Longitudinal variability of diazotroph abundances in the subtropical North Atlantic Ocean

Longitudinal variability of diazotroph abundances in the subtropical North Atlantic Ocean MAR BENAVIDES1,2*, PIA H. MOISANDER3, MEAGHAN C. DALEY3, ANTONIO BODE4AND JAVIER ARI ´STEGUI2 1 AIX MARSEILLE UNIVERSITE ´,CNRS/INSU,UNIVERSITE ´DE TOULON,IRD,MEDITERRANEAN INSTITUTE OF OCEANOGRAPHY (MIO)UM 110,98848 NOUME ´A, NEW CALEDONIA, 2 INSTITUTO DE OCEANOGRAFI ´AYCAMBIO GLOBAL,UNIVERSIDAD DE LAS PALMAS DE GRAN CANARIA,35017 LAS PALMAS DE GRAN CANARIA, SPAIN, 3 DEPARTMENT OF BIOLOGY,UNIVERSITY OF MASSACHUSETTS DARTMOUTH,NORTH DARTMOUTH,MA 02747,USA AND 4 CENTRO OCEANOGRA ´FICO DE A CORUN ˜A,INSTITUTO ESPAN ˜OL DE OCEANOGRAFI ´A,APDO 130,15080 ACORUN ˜A,SPAIN *CORRESPONDING AUTHOR: mar.bena[email protected] Received September 17, 2015; accepted December 11, 2015 Corresponding editor: John Dolan Diazotrophy-related studies in the North Atlantic have largely focused on its western tropical area, leaving the subtropics and the east undersampled. We studied the longitudinal distribution of Trichodesmium, UCYN-A, UCYN-B, the putative Gammaproteobacterium g-24774A11 and Richelia (Het1) along 24.58N, using quantitative polymerase chain reaction on different size fractions (10, 10–3 and 3–0.2 mm) and additional filament counts for Trichodesmium. Trichodesmium was the most abundant phylotype, followed by UCYN-A, g-24774A11 and Het1, with maximum abundances of 8.8 !10 5 , 2.0 !10 5 , 3.3 !10 3 and 3.4 !10 2 nifH copies L 21 , respectively, whereas UCYN-B was mostly undetected. A clear shift in the diazotroph community was observed at !308W, coinciding with the transition between the North Atlantic Subtropical Gyre boundary and inner core. This transition zone divided the transect into an eastern half dominated by UCYN-A and western half dominated by Trichodesmium and g-24774A11. g-24774A11 was only detected in the 10–3 mm fraction, suggesting their association with larger microbes or aggregates. Our results indicate that typical size fractionation by 10 mm is not optimal for reconciling diazotroph phylotypes to N 2 fixation rates and that non-cyanobacterial diazotrophs may contribute importantly to bulk diazotrophic activity in the western subtropical North Atlantic. KEYWORDS: nifH; qPCR; N 2 fixation; upwelling; North Atlantic Subtropical Gyre available online at www.plankt.oxfordjournals.org #The Author 2016. Published by Oxford University Press. All rights reserved. For permissions, please email: [email protected] Journal of Plankton Research plankt.oxfordjournals.org J. Plankton Res. (2016) 0(0): 1–11. doi:10.1093/plankt/fbv121 JPR Advance Access published January 18, 2016 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from INTRODUCTION The biological fixation of atmospheric dinitrogen (N 2 ) provides significant inputs of reactive nitrogen to aquatic ecosystems and hence contributes to phytoplankton primary production (Karl et al., 2002). This process is performed by specialized prokaryotes called diazotrophs, distributed among broad phylogenetic groups from cyanobacteria and other bacteria to archaea (Zehr et al., 2003). The occurrence of N 2 fixation in the ocean was first described in the North Atlantic basin and attributed to the filamentous cyanobacterium Trichodesmium (Dugdale et al., 1961). In the following decades, it has become apparent that the diazotrophic community is also composed of diatom-diazotroph symbioses (DDAs), unicellular cyanobacteria (UCYN, divided into three groups UCYN-A, -B, -C) and non-cyanobacterial diazotrophs (including both bacteria and archaea) (Zehr et al., 2003). N 2 fixed by UCYN and DDAs at times locally exceeds that of Trichodesmium, highlighting the importance of these diazotrophs in N 2 fixation studies (e.g. Ratten et al., 2015). The different diazotrophic phylotypes are heterogeneously distributed in the North Atlantic Ocean because of geographical differences in the environmental factors that control them (Benavides and Voss, 2015). Trichodesmium is known for needing nitrate-poor, warm, stratified waters and is abundant in the tropical western Atlantic where these conditions persist (Capone et al., 2005). UCYN-A are small photoheterotrophic cyanobacteria (,1mm) not capable of carrying out the oxygen-evolving steps of photosynthesis or fixing inorganic carbon (Zehr et al., 2008), and they are commonly found in symbiosis with a eukaryotic alga (Thompson et al., 2012). UCYN-A seem to be able to fix N 2 at relatively low temperatures and high inorganic nitrogen concentrations (e.g. ,228C and .1.5 mM)(Turk et al., 2011;Ratten et al., 2015). UCYN-B are restricted to warm tropical waters (e.g. Moisander et al., 2010;Turk et al., 2011), whereas UCYN-C are often associated with coastal areas (Langlois et al., 2008;Turk et al., 2011). DDAs need a source of silica for the synthesis of the frustule of the diatom host; thus silica in the river plumes of the Amazon and Congo rivers may promote them (Foster et al., 2007;2009;Subramaniam et al., 2008). The distribution of non-cyanobacterial diazotrophs in the North Atlantic seems to be extensive, spanning a wide latitudinal range (Luo et al., 2012), although their ecology and environmental constraints are unknown (Farnelid et al., 2011), partly because of the difficulties in discerning cyanobacterial from non-cyanobacterial N 2 fixation rates in 15 N 2 tracer incubations (Benavides and Voss, 2015). Diazotrophic bacteria from the open ocean cluster with a wide range of phylogenetic groups, including Firmicutes, Alpha-, Beta-, Gammaand Deltaproteobacteria (Zehr et al., 2003). The Gammaproteobacteria-affiliated phylotype g-24774A11 was recovered from the South China Sea (Moisander et al., 2008), and the same diazotroph phylotype (also called Gamma A or UMB) has been recovered at high frequencies from several environments such as the Atlantic, Pacific and Indian Oceans and the Mediterranean Sea (Bird et al., 2005;Church et al., 2005a; Turk et al., 2011;Moisander et al., 2014;Shiozaki et al., 2014;Langlois et al., 2015). At least three different published quantitative polymerase chain reaction (qPCR) primer sets target this phylotype cluster that has some microdiversity (Moisander et al., 2014), and these primers have been termed Gammaproteobacteria (Church et al., 2005a), Gamma-A (Langlois et al., 2008) and g-24774A11 (Moisander et al., 2008). Each of these three primer sets targets a slightly different variant in the cluster (Moisander et al., 2014). The distribution and abundance of all of these diazotrophic phylotypes have been previously reported in the North Atlantic by qPCR assays of the nifH gene (which encodes for the nitrogenase iron protein component of the nitrogenase enzyme complex) (e.g. Langlois et al., 2008;Foster et al., 2009;Goebel et al., 2010). There is a geographic bias in the distribution of these studies, with more samples taken in the western than in the eastern North Atlantic Ocean (Montoya et al., 2007). Moreover, more sampling has been conducted in the tropics compared with higher latitudes, leaving the subtropical and temperate latitudinal zones with scarce data (Benavides and Voss, 2015). Most nifH data available in the North Atlantic are restricted to the latitudinal band located between the Equator and !158N, and there is a wide gap in observations in the areas comprised by 20–408N and 40–808W, as can be observed in Fig. 3 of Luo et al. (2012). This gap impedes a comprehensive assessment of the spatial distribution of different diazotroph species in the North Atlantic. With the aim of contributing to the few published studies on the longitudinal variability of diazotrophy in the North Atlantic (Vos s et al.,2004;Montoya et al.,2007), we quantified the abundance of the main diazotrophic phylotypes along 24.58Nfrom10to808Wandexaminedtheir geographical distributions in parallel with different environmental conditions and phylotype-specific constraints. METHOD Hydrography, nutrients and chlorophyll a The leg 8 of the Malaspina circumnavigation cruise took place from 27 January to 15 March 2011 onboard the JOURNAL OF PLANKTON RESEARCH j VOLUME 0 j NUMBER 0 j PAGES 1–11 j 2016 2 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from R/V Sarmiento de Gamboa,sailingwestwardsfromthe Northwest African coast to the North American coast along 24.58N(Fig.1). Temperature, salinity and chlorophyllfluorescence data were recorded with a SeaBird 911 plus CTD equipped with a Sea-Tech fluorometer. The CTD was mounted on a General Oceanics rosette frame equipped with 24 Niskin bottles (12 L volume). Samples for nutrient analysis were taken with the 12 L Niskin bottles, stored in 15 mL polypropylene tubes and immediately frozen at 2208C until analysis ashore. The concentrations of nitrate plus nitrite, phosphate and silicate were determined with a Technicon segmented-flow autoanalyser. Standard methods were modified to obtain a detection limit of 2 nmol L 21 (Raimbault et al., 1990). Details on the concentrations and distributions of nutrients can be found elsewhere (Benavides et al., 2013b; Mompea ´net al., 2013) and will be used here only with the purpose of interpreting the distribution of diazotrophs. The mixed layer depth (MLD) was estimated from an increase in water column density ( s t ) of 0.125 kg m 23 with respect to surface values (Benavides et al., 2013a). Chlorophyll a(Chl a;mgm 23 ) data were obtained from the National Aeronautics and Space Administration (NASA) Goddard Earth Sciences Data and Information Services Center Giovanni (NASA GES DISC) online database and averaged for the months of February and March 2011. DNA collection, extraction and qPCR assays Samples for DNA were taken at 15 stations along the transect (Table I). Surface seawater samples (!5 m) were collected using a 30 L Niskin bottle. Ten litres were transferred to acid-washed darkened carboys with a barbed faucet. Acid-washed vacuum tubing was connected to the faucet, and three separate filter holders (Whatman) were connected in-line containing 47 mm diameter white 10, 3 and 0.2 mm polycarbonate filters (GE-Osmonics Poretics). The end of the line was connected to a water jet aspirator vacuum pump (Eyela). The fractions retained were thus .10, ,10– .3 (hereafter abbreviated as 10–3 mm) and ,3–.0.2 mm (hereafter abbreviated as 3–0.2 mm), respectively. After filtration, the filters were transferred to sterile screwcap cryovials and stored at 2808C until analysis. DNA was extracted using the