Lack of P-limitation of phytoplankton and heterotrophic prokaryotes in surface waters of three anticyclonic eddies in the stratified Mediterranean Sea

We investigated the identity of the limiting nutrient of the pelagic microbial food web in the Mediterranean Sea using nutrient manipulated microcosms during summer 2008. Experiments were carried out with surface waters at the center of anticyclonic eddies in the Western Basin, the Ionian Basin, and the Levantine Basin. In situ, the ratio of N to P was always higher in both dissolved and particulate organic fractions compared to the Redfield ratio, suggesting a relative P-starvation. In each experiment, four different treatments in triplicates (addition of ammonium, phosphate, a combination of both, and the unamended control) were employed and chemical and biological parameters monitored throughout a 3–4 day incubation. Temporal changes of turnover time of phosphate and ATP, and alkaline phosphatase activity during the incubation suggested that the phytoplankton and heterotrophic prokaryotes (Hprok) communities were not P-limited at the sites. Furthermore, Correspondence to: T. Tanaka (tsuneo.tanaka@obs-vlfr.fr) statistical comparison among treatments at the end of the incubation did not support a hypothesis of P-limitation at the three study sites. In contrast, primary production was consistently limited by N, and Hprok growth was not limited by N nor P in the Western Basin, but N-limited in the Ionian Basin, and N and P co-limited in the Levantine Basin. Our results demonstrated the gap between biogeochemical features (an apparent P-starved status) and biological responses (no apparent P-limitation). We question the general notion that Mediterranean surface waters are limited by P alone during the stratified period.


Introduction
A large portion of oceanic pelagic waters are characterized as "low nutrient and low chlorophyll" (LNLC), i.e. oligotrophic. In such waters, pelagic productivity is generally limited by the availability of inorganic nutrients (e.g. nitrogen (N), phosphorus (P)), and the microbial food web plays Published by Copernicus Publications on behalf of the European Geosciences Union.

526
T. Tanaka et al.: Lack of P-limitation of phytoplankton and heterotrophic prokaryotes in surface waters a significant role in carbon flux through the pelagic plankton food web (reviewed by Karl, 2007). In order to better understand the biogeochemical cycling of carbon (C) in the ocean, it is important to identify which nutrient is the most limiting factor of pelagic productivity and how the structure and function of the plankton food web is affected by the availability of the limiting nutrient.
The Mediterranean Sea is a unique marine system in terms of its hydrography and biogeochemistry (reviewed by Krom et al., 2003). Nutrient concentrations, integrated chlorophyll and primary production in the epipelagic layer, and the export of particulate organic carbon (POC) from the epipelagic layer, all decrease in the Mediterranean Sea from west to east on the basin scale . The low nutrient status is caused by anti-estuarine circulation in which nutrient-poor surface water flows eastward through the Straits of Gibraltar and Sicily getting progressively more saline to the east, while a counter-current of nutrient-rich water flows out of the basin. The waters below the epipelagic layer have a nitrate (NO 3 ) to phosphate (PO 4 ) ratio of 20-25:1 in the Western Mediterranean Sea (Béthoux et al., 1992;Marty et al., 2002;Pujo-Pay et al., 2010) and 25-30:1 in the Eastern Mediterranean Sea (Krom et al., 1991;Kress and Herut, 2001;Pujo-Pay et al., 2010). The biogeochemical parameter of the N to P ratio suggests that the Mediterranean Sea is P-starved compared to other oceanic regions whose ratios of NO 3 to PO 4 conform to what is known as the Redfield ratio of 16.
Nutrient enrichment studies have shown that growth of phytoplankton and heterotrophic prokaryotes (hereafter, Hprok) is often limited by the availability of PO 4 in Mediterranean surface waters during the stratified period (Jacques et al., 1973;Fiala et al., 1976;Zweifel et al., 1993;Dolan et al., 1995;Vaulot et al., 1996;Thingstad et al., 1998;Zohary and Robarts, 1998;Sala et al., 2002;Pinhassi et al., 2006;Tanaka et al., 2009). The specific affinity for PO 4 uptake by phytoplankton and Hprok, which is the specific PO 4 uptake rate normalized to biomass, is a useful diagnostic tool to measure the extent of PO 4 availability (Thingstad and Rassoulzadegan, 1999;Tanaka et al., 2006). In Mediterranean surface waters during the stratified period, specific PO 4 affinities are close to their respective theoretical maxima which are predicted from the assumption that molecular diffusion to the cell surface is the rate-limiting step Tanaka et al., 2003Tanaka et al., , 2004Flaten et al., 2005). These results would then indicate that the biologically available PO 4 concentrations are so low that molecular transport by diffusion towards cells is the rate-limiting process for PO 4 uptake in Mediterranean surface waters.
