Drivers of Microbial Carbon Fluxes Variability in Two Oligotrophic Mediterranean Coastal Systems

The carbon fluxes between phytoplankton and heterotrophic bacterioplankton were studied in two coastal oligotrophic sites in the NW Mediterranean. Phytoplankton and bacterial production rates were measured under natural conditions using different methods. In the Bay of Villefranche, the temporal variability revealed net heterotrophy in July-October and net autotrophy in December-March. The spatial variability was studied in the Bay of Palma, showing net autotrophic areas in the west and heterotrophic areas in the east. On average bacterial respiration, represented 62% of the total community respiration. Bacterial growth efficiency (BGE) values were significantly higher in autotrophic conditions than in heterotrophic ones. During autotrophic periods, dissolved primary production (DPP) was enough to sustained bacterial metabolism, although it showed a positive correlation with organic carbon stock (DOC). Under heterotrophic conditions, DPP did not sustain bacterial metabolism but bacterial respiration correlated with DPP and bacterial production with DOC. Temperature affected positively, DOC, BGE, bacterial respiration and production when the trophic status was autotrophic. To summarize, the response of bacterial metabolism to temperature and carbon sources depends on the trophic status within these oligotrophic coastal systems.

the productivity of the system. Bacterial carbon demand includes all the organic carbon assimilated by heterotrophic bacteria, which can be allocated into anabolic (bacterial production) and catabolic processes (bacterial respiration).
Currently, there is some debate about the dependence of bacterial carbon demand on phytoplankton production, mainly within oligotrophic systems 20,21 . In offshore areas not influenced by carbon inputs from land, phytoplankton primary production was reported to sustain bacterial carbon demand 18 , but the generality of this observation for inshore regions has not yet been confirmed 22 . Other authors 21 suggested that bacterial carbon demand always exceeds contemporaneous primary production in all aquatic systems. Bacterial carbon demand is governed by bacterial growth efficiency (BGE = bacterial production/bacterial carbon demand), however there is a large range of variation in BGE (5-60% 23 ;). In many instances, BGE is not directly measured but inferred from literature models 24,25 , which are generally biased with data from productive areas. Therefore, there is a lack of data about the BGE regulation in oligotrophic coastal waters 14,24,[26][27][28] .
Ideally, direct measurements of net microbial carbon production and consumption should be used 29 but bacterial production is usually measured in unfiltered samples and with short incubation times whereas bacterial respiration is measured in filtered samples with longer incubation times. In many studies bacterial respiration is derived from biomass and production data [30][31][32] . Moreover, disentangling the factors that control the carbon fluxes dynamics of marine microbial communities is challenging. The temperature dependence of biological processes has been traditionally expressed using the Van't Hoff-Arrhenius equation also known as Boltzmann's factor 33-35 . According to this expression, the slope of the natural logarithm of a metabolic rate against the temperature function 1/kT, where k is Boltzmann's constant and T is the absolute temperature (Kelvin degrees), represents the activation energy (Ea) of that particular metabolic process [33][34][35] . Temperature has been described to play a major role regulating the metabolism of Mediterranean plankton communities 36,37 . Here we test how the metabolic status or balance of the plankton community can affect the temperature-dependence of microbial metabolism in these ecosystems.
To improve our knowledge of the trophic status in oligotrophic bays, it is necessary to focus the studies on temporal and spatial variability, despite the given idea that these environments as steady state systems. To reach this goal, we performed simultaneous measurements of net community production, particulate and dissolved primary production, bacterial production and respiration, thus allowing direct BGE, bacterial carbon demand and conversion factor values. The study was carried out in two oligotrophic coastal ecosystems in the NW Mediterranean (Fig. 1). The temporal variability was studied within the Bay of Villefranche (France) and the spatial variability was focused in July within the Bay of Palma (Spain), during the period most limited by nutrients. The comparison and join, of both bays is based on the environmental similarities of these regions 2,10 . The results will improve further understanding of the link between carbon fluxes and the metabolic trophic status, of these marine coastal ecosystems.

