Metagenomic Analysis of Healthy and White Plague-Affected Mussismilia braziliensis Corals

Sample Collection Two archipelago reef sites were chosen for coral sampling: (1) Sebastião Gomes (17°54′42.49′′/39°7′45.94′′), near shore (∼20 km), and (2) Parcel dos Abrolhos (17°57′32.7′′/ 38°30′20.3′′), located on an off shore area (∼70 km), inside the Abrolhos Marine National Park (Fig. 1). Sampling was performed on February 22, 2010 by scuba diving, on both sites, in a 30-m2 area and 3–8-m depth. Six fragments of the stony coral M. braziliensis were collected with a hammer and a chisel: three healthy and three presenting signs of white plague disease, on each site, amounting to 12 coral specimens. Each fragment was immediately stored on poly- propylene tubes, identified, and frozen in liquid nitrogen. After transportation to the lab, all samples were preserved, at −80 °C, until processing.

Coral Holobiont DNA Extraction Several variations of the DNA extraction protocol were tested (schematized in Fig. 2), since there is no consensus protocol in the literature. After a quick removal of calcium carbonate (CaCO3) skeletons with sterile spatulas, about 200 mg of each coral sample was crushed with sterile mortar and pestle, in the presence of liquid nitrogen. The resulting slurry received one of four lysis buffers: (i) 1 ml Lysis G buffer [4 M guanidine hydrochloride (Sigma-Aldrich), 200 mM Tris–HCl (pH8.0), 50 mM EDTA, and 0.5 % (m/v) sodium N-lauryl sarcosine (Sigma-Aldrich)]; (ii) 1 ml cetyl trimethylammonium bromide (CTAB) buffer [2 % (m/v) CTAB (Sigma-Aldrich), 1.4 M NaCl, 20 mM EDTA, 100 mM Tris–HCl (pH 8,0), and freshly added 5 μg proteinase K and 0.5 (or 1)% (v/v; Invitrogen) 2-mercaptoethanol (Sig- ma-Aldrich)]; (iii) Gianni’s buffer, 250 μl ice-cold stabilizing solution [20 % (m/v) sucrose, 50 mM Tris–HCl (pH8,0), 50 mM EDTA] plus 500 μl ice-cold lysis solution [50 mM NaCl, 1 % (m/v) sodium N-lauryl sarcosine (Sigma-Aldrich), and freshly added 25 μg proteinase K]; and (iiii) 1 ml Trizol solution (Invitrogen). Alternative solutions containing in- creased concentrations of EDTA (50, 75, or 100 mM) were also tested. We compared the usefulness of each of these four buffers. Two lysis conditions were tested with buffers (i), (ii), and (iii). Tubes were alternatively frozen (−80 °C) and heated (water bath 65 °C) in water bath heating (56 °C/1–2 h) and freeze–thawing cycles. Three freeze–thaw cycles were per- formed, in about 3 min per step, with mixing by inversion in between. For deproteinization, phenol–chloroform–isoamilic alcohol (25:24:1) and chloroform–isoamilic alcohol (24:1) washes were performed (one each) in all cases [28]. DNA purification was obtained by two protocols: (a) isopropanol precipitation, with 3 M ammonium acetate, at −20 °C imme- diately/for 1 h/for 4 h/overnight, followed by washing/desali- nization with 70 % ethanol, air drying, and resuspension in Tris–EDTA solution (10:1) and (b) addition of solution C4 and loading in the purification column of the PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA). Washings with solution C5 and elution with solution C6 followed manufacturer’s instructions. The PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA) was also fully employed, following manufacturer’s instructions. The alternative lysis methods suggested in the manual were tested. The quality and size of the extracted DNAs were evaluated by electrophoresis on 1 % agarose gels stained with GelRed (Uniscience). Further confirmation of the purity of the extraction and quantification was made on a K and 0.5 (or 1)% (v/v; Invitrogen) 2-mercaptoethanol (Sig- ma-Aldrich)]; (iii) Gianni’s buffer, 250 μl ice-cold stabilizing solution [20 % (m/v) sucrose, 50 mM Tris–HCl (pH8,0), 50 mM EDTA] plus 500 μl ice-cold lysis solution [50 mM NaCl, 1 % (m/v) sodium N-lauryl sarcosine (Sigma-Aldrich), and freshly added 25 μg proteinase K]; and (iiii) 1 ml Trizol solution (Invitrogen). Alternative solutions containing in- creased concentrations of EDTA (50, 75, or 100 mM) were also tested. We compared the usefulness of each of these four buffers. Two lysis conditions were tested with buffers (i), (ii), and (iii). Tubes were alternatively frozen (−80 °C) and heated (water bath 65 °C) in water bath heating (56 °C/1–2 h) and freeze–thawing cycles. Three freeze–thaw cycles were per- formed, in about 3 min per step, with mixing by inversion in between. For deproteinization, phenol–chloroform–isoamilic alcohol (25:24:1) and chloroform–isoamilic alcohol (24:1) washes were performed (one each) in all cases [28]. DNA purification was obtained by two protocols: (a) isopropanol precipitation, with 3 M ammonium acetate, at −20 °C imme- diately/for 1 h/for 4 h/overnight, followed by washing/desali- nization with 70 % ethanol, air drying, and resuspension in Tris–EDTA solution (10:1) and (b) addition of solution C4 and loading in the purification column of the PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA). Washings with solution C5 and elution with solution C6 followed manufacturer’s instructions. The PowerSoil® DNA Isolation Kit (MO BIO Laboratories, Carlsbad, CA, USA) was also fully employed, following manufacturer’s instructions. The alternative lysis methods suggested in the manual were tested. The quality and size of the extracted DNAs were evaluated by electrophoresis on 1 % agarose gels stained with GelRed (Uniscience). Further confirmation of the purity of the extraction and quantification was made on a NanoDrop spectrophotometer (Thermo Fisher Scientific Inc.). PCR amplification of 16S rRNA gene (with bacteria- specific primer 27F and the universal primer 1492R) was performed on holobiont DNA samples, according to [29], in order to assess the quality of the samples for further molecular biology manipulations.

