Dynamics and controls of heterotrophic prokaryotic production in the western tropical South Paciﬁc Ocean: links with diazotrophic and photosynthetic activity

. Heterotrophic prokaryotic production (BP) was studied in the western tropical South Paciﬁc (WTSP) using the leucine technique, revealing spatial and temporal variability within the region. Integrated over the euphotic zone, BP ranged from 58 to 120 mg C m − 2 d − 1 within the Melanesian Archipelago, and from 31 to 50 mg C m − 2 d − 1 within the western subtropical gyre. The collapse of a bloom was followed during 6 days in the

F. Van Wambeke et al.: Dynamics and controls of heterotrophic prokaryotic production terial carbon demand (BCD; i.e., the sum of heterotrophic bacterial production (BP) and bacterial respiration (BR)) is compared to primary production (PP). The metabolic state of the ocean, and in particular the status of net heterotrophy within oligotrophic systems, has been largely debated in the last decade (see for example the review in Duarte et al., 2013;Ducklow and Doney, 2013;Williams et al., 2013).
The South Pacific Gyre (GY) is ultra-oligotrophic, and is characterized by deep UV penetration, by deep chlorophyll maximum (dcm) depth as great as 200 m, and by a 0.1 µM nitrate (NO 3 ) isocline at 160 m Halm et al., 2012). Our knowledge of the South Pacific Ocean's metabolic state based on C fluxes is fragmentary, since only few primary production data have previously been reported, and never simultaneously with BP (see references in Table 1). The exception is the BIOSOPE cruise conducted in the GY and eastern tropical South Pacific (ETSP) in November/December 2004, where both PP and BP were estimated simultaneously (Van Wambeke et al., 2008b).
The waters coming from the GY are essentially transported by the South Equatorial Current toward the Melanesian archipelagos in the western tropical South Pacific (WTSP). Interest in this region has increased due to field and satellite observation showing intermittent phytoplankton blooms in the area associated with Trichodesmium (Dupouy et al., 1988(Dupouy et al., , 2011Tenorio et al., 2018). The WTSP is a highly dynamic region (Rousselet et al., 2017) where patches of chlorophyll blooms can persist for up to a few weeks . The WTSP is a hotspot for biological nitrogen fixation (N 2 fix, Bonnet et al., 2017), extending to this whole oceanic region what was already measured locally near New Caledonia (Garcia et al., 2007). Based on nitrogen budgets, such blooms can sustain significant new production and export in this area . The development of these blooms is explained by different hypotheses, including temperature thresholds (in particular regulating Trichodesmium blooms); increased light providing more energy; the stratification of surface waters favoring depletion of nitrate and reducing competition with non-fixing primary producers; and increased availability of iron and phosphate (DIP) due to island mass effects, volcanic activities or atmospheric nutrient deposition (Moutin et al., 2005(Moutin et al., , 2008Luo et al., 2014;Martino et al., 2014;Shiozaki et al., 2014.
While the dynamics of heterotrophic prokaryotes coupling with primary producers has been explored in many regions of the ocean, these processes have not been studied in the WTSP. Because most oligotrophic oceans are nitrogen limited, PP and N 2 fix have already been sampled simultaneously in diverse studies and their relationships examined. Taking a Redfield ratio of 6.6, the contribution of N 2 fix rates to PP, integrated over the euphotic zone, has been found to range from 1 to 9 % in diverse provinces of the Atlantic (Fonseca-Batista et al., 2016). The ratio is 15-21 % in the WTSP and 3-4 % in the center of the GY Caf-fin et al., 2018). Few studies have attempted to examine how the variability of N 2 fix can be linked to that of heterotrophic activity, or to identify the contribution of N 2 fix rates to heterotrophic prokaryotic N demand. Yet, recent genomic analyses exploring the diversity of the nitrogenase reductase (nifH) gene have revealed the importance of non-cyanobacterial nitrogen fixers (Gradoville et al., 2017, and references therein). Owing to the fact that a great abundance of nifH gene copies does not imply that N 2 fix is occurring (see for example Turk-Kubo et al., 2014), diverse tests have been conducted to assess heterotrophic N 2 fixation indirectly. For example, in the oligotrophic eastern Mediterranean Sea, aphotic N 2 fix can account for 37 to 75 % of the total daily integrated N 2 fix rates (Rahav et al., 2013). In the Red Sea, N 2 fix rates are correlated with BP but not with PP during the stratified summer season, while during a Trichodesmium bloom in winter, both PP and BP increased with N 2 fix rates, although the correlation was still insignificant with PP (Rahav et al., 2015). In the South Pacific, the presence of non-cyanobacterial nitrogen fixers has been detected in the dark ocean as well as in the euphotic layer, with detectable levels of nifH gene expression, as measured by qPCR or N 2 fix activity determined in darkness (in the GY: Halm et al., 2012;Moisander et al., 2014; in the eastern tropical South Pacific Ocean:  in the Bismarck and Solomon seas: . The addition of selected organic molecules such as glucose  or natural organic matter such as transparent expolymer particles, can also influence N 2 fix rates . Finally, recent experiments based on incubation with 15 N-labeled N 2 coupled to nano-SIMS analyses also demonstrated that a rapid transfer, at the scale of 24 to 48 h, can occur between N 2 fixers, non-fixing phytoplankton and heterotrophic prokaryotes . In this study, we examined the horizontal and vertical distributions of heterotrophic prokaryotic production alongside photosynthetic rates, N 2 fix rates and phosphate turnover times across the WTSP, in order to relate these fluxes to bottom-up controls (related to nitrogen, phosphorus and labile C availability). Particular attention was given to determine the coupling between BP and PP or N 2 fix rates, to examine the variability of bacterial carbon demand (BCD) in comparison to gross primary production (GPP) ratios and to discuss the metabolic state of this region.

Sampling strategy
The OUTPACE cruise (https://doi.org/10.17600/15000900) was conducted in the WTSP region, from 18 February to 3 April 2015, along a transect extending from the north of New Caledonia to the western part of the South Pacific Gyre (WGY) (25 • S, 115 • E-15 • S, 149 • W, Fig. 1). For details on the strategy of the cruise, see Moutin et al. (2017). Stations of short duration (< 8 h, 15 stations named SD1 to SD15, Fig. 1) and long duration (6 days, 3 stations named LDA to LDC) were sampled. Generally, at least three CTD casts going down to 200 m were conducted at each short station, except at SD5 and SD9 (two casts) and at SD13 (one cast). The LD stations were abbreviated LDA (situated north of New Caledonia), LDB (in the Vanuatu area) and LDC (oligotrophic reference in the GY area). LD stations were selected based on satellite imagery, altimetry and Lagrangian diagnostics , and by the abundance of selected diazotrophs nifH gene copies analyzed by quantitative polymerase chain reaction (qPCR), in real time on board . At these LD stations, CTD casts were performed every 3 h following the water mass using a Lagrangian sampling strategy during at least 5 days. All samples were collected from a CTD-rosette system fitted with 20 12 L Niskin bottles and a Sea-Bird SBE9 CTD. At the SD stations, water samples used for measuring in situ simulated primary production (PP deck ), dissolved inorganic phosphate turnover times (the ratio of DIP concentration to DIP uptake rate, T DIP ) and N 2 fix rates came from the same rosette cast as water used for measuring BP. At the LD sites, we also conducted biodegradation experiments to determine bacterial growth efficiency (BGE), along with enrichment experiments to explore the factors limiting BP.
In addition to the measurements of chlorophyll a, BP, PP, T DIP and DOC described below, other data presented in this paper include hydrographic properties, nutrients and N 2 fix, for which detailed protocols of analysis and considerations for methodology are available in Moutin et al. (2017Moutin et al. ( , 2018 and Bonnet et al. (2018). Briefly, DIP and nitrate concentrations were measured using standard colorimetric procedures on a AA3 AutoAnalyzer (Seal-Analytical). The quan-tification limits were 0.05 µM for both nutrients. N 2 fixation rates were measured using the 15 N 2 tracer method in 4.5 L polycarbonate bottles inoculated with 5 mL of 15 N 2 gas (99 atom % 15 N, Eurisotop). Note that the risk of underestimation by this bubble method was checked by subsampling and fixing 12 mL of each bottle after incubation and to analyze the dissolved 15 N 2 with a Membrane Inlet Mass Spectrometer.

