Enhancement of diatom growth and phytoplankton productivity with reduced O2 availability is moderated by rising CO2

Many marine organisms are exposed to decreasing O2 levels due to warming-induced expansion of hypoxic zones and ocean deoxygenation (DeO2). Nevertheless, effects of DeO2 on phytoplankton have been neglected due to technical bottlenecks on examining O2 effects on O2-producing organisms. Here we show that lowered O2 levels increased primary productivity of a coastal phytoplankton assemblage, and enhanced photosynthesis and growth in the coastal diatom Thalassiosira weissflogii. Mechanistically, reduced O2 suppressed mitochondrial respiration and photorespiration of T. weissflogii, but increased the efficiency of their CO2 concentrating mechanisms (CCMs), effective quantum yield and improved light use efficiency, which was apparent under both ambient and elevated CO2 concentrations leading to ocean acidification (OA). While the elevated CO2 treatment partially counteracted the effect of low O2 in terms of CCMs activity, reduced levels of O2 still strongly enhanced phytoplankton primary productivity. This implies that decreased availability of O2 with progressive DeO2 could boost re-oxygenation by diatom-dominated phytoplankton communities, especially in hypoxic areas, with potentially profound consequences for marine ecosystem services in coastal and pelagic oceans.

H ypoxic waters (defined as having dissolved O 2 < 63 μM or 2 mg L −1 ) occur naturally in both open ocean and nearshore waters, and global warming, as well as anthropogenic eutrophication, have been increasing in their spatial extent and severity [1][2][3][4] . While hypoxia has often been considered exclusive to deeper waters, near-surface hypoxic waters (< 20 m) are often observed in estuaries 5 , coastal waters 6 , and upwelling regions 7 . Deoxygenation (DeO 2 ) in these areas is predicted to accelerate with progressive ocean global changes, mainly due to ocean-warming 8 . Decreases in the dissolved O 2 content of coastal seawaters are principally due to the heterotrophic degradation of dissolved organic matter associated with coastal eutrophication, resulting in low O 2 , low pH, and high CO 2 conditions [9][10][11] . While such changes are measured in bulk seawater, their levels are not the same as those in the diffusion boundary layer (DBL) at the photosynthetic cell surface, but nonetheless modeling and direct measurement suggest that changes in the DBL exhibit the same trends, maintaining higher CO 2 (lower pH) under elevated CO 2 conditions or lower O 2 under reduced O 2 conditions 12,13 . Therefore, reduced O 2 availability and increased CO 2 (lowered pH) in seawater are co-varying drivers in the context of DeO 2 and ocean acidification (OA) 14 . This combination has the potential to disturb the balance between photosynthetic energy supply and respiratory energy consumption in marine ecosystems, and can thus disrupt ecological services 3,15 .
Photosynthesis of phytoplankton is a major biogeochemical process that oxidizes the oceans, especially by the diatoms that have been estimated to contribute up to 52% of marine O 2 production 16 and that dominate the phytoplankton communities in hypoxic regions 17 . Photosynthesis of some diatoms appears to decrease with increased ratios of O 2 to CO 2 availability 18 , because carboxylation and oxygenation are catalyzed simultaneously by the central enzyme of photosynthesis ribulose-1,5-bisphosphate carboxylase/oxygenase (Rubisco), and these two reactions compete with each other at the active site of the enzyme to fix CO 2 and to consume O 2 , respectively 19 . In common with most other phytoplankton, diatoms use energy-costly CO 2 concentrating mechanisms (CCMs) 20 to increase intracellular CO 2 around the active site of Rubisco, minimizing competition from O 2 and favoring efficient carboxylation 19 . It has been shown that increased seawater pCO 2 at the levels projected for the end of this century can decrease CCM activity in diatoms and other microalgae 21,22 and repress expression of CCM-related genes 23,24 . The energy savings and resources freed up from downregulation of the CCMs under elevated CO 2 conditions could potentially increase primary production under low light levels 20,21,25,26 . However, under high light levels, excess photochemical energy has been suggested to act with acidic stress to enhance photoinhibition and therefore decrease primary productivity in surface phytoplankton communities 26 .
It is usually accepted that higher levels of chlorophyll a (Chl a) abundance are positively correlated with high primary productivity 27 . However, primary productivity per volume of water does not reflect photosynthetic activity or light use efficiency per Chl a, since higher O 2 and low pCO 2 are often found in waters of high Chl a concentrations 28 , and are supposed to reduce carboxylation or photosynthetic efficiency as aforementioned. These previous theoretical inferences 18 along with our own fieldwork shown here and other observations showing higher levels of phytoplankton photosynthetic efficiency or biomass density in low O 2 waters 29 led us to hypothesize that a decreased pO 2 :pCO 2 ratio in estuarine and coastal waters could enhance marine productivity and, that this effect of deoxygenation is due to differentially influenced physiological performances of CCMs and photorespiration, which together could act to increase diatom growth rates. OA entails both increased CO 2 availability and acidic stress, and so may either decrease or increase photosynthetic efficiency and growth in diatoms, depending on taxonomic differences and environmental conditions 18,26,[30][31][32] . In contrast, the interactions of increased CO 2 and reduced O 2 on phytoplankton have rarely been considered 18 . We present here a test of our hypothesis using a series of mesocosm and laboratory experiments that determined the combined effects of elevated CO 2 and decreased O 2 availability on diatom growth and photosynthesis.

