Abundance and Distribution of Dimethylsulfoniopropionate Degradation Genes and the Corresponding Bacterial Community Structure at Dimethyl Sulfide Hot Spots in the Tropical and Subtropical Pacific Ocean

Abstract

Dimethylsulfoniopropionate (DMSP) is mainly produced by marine phytoplankton but is released into the microbial food web and degraded by marine bacteria to dimethyl sulfide (DMS) and other products. To reveal the abundance and distribution of bacterial DMSP degradation genes and the corresponding bacterial communities in relation to DMS and DMSP concentrations in seawater, we collected surface seawater samples from DMS hot spot sites during a cruise across the Pacific Ocean. We analyzed the genes encoding DMSP lyase (dddP) and DMSP demethylase (dmdA), which are responsible for the transformation of DMSP to DMS and DMSP assimilation, respectively. The averaged abundance (±standard deviation) of these DMSP degradation genes relative to that of the 16S rRNA genes was 33% ± 12%. The abundances of these genes showed large spatial variations. dddP genes showed more variation in abundances than dmdA genes. Multidimensional analysis based on the abundances of DMSP degradation genes and environmental factors revealed that the distribution pattern of these genes was influenced by chlorophyll a concentrations and temperatures. dddP genes, dmdA subclade C/2 genes, and dmdA subclade D genes exhibited significant correlations with the marine Roseobacter clade, SAR11 subgroup Ib, and SAR11 subgroup Ia, respectively. SAR11 subgroups Ia and Ib, which possessed dmdA genes, were suggested to be the main potential DMSP consumers. The Roseobacter clade members possessing dddP genes in oligotrophic subtropical regions were possible DMS producers. These results suggest that DMSP degradation genes are abundant and widely distributed in the surface seawater and that the marine bacteria possessing these genes influence the degradation of DMSP and regulate the emissions of DMS in subtropical gyres of the Pacific Ocean.

INTRODUCTION

Dimethylsulfoniopropionate (DMSP), the precursor of dimethylsulfide (DMS), is mainly produced by marine phytoplankton, marine macroalgae, and a few angiosperms in the ocean (1–3) and is an important carbon and sulfur source for marine bacteria (4). After DMSP has been released, it is mainly assimilated and degraded by marine bacteria (5, 6). Phytoplankton and their predators also degrade DMSP to a certain extent (7, 8). Once incorporated into bacterial cells, DMSP is degraded via two major pathways: a demethylation pathway involving DMSP demethylase, encoded by dmdA (9), and a cleavage pathway involving several different ddd (DMSP-dependent DMS) (dddD, dddL, dddP, dddQ, dddY, and dddW) genes (10–15). dmdA, the first DMSP degradation gene identified, is the most widely distributed DMSP degradation gene. It was reported that approximately 60% of marine bacteria in the open ocean and coastal waters contain this gene (16). dmdA genes, which are found mainly in members of the SAR11, SAR116, Gammaproteobacteria, and Roseobacter clades (16–19), can be grouped into five clades and 14 subclades based on the genes' nucleotide sequences (16, 20).

In the cleavage pathway, bacteria transform DMSP to DMS. Aerosols formed from the oxidation of DMS increase the cloud cover over the ocean, thus creating the albedo that indirectly controls global heat fluxes (8, 21–23). DMS was found to be universally present in seawater and makes up a large proportion of the organic sulfur compounds emitted from the sea surface to the atmosphere (5, 23). This sea-to-air DMS flux represents about half of the global biogenic sulfur flux (24). dddP and dddQ genes are the most frequently detected ddd genes in marine bacteria and are mainly found in the Roseobacter clade (12, 13, 25).

Recent metagenomic and metatranscriptomic analyses have increased our understanding of DMSP cycling. These studies showed that the temporal variability in the abundance of DMSP degradation genes in the Sargasso Sea (18) and the North Pacific Ocean (17) was strongly influenced by seasonal changes in primary production, UV radiation, and depth-related environmental parameters, such as particulate DMSP (DMSPp) and DMS concentrations (17, 18, 26). However, their findings were limited to a few small, localized sampling points in the study area. The main objectives of this study were as follows: (i) to describe the distribution and abundances of DMSP degradation genes in relation to surface water DMS concentrations along a large-scale Pacific Ocean transect, (ii) to identify the bacterial communities potentially associated with these genes, (iii) to identify the environmental factors influencing the distribution of bacteria possessing DMSP degradation genes, and (iv) to explore the relationship between DMS concentration and DMSP degradation gene abundance on a larger spatial scale. In this study, we targeted dmdA and dddP genes and collected 20 seawater samples corresponding to an increase of DMS concentrations recorded with the use of an on-board real-time DMS monitoring system during a cruise across the Pacific Ocean.

MATERIALS AND METHODS

Real-time DMS monitoring and seawater sample collection.

In order to monitor the real-time dissolved DMS concentration on board, we continuously drew seawater from a 5-m depth with the use of a shipboard built-in pumping system. The outlet of the pumping system was directly connected to the equilibrator inlet-proton transfer reaction-mass spectrometry (EI-PTR-MS) system throughout the cruise, which allowed us to continuously measure the seawater DMS concentration and pinpoint “DMS hot spots” (27). Linear regression analysis showed that the dissolved DMS concentrations obtained using EI-PTR-MS were consistent and within the range of dissolved DMS concentrations measured using gas chromatography-mass spectrophotometry during the research vessel (R/V) Hakuho-maru cruise from July to August 2008 (slope, 0.90 ± 0.02 [mean ± standard deviation]; intercept, −0.03 ± 0.30; and R² = 0.99) (27). Seawater samples for DNA extraction were collected from an alternative outlet of the pumping system when we detected any significant increase (>1 nM) in DMS concentrations. Once collected, the samples were immediately filtered onto 47-mm-diameter, 3.0-μm-pore-size Nuclepore filters (Whatman, Clifton, NJ, USA) and sequentially filtered onto 0.22-μm-pore-size Sterivex filter units (Millipore, Bedford, MA, USA) to collect particle-associated (>3 μm) and free-living (0.22- to 3-μm) bacteria, respectively. The filters were immediately stored at −80°C until further analysis. We collected 20 different surface seawater samples from DMS hot spot sites, which matched to known DMS concentrations, from different sites during the cruise of R/V Hakuho-maru from December 2011 to March 2012 (Table 1; see also Fig. S1 in the supplemental material). In total, there were eight samples from the subtropical North Pacific Ocean (NP), seven samples from the subtropical South Pacific Ocean (SP), and five samples from the equatorial Pacific Ocean (EP). Among the samples from the subtropical South Pacific Ocean, three samples (S3p1, S3b, and S3p2) were collected from the Peruvian upwelling area.