Chlorophyll a
For chlorophyll a (Chl a), a sample of 288 mL of seawater was filtered through 25 mm Whatman GF/F filters immediately after sampling and placed at −80 • C in Nunc tubes until analysis. At the laboratory (3 months after the cruise), after grinding the GF/F filter in 5 mL methanol, pigments (Chl a and phaeophytin) were extracted in darkness over a 2 h period at 4 • C and analyzed with a Trilogy Turner 7200-000 fluorometer according to Le Bouteiller et al. (1992). Sampling for Chl a analysis started only at site LDA (Dupouy et al., 2018). In vivo fluorescence was performed with an Aqua-Traka III (ChelseaTechnologies Group Ltd) sensor mounted on the CTD.
The overall correlation between in vivo fluorescence (Chl iv) and Chl a was very patchy (Chl a = 1.582 × Chl iv + 0.0241, n = 169, r = 0.61). This is due to the heterogeneity at the time of sampling and the nature of the populations present, i.e., essentially different fluorescence yields over depth and species (Neveux et al., 2010). Thus in vivo fluorescence was used only to track high-frequency variability at the LD sites, the shape of a vertical profile's distributions and the location of the dcm, as well as longitudinal trends. Fluorometric discrete data (Chl a) were used for calculating and comparing integrated Chl a stocks. Bacterial production (BP, sensus stricto referring to heterotrophic prokaryotic production) was determined onboard using the microcentrifuge method with the 3 H-leucine ( 3 H-Leu) incorporation technique to measure protein production (Smith and Azam, 1992). Triplicate 1.5 mL samples and a killed control with trichloracetic acid (TCA) at 5 % final concentration were incubated with a mixture of [4,5-3 H]leucine (Amersham, specific activity 112 Ci mmol −1 ) and nonradioactive leucine at final concentrations of 7 and 13 nM, respectively. Samples were incubated in the dark at the respective in situ temperatures for 1-4 h. On nine occasions, we checked that the incorporation of leucine was linear with time. Incubations were ended by the addition of TCA to a final concentration of 5 %, followed by three runs of centrifugation at 16 000 g for 10 min. Bovine serum albumin (BSA, Sigma, 100 mg L −1 final concentration) was added before the first centrifugation. After discarding the supernatant, 1.5 mL of 5 % TCA was added before the second centrifugation. For the last run, after discarding the supernatant, 1.5 mL of 80 % ethanol was added. The ethanol supernatant was then discarded and 1.5 mL of liquid scintillation cocktail (Packard Ultimagold MV) was added. The radioactivity incorporated into macromolecules was counted using a Packard LS 1600 Liquid Scintillation Counter onboard the ship. A factor of 1.5 kg C mol leucine −1 was used to convert the incorporation of leucine to carbon equivalents, assuming no isotopic dilution (Kirchman, 1993). Indeed, isotopic dilution ranged from 1.04 to 1.18, as determined in five experiments where we checked the saturating level of 3 H-leucine. Standard deviation associated with the variability between triplicate measurements averaged 12 and 8 % for BP values lower and higher than 10 ng C L −1 h −1 , respectively. At the LD sites, BP was sampled every day at 12:00 local time.

Primary production and phosphate turnover times
Primary production (PP) and dissolved inorganic phosphate turnover times (T DIP ) were determined using a dual 14 C-33 P labelling technique following Duhamel et al. (2006) and described in Moutin et al. (2018). Briefly, after inoculation with 10 µCi of 14 C sodium bicarbonate and 4 µCi of 33 Porthophosphoric acid, 150 mL polycarbonate bottles were incubated in on-deck incubators equipped with blue screens (75,54,36,19, 10, 2.7, 1, 0.3 and 0.1 % incident light, https://outpace.mio.univ-amu.fr/spip.php?article135, last access: 27 April 2018) and flushed continuously with surface sea water. Incubation times lasted 4 (western stations) to 24 h (South Pacific Gyre area) and were chosen according to expected T DIP . Samples were then filtered through 0.2 µm polycarbonate membranes, with radioactivity retained by the filters being assessed by liquid scintillation counting directly on board and after 12 months in the laboratory. Rates of daily primary production were computed using the conversion fac-tors τ (T i;T ) according to Moutin et al. (1999) to calculate normalized (dawn-to-dawn) daily rates from the incubation period measured in the on-deck incubators (PP deck ).
Measurements of PP using the JGOFs protocol (in situ moored lines immerged for 24 h from dawn-to-dawn, IPP in situ ) were also performed at each long station on days 1, 3 and 5 (see Caffin et al., 2018, for details). Integrated rates within the euphotic zone were estimated by trapezoidal integrations, assuming the same rate between 0 m and the shallowest sampled depth and considering PP to be zero at 20 m below the deepest sampled depth.

Bacterial growth efficiency and dark community respiration
Bacterial growth efficiency (BGE) and DOC lability were estimated at the three LD sites using dilution experiments with seawater sampled in the mixed layer. The seawater used for these experiments was sampled from Niskin bottles (9 m at LDA, 7 m at LDB and 16 m at LDC) from a CTD cast done at 12:00 local time on the first day of occupation at each LD site. A 1/5 dilution culture was established by mixing a bacterial inoculum from the same seawater sample (0.4 L of a < 0.8 µm filtrate) with 1.6 L of < 0.2 µm filtrate, in a borosilicate bottle. Samples were incubated in the dark, for up to 10 days, in a laboratory incubator set at in situ temperature. Periodically, for up to 10 days, subsamples were taken to estimate DOC concentrations and bacterial production. The BGE was estimated from DOC and bacterial production estimates on a given time interval corresponding to the exponential phase of BP following Eq. (1): where BP int is the trapezoidal integration of BP with time for the period considered, and DOC cons the dissolved organic carbon consumed during that period, corresponding to the difference in DOC concentration between initial (DOC initial ) and minimal DOC (DOC min ). From these experiments, the labile fraction of DOC (LDOC) was determined following Eq. (2): Samples for DOC concentration were filtered through two precombusted (24 h, 450 • C) glass fiber filters (Whatman GF/F, 25 mm) using a custom-made all-glass/Teflon filtration syringe system. Samples were collected into precombusted glass ampoules and acidified to pH 2 with phosphoric acid (H 3 PO 4 ). Ampoules were immediately sealed until analyses by high-temperature catalytic oxidation (HTCO) on a Shimadzu TOC-L analyzer (Cauwet, 1999). Typical analytical precision is ±0.1-0.5 (SD) or 0.2-0.5 % (CV). Consensus reference materials (http://www.rsmas.miami.edu/ groups/biogeochem/CRM.html, last access: 27 April 2018) were injected every 12 to 17 samples to ensure stable operating conditions. Rates of dark community respiration (DCR) were used to estimate bacterial growth efficiency (see the discussion). Briefly, DCR was estimated from changes in the dissolved oxygen (O 2 ) concentration during dark incubations of unfiltered seawater (24 h) carried out at LD stations, in situ on the same mooring lines used for PP in situ . Quadruplicate biological oxygen demand bottles were incubated in the dark at each sampled depth. The concentration of oxygen was determined by Winkler titration (Williams and Jenkinson, 1982). DCR was calculated as the difference between initial and final O 2 concentrations. The mean standard error of volumetric DCR rates was 0.28 µmol O 2 L −1 d −1 .

