Previously uncharacterized rectangular bacterial structures in the dolphin mouth

To maximize reproducibility, a list of the reagents and resources used in this study, as well as their source and identifier, is provided in Supplementary Table 5.

Experimental model and subject details

Oral swab samples were obtained from bottlenose dolphins (Tursiops truncatus) managed by the U.S. Navy MMP Biosciences Division, Space and Naval Warfare Systems Center Pacific, San Diego, USA. The earliest sample containing RBSs was collected on April 1, 2012, and the latest on March 24, 2022. Swab samples were obtained using sterile foam Catch-All sample collection swabs (Epicenter, WI, Cat. #QEC091H). Samples collected in 2012 were obtained by swabbing the left gingival sulcus. Samples collected in 2018 were obtained by swabbing the palate, the tongue, and the left gingival sulcus (all three surfaces for each swab). Samples collected in 2022 were obtained from the palate, buccal surface, or left gingival sulcus. Of the 2022 samples of the gingival sulcus, 5 were stored in 20% glycerol. All other swab samples were dry frozen. The swabbing protocol adhered to the guidelines described in the CRC Handbook of Marine Mammal Medicine.

The MMP is accredited by the Association for Assessment and Accreditation of Laboratory Animal Care International and adheres to the national standards of the United States Public Health Service Policy on the Humane Care and Use of Laboratory Animals and the Animal Welfare Act. As required by the U.S. Department of Defense, the MMP’s animal care and use program is routinely reviewed by an Institutional Animal Care and Use Committee (IACUC) and by the U.S. Navy Bureau of Medicine and Surgery. The animal use and care protocol for MMP dolphins in support of this study was approved by the MMP’s IACUC and the Navy’s Bureau of Medicine and Surgery (IACUC #92-2010, BUMED NRD-681).

Microscopy sample preparation

To separate cells from swabs, swabs were immersed in 1X PBS (~50–100 µL, depending on cell density) in microcentrifuge tubes. Tubes were vortexed vigorously for ~10 s and lightly centrifuged to remove liquid from tube caps. The resulting solution was used for microscopy.

Light microscopy

Approximately 1 µL of cell solution in PBS was spotted onto an agarose pad (1% agarose in PBS) and imaged with an Eclipse Ti-E inverted microscope with a 100X (NA: 1.4) objective (Nikon, Tokyo, Japan). To determine DNA localization, cells were stained with DAPI at a final concentration of 0.5 μg mL−1 for 5 min prior to imaging using emission/excitation spectra of 340/488 nm. High-throughput, automated imaging of dolphin oral samples via phase contrast microscopy was achieved using the Strain Library Imaging Protocol38 to capture 226 fields of view for each of the samples collected in 2018, and 100 fields of view for those collected in 2022.

Gram and FM4-64 staining

Gram staining was performed using a Gram Staining Kit (Sigma Aldrich, cat. #77730-1KT-F) following the manufacturer’s protocol. Cells were imaged using a bright field microscope with a 100X objective (Nikon). FM4-64 dye (ThermoFisher Scientific, cat. #T13320) was applied directly to dolphin oral swab samples following the manufacturer’s protocol. The FM4-64 dye did not stain any part of the RBSs.

RBS-A cryofixation and cryoEM/ET data acquisition

A solution of cells in PBS (2.5 µL) was applied to glow-discharged 200-mesh copper, holey-carbon Quantifoil grids (Quantifoil, Großlöbichau, Germany, Cat. #Q2100CR1) or gold GridFinder Quantifoil grids (Quantifoil, Großlöbichau, Germany, Cat. #LFH2100AR2), followed by application of 2 µL of 15 nm gold fiducial solution to both sides of each grid. Grids were blotted for 5 s and plunge-frozen in liquid ethane cooled by liquid nitrogen to approximately −195 °C using an EM GP Plunge Freezer (Leica, Wetzlar, Germany).