TABLE 1.

Sampling locations and environmental data

Site	Description	Location	Date (mo/day/yr)a	Time	Location		Depth (m)	Value for indicated environmental factorb
					Latitude	Longitude		DMSc (nM)	Temp (°C)	Salinity (ppt)	Chlorophyll a (μg liter−1)	DMSc (nM)	DMSPd (nM)	DMSPp (nM)	DMSPt (nM)
N1p	Peak	North Pacific Ocean	12/12/2011	10:36	23°0′6″N	177°48′22″E	5	4.49	26.76	35.32	NA	NA	NA	NA	NA
N1b1	Baseline	North Pacific Ocean	12/12/2011	14:04	22°59′56″N	178°43′11″E	5	1.39	26.82	35.33	NA	NA	NA	NA	NA
N1b2	Baseline	North Pacific Ocean	12/13/2011	20:30	22°59′54″N	179°59′54″E	5	1.00	26.50	35.33	NA	NA	NA	NA	NA
N2p	Peak	North Pacific Ocean	12/28/2011	19:45	11°56′49″N	146°55′35″W	5	4.25	25.35	33.89	0.07	NA	NA	NA	NA
N2b	Baseline	North Pacific Ocean	12/28/2011	22:03	11°32′47″N	146°31′54″W	5	3.14	25.36	33.58	0.07	NA	NA	NA	NA
S1p	Peak	South Pacific Ocean	01/06/2012	20:26	19°59′58″S	120°0′51″W	5	5.66	25.82	36.21	0.08	NA	NA	NA	NA
S1b	Baseline	South Pacific Ocean	01/07/2012	10:29	22°31′14″S	119°59′59″W	5	3.37	25.66	36.35	0.10	NA	NA	NA	NA
S2p1	Peak	South Pacific Ocean	01/20/2012	11:26	20°0′2″S	100°0′3″W	5	6.08	24.48	35.86	0.11	NA	NA	NA	NA
S2p2	Peak	South Pacific Ocean	01/20/2012	11:26	20°0′0″S	100°0′6″W	5	6.31	24.73	35.86	0.07	NA	NA	NA	NA
S3p1	Peak	South Pacific Ocean	01/23/2012	23:13	15°24′16″S	86°43′20″W	5	4.50	24.04	35.49	0.07	NA	NA	NA	NA
S3b	Baseline	South Pacific Ocean	01/23/2012	18:09	15°0′36″S	85°35′54″W	5	3.21	24.51	35.42	0.12	NA	NA	NA	NA
S3p2	Peak	South Pacific Ocean	01/24/2012	01:25	14°45′15″S	84°52′41″W	5	5.29	24.10	35.37	0.19	NA	NA	NA	NA
E1p	Peak	Equatorial Pacific Ocean	02/07/2012	03:25	0°3′22″N	115°2′2″W	5	4.80	24.60	34.88	0.22	3.56	4.37	18.79	23.16
E1b	Baseline	Equatorial Pacific Ocean	02/08/2012	15:05	0°0′4″N	115°0′4″W	5	2.86	24.27	34.90	0.19	2.38	2.05	20.20	22.25
E2p1	Peak	Equatorial Pacific Ocean	02/09/2012	21:31	0°0′8″S	121°5′35″W	5	4.16	26.43	34.24	0.16	5.14	6.84	14.80	21.63
E2p2	Peak	Equatorial Pacific Ocean	02/10/2012	02:59	0°0′10″S	122°43′34″W	5	4.55	26.73	34.12	0.21	5.80	6.37	17.70	24.07
E2b	Baseline	Equatorial Pacific Ocean	02/10/2012	09:25	0°0′9″S	124°39′30″W	5	3.11	23.86	34.81	0.22	3.70	5.35	19.86	25.21
N3b1	Baseline	North Pacific Ocean	02/16/2012	18:30	8°25′11″N	147°3′13″W	5	2.91	26.33	34.79	0.15	3.23	1.02	6.00	7.02
N3p	Peak	North Pacific Ocean	02/17/2012	09:05	10°53′13″N	149°8′19″W	5	4.71	25.35	34.02	0.19	5.34	0.72	8.78	9.49
N3b2	Baseline	North Pacific Ocean	02/18/2012	00:17	14°3′23″N	151°50′54″W	5	3.19	25.36	34.22	0.11	3.42	0.47	6.59	7.06

^aGreenwich mean time.

^bNA, not available.

^cDMS concentrations were detected by gas chromatography for several stations; therefore, two sets of DMS concentration data were available for these stations.

DMS and DMSP quantification by gas chromatography.

The DMS and DMSP (dissolved DMSP [DMSPd], particulate DMSP [DMSPp], and total DMSP [DMSPd + DMSPp] [DMSPt]) concentrations were measured by gas chromatography at the same locations where DNA samples were collected during the latter half of the cruise (at sites E1p to E2b and N3b1 to N3b2). The method was previously described in detail (28). For DMS measurement, 30 ml of seawater collected from the outlet of the shipboard built-in pumping system was directly filtered through a GF/F (47-mm) filter (Whatman, USA). As soon as possible after sampling, the DMS concentration was measured by a gas chromatography-flame photometric detector (GC-FPD) (GC-2014; Shimadzu, Kyoto, Japan) with a 25% TCEP [1,2,3-Tris(2-cyanoethoxy)propane]-packed column (GL Science, Tokyo, Japan) combining a purge-and-trap extraction/preconcentration system. For DMS plus DMSPt measurement, 30 ml of seawater was collected directly into a glass vial without prior filtration and was subjected to an alkali treatment to permit cleavage into gaseous DMS. After the alkali treatment, the sample bottles were stored at 4°C for at least 24 h to allow completion of the cleavage process before further analysis. For DMS plus DMSPd measurements, 30 ml of seawater was directly filtered through a GF/F (47 mm) filter (Whatman) and subjected to an alkali treatment to cleave DMS in a manner similar to that used for DMS plus DMSPt measurement. The DMSPp concentration was calculated as follows: DMSPp = (DMS + DMSPt) – (DMS + DMSPd).