Metagenome Pyrosequencing Metagenome sequencing of the 12 samples was performed using 454 pyrosequencing technology [30]. Shotgun libraries were individually generated with 0.5 μg of whole DNA and sheared into fragments by nebulization. End-repair and adap- tor ligation were performed with the use of GS FLX Titanium kit (Roche), following manufacturer’s specifications. Quality control and quantification were done with the use of Agilent 2100 Bioanalyzer (Agilent Technologies) and TBS 380 Fluo- rometer (Turner Biosystems), respectively. After library con- struction, 106 molecules of each sample were denatured and amplified by emulsion PCR. Sequencing was done using the GS Junior System (Roche).

Dinucleotide Composition Analysis The dinucleotide composition of the Mussismilia metage- nomes was compared to other metagenomes obtained from and freely available in the databases, namely Porites com- pressa corals (field and lab samples), Acropora millepora corals (healthy and bleached), seawater from the Abrolhos Bank, and soil from Black Soudan mine. Frequency tabula- tion of the sequence data was performed according to [31], using homemade Perl scripts, and the principal compo- nent analysis (PCA) analysis (using covariance) of the tabulated data was performed in the STATISTICA software (StatSoft®).

Metagenomic Analysis The taxonomic composition of the coral metagenomes was first evaluated through local nucleotide blast (blastall 2.2.18) against a Coral database (Table S1). Data were organized using the MEGAN4 software [32]. Sequences with no hits against this database (e-value 10−5) were further annotated on the MG-Rast v3 server [33], under taxonomic classification (GenBank database, e-value cutoff 10−5), and both results were pooled together. The functional analysis was performed directly on MG-Rast (SEED database, e- value cutoff 10−3). Databases were last updated in Septem- ber 2012. Metagenome data are available on the MG-Rast server.

Statistical Analysis The taxonomic annotation was analyzed using the R statis- tical package ShotgunFunctionalizeR [34], to evaluate com- positional differences between healthy and white plague- affected corals. Gene-centric regression analysis was per- formed in the data using the Poisson model. Abundance percentage values were normalized based on the number of identified sequences in each domain within each metagenome. The R statistical package ShotgunFunctionalizeR [34] and the STAMP bioinformatics software [35] were also used for statistical analysis of the functional annotation of MG- RAST [33]. Additionally, the “All annotations” tool of MG- RAST was used to classify Mussismilia sequences (both healthy and WP) and also Porites [27] and Acropora (healthy and bleached) [26] sequences, both available in public databases. Manual curation of this automatic annota- tion was subsequently performed and resulted in a compar- ative table of the functions represented in these corals.