Samples were loaded into one of two microscopes: a Titan Krios G3 operated at 300 kV with an energy filter (20-eV slit width), or a Titan Krios G4 operated at 300 kV without an energy filter. Both microscopes were equipped with a K2 Summit direct electron detection device (Gatan, Pleasanton, USA) used to record micrographs. Data were acquired semi-automatically in counting mode using SerialEM (v. 3.8)61. CryoEM/ET imaging parameters are provided in Supplementary Table 6.

CryoEM/ET data processing

Montages were blended and binned 4-fold or greater using the IMOD v. 4.12.9 “blendmont” algorithm62 and normalized, band-pass filtered, rotated, and cropped for display purposes using EMAN2 v. 2.3963. Fifteen out of sixteen tilt series were suitable for tomographic reconstruction in IMOD v. 4.12.9. Tilt series with sampling at 7.5 Å pixel−1 were down-sampled by 2-fold and those with sampling at 3.48 or 3.75 Å pixel−1 were down-sampled by 4-fold. Images with artifacts such as excessive charging, drifting, large ice contamination creeping in at high tilts, or excessive thickness at high tilt were excluded from 12 of the tilt series prior to manual gold-fiducial-based alignment; up to 13 images were removed out of the 41 images in the original raw tilt series.

Tomograms were reconstructed using standard weighted back-projection and a SIRT-like filter (mimicking 16 iterations) and were band-pass-filtered and further binned by 2-fold in most cases for feature annotation, segmentation, movie production, and other display purposes. Tomogram thickness was estimated by visually identifying the smallest and largest z-slices with visible RBS-A or ice contamination densities and converting the number of slices to nanometers. Subtomogram averaging was attempted using EMAN2 v. 2.3964,65 for globular densities suspected to be ribosomes, matrix densities under the outer membrane, patches of the periodic surface covering, and regions of pilus-like appendages, but no interpretable structures with resolution better than ~50 Å were obtained. The ranges of thickness and length for the pilus-like appendages were derived by visually scanning the slices in the tomograms for the thinnest individual filaments and thickest bundles perceptible to the naked eye and measuring their dimensions in binned-by-4 tomographic slices using the measuring tape tool of EMAN2 e2display.py. The repeat distance of the periodic surface covering was measured manually in a similar fashion as the pilus-like appendages from tomographic slices, with ~10–20 measurements from each of the three tomograms displaying at least small regions where the repeat was discernible. This quantification yielded a range between ~6 and 10 nm, suggesting that either the layer components are flexible or that the underlying structure can yield different apparent distances between its subunits depending on the angle at which it is sliced. In addition, regions showing the pattern much more clearly in higher-magnification montage two-dimensional projection images were cropped out, rotated to lie in a horizontal plane, filtered, and masked to compute line-density profiles parallel to the outer membrane.

We initially carried out tomographic annotation of three features (periodic surface covering, lipid membranes, and pilus-like appendages) for three tomograms using EMAN2’s semi-automated two-dimensional neural network-based pipeline66 and performed manual clean-up of false positives in UCSF Chimera v. 1.1667. The output annotation probability maps from EMAN2 v. 2.39 were turned into segmentations by applying a visually determined threshold and multiplying the contrast-reversed tomograms by the thresholded annotation map. The segmentations were low-pass-filtered with EMAN2 v. 2.39 to smooth out noise. However, since the complexity of subcellular structures was not captured by the semi-automated annotations, we applied a similar process to generate segmentations of five features (pilus-like appendages, periodic surface covering, outer membrane, matrix, and inner membranes) using manual annotations performed with IMOD v. 4.12.9, following a recent protocol that increases manual annotation efficiency68. Snapshots for Fig. 6 displaying RBS-A features in color as well as Supplementary Movies 2 and 3 showing segmentation results were produced with UCSF Chimera v. 1.16.