Effect of filtration and incubation time on bacterial metabolism estimations.
To estimate the effect of filtration on bacterial metabolism, bacterial production, bacterial abundance and DOC concentration were compared between filtered and unfiltered samples. DOC concentration in unfiltered (83 ± 3 µmol l −1 ) and filtered (97 ± 5 µmol l −1 ) samples were significantly different (paired t-test; t 21 = 3, p < 0.01, n = 22). Bacterial abundance was not significantly different before (6.92 ± 0.48 × 10 5 cell ml −1 ) and after filtration (6.24 ± 1.02 × 10 5 cell ml −1 ). Bacterial production was significantly lower in filtered (17.59 ± 4.03) than in unfiltered (66.45 ± 11.41) samples (paired t-test; t 83 = 4.5, p < 0.001, n = 84). Bacterial production was estimated after different periods of in situ incubation times (no incubation, 6 and 12 h). After 12 h incubation bacterial production showed values   Fig. 2). Also, a significant effect of incubation time was observed in bacterial respiration rates ( Fig. 2; F 2,21 = 4.1, n = 24, p = 0.03), with higher values after 48 h than after 12 and 24 h (Fig. 2). However, the duration of incubation had no significant effect on unfiltered samples. BGE and bacterial carbon demand are usually estimated using bacterial production and respiration rates which are commonly incubated during different time lengths. In order to minimize possible bias, BGE 12 and BCD 12 in our experiments were estimated using bacterial respiration and production measurements from pre-filtered samples both incubated during 12 hours (see material and methods). The average percentage of oxygen consumption in filtered seawater samples (12 and 24 h) compared to whole seawater samples, were enabled to estimate bacterial respiration.
Temporal variability in the Bay of Villefranche. As expected, irradiance showed a seasonal pattern with the highest values found during July followed by October, December and March (Table 1). The lowest temperatures were recorded in March (13.2 °C throughout the water column) and the highest in July, coincident with the strongest stratification (Table 1). Changes in the percentage of O 2 saturation were similar to those of temperature, with oxygen-saturated waters during July and below 100% during the rest of the year (Table 1, Fig. 3). The concentrations of NO − 3 and Si (OH) 4 were lower than 1 µmol l −1 except in March (Table 1). The highest DOC concentrations were during July (70.7 ± 2.2 µmol l −1 ) and the lowest during March (63.8 ± 0.6 µmol l −1 ; Table 1).
Chlorophyll-a concentrations varied considerably (from 0.1 to 1.1 mg m −3 ) across depths and months, with mean values increasing from summer to spring (Fig. 3). The picophytoplankton community composition was dominated by Synechococcus cyanobacteria. The highest abundances of Synechococcus were found in October and July, while Prochlorococcus peaked in December and October and photosynthetic picoeukaryotes in March. Heterotrophic prokaryotes abundances ranged from 6.1 to 12.3 × 10 5 cells ml −1 , with the highest abundances measured in December and the lowest in October (Fig. 3). Picoautotrophs biomass average contributed with 9.4 (±1.1) % of the total autotrophic biomass (Chl-a).