DNeasy Plant Mini Kit (Qiagen), as modified by Moisander et al. (2008). The abundance of diazotrophs was determined using TaqMan qPCR assays and previously published primer–probe sets for Trichodesmium,UCYN-A(Church et al.,2005a), UCYN-B (Moisander et al., 2010), the Gammaproteobacterium g-24774A11 (Moisander et al., 2008) and Het1 (RicheliaRhizosolenia DDAs) (Church et al., 2005b). Trichodesmium trichomes usually measure .80 mm and were therefore only analysed on the .10 mm fraction. Het1 on the .10 and 10–3 mm fractions, UCYN-A and UCYN-B were analysed in the 10–3 and 3–0.2 mm fractions, and the g-24774A11 were analysed in all three size fractions. The qPCR was run in 20 mL reactions that consisted of 10 mL ABI TaqMan Gene Expression Master Mix, 6.4 mL nuclease-free water, 0.5 and 0.25 mM final concentrations of primers and probe, respectively, and 1.6 mL DNA template. All samples were run in duplicate. Ten-fold dilutions of linearized plasmids containing the relevant nifH targets were used as standards and were included with each 96-well plate run. The reactions were run on a StepOnePlus Real-Time PCR system (Life Technologies). Standard curves were made by plotting linear regression of the threshold cycle (C t ) and log gene copies per reaction using duplicate standards ranging from 10 8 to 10 0 gene copies. Amplification efficiencies were .90% for all reactions. Duplicate no template control wells were included in all plates run and did not show amplification in any runs. Inhibition tests were carried out for all samples by combining the sample and 1.6 mL of 10 5 standard as a template. The efficiencies of inhibition tests ranged from 97.3 to 101.73%, and thus we consider that our samples were not inhibited. The limit of detection (LOD) and detected but not quantifiable (DNQ) limits used were one and eight gene copies per reaction, respectively. Samples that were below LOD were designated a value of 0 in the data set, whereas gene copies higher than LOD but less than DNQ were designated a conservative value of 1 nifH gene copy per litre. Fig. 1. Map of stations where CTD profiles were performed (“CTD stations”, labelled with an asterisk) and stations where samples for nifH analyses were taken (“nifH stations”, labelled with an open symbol and station number), superimposed on chlorophyll aconcentration data (Chl a; mg m 23 ). M. BENAVIDES ET AL. j DIAZOTROPHS IN THE SUBTROPICAL NORTH ATLANTIC OCEAN 3 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from Trichodesmium filament counts Samples were collected by towing a 40 mm mesh size plankton net from 200 m to the surface at 43 stations (Mompea ´net al., 2013), from which 6 stations (24, 45, 85, 91, 97 and 104; Fig. 1) were coincident with those where DNA was sampled. The volume of seawater filtered at each station was 14 m 3 .Trichodesmium filament counts were made on aliquots of the sample preserved in glutaraldehyde (25% final concentration) using a FlowCam system (Fluid Imaging Technologies), following the recommendations by A ´lvarez et al.(2014). Samples were run in the autoimage mode using a 4!lens (40!overall magnification) and a flow cell of 300 mm depth. Prior to analysis, the samples were screened with a 100 mm nylon mesh to prevent clogging of the FlowCam cell. Results are reported as number of trichomes per litre of seawater. N 2 fixation rates The fractionated (.10 and ,10 mm) N 2 fixation was assayed with the dissolved 15 N 2 method according to Mohr et al. (2010), as described in Benavides et al. (2013b). RESULTS Hydrography, nutrients and Chl a data Temperature and salinity were low near the coastal upwelling off Northwest Africa (!198C and 36.8, respectively; Table I). From east to west, temperature increased from ca. 20 to ca. 258C at 628W and then decreased slightly (to !238C) until the westernmost station. Fluorescence values were highest off the Northwest African coast (Table I) and decreased westwards mirroring the distribution of Chl a(Fig. 1). The surface concentrations of nitrate plus nitrite, phosphate and silicate were maximal off the Northwest African coast (12.61, 0.73 and 19.34 mM, respectively; Table I) and also decreased westwards along the transect. Further details on hydrographic variables measured along the transect can be found elsewhere (Benavides et al., 2013b;Mompea ´net al., 2013). The MLD was !80 m in the eastern end of the transect, and it deepened in the central part of the North Atlantic Subtropical Gyre (NASG), shoaling towards the western end of the transect with values !60 m (Table I). N 2 fixation rates (from Benavides et al., 2013b) were in a similar range in the .10 mm fraction (0.009– 0.385 nmol N L 21 h 21 ) and in the ,10 mm fraction (0.011–0.335 nmol N L 21 h 21 ), although the former were generally higher than the latter along the transect. N 2 fixation rates were highest off the Northwest African coast and maintained values !0.3 nmol N L 21 h 21 Table I: Measured values of core parameters (temperature, salinity and fluorescence) and inorganic nutrient concentrations (nitrate plus nitrite -NO 3"þNO2"-;silicate - SiO2";and phosphate - PO43"-;Benavides et al., 2013b;Mompea ´net al., 2013) at stations where diazotroph phylotypes were surveyed Station Date (dd/mm in 