CryoFIB-SEM

We first identified a grid that contained RBS-As using light microscopy. Using an Aquilous CryoFIB (ThermoFisher Scientific, MA, USA), we created five lamellae using 30 kV, 30 pA current and 5 µs duration time. The samples were then loaded into a cryoTEM for sample observation and data collection. We were unable to identify RBSs in the lamellae.

Fluorescence in situ hybridization

Cell cultures of axenic Escherichia coli MG1655, non-axenic Skeletonema costatum LB 2308 (UTEX Culture Collection of Algae at the University of Texas at Austin, Austin, TX, USA), Caulobacter crescentus CB15N, and Simonsiella muelleri ATCC 29453 were prepared as controls. E. coli was cultured in LB broth and grown at 37 °C, S. costatum was cultured in Erdschreiber’s Medium at 20 °C with a ~12 h light and ~12 h dark cycle, C. crescentus was cultured in PYE medium at 30 °C, and S. muelleri was cultured in BSTSY medium (2.75% (w/v) Tryptic Soy Broth, 0.4% (w/v) yeast extract, 10% bovine serum) at 37 °C.

All FISH probes were ordered from Integrated DNA Technologies (Coralville, USA) with HPLC purification. Probe sequences and fluorescence labels are as follows: Euk-1209: 5’-GGGCATCACAGACCTG-/3Alx660/−3’, Bact338: 5’-GCTGCCTCCCGTAGGAGT-/Alx488/−3’, BET42a: 5’-/Alx594/-GCCTTCCCACTTCGTTT-3’, GAM42a: 5’-/Alx488/-GCCTTCCCACATCGTTT-3’, nonEUB: 5’-/Cy5/-ACTCCTACGGGAGGCAGC-3’.

Cells from controls and RBS-As were collected in microcentrifuge tubes. To ensure sufficient biomass from dolphin oral swabs, cells from four swabs were condensed into a single tube. The FISH protocol was adapted from ref. 69. Cells were fixed in 1 mL of 3.7% formaldehyde solution (800 μL of DEPC-treated water, 100 μL of 10X PBS, 100 μL of 37% formaldehyde) for 30 min with gentle shaking at 700 rpm. Cells were then washed twice in 1 mL of 1X PBS, and permeabilized in a mixture of 300 μL of DEPC-treated water and 700 μL of 200-proof ethanol with gentle shaking at 700 rpm for 2 h. Probes were added to 50 μL of hybridization solution to a final concentration of 1 μM per probe set. For BET42a and GAM42a, the hybridization solution contained 55% formamide solution (4 mL of DEPC-water, 1 g of dextran sulfate, 4.85 mL of formamide, 1 mL of 2X SSC, brought to a total volume of 10 mL with DEPC-treated water); for other probes, the hybridization solution contained 40% formamide (5 mL of DEPC-water, 1 g of dextran sulfate, 3.53 mL of formamide, 1 mL of 2X SSC, brought to a total volume of 10 mL with DEPC-treated water). For BET42a and GAM42a, cells were incubated in 50 µL of hybridization buffer with probes at 46 °C for 1 h; for other probes, cells were incubated overnight in 50 µL of hybridization buffer with FISH probes at 30 °C. Cells were washed twice using a wash solution (2 mL of 20X SSC buffer, 7.06 mL of formamide, 10.94 mL of DEPC-treated water) and resuspended in 2X SSC buffer. One microliter of cells was mounted onto 1% agarose pads containing PBS and 5 µg mL−1 of DAPI for imaging. Imaging data were processed using FIJI v. 2.0.070.