However, there are indications of the limiting nutrient being other than PO 4 . Data on concentration ratios of NO 3 + NH 4 to PO 4 and assimilation ratios of organic C to inorganic N suggested a potential N limitation of phytoplankton production in the NW Mediterranean Sea, although this study was done during the winter overturn (January 1989) in the region of the plume of the Rhone River (Owens et al., 1989). Van  reported the results of Mediterranean transect study in which prokaryotic heterotrophic production (PHP) was stimulated by addition of PO 4 at 13 sites but also stimulated by addition of NO 3 at 2 sites and by addition of organic C at 5 sites of the 18 sites. These studies suggest that phytoplankton and Hprok may experience growth limitation by substrates other than P in the Mediterranean Sea.
PO 4 addition to surface waters of the P-starved ultraoligotrophic Eastern Mediterranean Sea in a Lagrangian experiment resulted in unexpected ecosystem responses : the added PO 4 rapidly disappeared, resulting in an increase of particulate P (i.e. biological P uptake). A decline in chlorophyll-a, no significant change in biomass of Hprok, but an increase in PHP, ciliate biomass, and copepod egg abundance were observed after a short lag (ca. 2 days). To explain these responses,  proposed two, not mutually exclusive, mechanisms: because of P-limited Hprok and N and P co-limited phytoplankton, (i) the added PO 4 was transferred "around" the phytoplankton compartment to copepods (bypass mechanism), and (ii) the added PO 4 was rapidly taken up into Pstarved phytoplankton and Hprok, by which the stoichiometry of prey organisms was rapidly changed from P-poor to P-rich (i.e. luxury uptake) (tunneling mechanisms). Results from a microcosm experiment, which was done in parallel with the Lagrangian experiment, support P-limited Hprok and N and P co-limited phytoplankton . It is generally considered that element transfer from lower to higher trophic levels is accompanied by biomass oscillations between prey and predator, and ecological efficiency becomes lower in more oligotrophic system due to increased numbers of the trophic levels. However, the above counterintuitive observation suggests that, once the limiting nutrient (P) is added to a P-starved ultra-oligotrophic system, the microbial food webs can immediately contribute to copepod production, thus retrieve part of the primary production which would otherwise be lost through dissipation inside a microbial loop. The study of  showed that it is necessary to investigate the multiple levels of the food web in response to nutrient manipulation, rather than the osmotroph level (phytoplankton and Hprok) alone, to understand the structure and function of the plankton food web in P-starved oligotrophic systems.
The objectives of this study were to identify which nutrient is the most limiting for the pelagic microbial food web in the Mediterranean Sea, and determine how the structure of the pelagic microbial food web responds to enrichment of the most limiting nutrient. We performed on-board microcosm experiments that manipulated availability of inorganic N and P in surface offshore waters collected at three Mediterranean basins. A suite of chemical and biological variables was measured during the experiments to determine the effect of the nutrient manipulation.  . These stations are located at the center of anticyclonic eddies. Water samples for the experiments were collected at 8 m depths using a multi-sampler/carousel rosette system equipped with Niskin bottles and a CTD. The sampling depth located at the lower part of the surface mixed layer (13.5 m at Stn A, 8.5 m at Stn B, 11.5 m at Stn C; Moutin et al., 2011). The collected waters were poured into polycarbonate carboys (20 l, Nalgene). The four treatments in triplicate microcosms consisted of the unamended treatment, N enrichment, P enrichment, and N and P enrichment (hereafter, Control, +N, +P, and +NP, respectively). +N received 1600 nM-N at Stn A and Stn B and 3200 nM-N at Stn C, +P received 100 nM-P at all stations, and +NP received 1600 nM-N and 100 nM-P at Stn A and Stn B and 3200 nM-N and 100 nM-P at Stn C at the start of the experiment. N and P were added as NH 4 Cl and KH 2 PO 4 , respectively. The level of nutrient amendments were chosen with the aim of satisfying the requirement of either N or P or both by the natural Hprok and phytoplankton community during the experimental period (i.e. 3-4 days). The N:P ratios (16 and 32) were used as a rough approximation of those below the epipelagic layer of the Western and the Eastern Basins, respectively (see Introduction).
Since previous studies have shown that Hprok and small phytoplankton (pico-and nanophytoplankton) use ammonium (NH 4 ) rather than NO 3 (e.g., Lipschultz, 1995;Wheeler and Kirchman, 1986), we used NH 4 as an inorganic N source in this study. The microcosms were then incubated for 3 or 4 days in on-deck flow-through incubators, which were covered by a screen to reduce the incident light by ca. 50%. During the incubation, samples were taken from each of 12 microcosms to measure a suite of chemical and biological variables (see below for the sampling frequency of each parameter). The carboys used as microcosms were washed with 10% of HCl, and rinsed thoroughly with Milli-Q water and three times with the water samples from the Niskin bottles. To minimize contamination, clean gloves were always used during the experimental set up and sampling.