Environmental factors.

The salinity, temperature, and chlorophyll a concentrations were continuously monitored with the automated environmental monitor for biological oceanography (AMEMBO) system. The system consisted of a bubble trap, a Seabird SBE-21 thermosalinograph, and a Turner Design 10-R in vivo fluorometer. An outlet of the shipboard built-in pumping system was directly connected to the AMEMBO system. The chlorophyll a concentrations determined by the AMEMBO system were calibrated against the chlorophyll a extracted from filter samples.

DNA extraction.

Genomic DNA was extracted using the ChargeSwitch Forensic DNA purification kit (Invitrogen, Carlsbad, CA, USA) according to the manufacturer's instructions, with slight modifications. After breaking the Sterivex filter unit that was used to collected free-living bacteria (0.22 μm to 3 μm), the membrane was cut into pieces with sterile surgical blades. The membrane pieces were crushed by bead beating in a Micro Smash MS-100 (Tomy Digital Biology Co., Ltd., Tokyo, Japan) at 5,000 rpm for 30 s to physically break the cell walls of bacteria on the membrane. After the bead beating, we followed the manufacturer's instructions for using the ChargeSwitch Forensic DNA purification kit (Invitrogen). For each sample, initial steps of bead beating until elution were repeated twice to increase yield. The DNA concentration was determined using a Quant-iT PicoGreen double-stranded DNA (dsDNA) kit (Invitrogen). Nuclepore filters (3 μm) were subjected to the same extraction steps as the Sterivex filter units to collect DNA from particle-associated bacteria (>3 μm).

Abundances of dddP, dmdA, and 16S rRNA genes.

The copy numbers of DMSP degradation genes were analyzed by quantitative PCR (qPCR) on a Bio-Rad Chromo4 system. Eight dmdA primer sets designed to target the different subclades A/1, A/2, B/3, B/4, C/2, D/1, D/3, and E/2 (20) and a dddP primer set targeting the Roseobacter clade (18) were used to detect DMSP degradation genes. Among the dmdA primer sets, A/1 and A/2 target the Roseobacter clade (dmdA genes from clade A), B/3 and B/4 target dmdA genes from clade B bacteria (SAR116 member “Candidatus Puniceispirillum marinum,” IMCC1322-like marine bacterial clade), C/2 targets dmdA genes from clade C bacteria (Pelagibacter ubique HTCC7211 [PB7211-1421]-like clade), D/1 and D/3 target dmdA genes from clade D bacteria (Pelagibacter ubique HTCC7211 [PB7211-770] and Pelagibacter ubique HTCC1062 [SAR-0264]-like clade), and E/2 targets dmdA genes from HTCC7211 (PB7211-1421) clade E bacteria (marine gammaproteobacterium HTCC2080-like clade) (9, 16, 20). Environmental samples (one sample each from the North, South, and equatorial Pacific Ocean) were subjected to Sanger sequencing to confirm dmdA and dddP gene specificity. The annealing temperatures of the dmdA primer sets, except for those targeting the B/4 and E/2 subclades, were as described by Varaljay et al. (20). The annealing temperature of the dddP primer set was as described by Levine et al. in 2012 (18). In this study, the annealing temperature of the primers targeting the E/2 subclade was optimized to 61°C, because PCR amplifications with the previously described annealing temperature (57°C) yielded multiple bands from our environmental samples. PCR amplification of B/4 subclade genes also yielded multiple bands, and so we omitted this primer set from this study. The abundance of 16S rRNA genes was determined using the BACT1369F and PROK1492R primer set (29). qPCR standards were made from PCR products amplified from environmental samples using the TOPO TA cloning kit (Invitrogen), and plasmid DNA was purified using a PureLink quick plasmid miniprep kit (Invitrogen). The purified plasmid DNA was digested with NotI and subjected to agarose gel insert purification (Qiagen, Hilden, Germany). The concentration of the qPCR standards was determined using a Quant-iT PicoGreen dsDNA kit (Invitrogen).

We used the SYBR premix Ex Taq (Tli RNaseH Plus) kit (TaKaRa Bio, Inc., Otsu, Japan) for qPCR detection. qPCR was performed in 20-μl reaction mixtures containing 10 μl 2× SYBR premix Ex Taq, 0.4 μl 50× ROX reference dye, 0.2 mM each primer, 2 μl 1/10-diluted template DNA, and 6.8 μl water. All qPCRs were run in triplicate for each sample. The qPCR conditions were as follows: initial denaturation for 30 s at 95°C, followed by 35 cycles of denaturation at 95°C for 5 s, annealing at the primer-specific annealing temperature for 30 s, and extension at 72°C for 30 s. The fluorescence intensity was read, and then the temperature was raised from 65°C to 95°C and the fluorescence intensity read 2 s after every 0.2°C increase in temperature. A 10-fold serially diluted standard and no-template control were run in triplicates for each reaction. The presence of a single band was verified by agarose gel electrophoresis.

Bacterial 16S rRNA gene analysis.

