We have investigated here the structure of the pelagic microbial food web and quantified the carbon fluxes from viruses to microplankton along trophic gradients in the Mediterranean Sea. To explore the complex trophic pathways of the pelagic food web, we conducted independent and replicated experiments to measure (i) predation on prokaryotes by microzooplankton, (ii) predation on prokaryotes by heterotrophic nanoflagellates, (iii) virus-induced prokaryotic mortality, and (iv) microzooplankton grazing on nanoplankton and microphytoplankton. Our study covered more than 5000 km, from the Atlantic Ocean to the Levantine basin, and from conditions of high primary production and nutrient availability to ultraoligotrophic and phosphate-limited waters. Microphytoplankton abundance and biomass were typically scarce across the entire Mediterranean basin, with almost negligible levels in the eastern part. Also, nanoplankton biomass was typically low. Conversely, prokaryotes, and particularly the heterotrophic components, were abundant and represented the only significant food source for both nanoplankton and microplankton grazers. Viral infections were not the primary agents of prokaryotic mortality, but in some areas, such as the Ligurian Sea, they had a key role in prokaryotic dynamics. The scenario depicted in this study in summer reveals the pivotal role of microzooplankton in the pelagic food web of the Mediterranean Sea, with a key role in the potential transfer of biomass to higher trophic levels. We also show that converse to theoretical expectations, the microbial food web was relatively complex under the mesotrophic conditions (Atlantic and western Mediterranean) and was much more simplified in the ultraoligotrophic conditions of the eastern Mediterranean.
 Moving from the western to the eastern sectors, the Mediterranean basin has a major decreasing gradient of primary productivity, autotrophic biomass, and export of primary production to the seafloor [Danovaro et al., 1999; Dolan et al., 1999; Turley et al., 2000]. This gradient is linked to a significant decrease in nutrient concentrations. This is particularly evident for phosphorus, which can regulate primary production and can influence, to various extents, the food web and the functioning of the entire basin [Krom et al., 1991; Thingstad et al., 2005].
 Picoplankton are a class of organisms of size ranging from 0.2 µm to 2.0 µm, which includes heterotrophic prokaryotes, autotrophic prokaryotes [mainly Synecchococcus, Johnson and Sieburth, 1979; and Prochlorococcus, Chisholm et al., 1988], and some small eukaryotes. In the nutrient-poor and stratified pelagic systems, the planktonic biomass is dominated by picoplankton [Azam and Malfatti, 2007]. These microbial components are expected to regulate the energy flow through the food web within the oligotrophic pelagic systems [Azam et al., 1983; Turley et al., 2000], limiting the export of primary production to the benthos [Kiørboe, 1996]. Prokaryotes are the numerically dominant component of the picoplankton, which are at the base of the microbial food web, as they use dissolved organic matter to convert it into biomass that is then potentially available for the consumers [Azam et al., 1983]. Within the microbial food webs, predation is size selective [Chrzanowski and Šimek, 1990; Gonzalez et al., 1990; Simek and Chrzanowski, 1992; Hahn and Höfle, 2001]. Indeed, picoplankton are preyed upon by heterotrophic nanoflagellates, which include phagotrophic protists [and especially bacterivorous heterotrophic nanoflagellates smaller than 5 µm; Strom, 2000], and/or by microzooplankton [Fenchel, 1982; Sherr and Sherr, 1994]. Predation is one of the key factors that control the prokaryotic assemblages (in terms of both biomass and structure) and nutrient regeneration [Simek et al., 1999]. Microzooplankton (10–200 µm) account for a large portion of the total zooplankton biomass in a variety of oceanic environments [Odate and Maita, 1988; Tsuda et al., 1990; Booth et al., 1993], and they include several taxa (such as copepod nauplii and rotifers, and ciliates that are either loricate or aloricate) and heterotrophic dinoflagellates that can be preyed upon by mesozooplankton. Microzooplankton thus have two major roles in marine ecosystems: (1) they represent a trophic link between piconanoplankton and mesozooplankton [Gifford, 1991; Gifford and Dagg, 1991], and (2) they remineralize the particulate organic matter in the surface waters [Ducklow et al., 1986; Sanders and Wickham, 1993].
 As well as predation in a strict sense (i.e., larger predators), viral infections might also be responsible for a significant portion of prokaryotic mortality [Weinbauer et al., 2003]. Viral infections are apparently more important in meso-eutrophic systems that are characterized by high prokaryotic metabolism [Corinaldesi et al., 2003] and, in certain systems, including deep-sea sediments; they can be the main cause of prokaryotic mortality [Corinaldesi et al., 2007; Danovaro et al., 2008]. The viral shunt is antagonistic to the microbial food web, as virus-induced prokaryotic mortality converts microbial biomass into dissolved organic matter, thus reducing the flow of energy and biomass to the higher trophic levels [Wilhelm and Suttle, 1999]. Indeed, the effects on the functioning of pelagic ecosystems and biogeochemical processes in these systems can be completely different, depending on whether prokaryotic mortality is predominantly caused by nanomicropredators or by viral infection. Moreover, since nanosized and microsized predators can feed on both heterotrophic and autotrophic nanoplankton and picoplankton, it is evident that there are several parallel pathways for transferring microbial production into the higher trophic levels [Tanaka and Rassoulzadegan, 2002].