Dissolved and particulate nutrients
Samples for inorganic nutrients were collected on Day 0, Day 2, and at the end of the experiment. Concentrations of nitrate + nitrite (NO 3+2 ) and soluble reactive phosphorus (SRP) were measured with an autoanalyzer (Bran + Luebbe) (Tréguer and Le Corre, 1975), and those of NH 4 were measured with a fluorimeter (Fluorimeter Jasco FP-2020) (Holmes et al., 1999). Concentrations of dissolved inorganic nitrogen (DIN) were calculated by summing concentrations of NO 3+2 and NH 4 . SRP does not necessarily represent dissolved inorganic phosphorus or phosphate (e.g., Murphy and Riley, 1962), however we use the term phosphate (PO 4 ) in this paper. The detection limit and the precision were, respectively, 3 and 2 nM for NH 4 , 0.02 and 0.02 µM for NO 3+2 , and 0.01 and 0.01 µM for PO 4 .
Samples for organic nutrients were collected at the start and the end of the experiment. Concentrations of DOC were measured with a Shimadzu TOC-V analyzer by high temperature catalytic oxidation (Cauwet, 1994(Cauwet, , 1999. Samples for dissolved organic nitrogen and phosphorus (DON and DOP, respectively) were oxidized and measured spectrophotometrically (Pujo-Pay and Raimbault, 1994;Pujo-Pay et al., 1997). Particulate organic carbon (POC) were collected on pre-combusted glass fiber filters (Whatman GF/F), dried, and measured with a CHN analyzer (Perkin Elmer 2400). Particulate organic nitrogen (PON) and phosphorus (POP) were collected on 25 mm GF/F filters. After oxidization of PON and POP, liberated N and P were measured spectrophotometrically (Pujo-Pay and Raimbault, 1994). See also Pujo-Pay et al. (2010) for details.

Chlorophyll-a (Chl-a) concentration
Samples for Chl-a were collected on Day 0, Day 2, and at the end of the experiment. Chl-a was measured fluorometrically, according to Yentsch and Menzel (1963). For each sample, 500 ml were filtered through 0.2 µM polycarbonate filters. Filters were kept frozen in the dark until extraction in 90% acetone solution overnight. Measurements were performed on board with a Shimadzu RF5301 spectrofluorometer.
equipped with an air-cooled argon laser (488 nM, 15 mW) and a standard filter setup. Thus, the term Hprok used in this paper does not include cyanobacterial cells such as Synechococcus and Prochlorococcus. Using the same model of the flow cytometer mentioned above, three different groups of small phytoplankton (Synechococcus, picoeukaryotes, and autotrophic nanoplankton (ANP)) were discriminated and counted based on right-angle light scatter properties and orange and red fluorescence.
Samples were filtered onto black Nuclepore filters (pore size, 0.8 µM) and stained with 4 6-diamidino-2-phenylindole (DAPI) (Porter and Feig, 1980) within 5 h of sampling and stored at −20 • C until counting. HNF were enumerated using an epifluorescence microscope (Leica FW4000) at 1000×. To distinguish between ANP and HNF, autofluorescence (chlorophyll) was determined under blue light excitation. For ciliate enumeration, samples (500 ml) were fixed with acid Lugol's solution (final conc. 2%) and stored at 4 • C in the dark until analysis. Ciliates were pre-concentrated by gravity in the sample bottles, and then settled in Utermöhl chamber. Ciliate enumeration was done by an inverted microscope (Nikon Eclipse TE2000-S) at 400×.

Turnover times of PO 4 and adenosine 5 -triphosphate (ATP)
Samples for turnover times of PO 4 and ATP were collected every day and measured using 33 P-orthophosphate and adenosine 5 -[γ -33 P]triphosphate (AT 33 P), respectively (Thingstad et al., 1993). Carrier-free 33 P-orthophosphate (PerkinElmer, 370 MBq ml −1 ) was added to samples at a final concentration of 20-79 pM. Samples for the subtraction of the background and abiotic adsorption were fixed with 100% trichloroacetic acid (TCA) (final conc. 0.5%) before isotope addition. Samples were incubated under subdued (laboratory) illumination. The incubation time varied between 15 and 20 minutes: short enough to assure a linear relationship between the fraction of isotope adsorbed vs. the incubation time but it was long enough to reliably detect isotope uptake above background levels. The radioactivity from blanks accounted for on average 10.7% (n = 141) of that from live samples. Incubation was stopped by a cold chase of 100 mM KH 2 PO 4 (final conc. 1 mM). Subsamples on Day 0 were filtered in parallel onto 25 mm polycarbonate filters with 2, 0.6, and 0.2 µm pore sizes, and subsamples on Days 1-4 were filtered onto 25 mm polycarbonate filters with 0.2 µm pore size. All filters were placed on a Millipore 12 place manifold with Whatman (GF/C) glass fiber filters saturated with 100 mM KH 2 PO 4 as support. After filtration, filters were placed in polyethylene scintillation vials with Ultima Gold (Packard), and radio-assayed. After the radioactivities of the filter were corrected for those of the blank filter obtained from fixed samples, phosphate turnover time is the fraction (no dimension) of added isotope collected on the 0.2 µM filter after the incubation time (t:h). AT 33 P (PerkinElmer, specific activity 111 TBq mmol −1 ) was added to samples at a near-tracer concentration of 5-25 pM and incubated under laboratory temperature and illumination for 1-2 h. Incubations were terminated by filtration through 25 mm polycarbonate filters with 0.2 µm pore size and washed with 2 ml of 0.2 µm filtrate of seawater. Samples for the subtraction of the background and abiotic adsorption were immediately (within 30 s) filtered after AT 33 P addition (i.e. blank filter). The radioactivity from blanks accounted for on average 8.5% (n = 78) of that from live samples. Filters were placed in scintillation vials with Ultima Gold (Packard) scintillation cocktail, and radio-assayed. After the radioactivities of the filter were corrected for those of the blank filter, ATP turnover time (T [ATP] :h) was calculated as mentioned above for T [PO 4 ] . Since the sample for the initial value of ATP turnover time at Stn B was lost, a datum obtained from the surface water (5 m), which was chronologically closest to the start of the microcosm experiment, was used in this study.