2011) Latitude (8N) Longitude (8W) Surface temperature (8C) Surface salinity Surface fluorescence (relative units) MLD (m) NO 3 2 þNO 2 2 (mM) PO 4 32 (mM) SiO 2 (mM) .10 mmN 2 fixation rates (nmol N L 21 h 21 ) a ,10 mmN 2 fixation rates (nmol N L 21 h 21 ) a 1 28/01 27.78 13.34 19.59 36.77 1.61 75.50 12.47 0.73 19.34 0.385 0.335 13 29/01 27.02 15.70 20.70 36.96 1.20 87.41 0.05 0.58 0.66 0.210 0.167 24 01/02 25.63 20.00 21.73 37.06 1.19 123.16 0.07 0.65 0.58 0.191 0.143 36 05/02 24.50 26.54 21.57 37.24 1.05 129.12 0.11 0.20 0.12 0.261 0.237 41 07/02 24.50 29.59 22.44 37.42 0.93 123.16 0.01 0.58 0.86 0.302 0.181 45 09/02 24.50 32.04 22.81 37.45 0.92 99.33 0.01 0.65 0.79 0.271 0.151 58 13/02 24.50 38.49 23.07 37.53 0.85 117.21 0.13 0.39 0.74 0.178 0.018 77 18/02 24.50 47.27 23.41 37.39 0.86 93.37 0.05 0.42 0.20 0.243 0.116 85 21/02 24.50 50.96 23.99 37.24 0.76 81.46 n/a n/a n/a 0.093 0.023 91 23/02 24.50 54.03 24.26 36.99 0.79 57.62 n/a n/a n/a 0.104 0.041 97 25/02 24.50 57.72 23.99 36.73 0.80 39.74 n/a n/a n/a 0.108 0.051 104 27/02 24.50 62.01 24.92 36.28 0.83 63.58 n/a n/a n/a 0.142 0.065 110 01/03 24.50 65.68 23.82 36.73 0.76 57.62 n/a n/a n/a 0.168 0.088 129 07/03 26.20 73.82 23.81 36.78 1.07 51.66 0.19 0.16 0.12 0.024 0.033 139 09/03 26.20 76.15 22.69 36.77 0.89 75.50 0.12 0.17 n/a 0.089 0.061 n/a indicates data not available. a From Benavides et al. (2013b). JOURNAL OF PLANKTON RESEARCH j VOLUME 0 j NUMBER 0 j PAGES 1–11 j 2016 4 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from westwards until !328W and then decreased steadily towards the western end of the transect (Table I). Longitudinal distribution of diazotrophic phylotypes The different diazotrophic phylotypes showed variability in their longitudinal distribution, with a shift in their patterns occurring at !308W (Fig. 2). Trichodesmium (only assayed in the .10 mm fraction) had the highest nifH copy L 21 numbers detected overall in this study, showing low abundances east of !308W, but increasing from !10 4 nifH copies L 21 to a maximum of 8.8 !10 5 nifH copies L 21 towards the western end of the transect (Fig. 2a). Trichodesmium trichomes (as counted by the FlowCam from samples recovered with a plankton net, see Methods) were absent between station 1 and !158W, peaked at !288W with !3!10 5 trichomes L 21 and then decreased steadily until 7–9 !10 3 trichomes L -1 at the westernmost stations of the transect (Fig. 2a). The putative Gammaproteobacterium g-24774A11 was detected in the .10 and 10–3 mm fractions, with maximum abundances in the former up to 3.3 !10 3 nifH copies L 21 , although their longitudinal distribution and abundance were very similar in both size fractions (Fig. 2a and b). In the .10 mm fraction, g-24774A11 showed a stable longitudinal pattern at 10 2 nifH copies L 21 , whereas in the 10–3 mm fraction, their distribution was similar to that of Trichodesmium, with non-detectable abundances at longitudes east of !308W and an increasing pattern west of this point, stabilizing at 10 2 nifH copies L 21 until 808W (Fig. 2b). g-24774A11 were not detected in the 3–0.2 mm fraction (Fig. 2c). Het1 were detected at low abundances (6.2 !10 1 –3.4 !10 2 nifH copies L 21 ; Fig. 2a and b) at stations located east of !308W, with the exception of station 104 at 628W, where they were detected at 3.8 !10 1 nifH copies L 21 (Fig. 2a). The abundance of UCYN-A decreased longitudinally from east to west (Fig. 2b and c). They were detected from the easternmost station until !508W in the 10–3 mm fraction (Fig. 2b), but only until !308W and at higher abundances (!10 4 nifH copies L 21 ) in the 3–0.2 mm fraction (Fig. 2c). UCYN-B were only detected in one sample (station 91, 548W) at a low abundance (!10 2 nifH copies L 21 ). Correlations with environmental variables and east-west differences The significant Spearman correlations (P,0.00042 after applying a Bonferroni correction for multiple comparisons; McDonald, 2014) between the abundance of diazotrophic phylotypes with environmental parameters (temperature, salinity, oxygen, fluorescence and nutrients) are shown in Table II. Some significant correlations were found between different diazotrophs. UCYN-A from the 3–0.2 mm fraction were negatively correlated with g-24774A11 from the 10–3 mm fraction and with Trichodesmium nifH copies (analysed only in the .10 mm fraction). Interestingly, Trichodesmium nifH copies were not significantly correlated with their corresponding trichome counts, although visually some correspondence was apparent in the two methods (Fig. 2a). The latter were instead correlated positively with g-24774A11 (in both the .10 and 10–3 mm fractions). Some diazotrophs also showed significant correlations with environmental parameters. A negative correlation was observed between temperature and the nifH gene copies of UCYN-A from the 3–0.2 mm fraction. Positive correlations were also observed between salinity and Trichodesmium trichome counts, and hence the correlation of the latter was negative with temperature (temperature and salinity were negatively correlated). UCYN-A recovered from the 3–0.2 mm fraction were positively correlated with oxygen and fluorescence (note that fluorescence and oxygen were correlated). No significant correlations were