 The high number of players within the microbial food web and the large number of variables that can affect each biotic component makes it extremely difficult to quantify the importance of each component and, thus, to clarify the pathways of biomass transfer across the trophic levels. On this basis, there are several questions that still remain to be answered, such as How does the nanoplankton and microzooplankton predation on prokaryotes change across the gradient of trophic conditions? Do nanoplankton compete with microzooplankton for the available resources? Are nanoplankton preyed upon by microzooplankton, thus fueling the microbial food web? Does the viral shunt impact on the microbial food web?
 In this study, we quantified the carbon flux throughout the microbial food web along a gradient of trophic conditions across the Mediterranean basin, from the Atlantic Ocean to the Aegean Sea, to determine the quantitative relevance of the trophic interactions among the different biotic components. To explore the complex trophic pathways of the pelagic microbial food web, we conducted independent and replicated experiments to measure (i) predation on prokaryotes by microzooplankton, (ii) predation on prokaryotes by heterotrophic nanoflagellates alone, (iii) virus-induced prokaryotic mortality, and (iv) consumption of nanoplankton and microphytoplankton by microzooplankton.
2.1 Sampling Area
 The Mediterranean Sea covers 2542 × 106 km2, with an average depth of ~1500 m. The average salinity is close to 38.5, although it is variable in the surface layers. The temperature of the superficial layers typically varies between ~13°C in winter and ~26°C in summer, and the water is well oxygenated throughout [Margalef, 1985]. The oligotrophy increases along both the west-east and the north-south directions. The nutrient availability is low, especially for phosphorous (N:P, up to 60), although this limitation can be buffered by the input from highly populated coasts and from the atmosphere [Siokou-Frangou et al., 2010]. Phosphorus is the most limiting factor for primary production [Thingstad and Rassoulzadegan, 1995]: its mean concentration in the deep waters is only 0.2 µm, while in the surface layers, this can be much less, and it can even be undetectable. Indeed, the eastern Mediterranean is one of the most nutrient-depleted areas in the world [Krom et al., 1991; Turley et al., 2000].
 During an oceanographic cruise carried out from 28 May to 28 June 2007, nine sites from the Moroccan Atlantic coast to the eastern Mediterranean Sea were identified (Figure 1) that follow a longitudinal trophic gradient (from west to east). The sampling locations and depths are reported in Table 1.
Table 1. Sampling Locations, Temperature (T), Salinity, Chlorophyll a (Chl a), Primary Production (PP), and Inorganic Nutrients for the Different Sampling Sites of the Mediterranean Sea
2.1.1 Environmental Variables, Primary Production, and Chlorophyll a
 The temperature, salinity, oxygen, and fluorescence were measured using an SBE911 multiparametric probe that was coupled to an Sea Bird Electronics (SBE) sampler carousel. This was equipped with a 24-Niskin-bottle rosette; each 12 L Niskin bottle was fitted with silicon elastic ribbon and red silicon O-rings.
 Samples were taken to determine the dissolved inorganic nutrients (NO3, NO2, SiO2, and PO4), which were stored at −20°C until laboratory analysis.
 The primary production was measured in situ along the water column, according to the Steemann-Nielsen  method. Seawater (500 mL) was collected using glass bottles that were lowered into the seawater fastened to a graduated rope (with surface buoys). The seawater samples were inoculated with 14C and incubated in situ for 4 h. Then, the seawater was filtered according to different sizes through Whatman GF/F filters (diameter, 25 mm) and polycarbonate filters and then stored at −20°C until analyzed. Samples for chlorophyll a analysis were taken as for the primary production, with 5 L samples collected and filtered (in three replicates) onto glass fiber filters (GF/F) and polycarbonate filters, according to the different sizes. These samples were immediately stored in liquid nitrogen.
 Qualitative and quantitative analyses of the microphytoplankton and microzooplankton were carried out using a Leitz Diavert inverted microscope, at a magnification of 320X [Utermöhl, 1958]. The original samples of seawater (2 L) were presedimented for 72 h and then concentrated to ~200 mL by gentle aspiration of the supernatant. After homogenization, 100 mL of the sample was settled in a sedimentation cylinder for at least 72 h (3 h of sedimentation × 1 cm length). The analysis was performed on three replicates, with half of the sedimentation chamber analyzed for the autotrophic fraction and the whole of the sedimentation chamber analyzed for the heterotrophic fraction.
 Tintinnids were identified on the basis of the works of Kofoid and Campbell [1929, 1939] and Marshall ; aloricate ciliates, foraminiferans, radiolarians, and acantharians were identified following Brandt and Apstein . The identification of the metazoan larvae and copepod nauplii was based on the work of Trègouboff and Rose , with the descriptions of Hasle and Syvertsen  and Steidinger and Tangen  adopted for the dinoflagellates. Diatoms were identified according to Rampi and Bernhard  and Hasle and Syvertsen . The term nanociliates is used in this study to describe ciliates <20 µm [Pitta et al., 2001]. Carbon biomass was estimated after determination of the biovolumes, based on standard geometric shapes, and using the specific conversion factors that are available in literature.