The specific affinity for phosphate uptake was calculated by normalizing specific phosphate uptake rates (inverse of phosphate turnover times) to the summed P-biomass of phytoplankton and Hprok (Tanaka et al., 2006). P-biomass of phytoplankton was estimated from Chl-a concentration with an assumption of C/Chl-a ratio of 50 and C/P ratio of 106, that of Hprok was estimated from abundance of Hprok with an assumption of cell carbon content of 15 fg cell −1 (Caron et al., 1995) and C/P ratio of 50 (Fagerbakke et al., 1996). The POP data present a sum of living particles and detritus so that we estimated P-biomass based on Chl-a and cell concentration. The estimated P-biomass was 45-98% (n = 7) of POP except for two occasions (105 and 115% of POP).

Alkaline phosphatase activity (APA)
Samples for APA were collected every day. APA was measured fluorometrically using 3-O-methylfluoresceinphosphate as substrate (Perry, 1972). Fluorescence in the samples mixed with the substrate (final conc. 0.1 µM) was measured immediately after the addition of the substrate solution and at three or four subsequent times according to the fluorescence increase. After correcting fluorescence values of samples to those of blank, APA (nM-P h −1 ) was calculated by using a linear regression of fluorescence values versus incubation time.

Primary production (PP) and prokaryotic heterotrophic activity (PHA)
Samples for PP and PHA were collected on Day 0, Day 2, and at the end of the experiment. PP and PHA were measured as a measure of growth rate of phytoplankton and Hprok, respectively. PP was measured by the 14 C incorporation method of Steemann Nielsen (1952). Three light and one dark 170 ml polycarbonate bottles were filled up with sample water from each microcosm. Each bottle was inoculated with 20 µCi of NaH 14 CO 3 , and all bottles were incubated in the on-deck incubator for 4 h around midday. After the incubation, samples were filtered through 0.2 µm polycarbonate filters under low vacuum pressure (< 200µmHg). Filters were put in scintillation vials, fumed with 1 ml of 1% HCl in order to remove excess 14 C-bicarbonate over night, and radioassayed with scintillation cocktail. PP (nM-C h −1 ) was calculated by subtracting the radioactivity in dark bottles from that in light bottles under the assumption that DIC is 2 × 10 6 nM-C and a correction factor for the lower uptake of 14 C as compared to 12 C is 1.05 (see Lagaria et al., 2010 for details). PHA (pM-leucine h −1 ) was measured as 3 H-leucine incorporation rate into TCA-insoluble fraction by the centrifuge method (Smith and Azam, 1992). For each sample, duplicate aliquots (1.5 ml) and one TCA-killed control were incubated with 22 nM of leucine (a mixture of 8 nM of 3 Hleucine and 14 nM of cold leucine) for 2 h at in situ temperature in the dark. The incorporation was stopped with the addition of TCA (final conc. 5%). Bovine serum album was added to each sample (final conc. 100 mg l −1 ) prior to the first centrifugation. After aspirating the supernatant, 5% TCA was mixed with pellet, and then the sample was centrifuged again. Supernatant was discarded, and a last centrifuge treatment was done after addition of 80% ethanol. After removing the supernatant, and addition of scintillation cocktail, the sample was radioassayed (see Van Wambeke et al., 2010 for details).

Statistical analysis
Student t-test was used to compare values of parameters between the start and the end of the incubation. One-way analysis of variance (ANOVA) was used to test the effect of different nutrient treatments on a given parameter on the last day of the incubation at each station. For comparison of each parameter between different nutrient treatments, statistical analysis was done by post hoc Tukey Honestly significant difference (HSD) test after ANOVA test. Before statistical comparison, data were log 10 transformed to meet the requirement of homogeneity of variance. When the dataset included zero values, data were log transformed using the equation: log 10 (x + 1) (x: data). Virus samples collected at the end of the experiment at Stn A and Stn B were lost; consequently, the statistical test for viral particles employed data from Day 2 for all three stations. All statistical analyses were done using the R software (http://www.r-project.org/).
The following four "responses" to nutrient additions were defined and used to interpret results of the statistical comparison of parameter values on the last day of the incubation: 1. No limitation: No statistical difference in concentration or rate of a given parameter detected among the four treatments.