Enrichment experiment
Enrichment experiments were performed along vertical profiles at the three LD sites LDA, LDB and LDC. Seawater was sampled at 12:00 local time on day 2 of occupation at each site (CTD cast numbers 33, 117 and 166, respectively). Nutrients were added in 60 mL transparent polycarbonate bottles at a final concentration of 1 µM NH 4 Cl + 1µM NaNO 3 in "N" amended bottles, 0.25 µM Na 2 HPO 4 in "P" amended bottles and 10 µM C-glucose in "G" amended bottles. The sum of all these elements was added in "NPG" amended bottles. Controls "C" were left unamended. Bottles were incubated on average for 24 h under simulated in situ conditions (in the same on-deck incubators as those used for PP deck ). Selected depths chosen encompassed the euphotic zone. At LDA: 9, 24, 35, 70, and 100 m were incubated under 54, 10, 3, 1, and 0.3 % incident light; at LDB: 7, 12, 27, and 42 m were incubated under 54, 36, 10, and 3 % incident light; and at LDC: 16, 60, 91, and 135 m were incubated under 54, 10, 3, and 1 % incident light, respectively. For depths deeper than the euphotic zone (200 m at LDA, 100 and 200 m at LDB and 200 m at LDC), flasks were incubated in the dark in a laboratory incubator set at in situ temperature. After 24 h of incubation, subsamples were taken from each flask to perform BP incubations as described for in situ samples (triplicate estimates, incubation in the dark), except that incubations lasted only 1 h. Results are presented as an enrichment factor relative to the unamended control.

Statistics
Relationships between variables were established using model II Tessier linear regressions, from log-transformed data. Multiple regressions were also used to study the simultaneous effects of PP and N 2 fix rates on BP variability. The effect of enrichments was tested by comparing BP obtained in the unamended control with BP obtained in the amended samples using a Mann-Whitney non-parametric test. This test was also used to estimate differences between geographic zones described in Table 2. Standard errors (SE) of integrated rates were calculated following the propagation procedures. Primary production (mg C m d ) Heterotrophic prokaryotic production (ng C l -1 h -1 ) 3 4 5 6 7 8 9 10 11 12 13 15 14 1

Section distance [km]
Figure 2. Distribution of primary production (a) and heterotrophic prokaryotic production (b) along the OUTPACE cruise transect; WMA (dark green), EMA (light green) and WGY (blue) stations are noted by colored rectangles. Interpolation between sampling points in contour plots was made with the Ocean Data View software (VG gridding algorithm, Schlitzer, 2004). In order to be homogeneous for the whole transect, for sites LDA, LDB and LDC the data plotted for PP were from a single profile, that of PP deck , while for BP we plotted all profiles. The white dots in (a) correspond to the average ± SD of the dcm depth at each station. The white rectangles mask abnormal extrapolation due to the absence of PP data.

Regional oceanographic settings
The longitudinal transect started northwest of New Caledonia, crossed the Vanuatu and Fiji Arcs and finished inside the western part of the ultra-oligotrophic South Pacific Gyre. It covered a vast region of the WTSP and the main gradient of biogeochemical and biological properties between the Melanesian Archipelago (MA) (stations SD1 to SD12 and LDA) and the western part of the South Pacific Gyre (WGY) (SD13 to SD15 and LDC) separated by the Tonga Volcanic Arc (Fig. 1). Temperature ranged from 19.7 to 30.2 • C within the 0-200 m layer (see Fig. 3a in Moutin et al., 2017). Density revealed shallow mixed layers, due to the sharp vertical temperature gradients. The transition between these MA and WGY areas is particularly evidenced by an enhanced degree of oligotrophy in the WGY area. The WGY area was characterized by dcm depth greater than 115 m (Table 2), a deep nitracline (130 m), nitrite peaks around 150 m and detectable amounts of phosphate at the surface (> 100 nM, . A detailed analysis of the vertical distribution of nutrients and organic matter made it possible to identify two groups of stations within the MA area, each having common biogeochemical characteristics: one group between 160 and Table 2. Physical and biological characteristics of main biogeochemical areas and long duration stations sampled during the OUTPACE cruise. Depth of the dcm (deep chlorophyll maximum, based on vertical profiles of in vivo fluorescence), σ t: sigma-theta at the dcm (kg m −3 ), Ichl a (integrated chlorophyll a from fluorometric discrete analyses on extracted samples), IN 2 fix (integrated N 2 fixation rates), IPP (integrated primary production), IBP (integrated bacterial production at the depth of the euphotic zone), DCR (dark community respiration integrated over the euphotic layer). WMA clustered stations SD1, 2, 3 and LDA, EMA clustered SD 6, 7, 9 and 10, WGY clustered SD13, 14, 15 and LDC. In order to encompass only spatial variability for WMA, EMA and WGY groups of stations, means and ranges of dcm depths and of σ t at the dcm depth were based on the averages values set individually at each SD or LD stations as more than one cast was sampled per station.  Distribution of integrated N 2 fixation rates and of ratio N 2 fixation rates to bacterial nitrogen demand (I N 2 fix/I-BND, assuming a bacterial C / N ratio of 5 and no nitrogen excretion) along the transect. Data were integrated down to the deepest sampled depth for N 2 fixation rates. Data plotted for sites LDA, LDB and LDC correspond to BP, PP deck and N 2 fix measured on day 5. Error bars are standard errors (SE) derived from triplicate measurements at each depth (BP, PP deck , N 2 fix rates). For BP, error bars also take into account the daily variability, and final SE were calculated after propagation of errors. PP obtained at SD13 was abnormally low (55 mg C m −2 d −1 ) and was excluded; BP and N 2 fix rates were not measured at this station.
170 • E called WMA for "Western Melanesian Archipelago" clustered SD1, 2, 3 and LDA and a second group south of Fiji called EMA for "Eastern Melanesian Archipelago" clustered SD6, 7, 9 and 10 ( Moutin et al., 2018). Main biogeochemical differences between these two groups of stations were related to shallower depths for the phosphacline (20 m), nitracline (76 m), and dcm (82 m) in the WMA group (see Table 2 and Fig. 5b, c in Moutin et al., 2018). The EMA group had depths for these parameters intermediate to those recorded at the WMA and WGY groups (phosphacline 44 m, nitracline 100 m and dcm 105 m). Although geographically included within the MA area, LDB corresponded to a particular bloom condition and is therefore presented and discussed separately.