 The nanoplankton were preserved in 1% glutaraldehyde and processed as described by Verity et al. . The samples were filtered onto black 0.8 µm polycarbonate filters (New Technology Group (NTG)) that were positioned on 1.2 µm nitrocellulose backing filters (Millipore), and the volumes were adjusted according to the dilutions. The cells were stained with 4'6-diamidino-2-phenylindole (DAPI, Sigma) at a final concentration of 1 µg mL−1 and stored at −20°C before being processed. The nanoplankton (three replicates) were counted using a 100X oil-immersion objective and an Olympus BX60 F5 epifluorescence microscope equipped with a 100 W high-pressure mercury burner (HPO 100 W/2). The total nanoplankton were fractionated according to dimensional sizes: <3 µm, 3 µm to 5 µm, and >5 µm. The abundance of the nanoplankton was determined under a UV filter set (365 nm), with at least 100 cells per filter counted. The carbon content for each size was calculated using a conversion factor of 183 fg C µm−3 [Caron et al., 1995].
 The picoplankton samples were preserved in prefiltered 0.2 µm formaldehyde (Acrodisc® syringe filters) at 2% final concentration. They were then filtered onto black 0.2 µm polycarbonate membrane filters (NTG) laid over prewetted 0.45 µm nitrocellulose backing filters (Millipore). The analysis was carried out following the method of Porter and Feig , with some modifications. The samples for the determination of the abundance of the heterotrophic prokaryotes were stained for at least 15 min in the dark with DAPI (final concentration, 1 µg mL−1); the samples for the determination of autotrophic picoplankton abundance were filtered separately. Nine replicates of heterotrophic prokaryote samples were filtered, while three replicates were filtered for the autotrophic picoplankton. The filters were maintained at −20°C until they were processed. The picoplankton enumeration was carried out using a 100X oil-immersion objective with an Olympus BX60 F5 epifluorescence microscope (HPO 100 W/2). The abundance of the autotrophic picoplankton was determined using blue (450–490 nm) and green (480–550 nm) excitation light, with counting of at least 150 cells; the abundance of the heterotrophic prokaryotes was determined under a UV filter (365 nm), with counting of at least 200 cells per filter. The cell numbers of the heterotrophic prokaryotes and autotrophic picoplankton were converted to carbon biomass using the factors of 20 fg C cell−1 [Ducklow and Carlson, 1992] and 200 fg C cell−1 [Caron et al., 1995], respectively.
2.5 Viral Abundance
 The viral counts were carried out as described by Patel et al. , without previous sample fixation to avoid underestimation of the viral abundance [Danovaro et al., 2001; Wen et al., 2004]. At each site, three independent samples were collected, frozen immediately, and stored at 4°C until laboratory analyses. For the slide preparation, subsamples (300–1000 μL, depending on the expected viral abundance) were filtered in triplicate using 0.02 µm Anodisc filters and then immediately stained with 20 μL SYBR Gold [stock solution diluted 1:5000; Chen et al., 2001]. The filters were incubated in the dark for 15 min and then mounted on glass slides with 20 μL sterile antifade made with 1:1 glycerol:phosphate-buffered saline, containing 0.5% ascorbic acid (wt/vol). The slides were stored at −20°C until the epifluorescence microscopy analyses (Zeiss Axioplan; magnification, 1000X). The viral counts were obtained by examining at least 10 fields; i.e., at least 200 viruses per replicate. The viral abundance is expressed as the number of viruses per liter of seawater.
2.5.1 Analysis of the Trophic Interactions Among the Different Microbial Components
 In the present study, to determine the relative quantitative importance of the different predators, and to provide an estimate of the main pathways of C flow through the pelagic microbial food webs (under the different environmental conditions), we determined (i) the abundance of the predators (i.e., microzooplankton, nanoplankton, and viruses) and their potential prey (i.e., heterotrophic prokaryotes, autotrophic picoplankton, nanoflagellates, and microphytoplankton), (ii) the specific growth and grazing rates of the microplankton and nanoplankton grazers and the specific growth and mortality rates of the prey, and (iii) the viral production rates and the virus-induced prokaryotic mortality.