2. N-limitation: concentration or rate of a given parameter is statistically higher in +N than the Control.
3. P-limitation: concentration or rate of a given parameter is statistically higher in +P than the Control.
4. N and P co-limitation: Concentration or rate of a given parameter is statistically higher in +NP than the others (i.e. the Control, +N, and +P). Moreover, no statistical difference in concentration or rate of a given parameter is detected between the Control, +N, and +P.

Initial characteristics of the study site waters
Concentrations of NO 3+2 and PO 4 were initially close to, or below, the detection limit of conventional analytical methods at all stations (Table 1) (Table 1). PO 4 turnover time ranged from 4.1 h to 6 h. Uptake of 33 PO 4 was always dominated by the 0.6-2 µM fraction, and the dominance of this size-fraction increased from west to east (65% at Stn A to 87% at Stn C, data not shown). ATP turnover time was one to two orders of magnitude longer than PO 4 turnover time. Specific PO 4 affinity ranged from 0.017 to 0.029 l nmol-P −1 h −1 . APA was always at measurable levels in situ. Chl-a concentrations and Hprok abundances were initially 0.03-0.06 µg l −1 and 1.8 × 10 5 − 3.4 × 10 5 cells ml −1 , respectively.

Temporal variations of dissolved inorganic N and P, turnover times of PO 4 and ATP, and APA
NH 4 concentrations in +N and +NP decreased significantly from the start to the end of the experiments at all three stations ( Fig. 1a, b, c; t-test, P < 0.05). These decreases corresponded to 16-99% of the initial NH 4 concentration. Significant decreases of PO 4 concentrations in +P and +NP (19% and 44%, respectively, of the initial concentration) were observed at Stn A during the incubation ( Fig. 1d; ttest, P < 0.05). In contrast, a significant increase of PO 4 concentration (1.2-6 times) was detected in the Control, +N, and +P at Stn B, and in +N and +P at Stn C (Fig. 1e,  PO 4 turnover time decreased to 1-2 h in the Control and +N at Stn A and Stn C and in +N at Stn B during the incubation (Fig. 2a, b, c). In the Control at Stn B, turnover time oscillated between 1.7 and 6.5 h. The shortest turnover time at the end of the incubation was found in +N at Stn A, in +N and +NP at Stn B, and in the Control and +N at Stn C (Tukey HSD test, P < 0.05). PO 4 addition resulted in increase in turnover time in both +P and +NP during the first 1-2 days. The increase of turnover time after the PO 4 addition was smallest at Stn A. From Day 1 or 2 to the end of the incubation PO 4 turnover time in +P and +NP decreased. At the end of the incubation turnover time was longest in +P and +NP at Stn A and in +P at Stn B and Stn C (Tukey HSD test, P < 0.05). ATP turnover times in all nutrient treatments tended to change little or decrease during the incubation (Fig. 2d, e, f). The effect of the PO 4 addition alone resulted in, at most, a 1.6 time increase of turnover time of ATP, whereas we observed up to a 10 times increase of PO 4 turnover time. ATP turnover time in +N was shortest on Day 2 at Stn A and on Days 2 and 4 at Stn B and Stn C (Tukey HSD test, P < 0.05). APAs continuously increased in +N treatments but in the other treatments varied little at all stations (Fig. 2g, h, i). APAs were significantly higher in +N than all the other treatments on Days 1-3 at Stn A,

Responses of organic C, N, and P, and microbial components
Concentrations of POC were highest in the +NP treatments at all three stations (Fig. 3a, b, c), and significantly higher in +N than the Control at Stn B (Tukey HSD test, P < 0.05). An effect of nutrient addition on DOC concentration was detected only as a significantly lower concentration in +NP than the Control at Stn C (Fig. 3d, e, f). PON concentrations were significantly higher in +N than the Control and +P and highest in +NP at all stations (Fig. 3g, h, i; Tukey HSD test, P < 0.05). No significant difference of DON concentration was detected between the Control and the nutrient addition treatments at all stations (Fig. 3j, k, l). POP concentrations were highest in +NP at all stations (Fig. 3m, n, o; Tukey HSD test, P < 0.05). At Stn B and Stn C, POP concentrations in +P were significantly higher compared to those in the Control (Tukey HSD test, P < 0.05) but not compared to those in +N. Nutrient addition did not affect DOP concentration at all stations (Fig. 3p, q, r). Chl-a concentrations were higher in +N than the Control and +P, and highest in +NP at Stn A and Stn B, and higher in +NP than the others at Stn C (Fig. 4a, b, c; Tukey HSD test, P < 0.05). PP was significantly higher in +N than the Control and +P, and highest in +NP at all stations (Fig. 4d, e, f; Tukey HSD test, P < 0.05). No significant difference between the treatments was detected for Hprok abundance and PHA at Stn A (Fig. 4g, j). At Stn B, Hprok abundances were significantly lower in +N and +NP than the Control and +P, while PHA was significantly higher in +N than the Control and +P, and highest in +NP (Fig. 4h, k; Tukey HSD test, P < 0.05). At Stn C, Hprok abundances were significantly lower in +N than the other treatments, and PHA was significantly higher in +NP than the other treatments ( Fig. 4i, l; Tukey HSD test, P < 0.05).
