. botulinum C str. Stockholm (GCA_000219255) from Sweden (Figs. 2A and 2B).

Comparison of gut microbiome in different marine mammals

The composition of mammalian gut microbiota could be shaped by host diet, age and phylogenetic position (Hullar & Fu, 2014; Muegge, Kuczynski & Knights, 2011). To profile the unique gut microbiota of sperm whales, we compared the composition of gut microbiome between sperm whales and other marine mammals. A total of 122 bacterial Operational Taxonomic Units (OTUs) were identified in the sperm whale’s gut microbiota (Table S4). The majority of microbes were members of the phyla Bacteroidetes (51.49% of the total) and Firmicutes (31.18% of the total), as well as the phyla Euryarchaeota, Spirochaetes and Verrucomicrobia (Fig. 3A). The sperm whale’s sample had either very few or no reads assigned to Proteobacteria, which was comparatively common among six other marine mammals.

We also calculated these compositional dissimilarities using β-diversity (Whittaker, 1960), and built cluster trees using β-diversity to disentangle the effect of different factors shaping community assembly at different phylogenetic scales. The Bray–Curtis distance in mammalian gut microbiomes showed that the sperm whale was closest to the Beluga Whale (Delphinapterus leucas) and furthest from the manatee (Florida manatee), which is consistent with the biological evolutionary tree of these marine species (Fig. 3B) (Groussin et al., 2017).