2.5.2 Microzooplankton and Heterotrophic Nanoplankton Grazing
 The sampling was carried out using an SBE sampler carousel. Before all of the experiments, the equipment was washed in dilute HCl and rinsed with MilliQ water, the silicone tubing was cleaned and rinsed, and all of the incubation vessels were sterilized by autoclaving. The grazing of the microzooplankton and heterotrophic nanoflagellates on the heterotrophic prokaryotes and autotrophic picoplankton, as well as the grazing of the microzooplankton on the microphytoplankton and nanoplankton, were determined using separate dilution experiments [Landry and Hassett, 1982]. The dilution approach relies on the reduction of the encounter rates between the grazers and their prey. Varying proportions of 0.22 µm filtered seawater were added to natural water samples to create a dilution series, and the grazing rate was estimated as the increase in the apparent growth rate of the prey (i.e., the heterotrophic prokaryote abundance over time) with increasing dilution (i.e., decreasing heterotrophic nanoflagellate abundance). The grazing rate of the predators was estimated as the slope of a regression of the apparent growth of the prey in the various dilutions against the dilution factor. The growth rate of the prey was estimated as the apparent growth rate extrapolated to 100% dilution (growth in the absence of grazers) [Dolan and McKeon, 2005]. This approach is based on three restrictive assumptions: (1) The exponential model is assumed to apply in the broad sense. This makes allowance for the growth and mortality rates that might vary over short time scales but provides a framework for computing the average rates over the incubation periods of the order of a day. (2) The average specific growth rate of the prey is assumed to be constant, and density independent. To satisfy this assumption, the dissolved nutrients must remain nonlimiting, or equally limiting, to the growth at all of the dilutions during the incubation. (3) The average clearance rate of the individual consumers is assumed to be constant at all of the dilutions, whereby the ingestion rates of the grazers are directly related to the prey density. The variation of the density of the prey over a fixed period of time is represented by the following exponential equation, based on Landry and Hassett  [e.g., Landry, 1993]:
where C0 is the carbon concentration of the prey at the beginning of the experiment, Ct is the carbon concentration at the end (time t), k is the apparent growth coefficient, and g is the grazing coefficient. When the parameters such as the concentration of the prey at the beginning of the experiment (C0), the instantaneous coefficient of population growth (k), and the grazing coefficient (g) are measured, the ingestion rates can be calculated [Ruiz et al., 1998]. To avoid mesozooplankton grazing immediately after collection, we gently prefiltered the samples through a 200 µm mesh net (into polypropylene 20 L carboys). To obtain particle-free water, we filtered the water from the same samples on 0.22 µm pore size Millipore filters (diameter, 142 mm) using a low-vacuum peristaltic pump. Varying proportions of 200 µm filtered seawater were added to particle-free water to create a dilution series according to four levels (100%, 80%, 50%, and 20% whole water). For each dilution level, we performed three replicates using 12 polycarbonate bottles (2 L) and, in addition, three bottles of undiluted seawater (blanks; 100%) to verify the growth of the microzooplankton (secondary production). This incubation time was set at 48 h in a flowing seawater incubator, with the in situ conditions of temperature and light maintained. The set of 12 Nalgene bottles was incubated for 24 h using the same procedure. After the incubation, the samples were preserved in 2% buffered formaldehyde and stored in the dark at 5°C. The initial samples (C0) were filled with the same procedure indicated for the incubation bottles. For the picoplankton fraction only, the preservative was prefiltered using a 0.2 µm Acrodisc® syringe filter. Nutrients (i.e., 5 µm NaNO3 and 1 µm KH2PO4) were added equally to each incubation bottle. The composition and carbon biomass (µg C L−1) was compared at the beginning (C0) and the end (C48) of the experiment to determine the microzooplankton growth (secondary production); three replicates of the C48 samples (100% whole water) were treated as the initial samples (C0).
 To calculate the heterotrophic nanoflagellates grazing on the prokaryotes, and to measure the secondary production of this component, another set of dilution experiments (four dilution series × three replicates, plus three blanks containing undiluted water) was performed at C0 and C24, with mixing of the particle-free water with the same assemblage previously filtered through a 10 µm mesh nylon sieve to remove the larger-sized predators. All of the incubations were carried out without nutrient addition for 24 h in the same incubator. The samples for the heterotrophic nanoflagellate analysis were preserved in 1% buffered glutaraldehyde. Special care was taken to be certain of the performance of the dilutions. Regression analysis was carried out for each single variable determined at each dilution.
2.5.3 Viral Production and Virus-Induced Prokaryotic Mortality
 The viral production was determined by the dilution technique [Wilhelm et al., 2002], with some modifications. Three water samples (50 mL) were transferred in Whirl-Pak bags, mixed with 100 mL virus-free seawater (0.02 µm prefiltered using sterile Anotop filters), and incubated in the dark. Four subsamples (10 mL) were collected: at the beginning of the incubation (t0), and after 2 h (t1), 3 h (t2), and 6 h (t3). The rationale behind this approach is that by reducing the viral and host densities, the occurrence of new infections is also reduced. At the same time, sample dilution makes the effects of the predators on the viruses almost negligible and reduces the viral losses due to enzymatic degradation. As an advantage, the temporal changes in viral abundance can be directly estimated without using conversion factors [Wilhelm et al., 2002]. As the increases in the viral abundance were not linear over the first 6 h of incubation, the net viral production was determined as the maximum increment of viral abundance per liter of seawater in all of the samples [Dell'Anno et al., 2009]. The virus-induced prokaryotic mortality was calculated according to Fonda-Umani et al. , as follows:
where VIPM is the virus-induced prokaryotic mortality, VP is the viral production (expressed as viruses mL−1 d−1), BS is the burst size (i.e., the average number of viruses released by a single prokaryotic cell), and TPC is the total prokaryotic abundance. We assumed a mean prokaryotic burst size of 30, which was obtained from the burst size values reported for surface seawater of the Mediterranean Sea [Weinbauer et al., 2003]. The estimates of the carbon released by the killed prokaryotes were obtained at each station by multiplying the number of killed prokaryotes (obtained by the ratio between VP and BS; see above) by the prokaryotic cellular carbon content, which was assumed to be 20 fg C cell−1 [Corinaldesi et al., 2010].