Longitudinal distributions
The mixed layers for most of the cruise were ≤ 20 m de Verneil et al., 2018), except at SD13 and LDC, where the mixed layer depths were 27 and 34 m, respectively. The dcm depth ranged between 61 and 115 m in the MA and between 123 and 154 m in the WGY (Table 2; see Fig. 3d in Moutin et al., 2017). Integrated Chl a concentrations ranged from 13 to 23 mg Chl a m −2 in the WGY, and were significantly lower than those in the MA (20-38 mg Chl a m −2 , Mann-Whitney test, p = 0.013). The mean dcm depth within the EMA was slightly deeper (mean ± SD: 105 ± 10 m, Table 2) than in the WMA (82 ± 10 m, Mann-Whitney test, p = 0.03). Maximum primary production rates reached 20.8 mg C m −3 d −1 (Fig. 2a). PP rates greater than 10 mg C m −3 d −1 were obtained in the MA at SD1, SD7, and SD9 and also at sites LDA and LDB (see below), whereas stations in the WGY showed values less than 1.3 mg C m −3 d −1 . Bacterial production ranged from 0.8 to 138 ng C L −1 h −1 in the 0-200 m layer (Fig. 2b). Within the MA area, BP reached values greater than 100 ng C L −1 h −1 at SD1 and SD5 within the surface (5 m depth, Fig. 2b). High BP values were also found at LDB (see below). Within the WGY, maximum BP rates reached 27 ng C L −1 h −1 (at site LDC; see below).
Integrated primary production (IPP deck ) ranged from 178 to 853 mg C m −2 d −1 within the MA and from 104 to 213 mg C m −2 d −1 within the WGY (Fig. 3a). Integrated BP (IBP) over the euphotic zone ranged from 58 to 120 mg C m −2 d −1 within the MA and from 31 to 35 mg C m −2 d −1 within the WGY (Fig. 3a). Both integrated fluxes over the euphotic zone were statistically lower within WGY (Mann-Whitney test, p = 0.01 for IBP and p = 0.03 for IPP deck ). In contrast, for the WMA and EMA group of stations, integrated fluxes were not statistically different, neither for IBP (99 ± 15 versus 95 ± 12 mg C m −2 d −1 , Mann-Whitney test, p > 0.05) nor for IPP deck (481 ± 47 versus 471 ± 276 mg C m −2 d −1 , p > 0.05).
DIP turnover times (T DIP ) ranged over 4 orders of magnitude along the transect (from 2.1 up to 1000 h, Fig. 4). T DIP roughly increased with depth, coincident with the increase in DIP concentrations below the phosphacline. T DIP also showed a clear MA-WGY transition zone. Within the WGY mixed layers, T DIP ranged from 469 to 4200 h (Fig. 4a), coincident with detectable amounts of DIP (around 100 nM) in this area . T DIP were lower in the MA than in the WGY. However, T DIP ranged from 2 to 857 h in the mixed layers of the MA, with lower values associated with stations LDB-d5, LDA-d5, SD3, SD4 and SD6, and with elevated values being measured at SD2, SD5, SD7 and SD12. This T DIP range, encompassing 2 orders of magnitude, suggests a much higher range of DIP availability than DIP concentration alone would suggest. Below the phosphacline, T DIP increased with depth more sharply in the WMA (Fig. 4b) than for the EMA (Fig. 4c).

Daily variability at the long occupation sites
Site LDA showed variable dcm depth during the occupation time, with patches of in vivo fluorescence moving up and down the water column with time over a band of 40 m height (dcm depth varied between 63 and 101 m, Table 2). However, the dcm depth corresponded to a stable density horizon (σ t 23.55 ± 0.04 kg m −3 ), with its fluctuation corresponding to internal waves characterized by a periodicity of about 2 per day (Bouruet-Aubertot et al., 2018; Fig. 5). BP and PP peaked in shallower layers, at 10-25 m depth (range 47-68 ng C L −1 h −1 for BP and 3-6 mg C m −3 d −1 for PP), and sometimes presented a second, much less intense peak close to the dcm depth (Fig. 5). Overall, BP and PP showed parallel trends, increasing slightly on day 3 compared to days 1 and 5. Integrated Chl a was on average 26.0 ± 2.6 mg Chl a m −2 , IPP in situ 267 ± 79 mg C m −2 d −1  and IBP 98 ± 16 mg C m −2 d −1 (Table 2). DCR means (±SD) were 226 ± 43 mmole O 2 m −2 d −1 (Table 2). Site LDB, sampled inside a high chlorophyll patch, showed maxima of in vivo fluorescence between 10 and 77 m, the chlorophyll maximum depth showing a significant linear deepening with time (10.4 ± 0.8 m d −1 , r = 0.89, n = 45, p < 0.001). Contrarily to site LDA, the dcm depth did not correspond to a stable pycnocline horizon, as density associated with the dcm depth varied between 21.8 and 23.9 kg m −3 , and reached a plateau after day 4 (data not shown). Chl was distributed over a larger layer (between the surface and 80 m) during the first three days, and then presented a narrower and deeper zone of accumulation, with intensities increasing (Fig. 6). Integrated Chl a decreased from 53.2 to 23.9 mg Chl a m −2 between days 1 and 5, which corresponded to a chlorophyll biomass net loss of about 7.3 mg Chl a m −2 d −1 . The shape of BP and PP vertical profiles was particularly modified at day 5, showing a small decrease in subsurface values for BP (125 down to 100 ng C L −1 h −1 ) and a larger one for PP (15 down to 9 mg C m −3 d −1 ). In contrast, BP increased within the dcm depth at day 5. Integrated PP decreased by approximatively 145 mg C m −2 d −1 between days 3 and 5. PP was not measured at day 4, but a decrease of BP rates in sub-surface layers was already visible. Six profiles were available for BP from which we estimated a linear increasing trend of 7.2 mg C m −2 d −1 per day (n = 6, Site LDC, typical of the ultra-oligotrophic WGY area, presented a deeper dcm depth, ranging from 115 to 154 m, due to internal waves (Fig. 7). Similarly to site LDA, the dcm depth corresponded to a stable density horizon (σ t 24.59 ± 0.02 kg m −3 , n = 46). PP exhibited two peaks around 40-60 and 120 m, but they remained very low (max. 2.3 mg C m −3 d −1 ) compared to the other sites LDA and LDB. BP profiles paralleled those of PP, also reaching small maxima at 60 m and occasionally a second one at 120 m. The maximum BP rate was 27.7 ng C L −1 h −1 . IPP in situ ranged from 149 to 165 mg C m −2 d −1 , and IDCR ranged from 103 to 173 mmole O 2 m −2 d −1 . IBP values were also low (44 ± 5 mg C m −2 d −1 ) and the three integrated rates exhibited no trend with time.