Abundances of Synechococccus were significantly higher in +N than the Control and +P, and highest in +NP at Stn A and Stn B (Fig. 5a, b; Tukey HSD test, P < 0.05). At Stn C, Synechococcus abundance was significantly higher in +NP than the others (Fig. 5c). No significant differences in picoeukaryotes abundance between the treatments were detected at any station (Fig. 5d, e, f). A significant effect of nutrient addition on ANP abundance was detected at both Stn B and Stn C (Fig. 5h, i). ANP abundance was higher in +N than the Control and +P, and highest in +NP at Stn B (Tukey HSD test, P < 0.05). At Stn C, ANP abundance was significantly higher in +NP than the others. No significant difference in concentration of viral particles between the treatments was detected at any station (Fig. 6a, b, c). Significant differences in HNF abundance between the treatments were detected only on twice: higher in +N than the Control at Stn A, and higher in +NP than the Control at Stn B (Fig. 6d, e; Tukey HSD test, P < 0.05). Ciliate abundances were significantly higher in +NP than the others at Stn A and Stn C , primary production (nM-C h −1 ) (d, e, f), abundance of heterotrophic prokaryotes (Hprok) (cells ml −1 ) (g, h, i), and prokaryotic heterotrophic activity (pM-leucine h −1 ) (j, k, l) between the different treatments at the last day of the experiment. Values are shown as mean ± SD. C, +N, +P, and +NP denote Control, NH 4 addition, PO 4 addition, NH 4 +PO 4 addition, respectively. Columns labeled by different italic letters (a or b) are significantly different at P < 0.05. NS denotes no significant effect of different nutrient treatments on the parameter measured. See Materials and methods for details. (Fig. 6g, i; Tukey HSD test, P < 0.05), while no significant effect of nutrient addition was detected at Stn B (Fig. 6h).

Discussion
Our investigation, at three offshore stations in the Mediterranean Sea, is the first on-board study of nutrientmanipulated microcosms that examined simultaneously major biogenic elements (C, N, P), the extent of availability of PO 4 and labile DOP, abundances of major functional groups (viruses to ciliates) in the microbial food web, and growth of phytoplankton and Hprok. Our results are the first that demonstrate that a P-starved status, as indicated by relative in situ concentrations, may not correspond with P-limitation. We found no significant differences indicative of P-limitation  in any measured biological parameters of the microbial food web after P-addition in three anticyclonic eddies of the Western, the Ionian, and the Levantine Basins.
In all three experiments, the initial ratios of N to P were higher than the Redfield Ratio at all three stations with regard to both DON to DOP (111-217) and PON to POP (21-30) ( Table 1). In contrast, the ratio of DIN to PO 4 (1.8) at Stn C was much lower than the Redfield ratio. Chl-a concentrations, PP, Hprok abundance, and PHA in the waters used in this study were within the previously reported range in offshore surface waters of the Mediterranean Sea during the stratified period (see the compiled data in Siokou-Frangou et al., 2010). This suggests that the sites studied were initially oligotrophic or ultra-oligotrophic with a mixed signal of either N or P limitation. However, the estimated DIN to PO 4 ratio at Stn C includes certain elements of uncertainty. Although concentrations of NH 4 were low but well above the detection limit of the nanomolar analytical technique, concentrations of NO 3+2 and PO 4 were close to the detection limit of the conventional technique used in this study. The molybdenum blue reaction method measures not only PO 4 but also has a potential background such as from acid labile Values are shown as mean ± SD. C, +N, +P, and +NP denote Control, NH 4 addition, PO 4 addition, NH 4 +PO 4 addition, respectively. Columns labeled by different italic letters (a, b or c) are significantly different at P < 0.05. NS denotes no significant effect of different nutrient treatments on the parameter measured. See Materials and methods for details.