Primary production was generally low, with DPP and PPP ranging from 0.02 to 1.36 mg C m −3 h −1 and 0.03 to 2.10 mg C m −3 h −1 ; respectively, with the highest values observed in the spring bloom period during March (Fig. 4). Percentage of extracellular release was highly variable, from 5.0 to 86.6%, with the lowest values recorded in December and the highest ones in October (Table 1, Fig. 4). Heterotrophic bacterial production showed a minimum of 1.6 ± 0.4 pmol Leu l −1 h −1 in March (equivalent to 0.002 mg C m −3 h −1 ) and a maximum of 14.9 ± 6.6 pmol Leu l −1 h −1 in July (0.02 mg C m −3 h −1 ). Indeed, bacterial respiration (BR) was lowest in December (0.01 ± 0.004 µmol O 2 l −1 h −1 ) and highest in October (0.04 ± 0.01 µmol O 2 l −1 h −1 ) (Fig. 4) Fig. 3). Low dissolved inorganic nutrient concentrations characterized these oligotrophic waters, with NO − 3 and Si (OH) 4 always below or equal to 1 µmol l −1 ( Table 1). DOC concentration was higher at the western stations Posidonia and Bahia, (89.1 ± 4.4 and 85.5 ± 1.7; µmol l −1 , respectively (Fig. 1, Table 1). Chorophyll-a values were below 0.5 mg m −3 and followed the same pattern as DOC with the highest values measured at Posidonia and Bahia stations (Fig. 3). Picophytoplankton community composition was dominated by Synechococcus spp across stations, with highest abundances measured at Bahia station (Fig. 3). Prochlorococcus spp peaked at Station 4 and photosynthetic picoeukaryotes at Bahia station. Heterotrophic prokaryotes abundance ranged from 2.4 to 9.4 × 10 5 cells ml −1 , with the highest abundances at Bahia and lowest at Station 4 station (Fig. 3). The average picoautotrophs biomass accounted 5.7% ( ± 0.7) of the total autotrophic biomass. Total primary production ranged from 0.16 to 2.99 mg C m −3 h −1 , with DPP ranging from 0.11 to 1.36 mg C m −3 h −1 and PPP from 0.03 to 1.63 mg C m −3 h −1 . Highest values of both fractions, DPP and PPP, were measured at Posidonia station and the lowest at Enderrocat (Fig. 4) Fig. 4). In the Bay of Palma, the lowest bacterial production was 11.8 pmol Leu l −1 h −1 (0.01 mg C m −3 h −1 ) observed at Enderrocat station and the highest in Posidonia station of 252.9 pmol Leu l −1 h −1 (0.29 mg C m −3 h −1 ) (Fig. 4). Bacterial respiration showed the lowest value at Bahia station, with 0.06 µmol O 2 l −1 h −1 and the highest at Station 4 station with 0.24 µmol O 2 l −1 h −1 (Fig. 4). BGE 12 ranged from 0.3 to 40.6%, with the highest mean values at Posidonia station and the lowest observed at the Enderrocat station ( Table 1). All data were grouped into two categories regardless of the location and from an ecophysiological point of view: net autotrophic (NCP > 0) and net heterotrophic (NCP < 0). Coherently, results confirmed that DPP and    Fig. 5). BGE was also significantly higher (F 1,30 = 5.1, n = 32, p = 0.03) in autotrophic than in heterotrophic communities (8.5 ± 2.6% vs 1.9 ± 0.5%) (Fig. 5). However, percentage of extracellular release and bacterial carbon demand showed no significant differences between trophic status; respectively (F 1,30 = 1.7, n = 32, p > 0.05 and F 1,30 = 0.5, n = 32, p > 0.05) (Fig. 5).

Effects of environmental and biological variables on bacterial metabolism on both bays.
When differentiated according to NCP values, DPP and Chl-a were positive and significant linearly correlated only within autotrophic communities (R 2 = 0.28, n = 17, p = 0.03), but not within heterotrophic ones (p > 0.05; Fig. 6a). Similarly, the linear regression between PPP and chlorophyll-a was significant and positive for autotrophic communities (R 2 = 0.45, n = 17, p = 0.003), whereas no correlation was found in the heterotrophic ones (p > 0.05; Fig. 6b).
Bacterial production showed a significant relationship with DOC concentration independently of the trophic status of the community (Fig. 7a) while bacterial respiration and carbon demand were correlated with DOC concentration only with autotrophic communities (Fig. 7b,c). In contrast, bacterial respiration was only significantly related to DPP in heterotrophic communities (Fig. 7e), whereas no relationships of bacterial production and carbon demand were observed with DPP (Fig. 7d,f).   Bacterial production and respiration, BGE and DOC concentrations were positively correlated with temperature only under autotrophic conditions, but not under heterotrophic ( Fig. 8a-d). The slope or activation energy of the Arrhenius relationship for bacterial production was 1.42 eV, 0.67 eV for bacterial respiration and 0.83 eV for BGE.