found with inorganic nutrients. If we divide the transect into two halves (west and east of 27.158W or station 37) and pool the observations Fig. 2. Abundance of the diazotrophic phylotypes (nifH copies L 21 ) Trichodesmium,g-24774A11, Het1 and UCYN-A retained on (a) the .10 mm, (b)3mm and (c) 0.2 mm fractions. The abundance of Trichodesmium trichomes (log 10 trichomes L 21 ) is superimposed on (a) and scale shown on the right y-axis. 163 !183 mm 2 (200 !200 DPI). M. BENAVIDES ET AL. j DIAZOTROPHS IN THE SUBTROPICAL NORTH ATLANTIC OCEAN 5 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from within each of the halves, steep gradients in nifH abundances are demonstrated: the abundance of UCYN-A in the 3–0.2 mm fraction, g-24774A11 in the 10–3 mm fraction and Trichodesmium nifH and filament counts were significantly different between the eastern and western halves (Wilcoxon test P¼0.05, 0.02, 0.006 and 0.009, respectively). DISCUSSION Longitudinal variability of diazotroph phylotypes Our results suggest that the subtropical Atlantic Ocean has a clear east-west divide at !308W, where the productivity of surface waters decreases as depicted by Chl a concentrations in Fig. 1, the deep chlorophyll maximum deepens and the presence of measurable phosphate concentrations within the euphotic zone disappears (Table I; see also Fig. 2 in Mompea ´net al., 2013). This is a physical separation caused by the transition from the upwellinginfluenced waters off the Northwest African coast to the NASG, which affects the structure and activity of planktonic communities (e.g. Herna ´ndez-Leo ´net al., 1999). Our data show that this physical barrier also affects diazotrophs, in a phylotype-specific manner. Trichodesmium was the most abundant diazotroph phylotype detected by qPCR assays along the transect, followed by UCYN-A and g-24774A11, in agreement with previous studies in the North Atlantic (e.g. Langlois et al., 2008;Ratten et al., 2015). Trichodesmium was found at low abundances east of !308W estimated by either qPCR or filament counts (Fig. 2a), agreeing with previous studies (Herna ´ndez-Leo ´net al., 1999;Ferna ´ndez et al., 2012), and increased towards the west in parallel with increasing temperature and a shoaling MLD (Table I), consistent with its requirements for high light and oligotrophic conditions (Carpenter and Roenneberg, 1995). Trichodesmium is known for most frequently thriving in warm, stratified and nitrate-poor waters (Capone et al., 2005), whereas in the colder and nutrient-richer waters of the eastern subtropical and tropical North Atlantic, it is almost absent or present at very low abundances and mainly found as free trichomes instead of colonies (Ferna ´ndez et al., 2010; Goebel et al., 2010;Benavides et al., 2011). In this study, we determined the abundance of Trichodesmium in two ways: by quantification of nifH gene copies in seawater samples collected from the 5 m depth with Niskin bottles and filtered onto 10 mm filters (.10 mm fraction) and by vertically sampling the water column through the top 200 m using a 40 mm mesh net, followed by counting trichomes with a FlowCam (see Methods). Although the overall east-west trend was Table II: Significant (P,0.00042) Spearman correlations (positive or negative) between diazotroph abundances, environmental variables and nutrient concentrations UCYN-A 3mm UCYN-A 0.2 mm g-24774A11 10 mm g-24774A11 3mm Trichodesmium nifH Trichodesmium counts Het1 10 mm Het1 3mm Temperature Salinity Oxygen Fluorescence Nitrate þnitrite Phosphate Silicate UCYN-A 3 mm UCYN-A 0.2 mm22 2 þþ g-24774A11 10 mm þþ g-24774A11 3mm þ Trichodesmium nifH Trichodesmium counts þ Het1 10 mm Het1 3 mm Temperature 22 2 Salinity þ Oxygen Fluorescence Nitrateþnitrite Phosphate þ Silicate JOURNAL OF PLANKTON RESEARCH j VOLUME 0 j NUMBER 0 j PAGES 1–11 j 2016 6 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from similar, Trichodesmium nifH copy numbers did not correlate significantly with Trichodesmium trichome counts (Table II and Fig. 2a). When dividing Trichodesmium nifH copy numbers by trichome counts at stations where both types of data were available, each trichome would have contained between 0.7 and 4.5 nifH copies (with an exception of 19 nifH copies per filament at station 97; data not shown). This is one to two orders of magnitude fewer than would be expected, given that a 100 mm filament has approximately 70 cells (Tyrrell et al., 2003). The number of cells per trichome is variable (usually up to 100 cells, but finding broken filaments with fewer cells is common; Benavides et al., 2011). Parallel comparison of the two counting methods was only possible for six stations, and the differences between these methods may have affected the results considerably. Plankton nets are generally preferred over Niskin bottles for sampling Trichodesmium when its abundance is low (Chang, 2000), given the enormous difference in the volume of seawater filtered in each case (tens of litres from Niskin bottles versus hundreds to thousands of litres concentrated when using net tows). The differences observed between the two counting methods are also likely associated with the sampling depths covered (5 m versus integrated over 200 m), as well as the vertical