Seawater was collected across the euphotic zone (0–100 m) at 20 m depth intervals at an offshore station in the northern Gulf of Aqaba, Red Sea (29.47° N, 34.92° E). Sampling was conducted monthly from May to September 2023, spanning the transition from late spring to late summer. Measurements included a comprehensive set of microbial activity measurements, including particulate primary productivity (> 0.7 µm PP_TOT), dissolved primary production (PP_DOC, < 0.7 µm), bacterial production (BP), and community respiration (CR). Ancillary data comprised Conductivity, Temperature, Depth (CTD) and photosynthetically active radiation (PAR) profiles (Sea-Bird SBE 19plus), inorganic nutrients concentration, and Chlorophyll a measurement.

Autotrophic activity was assessed by measuring particulate primary productivity (PP_POC) and carbon fixed present in the dissolved fraction (PP_DOC), enabling a detailed partitioning of carbon flow pathways within the microbial food web. To this end, seawater was collected in triplicate 50 mL acid-washed Falcon tubes. The collected samples were spiked with NaH¹⁴CO₃ (Perkin Elmer, specific activity 56 mCi mmol⁻¹) at a final radioisotope dilution of 1:104 v:v (1 µCi / 10 mL) and incubated for 24 h under ambient temperature and light. Incubations were stopped by filtration of the ¹⁴C-enriched water samples through glass fiber filters (GF/Fs) (0.7 µm nominal pore size, PP_POC) using low vacuum pressure (< 50 mm Hg). Filtration was performed under gentle vacuum pressure to minimize cell disruption, intracellular carbon release, and cell breakthrough. Although GF/F filters have a nominal pore size of ∼ 0.7 µm, their depth-filtration properties allow efficient retention of small picocyanobacterial (e.g., Prochlorococcus) under low-pressure filtration conditions (Bertilsson et al., 2003). Filtrate samples (5–10 mL, PP_DOC) were also collected in 15 mL Faclon tubes. Filters were rinsed with 5 mL of sterile-filtered seawater collected from the same site (< 0.22 µm) and the samples were subsequently acidified with 50 µL of 37 % hydrochloric acid (HCl, lowering pH to < 2) and left open in a fume hood overnight to allow conversion of all DIC to carbon dioxide (CO₂) and complete degassing and removal of inorganic carbon as ¹⁴CO₂. Filters were transferred into glass vials and 5 mL of scintillation cocktail (Ultima Gold) was added. To determine PP_DOC, the filtrate (10 mL) was transferred to scintillation vials, acidified with 50 µL hydrochloric acid and 5 mL scintillation cocktail was added as above (Teira et al., 2001). Samples were counted using a TRI-CARB 4810 TR (Packard) liquid scintillation counter. Blank samples (for both PP_POC and PP_DOC) collected from each depth were spiked with NaH¹⁴CO₃ and filtered immediately without incubation. These blanks yielded negligible counts, and their reads were subtracted from their respective samples read. Added activity of the radiolabeled “working solution” was obtained in each campaign by removing 50 µL from random samples immediately after spiking and before incubation (usually one per sampling depth), placing it on a new GF/F filter, adding 50 µL ethanolamine and scintillation liquid, and counting immediately. The variability between the “added activity” measurements was usually negligible for the same work solution (prepared freshly before each campaign), and we therefore used the average value of all collected samples/depths on each cruise.

PP_POC and PP_DOC rates were calculated based on the Bermuda Atlantic Time-series Study (BATS) protocol (https://www.bco-dmo.org/dataset/893182/description#acquisition, last access: 2 September 2026) using the equation: 1Primary productivity=(DPM-blank)V×DIC×AAvolTDPM×f×1t DPM equals the disintegrations per minute, V = the filtered volume, DIC is the dissolved inorganic carbon in seawater calculated from total alkalinity; ∼ 25 µg C L⁻¹), AA vol = Added activity volume, TDPM = Total ¹⁴C disintegration per minute, t = incubation time, and f = factor correcting the uneven uptake of ¹⁴C due to fractionation (1.05).

The percentage of extracellular release (PER; %) was calculated as follows: 2PER%=PP_DOCPP_POC+PP_DOC×100

Triplicate seawater samples (1.7 mL) from each sampling depth were incubated in the dark under ambient temperature with 10 nmol L⁻¹ of ³H-leucine (Perkin Elmer; specific activity 123 Ci mmol⁻¹) for 4 h, at in situ temperature, following the protocol of Simon et al. (1990). Incubations were terminated by the addition of 100 µL of 100 % trichloroacetic acid (TCA), and samples were subsequently processed using the microcentrifugation method (Smith and Azam, 1992). After processing, 1 mL of scintillation cocktail (Ultima Gold) was added to each vial, and radioactivity was measured using a TRI-CARB 4810 TR (Packard) liquid scintillation counter. Random “killed” controls, containing ³H-leucine and TCA added prior to incubation were used to account for background activity. Leucine incorporation rates were converted to carbon production using a conversion factor of 3 kg C mol⁻¹ leucine, assuming an isotopic dilution factor of 2 (Simon and Azam, 1989).