16S rRNA amplicon sequencing and processing

Fifty-four dolphin oral samples were selected for 16S rRNA gene amplicon sequencing. Genomic DNA was extracted from dolphin oral samples and 32 negative (PBS) controls using the DNeasy UltraClean 96 Microbial Kit (Qiagen Cat. #10196-4) following manufacturer’s instructions. The 16S rRNA V4 region was amplified using 515F and 806rB primers using Platinum™ II HotStart PCR Master Mix (ThermoFisher Cat. #14000013). The PCR products were pooled at equal volume and gel-purified. Final purification was performed using Macherey-Nagel NucleoSpin Gel and PCR Clean-up, Mini Kit (Fisher, Cat. #740609). Amplicons were sequenced on the Illumina MiSeq platform with 250-bp paired reads at the Stanford Chan Zuckerberg Biohub Facility, resulting in a median read depth of 92,077 reads (min: 50,772, max: 265,108).

Demultiplexing was performed using Bcl2Fastq v. 2 (Illumina, CA, USA). ASVs were inferred using DADA271 v. 1.16.0, following guidelines in the “Big Data Workflow” (benjjneb.github.io/dada2/bigdata_paired.html). Taxonomic affiliations were assigned using the SILVA 138 SSU database72 as a reference. Forward and reverse reads were trimmed to 240 and 180 nt, respectively. This pipeline yielded a total of 1116 taxa and 5,339,751 reads across the 54 samples. ASVs were analyzed using phyloseq v. 1.28.072.

Spike-in experiment

To confirm that our DNA extraction, amplification, sequencing, and bioinformatic analysis pipeline was able to identify members of the family Neisseriaceae, we performed a spike-in experiment. We first selected a dolphin oral sample that contained RBS-As, as determined by phase-contrast microscopy. We then created two aliquots of this sample. To the first aliquot, we spiked in cells from an S. muelleri pure culture at a 1:1 ratio. The other aliquot was left untouched. These two samples, along with an aliquot of the S. muelleri pure culture (positive control), underwent DNA extraction, PCR amplification, and sequencing protocol as described above. The three samples were sequenced in a single Illumina MiSeq run, which was not in the same lane as the other amplicon samples from this study. Note that by PCR amplifying S. muelleri in the same lab environment as the negative control sample and then sequencing it on the same lane, some cross-contamination is to be expected. The first time the negative control sample was sequenced (in the absence of S. muelleri pure cultures in the laboratory), no family Neisseriaceae amplicons were detected.

Mini-metagenomics

To obtain candidate identities for RBS-As, we employed a mini-metagenomics approach. To limit contamination by foreign DNA, reagents, tubes, and PBS were treated with 11.4 J cm−2 of ultraviolet light following the guidelines in ref. 72. RBS-As were visualized using an Olympus IX70 inverted microscope (Olympus, Waltham, USA) with a 40X objective and Hoffman modulation optics. An Eppendorf TransferMan micromanipulator (Eppendorf, Hamburg, Germany, Cat. #5193000020) with a SAS-10 microinjector was used to capture RBS-As with Polar Body Biopsy Micropipettes (30° angled, beveled, and polished with an inner diameter of 13–15 μm) (Cooper Surgical, Målov, Denmark, Cat. #MPB-BP-30). After an RBS-A or chain of RBS-As was acquired, the micropipette tip was transferred to a collection tube containing 1X PBS and crushed into the tube to ensure the RBS-A(s) was deposited in the tube; this precaution was adopted because RBS-As frequently stuck to the glass micropipette and could not be dislodged. No dolphin cells were captured, although cell-free DNA and small, non-target cells from the sample were likely acquired as contaminants along with RBS-As based on the propensity of the latter to attach to other species (Fig. 5e). Four tubes of RBS-As were collected (sample names RBS1-4), along with four negative-control tubes (NEG1-4). Negative controls consisted of draws of PBS from the same sample that did not contain any visible cells and were otherwise treated identically to RBS-A-containing samples.