DNA from the free-living fraction of each environmental sample was targeted to determine the bacterial 16S rRNA gene community composition. Bacterial 16S rRNA genes were amplified for 20 cycles using the barcoded primer set 27F (5′-CCATCTCATCCCTGCGTGTCTCCGACTCAGXXXXAGAGTTTGATCMTGGCTCAG-3′) and the reverse primer 519R (5′-GWATTACCGCGGCKGCTG-3′), where Xs represent the barcode sequences, adapter sequences are in boldface, and primer sequences are underlined. This primer set targets the V1-to-V3 region of the bacterial 16S rRNA gene (30, 31). PCR products were prepared in triplicate and pooled for purification. The PCR products were purified using the Ampure system (Agencourt Bioscience Corporation), with modifications to the volume of purified PCR products (22.5 μl) and AMPure beads (72 μl). The prepared DNA libraries were sequenced on a 454 GS Junior instrument (Roche Applied Science, Indianapolis, IN) according to the manufacturer's instructions.

A total of 369,855 16S rRNA gene sequences for taxonomic analysis were processed using mothur version 1.33.3 (32) following the standard operating procedure proposed by Schloss et al. (33). Sequences were removed from the analysis if they had a read quality score under 25, contained ambiguous characters, contained more than two mismatches to the forward primer or one mismatch to the barcode, or were under 150 bp or over 550 bp. Sequencing noise was further reduced through the precluster method (34). Chimeras were identified and removed using chimera.uchime. The average read length was 186 bp after barcode and primer sequences were trimmed. The Greengene database (gg-13-5-99) was used to align and classify the sequences. A similarity cutoff of >97% was used for assigning the same operational taxonomic units (OTUs).

The dmdA genes detected in this study were from six different subclades, and previous studies showed that they were mainly found in the SAR11 clade (the Pelagibacteraceae). OTUs classified as Pelagibacteraceae and with a relative abundance of >0.5% were used for phylogenetic tree construction using the neighbor-joining method for further analysis.

Statistical analysis.

To investigate potential environmental drivers of DMSP degradation, principal component analysis (PCA) and redundancy analysis (RDA) were performed using R's BiodiversityR package (35, 36). The data for the relative abundances of DMSP degradation genes and for environmental factors were used to perform PCA and RDA. The environmental factors used in RDA were standardized by subtracting the mean value and then divided by the standard deviation of the variable. Hierarchical clustering was performed using R's Vegan package (37). Spearman's rank correlation coefficient (ρ) and significance (P) values were calculated using R's Hmisc package (38) and the PerformanceAnalytics packages (39).

The DMSP degradation gene abundance data used in this study were the copy numbers of DMSP degradation genes normalized to the bacterial 16S rRNA gene unless stated otherwise. DMSP (DMSPp, DMSPd, and DMSPt) data were only available for eight stations in this study; therefore, these variables were not included in the RDA. Chlorophyll a samples were not collected from the northernmost stations (N1p to N1b2), so these three stations were also excluded from the analysis. We used the Bray-Curtis dissimilarity distance matrix based on the data for relative abundance of 16S rRNA to construct the dendrogram. The dendrogram was constructed by using the unweighted-pair group method using average linkages (UPGMA) algorithm.

Nucleotide sequence accession number.

Bacterial 16S rRNA gene sequences and accompanying metadata have been deposited in the DDBJ (http://www.ddbj.nig.ac.jp/) Sequence Read Archive under the project number DRA002967.

RESULTS

DMS and DMSP concentrations.

The concentrations of DMS and DMSP in seawater samples are listed in Table 1 together with environmental data and sampling site information. In this study, we focused on samples corresponding to DMS peaks (i.e., those collected when increased DMS concentrations were detected) and DMS baselines (i.e., those collected when the DMS concentration was at least 1 nM lower than the corresponding peak DMS concentration) (see Fig. S1 in the supplemental material). The average DMS concentration was 3.95 ± 1.37 nM, with the lowest concentration detected in the North Pacific Ocean (1.00 nM at site N1b2 [where “N” denotes the North Pacific and “b” denotes a sample taken at baseline DMS concentration]) and the highest in the South Pacific Ocean (6.31 nM at site S2p2 [where “S” denotes the South Pacific and “p” denotes a sample taken at peak DMS concentration]). Regionally, the highest average DMS concentration was in the samples from the SP region (4.92 ± 1.16 nM) and the lowest was in the samples from the NP region (3.14 ± 1.29 nM). The average DMS concentration in the samples from the equatorial Pacific Ocean was 3.90 ± 0.78 nM.

DMSP concentrations were measured in all EP and three NP (N3b1 to N3b2) samples. The DMSP (DMSPd and DMSPp) concentrations were higher in the samples from the EP region than in those from the NP region. The average DMSPt (total DMSP, combined value of DMSPd and DMSPp) concentration was about three times higher in the samples from the EP region than in those from the NP region (EP, 23.26 ± 1.27 nM, and NP, 7.86 ± 1.16 nM). Most of the DMSPt (>81%) in seawater was in the form of DMSPp in both regions. The DMSPp concentrations in the Pacific Ocean correlated positively with the chlorophyll a concentrations (ρ = 0.75, P < 0.05), which agreed with previous results showing that marine phytoplankton were the main source of DMSPp production (3, 40).

Bacterial dmdA and dddP gene abundance.

Seven subclades (A/1, A/2, B/3, C/2, D/1, D/3, and E/2) of dmdA genes, dddP genes, and bacterial 16S rRNA genes were quantified using qPCR. The bacterial 16S rRNA gene copy numbers were used to normalize the copy numbers of dmdA subclade genes and dddP genes to evaluate the ratios of bacteria possessing DMSP degradation genes in the total marine bacterial community. The values obtained after normalization of each gene's copy number against the bacterial 16S rRNA gene copy numbers are indicated as the relative abundance.

DNAs from both free-living (0.22- to 3-μm) and particle-associated (>3-μm) bacteria were used to quantify DMSP degradation genes. However, the numbers of particle-associated bacteria possessing dddP genes or dmdA subclade C/2 and D/1 genes, which were the most abundant DMSP degradation genes detected in this study, accounted for less than 3% of the numbers of free-living bacteria possessing these genes (see Table S3 in the supplemental material). Therefore, in this paper, we mainly focused on the results from free-living bacteria.