Relationships between BP, PP, N 2 fix and T DIP
There are several limitations when comparing PP deck and PP in situ . Incubation on mooring lines for 24 h dawn to dawn is considered to be a good compromise by JGOFS recommendations (JGOFS, 1988), as conditions of temperature and light are close to in situ (except UV). Incubation on deck under simulated in situ conditions suffers from biases related to the use of artificial screens to mimic light attenuation with depth and also from biases related to temperature differences for deeper samples, as they are incubated at sea-surface temperature. During our cruise, at each LD site on day 5, we used both incubation methods, but unfortunately did not sample the same ctd cast: PP in situ was sampled at 03:00, while PP deck was sampled at 09:00. At site LDA, differences between the mean IPP in situ and IPP deck were particularly high. Besides artifacts related to screens and temperature described above, this difference could also be due partly to internal waves. For instance, at site LDA on day 5, the dcm depth changed from 69 to 87 m between the 03:00 ctd cast and the 09:00 ctd cast. At site LDB, the bloom collapsed rapidly and a trend with time was clearly detected, making the comparison between both methods impossible, even with a time lag of only 6 h. For this reason, and to keep relative comparisons consistent, we used only PP deck data when exploring relationships between BP, PP, N 2 fix and T DIP . Log-log relationships between BP and PP deck presented similar trends for all samples with T DIP values below or above 100 h (Fig. 8a). Values below 50-100 h are representative of a restricted access to DIP by microorganisms (Moutin et al., 2008). T DIP below 2 days was shown to be critical for Trichodesmium spp. growth (Moutin et al., 2005). The depth at which this threshold was reached varied from surface to 64 m in MA, although all T DIP values were higher than 100 h Table 3. Results of multiple regressions log BP = f (log PP, log N 2 fix). The BP unit before log transformation is ng C L −1 h −1 . Y int is the intercept with the Y axis. Partial coefficient of regression (part coeff), t value (t) and probability of the regression to be significant (p).

Units before log transformation mgC m
In contrast, log-log relationships linking BP and N 2 fix presented different trends for samples corresponding to depths where T DIP was below or above 100 h (Fig. 8b). For samples where T DIP was ≤ 100 h and > 100 h, relationships were, respectively, log BP = 0.752 log N 2 fix − 0.78, n = 39, r = 0.52, p < 0.001 (5) and log BP = 0.438 log N 2 fix − 0.31, n = 55, r = 0.43, p < 0.001. (6) This shows that BP was more correlated with N 2 fix than with PP deck in the P -depleted surface waters of the MA. However, as PP deck and N 2 fix could co-vary, a multiple regression BP = f (PP deck , N 2 fix) was tested (Table 3). The partial coefficient was not significant for N 2 fix for samples with T DIP > 100 h. The partial coefficients were both significant for N 2 fix and for PP for samples characterized by T DIP ≤ 100 h, but N 2 fix better explained the distribution of BP in the multiple regression analysis compared to PP (ttest, p = 0.024 for PP and p < 0.0001 for N 2 fix). Integrated N 2 fix accounted for 3.3 to 81 % of the bacterial nitrogen demand along the transect, assuming a stoichiometric molar C / N ratio of 5 for heterotrophic prokaryotic biomass (Fig. 3b). Among the three LD sites, the variability of this ratio was lower at sites LDA and LDC, with no particular temporal trend at any of these sites. The ratio ranged from 28 to 46 % at LDA, with a mean of 37 % ± 9 %, greater than that obtained at LDC (range 6-10 %, mean ± SD 8 % ± 2 %). The ratio exhibited no particular temporal trend, whereas a decrease was clearly observed at LDB, (68, 37 and 19 % on days 1, 3 and 5, respectively). This was due to a simultaneous decrease of N 2 fix rates (from 0.98 ± 0.058 mmole N m −2 d −1 on day 1 to 0.758 ± 0.058 and 0.38 ± 0.019 mmole N m −2 d −1 on days 3 and 5, respectively) and an increase in BP as described in Sect. 3.3. The mean contribution was 40 ± 20 % over all the transect, and including data from LD sites (n = 26 profiles).
We also examined relationships between T DIP and other biological fluxes using multiple regressions [logT DIP = f (log PP, log N 2 fix, log BP)], incorporating 91 samples for which the three rates were measured simultaneously (Table 4). The partial coefficients were significant for both N 2 fix (p < 0.0001) and BP (p = 0.003) but not for PP (p = 0.23). As all biological rates decreased with depth, we also examined this correlation using data within the mixed layer to avoid the depth effect. With this restricted dataset (47 samples) the partial coefficients were significant only for BP (p = 0.0024), just under the significance threshold for N 2 fix (p = 0.056) and still non-significant for PP.

DOC lability and BGE
In the three biodegradation experiments starting on day 1 at each LD site using sub-surface waters, BP increased significantly, with growth rates (determined from exponential phase of BP increase) ranging from 0.08 to 0.14 h −1 . DOC was slightly consumed, with DOC concentrations decreasing only 2 to 5 % over 10 days, and with the lowest percentage of labile fraction being measured at site LDC (Table 5). Bacterial growth efficiencies were 13, 6.3 and 6.7 % at sites LDA, LDB and LDC, respectively (Table 5).

Enrichment experiments
Conditions prevailing before enrichments are presented together with enrichment factors obtained at the different depths tested as vertical distributions of in vivo fluorescence, nutrients and BP sampled from a CTD cast starting at 12:00 on day 2 of each LD site (Fig. 9). Nitrate concentrations were below the detection limits of standard methods in upper layers. The depth of the nitracline varied with the dcm depth at LDA and LDC (100 m at LDA, 135 m at LDC), but not at LDB (a large peak of chlorophyll was observed within 20-70 m with a nitracline at 100 m). Slight peaks of nitrite also occurred in the vicinity of the nitracline. Phosphate concentrations exhibited more contrasted vertical profiles than did nitrate: DIP concentrations were greater than 100 nM in the surface layers of LDC, but presented a phosphacline shallower than the nitracline at LDA and LDB, with DIP reaching concentrations below the analytical detection limits in the mixed layer (i.e., < 50 nM; see Moutin et al., 2018, for more details on nutrient distribution). As DOC was not sampled on day 2, DOC data are presented for the whole site instead. DOC peaked near the surface at site LDA (77 ± 1 µM) and was more variable but higher near the surface at site LDB, with maximum values covering a larger surface layer down to 27 m, with average DOC of 84 ± 1 µM. Site LDC presented a large sub-surface maximum within 28-42 m, reaching 77 ± 2 µM. At site LDA, nitrogen was the first factor stimulating BP down to 100 m depth, which corresponded to the dcm depth and a nitrite regeneration layer. Although significant at 9, 24, 35 and 100 m in depth (Mann-Whitney test, p< 0.05), the response to N amendments was small, at best an enhancement factor of ×1.6. Glucose alone stimulated BP at 35 m, but only by a factor of ×1.4. However, below the euphotic zone, glucose was the first factor stimulating BP (enhancement factor ×2.4 at 200 m). NPG additions produced the largest increase in BP (×2.5-×3.0) along the profile, except at 9 m were NPG amendments did not significantly affect BP compared to the control.
At site LDB, between the surface and 42 m, both nitrogen alone and phosphate alone stimulated BP to a larger extent than at site LDA, but only by a factor ×1.5-3 for P, and ×1.8-3.7 for N. At 100 and 200 m, nitrogen continued to stimulate BP to a small extent (×3.0 and ×2.2, respectively), but the maximum enhancement was obtained after glucose addition alone (×59 and ×107, respectively). At these depths, the BP response after addition of NPG was also largely amplified compared to shallower layers (×120-132 compared to ×3.7-6.8, respectively).
At site LDC, BP reacted mostly to glucose alone, with enhancement factors increasing from ×2.6 at 16 m to ×24 at 200 m. Nitrogen alone also stimulated BP, but to a smaller extent than glucose, even within surface layers (×1.2 to ×9). In comparison to single amendments, the NPG addition particularly stimulated BP at 60 and 90 m depth.