DOP and arsenate (reviewed by Karl and Björkman, 2002). On the other hand, Marty et al. (2002) report that the NO 3 to PO 4 ratio in surface waters of the DYFAMED time-series station (an offshore site in the NW Mediterranean Sea) was highly variable in a range of 1-60 (observation period: 1991-1999), even when only the data on concentration of NO 3 and PO 4 that exceeds 0.1 µM (i.e. more than twice their analytical detection limits) were taken into account. The NO 3 to PO 4 ratios in Mediterranean surface waters which are lower than the Redfield ratio are a contrast with the NO 3 to PO 4 ratios in waters below the epipelagic layer (22-28, Krom et al., 1991;Béthoux et al., 1992;Kress and Herut, 2001;Pujo-Pay et al., 2010). This may suggest that surface offshore waters with low DIN to PO 4 ratios occur on limited spatial and temporal scales during the stratified period in the Mediterranean Sea. A P-starved status at all three sites was suggested by data on specific PO 4 affinity which is a measure of P availability for osmotrophs (cf. Thingstad and Rassoulzadegan, 1999). Tanaka et al. (2006) proposed that a specific PO 4 affinity > 0.02 l nmol-P −1 h −1 indicates P limitation, i.e. the growth rate of the existing osmotrophs (here, phytoplankton and Hprok community) is reduced because of low P availability. According to this criterion, the phytoplankton and Hprok communites were initially P-limited at all three stations, and the extent of P limitation tended to increase from west to east (Table 1). The P biomass estimate, which was used to determine specific PO 4 affinity, also includes elements of uncertainty (see Materials and methods). If specific PO 4 affinity is recalculated by assuming C/Chl-a ratio of 100 (e.g., Malone et al., 1993), and C/P ratio of 250 for phytoplankton, which is similar to POC/POP ratios in situ (Table 1), and C/P ratio of 150 for Hprok, which was found in P-limited conditions (e.g., Vrede et al., 2002), the estimate of specific affinity would increase by a factor of 2 at all three stations (i.e. an indication of enhanced P limitation). This suggests that the phytoplankton and Hprok community was initially P-limited at all three stations, despite the uncertainty in our estimate of specific PO 4 affinity. However, it should be noted that specific PO 4 affinity for the phytoplankton and Hprok "community" is a measure only of PO 4 availability. In other words, the extent of P availability based on specific PO 4 affinity values does not necessarily exclude the possibility that growth of the whole osmotroph community was colimited by P and other substrate nor the situation that specific osmotroph groups were limited by substrate(s) other than P in our study. Increases of PO 4 concentration in the Control and +N at Stn B during the incubation were unexpectedly high (up to ca. 6 times, Fig. 1e), and were not consistent with variations of organic P pools. PO 4 turnover times were relatively stable in the Control and decreased in +N at Stn B (Fig. 2b). Hence, we speculate that PO 4 contamination in these samples occurred outside rather than inside the microcosms. With the analytical detection limit and the precision for the NO 3+2 measurement in this study, significant changes of NO 3+2 concentration remained to be explained.
Our study demonstrated that no chemical and biological parameters measured during the microcosm experiments indicated P-limitation at any station. If the osmotroph community were limited by P alone, turnover times of PO 4 and ATP would have been similar between the Control and +N, and APA would not have been enhanced by NH 4 addition (Fig. 2). In a previous study, responses of PO 4 turnover time and APA were similar in the Control and the NH 4 addition treatments, when surface offshore waters of the Eastern Mediterranean Sea were manipulated by an in situ PO 4 addition in May 2002 . In that study, Hprok growth was P-limited and PP was N and P co-limited Zohary et al., 2005). Our results suggest that initially the availability of the PO 4 pool for osmotrophs was controlled by the availability of NH 4 (i.e. potential N-limitation) at the three stations. Hence, an NH 4 addition to the waters collected in this study enhanced the P requirement by the osmotroph community, by which turnover times of PO 4 and ATP decreased and APA increased in +N. In our experiments, whenever a significant difference between the different nutrient treatments was detected in a given parameter, the indication was almost always N-limitation or N and P co-limitation (Figs. 3-6).
We found that PP was consistently limited by N, while the limiting nutrient for Hprok growth was variable among the study sites: no nutrient limitation in the Western Basin, Nlimitation in the Ionian Basin, and N and P co-limitation in the Levantine Basin (Fig. 4). However, Ridame et al. (personal communication, 2011) found that P addition in their experiment significantly enhanced PP at 48 h at Stn B and Hprok growth at 24 h but not at 48 h at Stn A during 2-day incubation at the same cruise. While both our study and Ridame et al. study employed water collected at the same depth at the same sites, the experimental design and treatments were different (e.g., nutrient concentration and microcosm size). Hence, these differences in response of phytoplankton and Hprok between two studies remained to be explained. Ternon et al. (2010) found significant increases of PP in both the aerosol and the Saharan dust analog addition treatments compared to the Control at Stn A, B, and C in the same cruise. Since both analogs include N and P as well as other elements (e.g., Fe, Al), the limiting element was not evident in their study.
Responses to nutrient additions were also different among the stock parameters and between the study sites in our investigation (Figs. 3-6). The NH 4 addition resulted in increases of PON, Chl-a, Synechococcus, and HNF at Stn A, while at Stn B the NH 4 addition resulted in an increase of POC, PON, Chl-a, Synechococcus, and ANP but a decrease of Hprok abundance. At Stn C, the NH 4 addition alone resulted in an increase of PON and POP, but a decrease of Hprok abundance. A combined addition of NH 4 and PO 4 at Stn C resulted in an increase of POC, PON, POP, Chla, Synechococcus, ANP, and ciliates, but no change in Hprok abundance. Interestingly, despite N-limitation or N and P colimitation of Hprok growth, the abundance of Hprok was significantly smaller in the NH 4 addition compared to the Control (Fig. 4h, i). Hprok were dominated by a low DNA containing subpopulation at the start of the experiment, and the significantly smaller abundances of Hprok in +N and +NP compared to the Control were mostly because of little change or decrease of this subpopulation during the incubation (data not shown). From the same experiment done at Stn B, Talarmin et al. (2011) report that cell-specific PHAs of two high DNA containing subpopulations were 2 times higher in +N and +NP than the Control, but the proportion of these subpopulations to total Hprok abundance was similar between the Control, +N, and +NP at the end of the incubation. This suggests rapid top-down control on Hprok abundance. However, abundances of viruses, HNF, and ciliates were similar between the Control and +N at Stn B and Stn C (Fig. 6), while ANP abundance was significantly higher in +N than the Control at Stn B (Fig. 5h). This may indicate a tight trophic coupling in the heterotrophic compartments and a relative benefit for ANP (more growth and/or less grazing) over a short time.