The copy numbers of dmdA and dddP genes and their distributions in the Pacific Ocean varied greatly among sampling sites. Some sites showed particularly high copy numbers of dddP genes (sites around the upwelling area in the SP region and a few sites in the EP region). At most sites, the copy numbers of dddP genes were lower than the copy numbers of the total dmdA genes (the combined value of all dmdA gene subclades); however, around the upwelling area in the SP region, the copy numbers of dddP genes exceeded the copy numbers of total dmdA genes (see Fig. S2 in the supplemental material). The copy number ratios of these genes (dddP/total dmdA) ranged from 2% to 140%. The copy numbers of dddP genes in the samples from the NP region, especially the northernmost stations [the average copy number (±standard deviation) of samples N1p to N1b2 was (9.45 ± 0.65) × 10⁵ copies liter⁻¹], were lower than the copy numbers in samples from the SP and EP regions. In the samples from the SP region, the copy numbers of dddP genes were higher in areas closer to the Peruvian upwelling area. Samples from these stations (S3p1, S3b, and S3p2) showed the highest copy numbers of dddP genes, with an average of (1.03 ± 0.44) × 10⁸ copies liter⁻¹. dmdA subclade D/1 genes (SAR11 clade) were the most abundant throughout the collected samples [(3.49 ± 1.70) × 10⁷ copies liter⁻¹], except for those from the northernmost stations (N1p, N1b1, and N1b2). In these northernmost stations, the dominant dmdA subclade was C/2 (SAR11 clade), with an average of (1.35 ± 0.79) × 10⁷ copies liter⁻¹. The copy numbers of the dmdA subclade B/3 and E/2 genes were highly variable among samples and accounted for only about 5% of the total dmdA genes.

The relative abundances of DMSP degradation genes, the values after normalizing each gene's copy number to the bacterial 16S rRNA gene copy numbers, averaged 33% ± 12% (ranging from 16% to 60%) (Fig. 1). The averaged relative abundances of dddP genes in the samples from the NP, SP, and EP regions were 3% ± 2%, 13% ± 11%, and 15% ± 6%, respectively. Unusually high relative abundances of dddP genes were found in the samples from around the upwelling area of the SP region (S3p1, 23%, and S3p2, 31%) and also in the samples from the EP region (E1p, 22%, and E1b, 22%). The averaged relative abundance of total dmdA genes was 23% ± 5%; the maximum value was detected around the SP upwelling area (32%). C/2, D/1, and D/3 subclade genes from the SAR11 clade made up the greatest proportion of the total dmdA genes, with an average of 88% ± 6%. Bacteria possessing dmdA subclade A/1 and A/2 genes in the northernmost sites, N1p and N1b1, exceeded those possessing dddP, while at other sites, the bacteria with these genes made up about one-fifth of the bacteria possessing dddP genes.

FIG 1.

Relative abundances of DMSP degradation genes in the Pacific Ocean. (a) Relative abundances of seven different subclades of dmdA (bacterial 16S rRNA copy numbers were used for normalization). (b) Relative abundances of dddP genes (bacterial 16S rRNA copy numbers were used for normalization). Samples from the North Pacific Ocean are denoted by N, samples from the South Pacific Ocean by S, and samples from the equatorial Pacific Ocean by E; samples from peak DMS concentration sites are denoted by p and those from baseline DMS concentration sites by b. Arabic numbers correspond to sampling points in each region.

Bacterial community structure analysis.

After the removal of low-quality sequences, a total of 150,170 sequences were obtained from pyrosequencing, which were divided into 2,112 different OTUs (97% identity), and a maximum of 431 different OTUs were from DMS hot spot samples (see Table S4 in the supplemental material).

The bacterial communities in DMS hot spot samples were dominated by Alphaproteobacteria (52% ± 9%), followed by Synechococcophycideae (24% ± 11%), Gammaproteobacteria (11% ± 3%), and Flavobacteria (4% ± 1%). OTUs belonging to these four classes made up >80% of all observed OTUs. Generally, samples from NP regions showed lower relative abundances of Alphaproteobacteria and higher relative abundances of Synechococcophycideae compared with samples from the other two regions. In the two samples from the northernmost station, N1b1 and N1b2, the relative abundances of Synechococcophycideae (N1b1, 41%, and N1b2, 40%) were higher than the relative abundances of Alphaproteobacteria (N1b1, 38%, and N1b2, 36%). Within the Synechococcophycideae, over 99.9% of OTUs concentrated in the Synechococcaceae; the relative abundances of this family showed higher values in the samples from NP regions (35% ± 8%) and lower values in the samples from EP (16% ± 4%) and SP (17% ± 6%) regions (Fig. 2). Halomonadaceae made up 70 to 95% of the total abundance of Gammaproteobacteria in all samples collected. The samples from EP regions showed higher relative abundances of this family than did the samples from the other two regions. Within the Alphaproteobacteria, Pelagibacteraceae (or SAR11 clade) was the most abundant family, followed by the Rhodobacteraceae. OTUs belonging to these two families made up to 94% ± 2% of the total relative abundance of Alphaproteobacteria. Pelagibacteraceae was the most abundant family in all samples. This family was most abundant in the samples from the SP regions (53% ± 2% in total OTUs), followed by the samples from the EP (50% ± 4%) and NP (37% ± 5%) regions. Compared with the relative abundances of Pelagibacteraceae, Rhodobacteraceae exhibited much lower relative abundances among the total OTUs, ranging from 1 to 8%.

FIG 2.

Bacterial community compositions of DMS hot spot samples, classified on the family level based on 16S rRNA gene sequences. SAR11 groups were separated into subgroups Ia and Ib and others.

The phylogenetic tree of Pelagibacteraceae constructed from the OTUs with relative abundances of more than 0.5% yielded four different clades (see Fig. S5 in the supplemental material). OTU0001, the dominant OTU in all samples, was found to be confined to SAR11 subclade Ia. OTU0003, OTU0032, and OTU0106 belonged to SAR11 subclade Ib; OTU00005, OTU00016, and OTU00039 belonged to SAR11 subclade II; and OTU00028 belonged to SAR11 subclade IV. The combined relative abundances of SAR11 subgroups Ia and Ib were about 70 to 90% of the total relative abundance of the SAR11 clade (Pelagibacteraceae) and comprised 25 to 50% of the total bacterial OTUs detected.