Community respiration (CR) rates were estimated from dark triplicate incubations per depth under ambient temperature. Dissolved O₂ concentrations were measured using an optical oxygen meter (FireSting, PyroScience GmbH) equipped with fiber-optic sensors. Prior to the incubations, optical sensor spots (OXSP5, PyroScience) were affixed to the inner wall of Winkler bottles (300 mL) according to the manufacturer's instructions. Bottles were carefully filled to avoid bubble formation and sealed with ground-glass stoppers. Samples were incubated for 24 h under continuous, gentle stirring using magnetic stir bars to ensure homogeneous O₂ distribution without introducing air bubbles or disturbing the sensor spot. The system was calibrated prior to measurements using a two-point calibration (air-saturated seawater and zero-oxygen solution) following the manufacturer's guidelines. CR was calculated as the temporal change in O₂ concentration (T24 minus T0) normalized to the incubation duration (Cheung et al., 2024). The detection limit was 0.5 µmol O₂ L⁻¹. CR was converted into BR using the linear regression of Aranguren-Gassis et al. (2012): 3BR=0.3×CR1.22-0.013

A respiratory quotient of 1 was used to convert O₂ consumption into carbon respiration (del Giorgio and Cole, 1998). Bacterial growth efficiency (BGE) was calculated as follows (del Giorgio and Cole, 1998): 44BGE%=BPBP+BR×100

Samples were collected in acid-washed plastic Falcon tubes without prefiltration and kept frozen at -20 °C until analysis within a few months. One sample per depth was collected. Nitrate + nitrite (NO2+ NO₃), orthophosphate (PO₄) and silicic acid (Si(OH)₄) were measured with a Seal Analytical AA-3 system (Kress and Herut, 2001). The limits of detection (twice the standard deviation of the blank) were 0.03 µM for nitrate + nitrite, 0.008 µM for orthophosphate, and 0.05 µM for silicic acid. Ammonium (NH₄) concentrations were measured fluorometrically using the orthophthaldialdehyde (OPA) method (Holms et al., 1999) using a Turner Designs (Trilogy) fluorometer equipped with 365 nm excitation and 460 nm emission filters. The detection limit was 0.01 µM. The quality of the nutrient measurements is regularly confirmed by inter-comparison exercises (e.g., QUASIMEME program).

Seawater (300 mL) samples were filtered through GF/F and stored at -20 °C in a dark box. One sample per depth was collected. Samples were extracted overnight in 5 mL of 90 % acetone at 4 °C in the dark (Welschmeyer, 1994). Chlorophyll concentrations were determined using a Turner Designs (Trilogy) fluorometer with 436 nm excitation and 680 nm emission filters.

Pairwise relationships among environmental and biological variables were assessed using Pearson correlation analysis. All data points from the upper 0–100 m across the five sampling campaigns were included. Prior to analysis, data were screened for missing values, and only complete datasets were retained. Pearson correlation coefficients (r) and corresponding p-values were calculated to assess the strength and significance of pairwise relationships. Statistical significance was evaluated at two levels (p< 0.05 and p< 0.01). Correlation matrices were visualized using a lower-triangle format to improve readability, with color gradients representing correlation strength and overlaid values indicating correlation coefficients and their significance levels. Analyses and figure generation were performed using Python (version 3.13).

The Gulf of Aqaba exhibits strong seasonal hydrographic variability with deep winter mixing followed by stable stratification afterward (Biton and Gildor, 2011), leading to nutrient-depleted surface water throughout the summer months (Laiolo et al., 2014; Efrat Meeder et al., 2012). The shift from deeply mixed winter conditions to a strongly stratified summer water column substantially reshapes autotrophic production, heterotrophic activity, and the efficiency with which carbon is exported from the surface (Karl et al., 2021). Understanding how microbial communities adjust their carbon-use strategies under nutrient-poor conditions is therefore important for predicting the fate of newly fixed carbon in stratified and oligotrophic oceans.