2.6 Data Analysis
 The apparent growth rates of the picoplankton, nanoplankton, microphytoplankton, and microzooplankton, and the heterotrophic nanoflagellates grazing rates, were processed following Model I regressions of the apparent growth against the dilution factor, based on the work of Landry and Hassett  [e.g., Landry, 1993]. Heterotrophic nanoflagellate and microzooplankton growth rates were based on the assessment of their carbon concentrations in the whole-water samples (100%) at the beginning and the end of the incubations. We recalculated the rates of heterotrophic prokaryote mortality induced by microzooplankton and heterotrophic nanoflagellates for the effect of viral lysis by subtracting the heterotrophic prokaryotes killed by the viruses from the initial heterotrophic prokaryote standing stocks.
3.1 Environmental Variables
 The temperature ranged from 18.29°C to 24.33°C at sites V3 and V10, respectively, with an increasing trend from the western to the eastern sectors. The salinity showed a similar pattern, with a minimum at the Atlantic site (36.51) and a maximum at the easternmost site V10 (39.36) (Table 1). Both the low temperature and the high salinity registered at the surface of site V3 can be attributed to the wind events that mixed the whole water column, bringing the dense and cold water to the surface. The chlorophyll a concentration was typically low, and ranged from 0.04 µg L−1 to 0.22 µg L−1. The maximum was reported at the Ligurian site (site V1), where the highest primary production (2.19 µg C L−1 h−1) was also registered. As typical in this season, the surface water was nutrient depleted: the nitrates ranged from 0.04 µm to 0.78 µm at sites VA and V6, respectively, the nitrites were below the detection limit in almost the whole of the western basin, the silicates ranged from 0.46 µm to 1.92 µm at sites V4 and V1, and the phosphates were at the limit of detection, ranging from 0.01 µm to 0.18 µm at sites V7 and Viera, and at site V4, respectively.
3.2 Microzooplankton Variables
 Within the microzooplankton, there were aloricate ciliates, tintinnids, dinoflagellates, foraminiferans, radiolarians, acantharians, and micrometazoans (i.e., copepod nauplii, copepodites, and meroplanktonic larvae) (see Appendix S1 in the Supporting Information). The total microzooplankton abundance was, on average, always <150 individuals (ind.) L−1 in the Mediterranean basin, whereas at the Atlantic site, the abundances reached 207 ind. L−1 (Figure 2a). The most abundant were the aloricate ciliates, with abundances varying from a minimum of 27 ind. L−1 (site V2) to a maximum of 115 ind. L−1 (site V4). The second most important group was the heterotrophic dinoflagellates, with abundances ranging from 19 ind. L−1 to 35 ind. L−1 at sites V4 and VA, respectively. Most of the predator community was characterized by forms of medium (30–50 µm) and small (<30 µm) sizes. Tintinnids declined from the Atlantic site to the Mediterranean basin and showed low concentrations within the west-east transect (always <21 ind. L−1 recorded at site V1). In the western basin, the total microzooplankton biomass showed a minimum at site V2 (0.24 µg C L−1) and a maximum at site VA (0.84 µg C L−1), while in the eastern basin, the biomass was less variable (0.50–0.63 µg C L−1), although the differences between the basins were not significant (analysis of variance (ANOVA), not significant (n.s.)) (Figure 3a). The microzooplankton growth rates were determined for most of the taxa in the western Mediterranean, whereas in the eastern basin, significant growth rates were recorded only for the heterotrophic dinoflagellates, tintinnids, and other protozoans. The secondary production of microzooplankton was low: typically <0.10 µg C L−1 d−1 in the eastern basin and <0.80 µg C L−1 d−1 in the western basin (Figure 3a and Table 2 in Appendix S2 of the Supporting Information).
3.3 Microphytoplankton Variables
 The microphytoplankton assemblage was divided into six main taxonomic groups: diatoms, autotrophic dinoflagellates, coccolithophorids, euglenoids, silicoflagellates, and incertae sedis (see Appendix S3 in the Supporting Information for the taxonomic list). Along the trophic decreasing gradient, the diatoms shifted from centric to pennate forms of medium to small sizes. The microphytoplankton abundance was always very low: <4 × 103 cells L−1 (Figure 2b) in the western basin and below the detection limit in the eastern basin. The biomass of microphytoplankton in the western basin varied from 0.02 µg C L−1 to 0.47 µg C L−1 at sites V4 and VA, respectively (Figure 3b). Apparent growth for the total microphytoplankton was observed at five sites and varied between 0.02 µg C L−1 d−1 and 0.42 µg C L−1 d−1 (Figure 3b).
3.4 Heterotrophic Nanoflagellate Variables
 The heterotrophic nanoflagellate abundance varied from 0.3 × 106 cells L−1 (at site V10) to a maximum of 1.2 × 106 cells L−1 (site V1; Figure 2c). Cells <5 µm (the most important predators of the picoplankton) prevailed at all of the sampling stations. The heterotrophic nanoflagellate biomass showed a minimum at site V10 (1.05 µg C L−1) and a maximum at site V1 (8.7 µg C L−1) and followed the west-east decreasing trophic gradient, although the differences between the basins were only significant at p < 0.05 (ANOVA, p = 0.012). Considerable levels of secondary production were detected at seven out of the nine sites, and this varied from 0.11 µg C L−1 d−1 to 3 µg C L−1 d−1 (Figure 3c and Table 2 in Appendix S2 of the Supporting Information).