DNA from each tube was amplified via MDA using the Repli-g single-cell kit (Qiagen, Hilden, Germany, Cat. #150343) according to the manufacturer’s protocol. DNA was purified using a Zymo Clean and Concentrate Spin Column (Zymo Research Corporation, Irvine, USA, Cat. #D4013) and libraries were prepared using the Kapa Hyper Prep Kit (Kapa Biosystems, Wilmington, USA, Cat. #KK8504) at the W.M. Keck Center for Comparative Functional Genomics at the University of Illinois, Urbana-Champaign. The eight libraries were sequenced using the Illumina MiSeq 2 × 250 nt P2 V2 platform. RBS-A samples RBS1, RBS2, NEG1, and NEG2 were pooled and sequenced across a single lane that produced 11,371,243 read pairs, and samples RBS3, RBS4, NEG3, and NEG4 were pooled and sequenced across 1.5 lanes, collectively producing 19,615,690 read pairs. Sequencing adapters were computationally removed at the Keck Center.

Reads from all eight samples were co-assembled using SPAdes v. 3.11.173 with the single-cell (–sc) and careful (–careful) modes specified. A total of 61,973,866 read pairs were used for assembly, resulting in 1406 scaffolds ≥5 kbp long with a total length of 17,438,233 bp and an N50 of 14,592 for scaffolds ≥5 kbp long. Protein-coding genes were identified using Prodigal v. 2.6.274. Per scaffold average coverage was calculated by mapping reads per sample against the co-assembly using bowtie2 v. 2.2.475, using the samtools v. 1.6 depth function76 to calculate per-base read coverage, and a custom script to calculate average per-base read coverage per scaffold.

A search for the amiC2 gene in bins recovered from the mini-metagenomics experiment was performed by querying Pfam alignment PF01520 against each genome’s protein sequences, using HMMER suite v. 3.1b277.

To determine the taxonomic identity of sequenced cells, we employed a genome-resolved approach. Assignment of scaffolds to genome bins was performed using the tetranucleotide frequencies of all scaffolds ≥5 kbp long over windows of 5 kbp, as described in ref. 78. Results were computed and visualized using the Databionics ESOM Tools software v. 1.179, leading to the reconstruction of 18 genome bins (Supplementary Fig. 3). To refine bins, we removed scaffolds for which <50% of keys were assigned to the bin. Scaffolds <5 kbp long were not binned. The completeness and contamination per bin were assessed using CheckM v. 1.0.780. To evaluate how representative binning was of the genomes that were sequenced, we estimated the number of prokaryotic genomes expected to be recovered by searching the metagenome assembly for a set of 16 bacterial single-copy genes (bSCGs) assumed to be present in every genome in a single copy81, namely ribosomal proteins L2, L3, L4, L5, L6, L14, L15, L16, L18, L22, L24, S3, S8, S10, S17, and S19. Alignments for these proteins (PF00181, PF00297, PF00573, PF00281, PF00347, PF00238, PF00828, PF00252, PF00861, PF00237, PF17136, PF00189, PF00410, PF00338, PF00366, and PF00203) were obtained from the Pfam database82 (accessed March 2019) and queried against our dataset using HMMER suite v. 3.1b277. The median number of each bSCG was 10, suggesting ~10 prokaryotic genomes were represented in our sequencing dataset. In the case of the family Alcaligenaceae genome of interest, the 16S rRNA gene was manually extended from the end of a scaffold; using bowtie2 v. 2.2.4 we ensured that the reads supported the final sequence.

Taxonomic identification of bins posed a challenge since 16S/18S rRNA genes were not reliably amplified/sequenced/assembled, and genomes were partial with few phylogenetically informative bSCGs present in the dataset. Hence, we used BLAST v. 2.2.3083 to query all protein-coding genes from each genome against the NCBI non-redundant protein database using an e-value of 10−10 and taxonomic assignments were made based on the closest protein match. Genome bin taxonomic assignments were considered highly likely if ≥50% of the top BLAST hits originated from a single taxon and were considered plausible if <50% but ≥33% of the top BLAST hits originated from a single taxon.