To compare the bacterial communities of different sampling stations, UPGMA cluster analysis was performed with the OTUs with a relative abundance of more than 0.5% (see Fig. S6 in the supplemental material). The dendrogram separated the samples into three different clusters: the first cluster consisted of the three samples from the northernmost area (N1p, N1b1, and N1b2), the second cluster consisted of the remaining NP region samples and one SP region sample (S1b), and the final cluster consisted of all EP region samples and the remaining SP region samples. More specifically, most of the grouped samples (samples from the DMS peak site and corresponding baseline sites) clustered together, except for S1p-S1b and E2p1-E2p2-E2b.

Correlation analysis among bacterial 16S rRNA gene abundance, DMSP degradation gene abundance, and environmental factors.

dmdA subclade C/2, D/1, and D/3 genes (SAR11 clade) exhibited the highest abundances among the dmdA genes tested. The dddP genes quantified in this study were specific for the marine Roseobacter clade. Therefore, a correlation analysis of clade-specific dmdA and dddP gene abundances relative to the bacterial 16S rRNA gene abundances for Pelagibacteraceae and Rhodobacteraceae was carried out for all DMS hot spot samples. The relative abundances of the Rhodobacteraceae correlated significantly with the relative abundances of dddP genes (ρ = 0.92, P < 0.001) and dmdA subclade A/1 genes (ρ = 0.85, P < 0.001) but not with the relative abundance of dmdA subclade A/2 genes (ρ = 0.04, P > 0.1). The relative abundance of dmdA subclade D genes (the sum of dmdA subclade D/1 and D/3 genes) correlated significantly with the relative abundance of SAR11 subclade Ia (ρ = 0.57, P < 0.01), and the relative abundance of dmdA subclade C/2 genes correlated with that of SAR11 subclade Ib (ρ = 0.78, P < 0.001) (see Fig. S7 in the supplemental material). The relative abundance of Rhodobacteraceae correlated negatively with that of SAR11 subclade Ib (ρ = −0.91, P < 0.001) but positively with that of SAR11 subclade Ia (ρ = 0.74, P < 0.01). The relative abundance of SAR11 subclade Ia correlated negatively with that of SAR11 subclade Ib (ρ = −0.87, P < 0.001). The relative abundance of dddP genes correlated negatively with that of dmdA subclade C/2 genes (ρ = −0.75, P < 0.01) but positively with that of dmdA subclade A/1 genes (ρ = 0.95, P < 0.001). The relative abundance of dddP genes increased with that of dmdA subclade D genes, but this was not statistically significant (ρ = 0.6, P < 0.1).

The DMSP degradation gene abundance also correlated with some environmental factors. The relative abundances of total dmdA genes correlated positively with the DMSPd concentration (ρ = 0.71, P < 0.01) and also increased with the DMSPt concentration (ρ = 0.71, P < 0.1), although there was not a significant correlation. Among all of the dmdA subclades, the relative abundance of the SAR11 clade (combined value of relative abundances of dmdA subclade C/2, D/1, and D/3 genes) correlated with the DMSPd concentration (ρ = 0.6, P < 0.05). The relative abundance of dddP correlated with the DMSPp concentration (ρ = 0.9, P < 0.05).

Environmental factors affecting the distribution of bacteria possessing DMSP degradation genes.

Principal component analysis (PCA) was used to identify the overall variations in the relative abundances of DMSP degradation genes (Fig. 3a). Together, the first two PCA axes represented 96% of total variability in the relative abundances of DMSP degradation genes. The PCA results separated the 20 samples into two main clusters. The first cluster, designated group 1, consisted of samples from all NP and four SP stations (S1p to S2p2). The second cluster, designated group 2, consisted of samples from the EP region and those collected around the upwelling area in the SP region. Samples from the group 1 and group 2 sites were well separated on the PC1 axis, which represented 80% of the total variance. The scores for the individual genes indicated that this axis was mainly represented by the relative abundances of dddP genes and dmdA subclade D/1 genes. These genes were found in high relative abundances in group 2 sites and in low abundances in group 1 sites. In addition, PCA constructed with the relative abundances of total dmdA and dddP also separated the samples into two groups as described above (see Fig. S8 in the supplemental material).

FIG 3.

Ordination plots of principal component analysis (PCA) (a) and redundancy analysis (RDA) (b) results. PCA and RDA plots were constructed using the relative abundances of dmdA subclade genes and of dddP genes. Sampling sites are indicated in black, and the relative abundances of dmdA subclade genes and of dddP genes are indicated in red. Group 1 and group 2 sites are shaded in green and yellow, respectively. Environmental factors used in RDA are indicated by blue arrows. The two factors explaining the highest proportion of variability are shown in parentheses on the axes of the RDA plot.

Redundancy analysis (RDA) was used to evaluate the relationship between the relative abundances of DMSP degradation genes and environmental factors (chlorophyll a and DMS concentrations, salinities, and temperatures). Based on the RDA models, four environmental factors could be ranked according to their ability to explain variations in microbial functional gene data (Fig. 3b). From the strongest explanatory variable to the weakest, the ranking order for the RDA1 axis was as follows: temperature (°C) > chlorophyll a (μg liter⁻¹) > salinity (ppt) > DMS (nM). For the RDA2 axis, the order was chlorophyll a (μg liter⁻¹) > temperature (°C) > DMS (nM) > salinity (ppt). Analysis of variance (ANOVA) showed that the RDA1 axis, chlorophyll a, and temperature were statistically significant in the RDA model (P < 0.01), while the RDA2 axis, DMS, and salinity were not (P > 0.1). The patterns in the RDA ordination, which showed the relationships between DMSP degradation gene structures and the four environmental factors, were consistent with the patterns in the PCA ordination. The RDA1 axis represented 52% of the variation in bacterial communities possessing DMSP degradation genes, and most of the variation was explained by chlorophyll a and temperature. This axis (RDA1) correlated significantly with chlorophyll a (ρ = 0.62, P < 0.01) and temperature (ρ = 0.62, P < 0.01) but not with DMS (ρ = 0.11, P > 0.1) or salinity (ρ = 0.12, P > 0.1). The RDA1 axis correlated positively with the samples from the group 2 sites and negatively with the samples from the group 1 sites. RDA plots constructed with the relative abundances of total dmdA and dddP also showed results similar to those of the above-described RDA plots (see Fig. S8 in the supplemental material).

dddP gene abundance and DMS concentration.