3.5 Picoplankton Abundance and Production
 The autotrophic picoplankton ranged from 2.1 ± 0.5 × 106 cells L−1 (site Viera) to 4.4 ± 0.4 × 106 cells L−1 (site V2) (Figure 2d); the heterotrophic prokaryotes ranged from 2.44 ± 0.08 × 108 cells L−1 (site Viera) to 6.57 ± 0.3 × 108 cells L−1 (site V1) (Figure 2e). The autotrophic biomass varied from 0.42 µg C L−1 (site Viera) to 0.88 µg C L−1 (site V2) (Figure 3d), although there were no significant differences between the basins (ANOVA, n.s.). The heterotrophic biomass ranged from 4.87 µg C L−1 (site Viera) to 13.15 µg C L−1 (site V1) (Figure 3f). The differences between the basins were highly significant (ANOVA, p < 0.001). The heterotrophic prokaryote production was recorded at all sites except for V1, and it varied from 0.54 µg C L−1 d−1 to 24.11 µg C L−1 d−1 (Figure 3e). Autotrophic production was detected at three sites (VA, V2, and V3), and this ranged from 0.11 µg C L−1 d−1 to 0.29 µg C L−1 d−1 (Figure 3d).
3.6 Viral Abundance and Production
 The viral abundance in the surface waters ranged from 6.2 ± 0.99 to 7.1 ± 0.98 × 107 viruses L−1 (at sites V10 and V3, respectively), although no significant differences were observed among the sites (ANOVA, n.s.) (Figure 2e). Conversely, the viral production displayed a higher variability among the sites with the lowest values, at V4, V1, and V7, and the highest value in the western basin (6.2 ± 1.5 × 108 viruses L−1 d−1 at V3 site) (on average, 3.1 × 107 viruses L−1 d−1; ANOVA, n.s.) (Figure 3f).
3.6.1 Microzooplankton Grazing on Microphytoplankton
 The microzooplankton grazing on the microphytoplankton was estimated only in the western Mediterranean, as microphytoplankton abundances in the eastern basin were not sufficiently high to provide reliable results. The data from the dilution experiments are shown in Table 3 in Appendix S4 of the Supporting Information. At site VA, significant correlations were detected between the apparent growth rate and the dilution factor for most of the microphytoplankton assemblage and, respectively, for pennate diatoms (p < 0.001), centric diatoms (p < 0.01), and coccolithophorids (p < 0.001); a less significant correlation was displayed for armored dinoflagellates <20 µm and for other flagellates >20 µm (p < 0.05). At site V4, there was a significant correlation only for armored dinoflagellates <20 µm (p < 0.001), whereas at site V3, for all of the classes of the armored dinoflagellates and for the coccolithophorids (p < 0.001). At both sites V1 and V2, there was a positive correlation between the apparent growth rate and the dilution factor only for the diatoms (p < 0.05). The ingestion rates at the Atlantic site and in the western basin varied between 0.03 µg C L−1 d−1 and 0.60 µg C L−1 d−1 at sites V4 and VA, respectively (Figure 4 and Appendix S4 in the Supporting Information).
3.6.2 Microzooplankton Grazing on Nanoplankton
 At the Atlantic site (VA), there was a significant correlation between the apparent growth rate and the dilution factor only for the heterotrophic nanoflagellates in the size range of 3–5 µm (p < 0.01), while within the Mediterranean waters, significant correlations were found at sites V1 and V6 for all of the size classes (p < 0.01). At site V4, there was a significant correlation only for the heterotrophic nanoflagellates >5 µm (p < 0.01), and at sites V7 and V10, there was a significant regression for the two dimensional sizes (<3 µm and 3–5 µm, p < 0.01), whereas there was no significant regression detected for any of the three dimensional sizes at sites V3, V2, and Viera. When detected, the ingestion rates ranged from 0.79 µg C L−1 d−1 to 17.8 µg C L−1 d−1 at sites VA and V1, respectively (Figure 4 and Appendix S4 in the Supporting Information).
3.6.3 Microzooplankton Grazing on Prokaryotes
 Except for site V1, there were significant correlations between the heterotrophic prokaryote apparent growth and dilution factors at all of the sites. On the contrary, only at three sites (VA, V3, and V2) were there significant correlations for autotrophic picoplankton. In the eastern Mediterranean, the instantaneous growth coefficient of the prey (k) was higher than the grazing coefficient (g), which indicated that, despite the active grazing of the predators, there was no control on their prey. Conversely, in the western sector, there was a top-down control of both the heterotrophic prokaryotes and the autotrophic picoplankton. The ingestion rate of the heterotrophic prokaryotes was undetectable at site V1, while at the other sites, it ranged from a minimum of 3.52 µg C L−1 d−1 (site V3) to a maximum of 27.14 µg C L−1 d−1 (site V4). The ingestion rate on the autotrophic picoplankton was undetectable at sites V4, V1, V6, V7, V10, and Viera and varied from 0.31 µg C L−1 d−1 to 0.48 µg C L−1 d−1 at sites V2 and VA, respectively (Figure 4 and Appendix S4 in the Supporting Information). The recalculated ingestion rates for the effects of viral lysis did not significantly change, because of the low mortality due to viral infection.