An overview of BP and PP fluxes in the WTSP
Here, we provide a unique, coherent dataset with simultaneous estimates of PP, BP, T DIP and N 2 fix rates in the WTSP. Although recent interest has increased in describing fluxes and planktonic communities responsible for N 2 fix rates in diverse environments, particularly in oligotrophic open oceans, measurements in the tropical area of the South Pacific Ocean are rare (summarized in Gruber et al., 2016, and more recently in Bonnet et al., 2017). Moreover, few studies have attempted to simultaneously study the consequences of such activities on the functioning of the microbial food webs.
Stations in the western part of the transect along the Melanesian Archipelago (MA) generally displayed greater fluxes of PP, BP and N 2 fix rates than those in the WGY area, with a large degree of longitudinal variability. This was mirrored by T DIP variability spanning 2 orders of magnitude in the upper 40 m layers within MA (2 to 700 h, Fig. 4). The role of the submesoscale activity largely explained such variabilities in biogeochemical parameters and fluxes .
Previous in situ measurements of primary production in the tropical South Pacific, not directly focusing on coastal ar- Table 4. Results of multiple regressions log T DIP = f (log BP, log PP, log N 2 fix). The unit of T DIP before log transformation is h. Y int is the intercept with the Y axis. Partial coefficient of regression (part coeff), t value (t) and probability of the regression to be significant (p).
Units before log transformation ngC L −1 h −1 nmol N L −1 d −1 mgC m −3 d −1 independent variables BP N 2 fix PP Y int n r  , and phosphate (DIP)) and enrichment factors sampled from the 12:00 ctd cast on day 2 of occupation at each LD site.
As DOC was not sampled on this cast, we showed the data from all the other casts at the corresponding LD site (circles) and the average profile (line). Enrichment factors are the ratio of BP after a given enrichment (DIP: P in red; nitrate + ammonium: N in green; glucose: G in blue; and all components: NPG in black) compared to the unamended control, both measured after 24 h incubations. The error bar is standard deviation within triplicates, and a bar is shown only if BP is significantly greater than in the control (Mann-Whitney test, p < 0.05).
eas or within upwelling areas in the east, are scarce (Table 1). These daily particulate primary production rates, based on the 14 C or 13 C technique, confirm the trend that we observed in the WTSP, i.e., extremely low values in the central GY area ranging from 8 to 167 mg C m −2 d −1 . PP increased in the periphery of the GY, but rates remained typical of olig-otrophic to mesotrophic conditions, in the eastern region of the GY, in the south of the GY, and in the western part of the WTSP around New Caledonia and between New Caledonia and Australia (Table 1). Further northwest, in the Solomon Sea, PP increased to much greater values in an area of intense nitrogen fixation, up to 3000 mg C m −2 d −1 (Table 1).