migration of Trichodesmium colonies (Villareal and Carpenter, 2003), and/or their patchy distribution in the water column (Carpenter et al., 2004;Davis and McGillicuddy, 2006). UCYN-A detected in the 10–3 mm fraction were more abundant east of !308W and only detected in the 3–0.2 mm fraction from samples collected between !308W and the Northwest African coast (Fig. 2b and c). UCYN-A were the second most abundant phylotype in our study, highlighting their potential role in bulk N 2 fixation activity. UCYN-A usually appear in symbiosis with a prymnesiophyte from which it obtains organic carbon (Thompson et al., 2012), although it has also been found in association with larger plankton cells such as diatoms and dinoflagellates (Thompson et al., 2014), and at times as free-living cells, likely due to their loose association with the surface of their host (Thompson et al., 2012; Krupke et al., 2014). These different association possibilities, as well as detachment during sample handling, would explain the presence of UCYN-A in both the 10–3 and the 3–0.2 mm fractions in our study (Fig. 2b and c), which is consistent with previous reports (Agawin et al., 2014). Unidentified diazotrophic cells in the UCYN-A size range were also observed in aggregates using whole-cell hybridization techniques (Benavides et al., 2013a), which could contribute to their gene copies being found in the 10–3 mm size fraction in this study. The association of UCYN-A with phytoplankton cells also likely explains their apparent preference for colder and nutrient-rich waters compared with other diazotrophs (Moisander et al., 2010). In the North Atlantic, UCYN-A have been found in waters with temperatures ranging from !128C or lower to 238C(Langlois et al., 2008;Rees et al., 2009;Bentzon-Tilia et al., 2014;Krupke et al., 2014;Scavotto et al., 2015). UCYN-A have also been reported as the dominant diazotroph in the upwelling waters off the Northwest coast of Africa with temperatures as low as 178C and high nutrient concentrations (Agawin et al., 2014), as well as over the Equatorial upwelling waters where similar conditions are found (Foster et al., 2009). Although active UCYN-A (as nifH transcripts) were not found in these cold and nutrient-rich waters (e.g. Foster et al., 2009), the higher N 2 fixation activity off the Northwest coast of Africa (Benavides et al., 2013b) and in the Equatorial upwelling where they dominate (Subramaniam et al., 2013) suggests that they contribute importantly to fixed nitrogen inputs in these areas. In our study, UCYN-A were detected at .10 4 nifH gene copies L 21 in waters with temperatures ranging between !19 and 218C, with very high nutrient concentrations (e.g. nitrate plus nitrite .12 mM; Table I). However, although these diazotrophs have been detected in higher latitude coastal waters (off the Northeast American coast; Mulholland et al., 2012), their ubiquitous presence in oceanic waters at latitudes .30–408N was suggested to be unlikely (Krupke et al., 2014). The clear longitudinal segregation in UCYN-A abundances may also be driven by iron-stress alleviation via Saharan dust inputs, which are higher in this longitudinal range of the transect (Benavides et al., 2013b) and were reported to limit N 2 fixation by UCYN-A (Krupke et al., 2015). Non-cyanobacterial diazotrophs are cosmopolitan, and their activity may be driven by factors other than those controlling cyanobacterial diazotrophs (Farnelid et al., 2011). Interestingly, during our study, g-24774A11 nifH copies were detected at abundances comparable to those of UCYN-A (Fig. 2). The g-24774A11 have been shown to consistently express their nifH gene in the open ocean (Moisander et al., 2014), suggesting active N 2 fixation. Thus, if their per cell N 2 fixation activities are as high as with UCYN-A, with equal cell numbers they might have a comparable contribution to in situ N 2 fixation rates. The g-24774A11 were found exclusively in the .10 and 10–3 mm size fractions, implying the attachment of these diazotrophs to larger microbes, particulate material or a self-production of organic aggregates. The association of non-cyanobacterial diazotrophs with particles as a means of obtaining organic nutrients has been previously hypothesized, and gammaproteobacterial representatives have been found forming aggregates in oxygenated cultures (Bentzon-Tilia et al., 2015). M. BENAVIDES ET AL. j DIAZOTROPHS IN THE SUBTROPICAL NORTH ATLANTIC OCEAN 7 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from A particle-attached mode of life may provide several advantages for planktonic bacterial diazotrophs, such as oxygen-depleted microzones and the availability of organic matter (Riemann et al., 2010). However, the longitudinal distribution of Chl aalong the transect (Fig. 1) suggests that organic particles were more abundant east of !308W than on the western half of the transect, which is opposite to the distribution of g-24774A11 in the .10 and 10–3 mm fractions (Fig. 2a and b). This makes the self-production of organic aggregates the most plausible explanation for their longitudinal distribution in our study, which is consistent with a previous study (Bentzon-Tilia et al., 2015). Nevertheless, the association of diazotrophic