Discussion
Relying on actual system-specific measurements rather than on literature or models results, does not completely solve the coupling degree between phytoplankton and heterotrophic bacterioplankton because measuring the flow of carbon through bacteria is methodologically challenging. For instance, bacterial respiration and production estimates may still bring uncertainties due to factors such as the incubation time, bottle effects, bottle opaque to UVB radiation, filtration artefacts or the choice substrate-to-carbon conversion factors 29,37-39 .
Our first concern was to make sure that bacterial abundance was not significantly different after filtration, so that metabolic processes measured in the filtered samples represented realistically the rates of bacterial respiration in the no processed samples.
Filtering process by the 0.8 µm filters increases DOC concentration because some planktonic cells could have been broken, leaking part of their cytoplasmic content into the filtrate 40 . However, this process even if did not affect small bacteria it certainly had influence on attached bacteria 29,41 which are not detected by the flow cytometer. For this reason, bacterial production was higher before the filtering process and heterotrophic prokaryotes abundance did not show any significant difference.
Since bottle effects are to some extent unavoidable (e.g. 42 ), another factor contributing to uncertainty is the incubation time. Contrary to previous studies 27 , bacterial respiration and production did not show linear responses after 12 and 48 h, respectively. This observation is not trivial, as most of the previous studies did not consider as a key variable the incubation time or the bottle effect 14,36,43 . Bacterial production is typically measured in unfiltered samples and incubated for a short period (2-8 hours), whereas bacterial respiration is determined in filtered samples that are subsequently incubated for at least 24 or 48 h. Bacterial respiration could be stimulated during long incubations (48 h 42 ) by viral activity 44 or by changes in bacterial community composition 26,45 . Shorter incubations (i.e. <6 h) limit the bottle effect and provide the most accurate consumption rates of labile DOC 9 . On the contrary, samples incubated for longer times will be exposed to higher cumulative processes, such as cell lysis, increasing DOC release and consequently increased bacterial production and respiration 46 .
Despite the methodological challenge, in this study we were able to demonstrate coherent patterns in bacterial carbon fluxes in two NW Mediterranean embayment's. Our direct rather than indirect estimations and the simultaneous temporal and spatial studies, might provide a new insight of the bacterial carbon fluxes and their temperature and metabolic status dependence within oligotrophic systems.
In this study, we measured an expected significantly higher chlorophyll-a concentrations under autotrophic periods (stations which values of NCP are higher than 0) compare to those concentrations under heterotrophic conditions (NCP < 0). However, picophytoplankton biomass of the diverse groups measured, was not significantly different during heterotrophic and autotrophic periods. Indeed, regardless to the locations TPP, and consequently DPP and PPP, was significantly higher under autotrophic conditions compare to heterotrophic conditions. Contrary to previous works 16,18 and in agreement with others 19 differences in percentage of extracellular release average values between trophic states were not significant (41 ± 5% and 53 ± 5% in autotrophic and heterotrophic systems, respectively).  www.nature.com/scientificreports www.nature.com/scientificreports/ Our results showed a robust relationship between DPP and PPP with photosynthetic proxy biomass (Chl-a) in autotrophic waters, while they were not correlated in heterotrophic waters. At the Bay of Villefrance-sur-Mer, maximum rates of DPP and PPP were registered during March when nitrate and chlorophyll-a concentrations were higher and seawater temperature was lower. This situation corresponds with the spring phytoplankton bloom and is followed by an increase in bacterial production as a consequence of increased DPP 19,47 . However, the lack of correlation between DPP and PPP with chlorophyll-a under heterotrophic conditions, may be attributable to a strong nutrient limitation for phytoplankton growth 48 , due to process such as grazing 16 or to phytoplankton lysis 49 . Despite the general low nutrient concentrations measured, during the spring bloom of March, the high nitrate concentrations measured (>1 up to 4 µmol l −1 ) were similar to those registered in the same area 2 , but lower than other oligotrophic Mediterranean regions (up to 7.0 µmol l −1 for nitrate in 14 ).