There are numerous approaches by which one could assess whether a bin is “present” in a sample, each with largely arbitrary thresholds. We focused on the relative abundance of each bin per sample (Supplementary Table 4) as it accounts for the length of each bin and allows for comparisons between samples with different numbers of read pairs16.

Confirmation of the taxonomic identity of bin 16

Maximum likelihood phylogenies of the family Alcaligenaceae were inferred using the 16S rRNA gene (Supplementary Fig. 4a) and ribosomal protein S3 (Supplementary Fig. 4b) to confirm the taxonomic affiliation of bin 16, which was recovered from the mini-metagenomics experiment. Gene/protein sequences were acquired for each characterized genus in the family Alcaligenaceae, as shown on the NCBI Taxonomy Browser (accessed August 2022), when such sequences were available in the NCBI system (some genera have scant or no genomic representation). We additionally performed BLAST83 v. 2.2.30 queries of the bin 16 16S rRNA gene and rpS3 protein against the nr/nt and nr databases (accessed August 2022), respectively, and included the top 10 most similar sequences. 16S rRNA gene sequences were aligned using SINA84 v. 1.2.11, using the SILVA SSU database release 138.1 as a reference, and columns containing >3% gaps or rows with <50% sequence were removed. rpS3 protein sequences were aligned using Clustal Omega85,86 v. 1.2.4 and columns containing >5% gaps or rows with <50% sequence were removed. Both phylogenies were inferred using PhyML87 v. 3.1 with 1000 bootstrap replicates, with model selection performed using smart model selection88 (GTR+R for the 16S rRNA gene and Q.yeast+G+I for the rpS3 protein). Trees were visualized using iTOL89 v. 6.

Attempt at culturing RBSs

Four dolphin oral samples confirmed to contain RBSs were selected for culturing efforts. For each sample, one milliliter of sterile PBS was added to a 1.5-mL Eppendorf tube containing the oral swab sample. For liquid culturing, 600 µL of each sample were used to inoculate 3 mL of BSTSY90 (2.75% (w/v) Tryptic Soy Broth, 0.4% (w/v) yeast extract, 10% bovine serum), SHI91, or mSHI media (SHI supplemented with 0.9 g/L NaCl, 2.5 g/L K2PO4, 0.84 g/L NaHCO3, 0.17 g/L CaCl2, 0.04 g/L MgCl2•6H2O, and 5 g/L dextrose). BSTSY was selected for its use in successfully culturing bacteria (S. muelleri specifically) from the oral cavities of various mammals90. SHI was selected because this medium was designed to sustain high-diversity communities derived from the human oral microflora. mSHI (modified SHI) was included as a higher-salinity version of SHI in an attempt to further mimic the conditions that might be found in the oral cavity of dolphins. Inoculation was repeated under anaerobic conditions in an anaerobic chamber (COY Lab Products, Grass Lake, USA); note that all samples were unavoidably exposed to atmospheric oxygen prior to culturing. Cultures were incubated at 37 °C to mimic the body temperature of dolphins. No RBSs were detected in liquid media by visual screening under a microscope after ~24, ~48, ~72, and ~96 h. For solid-surface culturing, ~103−104 cells (verified by microscopy to contain RBS-As) were directly plated onto BSTSY or BHI-blood (BHI medium supplemented with 5% sheep blood) agar plates, and incubated at 37 °C with or without oxygen, respectively. No colonies grew on the BSTSY plates after 3 weeks of incubation; the colonies grown on the BHI-blood plates were screened using microscopy and no RBSs were visible. By contrast, a control of S. muelleri streaked onto BSTSY plates developed visible colonies after 1–2 days of incubation with oxygen at 37 °C, and the colonies were verified under microscopy to consist of cells with the morphology expected of S. muelleri.

Statistics and reproducibility

Given the exploratory nature of this descriptive study, most experiments to characterize the RBSs were performed a single time.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Read more here: Source link