The progressive shoaling of the mixed layer from early to late summer (Table 1), together with persistently low nutrient concentrations and reduced nutrient inventories (Fig. 1B–E), indicates increasing isolation of the euphotic zone from the nutricline, typically located at depths > 150 m in this region (Landou et al., 2023; Meeder et al., 2012; Rahav et al., 2015). As stratification intensified, depth-integrated primary production declined from early to mid-summer, while chlorophyll biomass became increasingly concentrated within a pronounced DCM (Fig. 2A). This decoupling between biomass and productivity is consistent with a community persisting under chronic low light and low nutrient availability (Marañón et al., 2010), representing a deep photo-acclimation maximum (DAM) rather than a deep biomass maximum (DBM) as found in some other oligotrophic settings (Hodges and Rudnick, 2004; Mignot et al., 2014). Additionally, the sharp decline in heterotrophic BP with depth (Fig. 3A) likely reduces microbial loss rates, thereby allowing photo-acclimated phytoplankton at the DCM (mainly Prochlorococcus, Lindell and Post, 1995; Reich et al., 2024) to persist despite low carbon fixation rates. In other words, the high chlorophyll does not imply high biomass but rather high chlorophyll content per cell. At the same time, the vertical distributions of PP_DOC and PER indicate a shift in carbon allocation within the water column (Fig. 2). PP_DOC remains concentrated in the upper water column and is relatively stable over the season, while PER increases both with depth and over time, reaching its highest values during periods of strongest stratification (Fig. 2). This pattern implies that an increasing fraction of newly fixed carbon was released as DOC in the upper 100 m, especially as summer progressed and as imbalances between carbon fixation and nutrient assimilation limited biomass synthesis and “favored” the exudation of photosynthates (Mueller et al., 2016; Ofaim et al., 2021). Under such conditions, excess carbon leaks or excreted as low-molecular-weight compounds, thus elevating PER and enhancing the supply of labile dissolved organic substrates to heterotrophic prokaryotes (Engel et al., 2017, 2014, 2004). Consequently, a larger share of PP_TOT was routed into the microbial loop rather than retained in particulate autotrophic biomass (i.e., PP_POC), enhancing carbon recycling and reducing the efficiency of export to depth (Table 1; Carlson et al., 2010).

The increasing release of newly fixed carbon as dissolved organic matter thus provides a key link between autotrophic production and sustained heterotrophic activity within the euphotic zone, even as primary production declines over the summer (Table 1, Fig. 2). In oligotrophic systems such as the Gulf of Aqaba, heterotrophic microbial communities are often tightly coupled to DOC supply derived from contemporaneous primary production as well as semi-labile DOC that accumulates over seasonal timescales (Hansell et al., 2009; Santinelli, 2015). The persistence of BP in our study suggests that microbial metabolism is supported by a combination of freshly released DOC, recycled organic matter, and/or external nutrient inputs such as those added from aerosol deposition (Paytan et al., 2009; Rahav et al., 2018). Similar accumulated DOC and associated impacts on the microbial community have been observed in other stratified regions (Hansell et al., 2009; Hansell and Carlson, 2002). In contrast, a recent study in the Eastern Tropical North Atlantic reported that BP and BR closely follow PP_TOT, with PP_DOC supply often exceeded bacterial carbon demand (Devresse et al., 2022). There, heterotrophic activity was largely controlled by and coupled to contemporaneous production, resulting in a production-driven carbon cycle. In the Gulf of Aqaba, however, the weaker coupling between PP_TOT and heterotrophic activity points to a recycling-dominated regime in which microbial communities increasingly depend on internally regenerated carbon and are not directly linked to contemporaneous production. Under sustained stratification, reduced nutrient supply constrains autotrophic growth while promoting DOC release and accumulation, thereby maintaining BR even as primary production declines (Hansell et al., 2009; Hansell, 2013). Interestingly, while absolute rates of PP_DOC significantly correlate with bacterial production (r=0.50, p< 0.01; Fig. 4), the PER does not exhibit a statistically significant correlation with BGE across the water column (r=0.106, p=0.59). This decoupling is primarily driven by opposing vertical gradients, as PER increases toward the base of the euphotic zone while BGE peaks in warmer surface layers, as well as potential bacterial overflow metabolism where carbon-rich exudates are preferentially respired rather than assimilated into biomass. Nevertheless, the low BGE observed (Table 1, Fig. 3C) further support the view that most of the consumed carbon is respired rather than converted into biomass, a characteristic of strongly recycling systems in oligotrophic ocean settings (del Giorgio and Cole, 1998). Indeed, the vertical structure of BGE closely mirrors that of BR, indicating that most of the organic carbon processed by heterotrophic prokaryotes was respired rather than incorporated into new biomass. These low growth efficiencies are consistent with the energetic constraints typical of oligotrophic waters, indicating that bacterial carbon use was directed mainly toward respiration rather than growth, with much of the organic carbon rapidly returned to dissolved inorganic form (del Giorgio and Cole, 1998). These low growth efficiencies are consistent with the energetic constraints typical of highly stratified oligotrophic waters, indicating that bacterial carbon use was directed mainly toward respiration rather than growth, with much of the organic carbon rapidly returned to dissolved inorganic form (del Giorgio and Cole, 1998). While low BGE is widely observed in unproductive open-ocean settings, a recent study from the oligotrophic Taiwan Strait reported elevated BGE values (Liu et al., 2026). Such contrasts suggest that oligotrophy alone does not dictate bacterial metabolic efficiency. Instead, BGE may be regulated by the substrate stoichiometry and lability, thermal regimes, and different physical oceanographic settings. In systems exhibiting higher BGE such as the Taiwan Strait, heterotrophic bacteria may access labile dissolved organic matter or nutrient-rich substrates through micro-scale spatial patchiness (e.g., phycosphere exudates, viral lysis), rapid biological turnover, or episodic physical events like upwelling, terrestrial runoff, and Kuroshio intrusions (Liu et al., 2026). In contrast, in the hyper-oligotrophic, warm, and highly isolated summer water column of the Gulf of Aqaba, severe nutrient limitation likely promotes the exudation of carbon-rich, nutrient-poor dissolved organic matter. This forces heterotrophic bacteria to respire excess carbon via overflow metabolism to meet their nutritional demands while sustaining high basal metabolic costs, thereby maintaining low to moderate BGE.