The relative abundances of dddP genes collected from the DMS peak site were higher than those of the corresponding DMS baseline site in the same sampling region, except for the samples from the northernmost area, N1p-N1b2, and one sample, E2b, from the EP region (Fig. 4). When all samples were evaluated, these relationships between the relative abundances of dddP genes and DMS concentrations did not meet the criteria for statistical significance (ρ = 0.32, P > 0.1).

FIG 4.

Relative abundances of dddP genes (bars) and DMS concentrations (filled circles) are shown. The relative abundances of dddP genes in grouped samples are marked with similar patterns and separated by dotted vertical lines.

The PCA results indicated that samples from the NP region and the non-upwelling area in the SP region (the group 1 sites) shared similar DMSP degradation gene structures (Fig. 3a), which were strongly influenced by the chlorophyll a concentrations and temperature (Fig. 3b). Therefore, samples from the group 1 sites and group 2 sites were analyzed separately. When data from group 1 sites were pooled, the relative abundance of dddP was positively correlated with the seawater DMS concentration (ρ = 0.70, P < 0.05) (see Fig. S9 in the supplemental material). In addition, the proportion of the relative abundance of dddP to the relative abundance of total DMSP degradation genes [dddP/(dddP + total dmdA)] was also positively correlated with DMS concentrations (ρ = 0.69, P < 0.05). For group 2 sites, the DMS concentration increased with the relative abundance of dddP genes, but this was not statistically significant (ρ = 0.44, P > 0.1).

DISCUSSION

We obtained four key results in this study. First, the distribution of both dddP and dmdA genes in the Pacific Ocean varied greatly among the sampling sites. In most sites, total dmdA genes were more abundant than dddP genes; however, in some sites, the dddP gene abundance exceeded that of the total dmdA genes (Fig. 1). Second, the distribution patterns of DMSP degradation genes in the oligotrophic North and South Pacific Ocean differed from the patterns in the equatorial and South Pacific upwelling area. These distribution patterns were influenced by the chlorophyll a concentrations and temperature (Fig. 3). Third, SAR11 subgroup I (Ia and Ib) bacteria possessing dmdA genes were the major potential DMSP consumers in the surface seawater. Fourth, the DMS concentrations increased with the relative abundance of marine Roseobacter clade bacteria possessing dddP genes (Fig. 4). In the samples from NP and non-upwelling SP regions, there were statistically significant positive correlations between the DMS concentrations and the relative abundances of dddP genes, as well as the proportions of the dddP genes in the total DMSP degradation genes (dddP/dddP + total dmdA) (see Fig. S9 in the supplemental material).

The range of average DMS concentrations in the subtropical and equatorial Pacific Ocean is moderate to low throughout the year, with higher concentrations during the summer (41). Our data possibly reflected this seasonal shift in DMS concentrations in both hemispheres. The average DMS concentration in the samples from the NP region (winter) was lower than that in the samples from the SP region (summer). The highest value was found at S2p2, which was higher than the average DMS concentration in summer in this region (South Pacific Subtropical Gyre [SPSG]). This high DMS concentration might be influenced by the permanent upwelling area of the Peru Current System (42).

The per-cell uptake of amino acid and/or dissolved ATP by particle-associated bacteria is faster than the uptake by free-living bacteria (43, 44). However, due to the low abundance of the particle-associated bacteria (<10% in total bacteria) (45–47), free-living bacteria accounted for 50% to 80% of the microbial turnover of simple organic compounds (45). According to the deep-sequencing results for samples from non-DMS hot spot sites collected from the same cruise, the relative abundances of SAR11 bacteria and Rhodobacteraceae in the particle-associated fraction (>3 μm) were 10 to 15% and less than 5%, respectively (S. Suzuki, personal communication). Our results indicated that the number of particle-associated bacteria possessing DMSP degradation genes was less than 3% of the number of free-living bacteria (see Table S3 in the supplemental material). Therefore, based on these results (the low abundance of particle-associated bacteria and their DMSP degradation genes), we assumed that SAR11 and Rhodobacteraceae populations, which are the main possessors of dmdA and dddP genes, respectively, contribute less to DMSP metabolism in the particle-associated bacterial fraction than in the free-living bacterial fraction.

The relative abundances of total dmdA genes ranged from 15% to 32% in the free-living bacteria within each sample. This value was lower than the values calculated from the 2007 Global Ocean Sampling (GOS) metagenomic database, normalized using well-known single-copy genes, which showed that more than half of marine bacteria possessed dmdA genes (16). We assumed several reasons to explain this discrepancy. First, the primer sets used in this study might not cover all known dmdA gene diversity, since the primer nucleotide sequences have to avoid redundancies for accurate qPCR amplification (20). Another possible reason was that the bacterial 16S rRNA gene was used as the normalizer in this study since the single-copy genes used in the Global Ocean Sampling Survey were hard to target. We predicted the 16S rRNA gene copy numbers within each sample according to the 16S rRNA gene community structures against strains with known genomic sequences using PICRUSt (48). The sequences used to predict 16S rRNA gene copy numbers covered over 80% of the total sequences obtained from bacterial community structures. The predicted 16S rRNA gene copy numbers ranged from 1.1 to 1.4 copies, with an average value of 1.3 ± 0.1 copies. This result agreed with previous results showing that the surface seawater free-living bacteria contained, on average, 1.8 copies of the 16S rRNA genes, with higher average 16S rRNA gene numbers in coastal ocean sites (2.8 copies) than in open ocean sites (1.3 copies) (49). The multicopy numbers of 16S rRNA genes contained in a bacterial cell could cause underestimation of the proportions of dmdA gene-possessing cells in the samples. Furthermore, nonbacterial 16S rRNA genes could have decreased the relative abundance of dmdA genes, as they contributed, on average, 4% ± 2% (range, 1 to 8%) to the total small-subunit rRNA pool.