3.6.4 Heterotrophic Nanoflagellates Grazing on Picoplankton
 The grazing rates of the heterotrophic nanoflagellates (by removing the microzooplankton) on the heterotrophic prokaryotes are reported in Table 4 in Appendix S5 of the Supporting Information. All of the regressions are significant (p < 0.001). There was no correlation for the autotrophic picoplankton. In five of the nine sites (i.e., V4, V3, V2, V10, and Viera), the apparent growth coefficient of the prey (k) was higher than the grazing coefficient (g), and even if they were actively feeding, the predators were not able to control the biomass of the prey. Conversely, in the remaining stations, g was higher than k, which suggested that heterotrophic nanoflagellates exerted top-down control on the heterotrophic prokaryotes. The ingestion rates showed a minimum at site V3 (0.03 µg C L−1 d−1) and a maximum at site VA (9.55 µg C L−1 d−1). The recalculated ingestion rates of the nanoplankton for the viral lysis also did not significantly change in this case.
3.6.5 Virus-Induced Prokaryotic Mortality
 The virus-induced prokaryotic mortality ranged from 0.2% to 3.8% killed prokaryotes per day (at sites V4 and V3, respectively) (Figure 5). At site V1, the prokaryotic mortality was due entirely to viral lysis.
4.1 Structure and Interactions Among the Different Components of the Microbial Food Web in Open Waters
 The surface waters of the Mediterranean basin during summer were characterized by low chlorophyll a concentrations and evident phosphate limitation. Our chlorophyll a data were always in the lower range of the values reported for summer 2008 by Christaki et al.  during the trans-Mediterranean Biogeochemistry from the Oligotrophic to the Ultraoligotrophic Mediterranean (BOUM) cruise. The prokaryotes dominated the microbial assemblages in terms of the biomass, and the analysis of the spatial distribution highlighted almost double the prokaryotic biomass in the surface waters of the western Mediterranean (~12 µg C L−1), when compared to the eastern basin (~6 µg C L−1). Christaki et al.  reported that in the summer of 2008, the abundances were of the same order of magnitude, although they did not report any significant differences between the two basins.
 Our data indicate that the naked ciliates (such as Strombidium spp. and Strombilidium spp.) accounted for about half of the microzooplankton biomass, whereas the tintinnids were quantitatively much less relevant (as also observed by Christaki et al.  during the BOUM cruise), and again, our data fall into the lower range of those reported for summer 2008. The importance of mixotrophic ciliates (e.g., Tontonia spp. and Strombidium conicum) increased in the ultraoligotrophic eastern Mediterranean Sea, as already seen by Dolan et al. , Pitta and Giannakourou , and Christaki et al. . The microzooplankton grazing is generally considered one of the major factors that control the microphytoplankton dynamics [e.g., Calbet and Landry, 2004; Calbet, 2008]. Conversely, in our study, the microphytoplankton represented a negligible fraction of the microzooplankton diet, accounting for <1% of the microzooplankton ingestion. Only at the Atlantic site and at one site of the western Mediterranean where the nutrient availability was higher, did the microphytoplankton accounted for ~10% of the total microzooplankton ingestion rate. Despite its minor relevance, the grazing of the microzooplankton on the microphytoplankton was selective toward the armored dinoflagellates (>20 µm). Moreover, the analysis of the coefficients of growth (k) and mortality (g) indicated that the microphytoplankton were always controlled by predation. Our incubation experiments revealed that the abundance of the microzooplankton increased with increasing incubation time (48 h) and that significant growth rates were recorded across the entire basin. Therefore, despite the scarce availability of prey, the microzooplankton grew, and maybe for some components, also using their mixotrophic ability. This microzooplankton growth was detected during summer experiments also in another Mediterranean area (Gulf of Trieste and northern Adriatic), which suggested higher reproduction rates at higher temperatures [Fonda-Umani et al., 2012].
 Previous studies have reported that small-sized ciliates (<30 µm) can ingest 72% of picoplankton and 28% of nanoplankton biomass per day, whereas larger ciliates (30–50 µm) feed preferentially on nanoplankton [Rassoulzadegan et al., 1988]. However, we found here that the nanoplankton contribution to the microzooplankton diet was relatively low (range, 0% to 16% d-1). Nonetheless, given the low abundance of the nanoplankton, the microzooplankton grazing controlled the nanoflagellate dynamics in the open waters across most of the areas of the Mediterranean basin. The picoplankton were the most important component (70%–100%) of the microzooplankton diet at almost all of the sites, due to the strong grazing over the heterotrophic prokaryotes, while the consumption of the autotrophic picoplankton was almost negligible.