The limiting nutrient in our sample waters was not the same as in the offshore water studied during the CYCLOPS Lagrangian P-addition experiment in the Eastern Mediterranean Sea (see Introduction, . Hence it was impossible to test in this study if the bypass and tunneling mechanisms for P exist in different sites of the Mediterranean Sea. On the other hand, the responses at the osmotrophs community level in +N treatment at Stn C (an increase of PP, no change of Chl-a and PHA, but a decrease of Hprok abundance) seem a mirror image of the results in the CYCLOPS experiment (an increase of PHA, no change of Hprok abundance and PP, but a decrease of Chl-a; Psarra et al., 2005). However, a rapid transfer of a limiting nutrient (N) to higher trophic level such as ciliates was not evident at Stn C. An existence of the bypass and tunneling mechanisms for N under N limited condition needs to be tested in future studies.
In the Eastern Mediterranean Sea, high amounts of NO 3 and PO 4 with N/P ratios of 22-28 is supplied to the epipelagic layer during the winter overturn, and annual phytoplankton bloom (November-March) results in an exhaustion of PO 4 and a NO 3 residual in the epipelagic layer when the water column stratification is established (Krom et al., 1991(Krom et al., , 2003(Krom et al., , 2010. In the same area, summer conditions are firmly established with a strong thermocline and a welldeveloped deep chlorophyll maximum by May . LNLC conditions develop in the euphotic zone above the chlorophyll maximum with concentrations of NH 4 (30-80 nM), NO 3+2 (< 1-10 nM), and PO 4 (< 2-4 nM) which are only detectable using nanomolar methods . Note again that Hprok were not limited by N likely because of the presence of measurable NH 4 and of bioavailable DON, while phytoplankton were N and P co-limited in surface offshore waters in May 2002 in the same area Zohary et al., 2005). The previous study which employed nutrient-manipulated bioassay experiments in July and September 1999 along the longitudinal transect in the Mediterranean Sea, found that Hprok growth is generally limited by P in surface offshore waters . Our study done in June/July indicated that PP was consistently limited by N, while Hprok growth was not nutrient limited in the Western Basin, Nlimited in the Ionian Basin, and N and P co-limited in the Levantine Basin (Fig. 4). Specific PO 4 affinities indicated Plimitation of the phytoplankton and Hprok community in the initial community (Table 1). A possible scenario for these seemingly disparate results would be that (1) as the stratified period progresses, the available pool of both inorganic N and P becomes very small in the surface mixed layer, (2) the plankton food web shifts to N and P co-limitation. Thus, growth of osmotroph groups is limited by different nutrients in a same system, and the most limiting nutrient for osmotroph growth shifts seasonally or sporadically among N, P and N + P (cf. Hecky and Kilham, 1988).
Our results show that while the C/N/P ratio of particulate organic matter consistently indicated P-starved status compared to the Redfield ratio, whenever nutrient limitation was detected, phytoplankton and Hprok never experienced P-limitation in surface waters in the center of anticyclonic eddies at the three Mediterranean Basins. These results suggest a large gap between biogeochemical features (P-starved status) and biological responses (no P-limitation). We question the general notion that Mediterranean surface waters are limited by P alone during the stratified period. However, surface waters in N-limitation or N and P co-limitation may create a niche for N 2 fixers that have recently been found in the Mediterranean Sea (Sandroni et al., 2007;Bar Zeev et al., 2008;Bonnet et al., 2011). If indeed organic matter in the surface layer is P-poor (skewed ratios of DON/DOP and PON/POP), but the microbial food web is either N-limited or N and P co-limited, are conditions which suggests a turnover time of organic N longer than that of organic P (cf. Thingstad and Rassoulzadegan, 1995). A required condition that permits N and P co-limited Hpok but N-limited phytoplankton in the same water is that N:P ratio in phytoplankton community is much higher than that in Hprok community (i.e. greater P requirement per cell volume by Hprok) (reviewed by Vadstein, 2000;Sterner and Elser, 2002) in surface waters. Further research is needed to investigate how the most limiting element for the pelagic plankton food web varies with time (spring phytoplankton bloom, development of the stratification, and decline of the stratification) and space (different depth layers in the euphotic zone, anticyclonic and cyclonic eddies, coast-offshore gradients, and different basins) in the Mediterranean Sea.