Contrary to the previous described situation 2,14,50 , during summer at the Bay of Palma, the lowest nitrate concentration co-occurred with elevated water temperature (23.4 °C) and high DOC concentrations (90 µmol l −1 ). Some authors 9,51 reported an accumulation of DOC during summer and autumn and attributed this to the inability or inefficiency of bacteria to use it under nutrient limitations, the so-called malfunctioning microbial loop 52 .
However, not only DOC, but also phytoplankton DPP has been considered an important carbon source to support bacterial metabolism 18,19 . In our study, we observed a positive correlation between DPP and BR, but exclusively when the system was heterotrophic. Under heterotrophic nutrient-deplete conditions, the fresh and labile carbon from phytoplankton production (DPP) may be firstly respired for bacterial basal metabolism 53 . Although, under the same heterotrophic scenario, bacterial production is correlated to DOC concentrations instead of DPP, probably due to the extra resources needed for bacterial growth or production, than those used to maintain basal bacterial respiration 31 . Therefore, at resource limiting conditions when phytoplankton production was too low www.nature.com/scientificreports www.nature.com/scientificreports/ to sustain BP, labile DOC from nearby seagrass meadows may be used as an organic carbon source [54][55][56] . Since abundant beds of Posidonia oceanica were present in both bays, a large fraction of the DOC used for bacterial production may come from the exported carbon from benthic organisms 56,57 .
On the other hand, under autotrophic conditions, bacterial metabolism, was positively correlated with DOC concentrations and DPP was potentially sufficient to sustain the bacterial carbon demand (DPP/BCD > 1). These results suggest that DPP contributed signifiicantly to the labile DOC pool under autotrophic conditions and bacteria were growing actively when DOC stock concentration was higher enough to maintain bacterial metabolism 2,9,28,58,59 . The highest DPP/BCD ratio was found in Villefranche´s Bay during March when the spring bloom occurred and picoeukaryotes biomass was higher than other picoautotrophs groups. Although picoeukaryotes only accounted 3% of the total autotrophic biomass, concurred with the higher DPP concentrations measured 60,61 during that period.
The previous reported BGE values depend, however, on the particular leucine-to-carbon conversion factor used 31 . In this work, bacterial production was estimated from in situ incubations and calculated with empirically determined leu-to-C conversion factors as emphasized in recent studies 27,62,63 . Bacterial production values (0.001 to 0.44 mg C m −3 h −1 ) were within the range reported in other Mediterranean coastal sites 14,36,64 , being more active during autotrophic periods, and leading to higher BGE values up to 4-fold compare to those periods where heterotrophic conditions dominated.
It has also been described that bacterial respiration contribute with a significant amount to total community respiration, frequently more than 50% 7,25,65 . In our study, bacterial respiration represented, on average, 62% of the total community respiration and did not depend on the metabolic balance of the system. However, bacterial respiration mean value at both coastal oligotrophic environments (1.9 ± 0.2 mmol C m −3 d −1 ) was substantially lower than the average for coastal regions (7.1 ± 12.2 mmol C m −3 d −1 ) given in the revision of Robinson 39 . Indeed, bacterial respiration was only slightly higher than open ocean value (1.3 ± 0.2 mmol C m −3 d −1 ), probably showing the general oligotrophic nature of Mediterranean coastal waters.
Our BGE values (0.3-40.6%) were within the range of published values for the Mediterranean Sea 2,14,36,58,66 . However, the median BGE (1.6%) was substantially lower than the value given for coastal regions (16%) and even the open ocean (8%) 39 , although it should be noted that this review 39 did not include any data from the coastal oligotrophic Mediterranean.