We acknowledge that the nominal ∼ 0.7 µm pore size of GF/F filters may raise concerns regarding the potential passage of small picoplankton, particularly Prochlorococcus. However, previous studies indicate that cell breakthrough is minimal under low-vacuum filtration conditions. For example, Bertilsson et al. (2003) showed that ∼ 95 % of cells in an axenic culture of Prochlorococcus marinus MED4 were retained on Whatman GF/F filters under low vacuum, comparable to the conditions and filters used in our study. In addition, although not specifically addressing Prochlorococcus, Chavez et al. (1995) reported equivalent recovery of total chlorophyll a using GF/F and 0.2 µm membrane filters in oligotrophic open-ocean samples, suggesting that GF/F filtration is unlikely to result in substantial loss of phytoplankton biomass (and thus production) under similar conditions. Moreover, the observed seasonal increase in PER (Fig. 2D) was associated with changes in nutrient availability and microbial metabolism rather than simply tracking chlorophyll biomass. These observations therefore support the interpretation that the increase in extracellular release reflects a physiological response rather than an artifact of size-selective filtration. Additionally, abiotic retention or adsorption of ¹⁴C-labeled dissolved organic matter onto the glass-fiber matrix may lead to biased PP_POC and, consequently, of PER (López-Sandoval et al., 2018; Maske and Garcia-Mendoza, 1994). Nevertheless, the shift toward elevated PP_DOC and PER as summer stratification progress, despite this potential matrix adsorption, suggest that enhanced extracellular release is a genuine physiological feature of nutrient-stressed autotrophs in the study area rather than a filtration artifact.

The observed vertical patterns in microbial carbon cycling are primarily structured by gradients in light availability/nutrient supply and stratification intensity, which all co-vary with depth (Fig. 4). Chlorophyll a peaks at the DCM, reflecting photo-acclimated cells at depth (DAM), while rates of PP_TOT, PP_DOC, and BP are highest in the upper, well-lit layers. Within this framework, a strong coupling between PP_POC and PP_DOC would suggest that dissolved carbon release scales directly with photosynthetic activity, consistent with extracellular release as an inherent component of primary production under oligotrophic conditions. However, the negative relationship between both PP_POC and PP_DOC with Chlorophyll a (Fig. 4) indicates that carbon fixation and release are not directly linked to standing stock of biomass, but rather to physiological activity in the upper water column (Fig. 2). Importantly, PP_DOC showed a stronger association with BP than PP_POC does, emphasizing the central role of dissolved organic carbon in sustaining heterotrophic metabolism. This supports the idea that the microbial loop in the Gulf of Aqaba is primarily fueled by recently produced dissolved substrates rather than by particulate pathways (abovementioned references). The absence of strong linear correlations between ambient nutrient concentrations and biological rates reflects the rapid uptake capacity of nutrient-starved microbial communities. This rapid biological drawdown maintains nutrient pools near detection limits, effectively masking statistical correlations. Together, these patterns indicate that carbon cycling in the Gulf of Aqaba is structured by two complementary processes: (i) vertical decoupling between biomass accumulation and carbon cycling activity driven by stratification, and (ii) tight coupling between dissolved carbon production and heterotrophic utilization (Fig. 4).