Gammaproteobacteria with pelagic particles has not been studied directly, and we cannot assume that the colonization of particles by these microbes is equal in space and time; it is possible that it changes with particle size, chemical composition, among other characteristics. The filtration method we used may have contributed to the observation of high abundances of g-24774A11 in the .10 and 10–3 mm size fractions. In our past studies in which a peristaltic pump system was used, this phylotype was not detected in the .10 mm size fraction (Moisander et al., unpublished data). It is possible that the water jet vacuum aspirator filtration method used here is more gentle than the peristaltic pump system and generates less bacterial detachment from particles or aggregates. The high abundance of g-24774A11 detected together with previous studies (Moisander et al., 2014;Langlois et al., 2015) underscores the current need to quantify the contribution of noncyanobacterial diazotrophs to global N 2 fixation rates, as well as to elucidate their ecology and physiological constraints. At present, it is our inability to discern cyanobacterial from non-cyanobacterial N 2 fixation activity that keeps us from further estimating the global importance of the latter (Benavides and Voss, 2015), although the measurable diazotrophic activity in aphotic waters (e.g. Bonnet et al., 2013), as well as the enhancement of N 2 fixation rates upon the addition of photosystem II-blocking agents to stop the activity of autotrophic diazotrophs (Rahav et al., 2015), suggests that their activity may be significant. Similar to UCYN-A, Het1 diazotrophs (Richelia) depend on the ecological constraints of their host (Rhizosolenia diatoms). Over the transect surveyed in this study, Het1 were found at low abundances east of !308W in the NASG boundary area (Fig. 2), coinciding with the highest phosphate and silicate concentrations of the transect because of the proximity of the Northwest African coastal upwelling (Table I). Het1 are usually restricted to a river plume or coastal runoff-affected areas such as the Amazon and the Congo River plumes (Foster et al., 2007,2009) and the North American coast at higher latitudes (e.g. .308N; Ratten et al., 2015). Their maximum abundances measured during our study (!10 2 nifH copies L 21 ; Fig. 2) are below those in the Amazon and Congo River plumes (up to 10 5 and 10 3 nifH copies L 21 , respectively; Foster et al., 2007,2009), and other studies performed in open-ocean waters of the North Atlantic have generally also found low abundances (,10 1 nifH copies L 21 ;Luo et al., 2012; also see Fig. 3 in Benavides and Voss, 2015). Overall, the low abundances and the low frequency of detection among samples over the transect suggest that their contribution to bulk N 2 fixation in this area is minimal. It also needs to be noted that although the nifH gene copies are detected in each cell of heterocystous diazotrophs, only some of the cells are heterocysts and thus N 2 -fixing. If only the number of N 2 -fixing cells is considered, the numbers reported here should be divided by the estimated vegetative cell: heterocyst frequency, which for Richelia has been reported to be approximately 5 (Foster et al., 2009). Taking into account, this conversion would reduce the contribution of this group to N 2 fixation further. Relationship between the diazotrophic community and N 2 fixation rates along 24.58N The longitudinal variability of N 2 fixation activity in the North Atlantic (encompassing both its eastern and western basins) has only been addressed in a few studies (Voss et al., 2004;Montoya et al., 2007;Benavides et al., 2013b;Ratten et al., 2015). The size-fractionated (.10 and ,10 mm) N 2 fixation rates corresponding to water samples investigated in this study have been published previously (Benavides et al., 2013b) and are also shown in Table I.N 2 fixation rates in both fractions decreased from east to west from !0.4 to 0.01 nmol N L 21 h 21 . This pattern is opposite to what would be expected from aTrichodesmium-dominated community; indeed, higher rates towards the western Atlantic have been observed previously in studies focusing on these cyanobacteria (e.g. Capone et al., 2005), but other bulk water N 2 fixation studies found an opposite longitudinal pattern (e.g. Voss et al., 2004). Benavides et al. (2013b) divided the transect into eastern and western halves at about !458W, coinciding with a change in sea surface height values. Despite this division that resulted in significantly different N 2 fixation rates between the eastern and western halves of the transect, the visual inspection of N 2 fixation rates (Table I) shows that the steepest change in N 2 fixation activity occurs at !328W, agreeing with the shift observed in the composition of the diazotrophic community in this study (Fig. 2). JOURNAL OF PLANKTON RESEARCH j VOLUME 0 j NUMBER 0 j PAGES 1–11 j 2016 8 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from Prior reports suggest that in tropical latitudes, UCYN-A dominate the N 2 fixation activity in the eastern North Atlantic basin, whereas Trichodesmium dominate in the western basin (Montoya et al., 2007;Benavides and Voss, 2015). This study suggests that a similar trend