The average relative abundance of dddP genes obtained in this study (10%) was about two times higher than the values in the GOS metagenomic data (8, 12) and, also, about three times higher than the values that were shown in the stations ALOHA and BATS (17, 18). The relatively high values (ca. 20 to 30%) were recorded in several EP and SP (around the Peruvian upwelling area) samples, such as S3p1, S3p2, E1p, and E1b. These values contributed to the relatively high average values obtained in this study. The dddP gene distributions in the Pacific Ocean varied significantly. This unevenness was also apparent in the GOS metagenomic data, with several higher values in coastal and hypersaline sites (12 to 70%) and low values in other coastal and open ocean sites (0%) (12).

The relative abundance of SAR11 subgroup Ia correlated positively with the relative abundance of dmdA subclade D genes, and SAR11 subgroup Ib correlated positively with dmdA subclade C/2 genes, implying that bacteria from SAR11 subgroups Ia and Ib possessed dmdA subclade D and C/2 genes, respectively. Genomic sequences for SAR11 subgroup Ia bacteria, such as Pelagibacter ubique HTCC 1062 and Pelagibacter ubique HTCC 1002, contained dmdA subclade D genes, which supported the hypothesis described above. However, no such data from SAR11 subgroup Ib were available to confirm the presence of dmdA subclade C/2 genes in this group. From our data, SAR11 subgroups Ia and Ib made up on average 38% of the total bacteria (based on 16S rRNA gene sequences) (Fig. 2), while dmdA subclade C/2 and D genes made up on average 20% of the total bacterial genes (based on 16S rRNA gene copy numbers) (Fig. 1a). Therefore, over half of the bacteria from SAR11 subgroups Ia and Ib possessed dmdA subclade C/2 or subclade D genes. Previous studies revealed that roughly half of the SAR11 bacteria were responsible for the assimilation of dissolved DMSP (50). Therefore, it can be inferred that SAR11 subgroup Ia and Ib bacteria possessing dmdA subclade C/2 and D genes could play an important role in DMSP assimilation pathways. In addition, based on the relatively high correlations between DMSPd concentrations and SAR11 subgroups Ia and Ib and dmdA genes from the SAR11 clade, as well as the abundance of the dmdA genes from the SAR11 clade in the total DMSP-related genes, we believed that DMSP was mainly assimilated by bacteria from SAR11 subgroups Ia and Ib, which possess dmdA genes. Previous studies also reported that SAR11 bacteria would compete for labile dissolved organic matter compounds, including DMSP, and one-carbon (C₁) compounds (51). This might explain why the relative abundance of SAR11 Ia correlated negatively with that of SAR11 Ib. Competition within SAR11 clades could also appear in the dmdA gene abundance data (as evidenced by the negative correlation of dmdA subclade C/2 and subclade D genes), although it was not statistically significant.

Rhodobacteraceae showed much lower abundances in all samples than SAR11 bacteria. Samples from the upwelling area of the SP region and a few samples from the EP region exhibited relatively high abundances of Rhodobacteraceae. The relative abundances of Rhodobacteraceae were highly correlated with the relative abundances of dddP genes; this was reasonable because the Roseobacter-specific primer set was used to quantify dddP gene abundance. The slope of this correlation line was 4.3, suggesting that each marine Roseobacter cell possessed about 4 copies of dddP genes. This amount does not agree with previous studies that reported a single copy of dddP per cell (7, 8). One of the reasons for this bias might be the overestimation of dddP genes with previously designed primer sets. The marine Roseobacter bacteria were the main populations possessing dddP genes; however, other organisms (e.g., Marinomonas, Burkholderia, and Aspergillus) were also reported to possess dddP genes (12). Therefore, non-Roseobacter dddP genes might be amplified during qPCR quantification and cause the high copy numbers of dddP per cell. The primer set used to quantify 16S rRNA genes could also cause the overestimation of dddP gene ratios in total bacterial populations, with the possible underestimation of total bacterial abundance.

Multidimensional analysis separated the DMS hot spot samples into group 1 and group 2. Based on this separation, we assumed that the different oceanographic provinces consisted of different DMSP degradation gene structures. RDA analysis revealed that one of the most explainable environmental parameters influencing these DMSP degradation gene structures was the chlorophyll a concentration, which was highly correlated with the RDA1 axis. Phytoplanktons were the main DMSP producers (3, 40), and thus, the DMSP that they produced might influence the bacterial DMSP degradation gene structure in that area.

At group 1 sites, the DMS concentrations correlated positively with both the relative abundance of dddP genes and the ratio of the dddP genes to the total DMSP degradation genes [dddP/(dddP + total dmdA)]. These results indicated a possible coupling between the presence of bacterial DMSP cleavage pathways and DMS emission in these sites. In addition, the DMS concentrations were to some degree connected with the relative abundances of dddP in group 2 samples. We assumed that one of the possible reasons for this weak connection was the small sample size, because there were just eight samples from group 2 sites, compared with 12 samples from group 1 sites. Another possible reason was that marine bacteria possessing other ddd genes (or other undiscovered genes) were responsible for DMS production in this group.

Overall, the distributions of dmdA and dddP genes in the Pacific Ocean showed great spatial variations. These distribution patterns were possibly influenced by the chlorophyll a concentrations and temperature. SAR11 bacteria, which possessed dmdA genes, were suggested to be the main potential consumers of DMSP. The marine Roseobacter clade, which possessed dddP genes, was responsible for DMS production in the North and South Pacific oligotrophic ocean. The results of our study imply that shifts in bacterial community structure in response to changes in environmental factors can influence the ability of bacterial assemblages to metabolize DMSP, as well as the extent to which they do so, and regulate DMS emissions in subtropical gyres of the Pacific Ocean.