Temperature, nutrient concentrations, dissolved organic matter lability and bacterial taxonomic composition have been regarded among the main factors that can influence BGE 14 . Robinson 39 suggested that most of the BGE variability within oligotrophic waters is explained by bacterial respiration. Other authors 31 found that BGE is not directly regulated by temperature, but by the availability of substrates for growth. According to our results, temperature explains a large part of BGE (47%), bacterial respiration (46%) and production variances (73%) when there is a not nutrient limitation. Indeed, the metabolic status of the system is a critical factor determining the temperature-dependence of bacterial metabolic rates. Consequently, the significant and positive relationships between temperature and bacterial metabolic rates (bacterial production and respiration, BGE) during periods of net autotrophy did not hold for heterotrophic periods. Our results confirm that substrate availability is a key factor governing the temperature dependence of heterotrophic bacterial metabolism 31,67 . Temperature is a relevant factor under autotrophic conditions, when our system is in the stage between the bottom-up and top-down control 68 . However, temperature is not a main driver of bacterial metabolism under heterotrophic conditions, due to the limited resource availability (bottom-up control).
The lower slope of the bacterial respiration regression (Ea BR = 0.67 eV) indicates a lower activation energy required for this process relative to that of bacterial production (Ea BP = 1.42 eV), what confirms their different temperature-dependence as an intrinsic characteristic of two independent biochemical processes 59 .
The bacterial respiration activation energy value estimated in this study (Ea BR = 0.67 eV) is very close to the theoretical Ea of 0.65 eV predicted by the Metabolic Theory of Ecology for respiratory processes 33,34 . Compared to previous works, our bacterial respiration Ea value is lower than the values reported by Mazuecos 69 for the dark ocean (0.90 eV) and higher than those values for the global ocean (Ea BR = 0.589 eV 31 ), (Ea BR = 0.57 eV 70 ).
Bacterial production activation energy (Ea BP = 1.42 eV) is more than twice the maximum value described by Morán 71 (from −0.14 to 0.67 eV), although it is within the range described by Arandia-Gorostidi 72 (−0.3 to 1.46 eV) both for the NE Atlantic coastal waters, and slightly higher than the in situ bacterial production measured at the Baltic Sea (1.24 ± 0.16 eV 73 ). The relationship of bacterial metabolic rates with temperature under autotrophic conditions, agrees with higher Eas values during the highest nutrient concentration 71 and productive bloom periods 67,72 .
We conclude that the metabolic status (autotrophic vs heterotrophic) of the oligotrophic coastal Mediterranean studied systems, determines the source of organic carbon where the bacterial communities are supported by, as well as the temperature dependence of microbial metabolic processes.

Material and Methods
Study site and sampling. The spatial variability was investigated in the Bay of Palma in June 2002 (Fig. 1a). www.nature.com/scientificreports www.nature.com/scientificreports/ Hydrography, irradiance and nutrients. Vertical profiles of temperature, salinity and oxygen saturation were obtained at each station using a SeaBird SBE19 CTD in the Bay of Palma or SBE25 in the Bay of Villefranche. Profiles of light penetration were performed with a LI-COR spherical sensor (LI-193SA) connected to a LI-1400 data-logger. Seawater was collected at four depths within the euphotic layer at each station (Table 1) using single 12 l Niskin bottles. Dissolved inorganic nutrients concentrations were measured using automated colorimetric techniques after filtration through Whatman GF/F glass fibre filters 74 . Dissolved silicate was measured after filtration through 0.45 µm polyvinylidene fluoride membrane filters (PVDF, Millipore). Chlorophyll-a was measured on GF/F filters that were stored frozen until extraction and analysis by high-performance-liquid chromatography 75 .
Dissolved organic carbon (DOC). Due to the low particulate carbon concentrations in the two sites, total organic carbon samples were collected and not filtered in order to minimize the risk of contamination. Samples of DOC concentration were collected in combusted (450 °C for 24 h) glass ampoules (Wheaton), acidified with 85% H 3 PO 4 flame-sealed immediately after collection and stored in the dark at ambient temperature pending analysis. DOC concentration was measured by high temperature combustion with a Shimadzu TOC-5000 total organic carbon analyser. All concentrations are reported as the average of three replicate injections.