BGE and metabolic state
Although the metabolic state of oligotrophic oceans is still controversial Ducklow and Donney, 2013;Williams et al., 2013;Serret et al., 2015;Letscher et al., 2017), a consensus emerges that in vitro estimates (involving O 2 derived rates or labelling with 18 O 2 , 13 C, or 14 C isotopes) tend to show net heterotrophy in oligotrophic environments. This is indeed what we obtained here, with negative Net Community Production (NCP) values at the three sites LDA, LDB and LDC (−97 to −198, −73 to −134, and −61 to −141 mmole O 2 m −2 d −1 , respectively, Dominique Lefevre, personal communication, 2018). In the WTSP, negative NCP were also obtained in the oligotrophic waters offshore of New Caledonia (Pringault et al., 2007). However, in vitro technique estimates suffer from many biases related to reproducibility, bottle effects, type of flasks used (selecting light wavelengths), condition of incubations and handling artifacts, and a lack of high-frequency measurements (Aranguren-Gassis et al., 2012a, b). For example, following the same water mass during 5 days, a substantial temporal variation in NCP by nearly 2-fold was observed at the LDA, LDB and LDC sites. In contrast, in situ based estimates, based on observations of mixed-layer net oxygen exchanges (O 2 / Ar technique), tend to favor slight net autotrophy (Williams et al., 2013). However, these results also suffer from biases related to the estimate of the mixed layer depth considered and the diffusive coefficients used for gases. Another approach based on the use of oxygen sensors in Argo floats recently showed annual NCP close to zero in the South Pacific Ocean . Recent models encompassing all seasons and a large areal basis find the global ocean to be net autotrophic, including all five oligotrophic subtropical gyres (Letscher et al., 2017). Estimating the metabolic balance by comparing different methods is thus of great interest but rare, especially in the South Pacific. For instance, in the center of the GY, between the Polynesian Archipelago and Easter Island, net heterotrophy was also obtained (although not statistically different from zero) using the in vitro O 2 technique (Van Wambeke et al., 2008b), whereas a non-intrusive bio-optical method showed metabolic balance . Simultaneous estimates of PP, BP and N 2 fix rates are almost absent in oligotrophic waters and to date, BP has not been measured in the WTSP. Here we analyse the contribution of primary production to bacterial carbon demand by comparing separate estimates from the WTSP using our Cbased discrete biological fluxes. The ratio of bacterial carbon demand (BCD) to gross primary production (GPP) is presented as an index of the coupling between primary producers and heterotrophic bacteria and of metabolic balance (Hoppe et al 2002;Fouilland and Mostajir, 2010): to sustain heterotrophy when BCD exceeds GPP, populations can be temporally non-synchronous and/or allochthonous sources of DOM may be required.
It is known that the in vitro 14 C method measures an intermediate state between net PP and GPP. However, Moutin et al. (1999) showed that GPP could be reasonably estimated from daily net PP determined from dawn-to-dawn as GPP = 1.72 × PP. On the other hand, dealing with the assumptions made to convert hourly leucine incorporation rates to daily BCD, there are many biases that have been largely debated, primarily those resulting from daily variability, assumptions about BGE or BR (Alonzo-Saez et al., 2007;Aranguren-Gassis et al., 2012b), carbon to leucine conversion factors (Alonso-Saez et al., 2010), and light conditions of incubations including UV (Ruiz-Gonzales et al., 2013). In this study, we provide direct and indirect estimation of BGE to discuss its variability. Daily variability will also be taken into account using results from previous experiments in the South Pacific Gyre (BIOSOPE cruise, Van Wambeke et al., 2008). Finally, we will also discuss one largely unexplored bias, related to the ability of Prochlorococcus to assimilate leucine in the dark.
Bacterial growth efficiencies (BGE) obtained from biodegradation experiments ranged from 6 to 12 %, with a small labile fraction of DOC (only 2-5 % of biodegradable DOC in 10 days). Thus, the bulk DOC was mainly refractory, although DOC concentration was high in the surface . Large stocks of DOC, with C / N ratios ranging from 16 to 23, have also been reported in the surface waters of the SPG . Both high C / N ratios and a small labile fraction suggests that this surface bulk pool of DOC is probably largely recalcitrant due to UV photodegradation or photooxidation (Keil and Kirchman, 1994;Tranvik and Stephan, 1998;Carlson and Hansel, 2015) or by action of the microbial carbon pump (Jiao et al., 2010). Small BGE and small labile fraction could also be due to strong resource dependence as low nutrient concentrations cause low primary production rates and low transfer across food webs. Indeed, Letscher et al. (2015) also observed surface DOC recalcitrant to remineralization in the oligotrophic part of the eastern tropical South Pacific. But as shown by these authors, incubation with microbial communities from the twilight zone, provided by addition of an inoculum concentrated in a small volume, allowed DOC remineralization. This was explained by relief from micronutrient limitation or by the co-metabolism of relatively labile DOC provided by the inoculum with more recalcitrant DOC. Results from our enrichment experiments effectively suggest nutrient limitation, although the second hypothesis cannot be excluded.
In order to better explain the variability of BGE measurements, we also estimated this parameter indirectly, using simultaneously measured dark community respiration (DCR) and BP data. We converted DCR to carbon units assuming a respiratory quotient RQ = 0.9, and computed BGE from Ze-integrated BP and DCR assuming either bacterial respiration (BR) to be within a range of 30 % of DCR (BGE = BP / (BP + DCR × 0.9 × 30 %), Rivkin and Legendre, 2001;del Giorgio and Duarte, 2002) or 80 % of DCR (BGE = BP / (BP + DCR × 0.9 × 80 %), Lemée et al., 2002;Aranguren-Gassis et al., 2012b). These indirect estimates of BGE ranges were similar to those obtained from the biodegradation experiments: 3-12 % at site LDA, 4-17 % at site LDB and 2-7 % at site LDC. Note, however, an increasing trend from day 1 to day 5 at site LDB: on average 8 % on day 1, 10 % on day 3 and 12 % on day 5. Including all direct and indirect estimates, the mean (±SD) BGE was 8 ± 4 % (n = 21).
Although bias introduced when converting hourly to daily BP rates was not studied here, we used a dataset obtained in the South Pacific Gyre (Van Wambeke et al., 2008) to estimate conversion errors. During the BIOSOPE cruise, vertical profiles of BP were acquired every 3 up to 72 h, using the leucine technique within the euphotic zone at three selected sites using a Lagrangian sampling strategy. For the three series of profiles, standard deviations of BPI with time were  Figure 10. Distribution of integrated bacterial carbon demand corrected for Prochlorococcus assimilation (I-BCD corr , grey bars) gross primary production derived from IPP deck (I-GPP, blue bars), along the transect. Error bars are standard errors (SE), calculated using propagation of errors. We included analytical SE (triplicate measurement at each depth for PP and BP), SE due to daily BP variability, SE from BGE, and SE due to leucine assimilation by Prochlorococcus. * Shows stations in which I-BCDcorr minus I-GPP was statistically different from zero.
13 % (n = 13), 16 % (n = 16) and 19 % (n = 9). Thus, standard errors represented 3.6, 4.2 and 6.1 % of the mean BPI, respectively. We used the average value of this percentage (5 %) to estimate the bias introduced by the conversion from hourly to daily BPI estimates of the OUPACE cruise. Finally, we considered the ability of Prochlorococcus to assimilate leucine in the dark. Using flow cytometry cell sorting of samples labelled with 3 H-leucine during the OUTPACE cruise, Duhamel et al. (2018) demonstrated the mixotrophic capacity of Prochloroccoccus, as this phytoplankton group was able to incorporate leucine, even under dark conditions, albeit at lower rates than under light conditions. This group was found to be able to assimilate ATP, leucine, methionine and glucose, a single C-containing molecule (Duhamel et al., 2018, and references therein). To date, few organic molecules have been tested, and those mainly include N, P or S sources. As leucine assimilation by Prochloroccoccus was significantly detected in dark incubations in all examined samples, it will affect BP measurements. We thus corrected BP (BP corr ) to represent the assimilation of leucine in the dark by heterotrophic bacteria alone. Based on Duhamel et al. (2018), leucine assimilation by HNA + LNA bacteria in the dark corresponded on average (±SD) to 76 ± 21 % (n = 5, range 44-100 %) of the activity determined for the community including Prochlorococcus (HNA + LNA + Proc). I-BCD corr was then calculated and its distribution presented with I-GPP (Fig. 10), assuming a mean 8 % for BGE in all our integrated data. Among 14 stations over the 17 presented the mean integrated I-BCD corr is greater than I-GPP. We tested whether the confidence interval 95 % of the difference I-BCD corr minus I-GPP is included or not zero (t-test) before concluding on the metabolic balance, and found that I-GPP is significantly lower than I-BCD corr at stations SD4, 5, 6, 12 and LDB (i.e., only 5 stations among 17), and greater at only one (SD9). The remaining stations (11 among 17) were in metabolic balance. These results confirm the necessity to include the variability of different conversion factors before making conclusions about whether GPP satisfies (or not) BCD and accounting for the bias introduced by Prochlorococcus assimilation of leucine in BP estimates.
Such comparisons between GPP and BCD would be more complex if the short-term temporal variability in conversion factors was considered. For example, site LDB illustrates how rapidly these relative fluxes changed during the collapse of the bloom. LDB was located inside a massive chlorophyll patch, which had been drifting eastwards for several months  and which collapsed at the time LDB was sampled. Considering the decreasing Chl a stocks, decreasing PP and increasing BP, I-BCD corr to I-GPP ratios would increase from 1.08 (day 1) to 1.83 (day 5), based on a constant BGE of 8 %. However, using the actual BGE increase measured from days 1 to 5, the ratio would increase only from 1.1 to 1.2, and the site would have been under metabolic balance during that period. Unfortunately, it is nearly impossible to assess all correction and conversion factors at the same resolution as BP and PP measurements, leading to unconstrained budgets (Gasol et al., 2008).