4.2 Methodological Limitations
 Although the dilution protocol is considered a “standard” method for determination of microzooplankton grazing [Dolan et al., 2000], it has been greatly debated over the past 20 years [see Bamstedt et al., 2000; Strom, 2000; Caron, 2001; Jürgens and Massana, 2008] due to the potential problems in the determination of the heterotrophic prokaryote mortality that is caused by both heterotrophic nanoflagellates and microzooplankton, mainly because of the enrichment in the dissolved organic matter due to the filtration steps. Sloppy feeding has also been shown to influence the trophic cascade by providing a source of nutrients, and this is greatly altered in these experiments. We are aware that the results of the dilution approach should be treated with caution, due to the intrinsic difficulty in reproducing experimentally natural conditions. However, they do allow the relative impact of the viruses, heterotrophic nanoflagellates, and microzooplankton on the prokaryotes to be disentangled, and the comparison of the potential predatory control in the different biogeographic regions, which was the main aim of the present study.
4.2.1 Food Web Complexity and Main Pathways of Biomass Transfer
 The spectrum of the prey consumed by the microzooplankton changed in moving across the west-to-east trophic gradient. In particular, while at the Atlantic site, the microzooplankton preyed upon different biotic components (e.g., microphytoplankton, nanoplankton, autotrophic, and heterotrophic picoplankton), in the Mediterranean Sea, and especially in the ultraoligotrophic conditions of the eastern basin, the microzooplankton relied on a few typologies of prey. This reaches the extreme situation at site Viera, where microzooplankton predation was observed only on the heterotrophic prokaryotes. The prokaryotic mortality rates due to heterotrophic nanoflagellate predation alone (i.e., after the removal of the microzooplankton grazers) was very high, particularly at the Atlantic site, and was apparently promptly transferred to microzooplankton biomass by predation of the microplankton over the nanoplankton component. The viruses also contributed to prokaryotic mortality, although to a lesser extent than for the microzooplankton grazing. On average, <2% of the prokaryotic abundance was killed per day by the viruses, which indicated that most of the prokaryotic biomass remained available for the larger predators (i.e., the nanoplankton and microzooplankton). Nonetheless, the impact of viral infections was important in certain regions, such as at the Atlantic site and in the western Mediterranean Sea. This was especially the case when the strong grazing by the microzooplankton over the heterotrophic nanoflagellates reduced the heterotrophic nanoflagellate grazing over the prokaryotes, which left room for a pivotal role of viruses in controlling the prokaryotic dynamics through viral lysis (up to 100% of prokaryotic mortality due to viral infections). The sites where the viruses become the key players in prokaryotic mortality are characterized by the upwelling phenomena that is associated with the presence of deep-sea canyons (e.g., the Ligurian Sea) or the intense mixing between the surface and deepwater masses (e.g., the Atlantic site), which might create different conditions. Future dedicated studies will allow us to clarify these aspects and the relevance of the physical regimes in the modulation of the importance of viral infections in picoplankton mortality.
 Overall, our findings suggest that the prokaryotes in surface waters of the Mediterranean Sea are tightly controlled by microzooplankton predation (Figure 6), which results in the “top-down” control of the prokaryotic assemblages. Therefore, most of the carbon flux at the investigated sites of the Mediterranean Sea passes through the microzooplankton, which control the abundance and dynamics of the smaller-sized autotrophic and heterotrophic components of the microbial food web through predation. The limited impact of the viral shunt in most regions allows this oligotrophic system to channel the prokaryotic biomass more efficiently into the higher trophic levels, thus reducing the conversion of biomass into dissolved organic matter through viral lysis.
 Converse to theoretical expectations, which predict an increase in the food web complexity in highly oligotrophic systems that are strongly dominated by the microbial food web, the analysis of the trophic interactions reported in the present study suggests that the microbial food web is more complex in the mesotrophic open waters (e.g., the Atlantic site and the western Mediterranean) than in the ultraoligotrophic conditions of the eastern Mediterranean, in which such interactions are oversimplified. We conclude that when the abundance of the organisms is severely limited by environmental conditions, the microbial food web becomes oversimplified.
 The present study was conducted during the second and third legs of the trans-Mediterranean cruise onboard the research vessels N/O Urania and Universitatis in the framework of the VECTOR project (Vulnerability of Coasts and Marine Italian Ecosystems to Climate Change and Their Role in the Mediterranean Carbon Cycles). This study was partially funded by the PRIN project OBAMA (Osservatorio off-shore per ricerche ecologiche a lungo termine (L-TER) sulla Biodiversità e funzionamento degli ecosistemi marini profondi in Mar Mediterraneo) and the Flagship Project RITMARE—The Italian Research for the Sea—coordinated by the Italian National Research Council and funded by the Italian Ministry of Education, University, and Research within the National Research Programme 2011–2013. We are grateful to A. Russo (Polytechnic University of Marche) for the temperature and salinity data, to E. Saggiomo (Zoological Station A. Dohrn, Naples) for the chlorophyll a and primary production values, and to S. Cozzi CNR-ISMAR (Trieste) for the nutrient concentrations. Thanks are also due to the crews of the two research vessels for their help in overcoming all of the logistic problems. We are also thankful to two anonymous reviewers, whose suggestions have greatly improved our manuscript.