Picoplankton abundance. The abundances of heterotrophic prokaryotes, Prochlorococcus spp, Synechococcus spp and photosynthetic picoeukaryotes were estimated using a bench-top flow cytometer (Becton and Dickinson FACScalibur) equipped with a laser emitting at 488 nm. Aliquots of 2 ml were fixed with 1% paraformaldehyde and 0.05% of glutaraldehyde (final concentration), frozen in liquid N 2 and stored at −80 °C pending analysis. The samples for heterotrophic prokaryotes abundance were previously stained with Syto-13 Molecular Probes (2.5 µM) and an aliquot of 0.96 µm fluorescent latex beads (Polysciences) was also added as internal standard to each tube. The beads were calibrated every day using a beads solution (TrueCount beads) for which the bead abundance is precisely known. Heterotrophic prokaryotes abundance was identified in plots of side scatter (SSC) versus green fluorescence (FL1) as described by 76 . Prochlorococcus spp and Synechococcus spp were identified in plots of SSC versus orange fluorescence (FL2) and red fluorescence (FL3). PA had a higher SSC and FL3 than both Prochlorococcus spp and Synechococcus spp and no FL2 signal. In order to convert cell abundance to carbon biomass it was used, the conversion factors proposed by 77 : 12 fg C cell −1 for heterotrophic prokaryotes, 1500 fg C cell −1 for picoeukaryotes, 32 fg C cell −1 for Prochlorococcus spp, and 100 fg C cell −1 for Synechococcus spp. The total phytoplankton carbon biomass was estimated using a carbon-to-chlorophyll-a ratio of 120 78 . primary production 14 c technique. The incorporation of carbon into the dissolved and particulate fractions was measured using the 14 C technique 79 . Water samples (30 ml) were transferred to one dark and three light bottles and incubated in situ at each depth and station. The samples were incubated in situ on a vertical line from sunrise to sunset after addition of 30 µCi (1110 Kbq) of sodium bicarbonate (NaH 14 CO 3 ). Upon completion of the incubation, two 5 ml aliquots of each sample were placed into 20 ml scintillation vials for determination of total labelled organic carbon (Total Primary Production = DPP + PPP). The unit for DPP and PPP were estimated by the incubation time, which was the daylight time (mg m −3 h −1 ). Two more 5 ml aliquots of each sample were filtered through 0.22-µm mixed cellulose esters membrane filters (Millipore) for determination of PPP and the filtrate was collected for determination of DPP. Filters were placed into vials and exposed to concentrated hydrochloric acid fumes for a minimum of 12 h before the addition of 3.5 ml Packard Ultima Gold (PerkinElmer) liquid scintillation cocktail. Liquid samples (total primary production and DPP) were acidified with 100 µl of hydrochloric acid (5N) and bubbled with air for 2 h before addition of 15 ml Ultima Gold XR (PerkinElmer) liquid scintillation cocktail. Radioactivity (dpm) was measured in a Packard Tri-Carb 4000 scintillation counter. Counts from the dark bottles were subtracted from counts measured in the light bottles to correct for non-photosynthetic 14 C-incorporation. The radioactivity of the NaH 14 CO 3 added to each sample was estimated from 20 µl aliquots.

Bacterial metabolism.
To estimate in situ community respiration, the first set of seawater samples were incubated in dark BOD (Biological Oxygen Demand) bottles (60 ml) on a vertical line in situ during 24 h. At each station and depth, 5 bottles were immediately fixed with Winkler reagents 80 (time 0) and five more bottles were fixed after 24 h. The second set of samples was incubated under dark condition in a temperature-controlled incubator simulating in situ temperature to calculate incubator community respiration. The third set of samples was filtered (142 mm diameter, 0.8 µm) to include just the bacterial fraction to estimate the incubator bacterial respiration. The filtration was carried out with a low vacuum pressure system at each station and depth. No more than 10 l of a given sample was filtered through the same filter. The percentage of oxygen consumption in filtered seawater samples compared to incubator whole seawater samples, enabled us to estimate incubator bacterial respiration. This percentage was applied to the in situ community respiration to calculate the in situ bacterial respiration.