Nutrient limitation and relationships with nitrogen fixers
Phylogenetic analyses of the functional gene nifH showed the prevalence of gamma-proteobacteria and unicellular cyanobacteria UCYN-A (presumably photo-heterotroph) in the surface layers of the ultra-oligotrophic center of the GY (Halm et al., 2012). However, quantifying gene transcripts in the GY and in the WTSP, Moisander et al. (2014) found nifH expression by UCYN-A to be 1-2 orders of magnitude greater than for a gamma proteobacterial diazotroph (γ -24774A11). Along the Australian Great Barrier Reef, as determined by qPCR, the abundance of nifH gene copies of the Gamma A group (with peaks of only 5.9 × 10 2 nifH copies L −1 ) were in general also 1-2 orders of magnitude less abundant than those of Trichodesmium (Messer et al., 2017). Unfortunately, to date, there are no data on the nifH gene diversity of heterotrophic nitrogen fixers, nor on their expression, in the euphotic layers for the OUTPACE cruise, but nifH gene presence in some heterotrophs has been detected in the aphotic zone . Using a qPCR analysis of nifH gene copies in selected diazotrophs, focusing on cyanobacteria (unicellular, filaments and heterocystous symbionts) during the OUTPACE cruise, Stenegren et al. (2018) showed that Trichodesmium dominated in surface layers (0-35 m) within MA, but was rare or undetected in the WGY. Within MA the second and third most abundant populations of cyanobacteria detected were, respec-tively, UCYN-B and heterocystous endosymbionts (diatomdiazotroph associations). Among the investigated cyanobacterial diazotrophs, this study showed a temperature-depth gradient separating two groups of cyanobacteria, with Trichodesmium occupying the warmest and shallowest waters, UCYN-A occupying the coldest and deeper waters and UCYN-B having a more widespread distribution. Regardless of the intermediary processes relating both fluxes, the amount of N 2 fixed corresponded to 15-83 % of heterotrophic bacterial nitrogen demand in MA and 3-34 % in the WGY (Fig. 3b). This fraction could be particularly high in the MA area, where N 2 fix rates reached maximum values for the cruise (1 mmole m −2 d −1 at station SD11). Besides this exception, the activity of N 2 fixers generally sustained less than half the heterotrophic bacterial N demand (mean ± SD: 34 ± 20 %, n = 26). Nevertheless, the activity of N 2 fixers was responsible for the injection of new N in the microbial food chain, enabling a cascade transfer through non-fixing autotrophs and heterotrophs, and possibly C and N export by sedimentation even in the oligotrophic areas of the WTSP . The transfer between N 2 fixers and heterotrophic prokaryotes occurs on a daily scale, as confirmed by nano-scale secondary ion mass spectrometry (nanoSIMS) experiments, which make it possible to track the fate of 15 N 2 at the individual cell level. These experiments have shown a rapid N transfer from Trichodesmium colonies to heterotrophic epiphyte bacterial cells (Eichner et al., 2017). Although it is not known from this study if the 15 N can reach free living heterotrophic prokaryotes or non-fixing phytoplankton, other experiments suggest that 15 N fixed by Trichodesmium reaches rapidly non-fixing diatoms (Foster et al., 2011;. Using diazotroph cultures inoculated in natural sea waters from the New Caledonia lagoon, Berthelot et al. (2016) also showed a rapid transfer (48 h) from a Trichodesmium (T. erythraeum) and a UCYN-B (Chrocosphaera. Watsonii) towards heterotrophic bacteria. Recent evidence also suggests a rapid transfer from the symbiotic photo-heterotrophic cyanobacterium UCYN-A, which have much greater growth rates than Trichodesmium, to its associated eukaryotic algae (Martinez-Perez et al., 2016).
Consequently, most N 2 -fixing cyanobacteria detected in the WTSP during this cruise have the potential to transfer N rapidly (1-2 days) towards strict heterotrophic bacteria or non-fixing phytoplankton. We found that the correlation between N 2 fix rate and BP was better within the mixed layers and when the T DIP is low (< 100 h), i.e., in areas of phosphate deficiency (Moutin et al., 2008). Greater correlation between BP and N 2 fix than between BP and PP would suggest that bacteria may have been more dependent on the availability of new sources of N than C, which is in agreement with results from enrichment experiments at LDA and LDB. Because T DIP was lower in areas of high N 2 fix rates, it is likely that DIP drawdown was due to diazotrophs which, while bringing new sources of N, reduced DIP availability. Indeed, at site LDB within the mixed layers, BP increased after N addition alone but also after P addition alone, which suggests a direct limitation of BP by N and potentially a cascade effect of P addition towards heterotrophic prokaryotes: P would directly stimulate N 2 fixers which rapidly would transfer new N and labile C available to stimulate BP. At sites LDA and LDB, the addition of the three-element NPG stimulated BP more than P alone or N alone, suggesting possible NP co-limitation of heterotrophic prokaryotes. Furthermore, if N was shown to be the first limiting nutrient during a short timescale experiment, addition of P stimulates N 2 fix, PP and export at longer timescales (Van Den Brock et al., 2004;Gimenez et al., 2016).
Below the surface layers, where T DIP increases, and where UCYN-A dominates, DIP becomes available and nitrate diffusing through the nitracline sustains primary production. BP and PP were correlated within these layers, suggesting a strong coupling between BP and PP through the release of labile organic C. Indeed, our enrichment experiments revealed that BP was limited first by labile C, secondarily by N within these layers (Fig. 9).
Finally, within the WGY area, where T DIP were elevated in the mixed layer (more than 100 h), with detectable DIP concentrations reaching 100 nM, the activity of N 2 fixers was extremely low. In this area, BP was limited mainly by the availability of energy or labile C. Indeed, within site LDC, glucose alone stimulated BP to a larger extent than did N alone, even when the latter was provided in the form of ammonium. This trend was also observed from enrichment experiments performed in the center and the eastern zones of the GY (Van Wambeke et al., 2008a). UCYN-B dominated the N 2 fixing populations in the WGY, which accounted for 81-100 % of the total detected cyanobacterial nifH gene copies . Among UCYN-B, Crocosphaera is one of the most studied representatives. One of its sub-populations is recognized to produce EPS (Bench et al., 2016), which could be a significant energy source for heterotrophic prokaryotes. In the North Pacific Subtropical Gyre, Wilson et al. (2017) hypothesized that Crocosphaera could fix N 2 in excess of its growth requirements and could release fixed N in seawater. They also showed a highly dynamic Crocosphaera growth and decay during diel cycles, suggesting rapid switch between cell growth and mortality processes, such as grazing and viral infection. The resources provided by leakage, lysis and grazing processes likely directed energy and N towards heterotrophic bacteria at a daily scale when N 2 fixation was favorable, as shown in the North Pacific. However, in the WGY, N 2 fix rates were very low and could only sustain a low percentage of bacterial and phytoplankton N demand.

Conclusions
Our results provide a unique set of simultaneous measurements of BP, PP and N 2 fix rates in the WTSP. BP obtained in the WTSP was in the same range as those previously mea-sured in the GY area east of 140 • W. BGE was low (8 % on average) and the bulk DOC was found to be refractory (labile DOC 2-5 % on average). We show that the interpretation of PP and BP fluxes based on instantaneous methods (radioisotopic labelling) needs regular tests to verify the major methodological biases and conversion factors hypotheses. In particular, to make conclusions about the metabolic state of oceanic regions, it is necessary to consider the variability of all conversion factors used to estimate carbon-based GPP and BCD. In addition, the use of the leucine technique to estimate BP should be used with caution in N-limited environments due to the potential mixotrophy by cyanobacteria. Taking these corrections into account and using propagation of errors, we compared I-BCD corr and I-GPP as indexes of the coupling between primary producers and heterotrophic bacteria. We found that the system was in metabolic balance at 11 over the 17 stations investigated. In the N and relatively DIP depleted surface waters, BP was more strongly correlated with N 2 fix than with PP, while the more traditional coupling of BP with PP occurred deeper in the euphotic zone. This suggests that in surface layers with elevated diazotrophic activity, BP was more dependent on the availability of new N from the activity of N 2 fixers than on the availability of fresh C from the activity of primary producers, as was also demonstrated through enrichment experiments.
Competing interests. The authors declare that they have no conflict of interest.
Special issue statement. This article is part of the special issue "Interactions between planktonic organisms and biogeochemical cycles across trophic and N 2 fixation gradients in the western tropical South Pacific Ocean: a multidisciplinary approach (OUTPACE experiment)". It is not associated with a conference.