in nifH phylotype distributions of UCYN-A and Trichodesmium is found also in subtropical latitudes. The distribution of nifH copies in the different size fractions measured suggests that in subtropical latitudes Gammaproteobacteria may also contribute importantly to .3mmN 2 fixation in the western North Atlantic basin. The presence of small diazotrophs such as Gammaproteobacteria in the .10 and 10–3 mm size fractions suggests that additional size fractionation of samples incubated with 15 N 2 with 3 mm filters could provide further information about groups contributing to the N 2 fixation rates. Over our transect, ,10 mmN 2 fixation rates contributed on average 40% to bulk N 2 fixation (Benavides et al., 2013b), suggesting an important role of the diazotrophs captured on the 10–3 and 3–0.2 mm fractions and quantified in this study. Moreover, ,10 mmN 2 fixation rates were significantly higher in the eastern half of the transect when compared with the western half, supported by the high abundance of UCYN-A detected (Fig. 2). In summary, N 2 fixation rates may be driven primarily by UCYN-A in the eastern part of the transect and by a combination of Trichodesmium and g-24774A11 in the western part. CONCLUSIONS This study covered two areas representative of the conditions at the boundary and centre of the NASG and provides a description of the diazotrophic community in an undersampled area of the North Atlantic Ocean. Our results indicate that the transition between the boundary and central NASG creates a separation in the composition of the diazotrophic community, with UCYN-A dominating the eastern half of the Atlantic and Trichodesmium and g-24774A11 Gammaproteobacteria dominating the western part. The detection of g-24774A11 in the .10 and 10–3 mm fractions and not in the 3–0.2 mm fraction indicates that these diazotrophs occur in association with either larger microbes, existing marine particles, or selfproduced mucilage matrices; to our knowledge, this has not been previously reported. The filtration method we used may be more gentle than a peristaltic pump system, having less bacterial detachment from particles. These results also suggest that size-fractionation studies are not optimal for identifying the different diazotroph phylotypes responsible for the measured N 2 fixation rates and support past observations that non-cyanobacterial diazotrophs are present at considerable abundances in openocean waters, thus deserving further study. ACKNOWLEDGEMENTS We thank the crew and technicians of the R/V Sarmiento de Gamboa and the chief scientist for facilitating the collection of samples for this study. We are grateful to J. Lorenzo and J. Varela for sampling and counting Trichodesmium filaments. FUNDING This work was supported by projects Consolider-Malaspina (CSD2008-00077), CAIBEX (CTM2007-66408CO2-02) and HOTMIX (CTM2011-30010-CO2-01) to J.A. M.B. was supported by a postdoctoral fellowship from the People Programme (Marie Skłodowska-Curie Actions) of the European Union’s Seventh Framework Programme (FP7/ 2007-2013) under REA grant agreement number 625185. P. H . M . w a s s u p p o r t e d b y N S F O C E 1 1 3 0 4 9 5 a n d f u n d s from the University of Massachusetts Dartmouth. REFERENCES Agawin, N., Benavides, M., Busquets, A., Ferriol, P., Stal, L. J. and Arı ´stegui, J. (2014) Dominance of unicellular cyanobacteria in the diazotrophic community in the Atlantic Ocean. Limnol. Oceanogr.,59, 623–637. A ´lvarez, E., Moyano, M., Lo ´pez-Urrutia, A., Nogueira, E. and Scharek, R. (2014) Routine determination of plankton community composition and size structure: a comparison between FlowCAM and light microscopy. J. Plankton Res.,36, 170–184. Benavides, M., Agawin, N., Arı ´stegui, J., Ferriol, P. and Stal, L. J. (2011) Nitrogen fixation by Trichodesmium and small diazotrophs in the subtropical northeast Atlantic. Aquat. Microb. Ecol.,65, 43–53. Benavides, M., Arı ´stegui, J., Agawin, N. S. R., Lo ´pez Cancio, J. and Herna ´ndez-Leo ´n, S. (2013a) Enhancement of nitrogen fixation rates by unicellular diazotrophs vs. Trichodesmium after a dust deposition event in the Canary Islands. Limnol. Oceanogr.,58, 267–275. Benavides, M., Bronk, D. A., Agawin, N. S. R., Pe ´rez-Herna ´ndez, M. D., Herna ´ndez-Guerra, A. and Arı ´stegui, J. (2013b) Longitudinal variability of size-fractionated N 2 fixation and DON release rates along 24.58N in the subtropical North Atlantic. J. Geophys. Res. Oceans, 118, 3406–3415. Benavides, M. and Voss, M. (2015) Five decades of N 2 fixation research in the North Atlantic Ocean. Front. Mar. Sci.,2, 1–40. Bentzon-Tilia, M., Severin, I., Hansen, L. H. and Riemann, L. (2015) Genomics and ecophysiology of heterotrophic nitrogen-fixing bacteria isolated from estuarine surface water. mBio,6, e00929-–15. Bentzon-Tilia, M., Traving, S. J., Mantikci, M., Knudsen-Leerbeck, H., Hansen, J. O. R. L., Markager, S. and Riemann, L. (2014) Significant N 2 fixation by heterotrophs, photoheterotrophs and heterocystous cyanobacteria in two temperate estuaries. ISME J.,9, 273–285. M. BENAVIDES ET AL. j DIAZOTROPHS IN THE SUBTROPICAL NORTH ATLANTIC OCEAN 9 at ULPGC - Universidad de Las Palmas de Gran Canaria on January 19, 2016http://plankt.oxfordjournals.org/Downloaded from