To estimate the effect of filtration on bacterial metabolism, bacterial abundance and production, and DOC concentration were compared at each station and bay between filtered and unfiltered samples at each depth.
To estimate the effect of incubation time on bacterial metabolism, at each station of the Bay of Palma (S4, Posidonia, Enderrocat and Bahia) and depth (surface and bottom) an extra set of filtered and unfiltered seawater samples were incubated in dark BOD bottles (60 ml) during 12, 24 and 48 hours to estimate community and bacterial respiration. For bacterial production rates, seawater samples were incubated in borosilicate bottles on a vertical line in situ. At each depth and station, three replicates were recovered at sunrise (0 h of incubation), midday (6 h of incubation) and sunset (about 12 h of incubation depending on the season). (2019) 9:17669 | https://doi.org/10.1038/s41598-019-53650-z www.nature.com/scientificreports www.nature.com/scientificreports/ Dissolved oxygen concentration was determined by titration with a potentiometric end-point detection 80 . Analyses to estimate respiration were performed with a redox electrode (9778-SC, Orion) and a custom built titrator. Bacterial production was estimated as the rate of 3 H-leucine incorporation using standard methods 81 in Eppendorf vials. Sea water samples were transferred to Eppendorf vials (1.5 ml). Three were spiked with 5 µl of cold leucine (20 nM final concentration) and 25 µl of 3 H-leucine (10 nM final concentration; Amersham, specific activity 170 µCi nmol −1 ). The fourth sample, to which 100 µl of 100% trichloroacetic acid was immediately added, served as a control.
The conversion factor required to express the incorporation of leucine into carbon units was estimated at stations Station 4 and Enderrocat at the surface (0.5 m) and bottom depth (33 and 16 m, respectively) following the procedure proposed in 82 . A sample (1.8 L) was filtered through a 0.2 µm polycarbonate membrane previously rinsed with milliQ water and with sample water and placed in a bottle with 0.2 l of unfiltered water. The incubation lasted for 2 d in a temperature-controlled incubator simulating in situ temperature. Aliquots for bacterial production (leucine incorporation) and abundance (cytometry) were processed every 6 h. The conversion factor (CF) was calculated using the relationship derived by Kirchman and Ducklow in 82 : Where µ is the growth rate (d −1 ) determined from the change in biomass over time, e B is the y-intercept of Ln (biomass in cell ml −1 ) vs time and e b is the y-intercept of Ln (leucine incorporation in pmol Leu −1 l −1 h −1 ) vs time. Four values of conversion factors were obtained and the average (1.15 ± 0.33 kg C mol −1 ; see Table 1 Supplementary Information) was used to convert leucine incorporation to carbon uptake for bacterial carbon demand and growth efficiency (BGE) calculations. BGE was estimated as the ratio of bacterial production to total bacterial carbon demand (BP + BR). A respiratory quotient (RQ) of 1 was assumed 24 .
Net community production (NCP). For estimating production of dissolved oxygen, seawater samples (10 replicates) were transferred into transparent BOD bottles (60 ml) and incubated on a vertical line in situ during 24 h. At each station and depth, incubations started at sunrise. Five bottles were immediately fixed with Winkler reagents 80 (time 0) and 5 other bottles were fixed at the following sunrise (24 h incubation). The dissolved oxygen concentration was determined as explained above. NCP was estimated by regressing O 2 against time.
Statistical analysis. One-way analysis of variance ANOVA (Statistica 5.0) was performed to evaluate differences among trophic status systems. Variance homogeneity was tested with Cochran's analysis. The Student-Newman-Keuls test (SNK) was used to discriminate among different treatments after a significant F-test. A Mann-Whitney U test was used when the data were not strictly parametric. The regression model type II was used to evaluate the relationship between variables when both variables were subject to measurement error.