Results
Field investigation. The aim of our field study was to examine whether photosynthetic activity correlates with levels of dissolved O 2 (DO), a factor that has seldom been considered in the context of potential effects on oceanic primary productivity. Accordingly, environmental parameters that may influence photosynthetic carbon fixation were investigated at eight different stations in the Pearl River estuary (Fig. 1a, details in Supplementary Table 1). Photosynthetic light use efficiency [PLUE, μmol C (μg Chl a) -1 h -1 (μmol photons m -2 s -1 ) -1 )] was derived from photosynthetic carbon fixation rates measured at low levels (photosynthesis-limiting, <100 and <60 μmol photons m −2 s −1 at 10 and 20 m, respectively) of incident sunlight. PLUE was significantly correlated with DO, CO 2 and pH ( Fig. 1b- Table 2, P = 0.6671, 0.0707, respectively, Pearson Correlation Analysis). There was an obvious significant increase in PLUE with decreased DO. However, this negative correlation might be attributed to the positive effects of increased CO 2 availability and other environmental factors ( Fig. 1 and Supplementary Table 1). Therefore, we employed Partial Correlation Analysis to further exclude disturbance from other environmental factors on the correlation between DO and PLUE (see "Data Analysis" for details). These results again indicate a significant correlation of higher PLUE with lower DO (Supplementary Table 2, P = 0.0035, r = −0.4736), suggesting that DO could be one of the key drivers altering in situ photosynthesis and primary production.
Natural phytoplankton assemblage mesocosm experiments. To test the responses of a natural coastal phytoplankton assemblage to different pO 2 :pCO 2 combinations, we conducted a 30-liter mesocosm experiment under natural levels of sunlight and temperature ( Supplementary Fig. 1a) with filtered (180 μm) seawater. While DO, CO 2 levels, and pH varied over time, DO and CO 2 remained significantly different between the low and high treatments ( Supplementary Fig. 1b-d, P < 0.0001). Macronutrients in the mesocosms were consumed rapidly and became depleted within 5 days ( Supplementary Fig. 2), with faster removal of the nutrients under low O 2 conditions. This was especially obvious for NO x and SiO 3 2- (Supplementary Fig. 2a, c). In contrast, concentrations of chlorophyll a (Chl a) in the mesocosms increased rapidly and peaked within 3 days then declined, with higher concentrations of Chl a under low O 2 /high CO 2 treatments at day 3 ( Supplementary Fig. 2d, P = 0.0414, 0.1547 for LOAC and LOHC, respectively).
During the mesocosm experiment, the net and gross photosynthetic rates were higher in the low O 2 (LO)-grown than in the ambient O 2 (AO)-grown phytoplankton assemblage under both ambient (AC) and high (HC) CO 2 levels (relative changes are presented in Fig. 2a-d, absolute values in Supplementary Table 3 and specific p values in Supplementary Table 4), and these enhancements increased with time when NOx stocks diverged between the LO and AO treatments in the mesocosms (Fig. 2a-d and Supplementary Fig. 2a). Under AC, reduced O 2 availability significantly enhanced the net photosynthetic rate per volume of seawater ( Fig. 2a) at day 3 and day 5 (P = 0.0102, 0.0124), and such significant enhancement was also observed under high CO 2 at day 1, 3, 5 (P = 0.0416, 0.0076, 0.0040). Similar trends were also found in Chl a-normalized net photosynthesis (Fig. 2b) under both LOAC and LOHC treatments, though significant enhancement was only observed at day 5 (P = 0.0145, 0.0359) and marginally significant enhancement at day 10 (P = 0.0529 for LOAC). Likewise, gross photosynthetic rates regardless of normalization units and CO 2 levels were higher under reduced O 2 levels (Fig. 2c, d). Elevated CO 2 and the associated pH drop appeared to run counter to the stimulating effects of reduced O 2 , with lower mean values of photosynthetic rate in the LOHC treatment compared with the LOAC treatment under both normalized units (per water volume or per Chl a), but this was not statistically significant ( Fig. 2a-  Reduced O 2 availability decreased nonphotochemical quenching (NPQ), an indicator of photosynthetic energy loss as heat dissipation and a signal of light stress (Fig. 2e), though only marginally significant changes were observed at day 4 (P = 0.0531 for LOAC and P = 0.086 for LOHC) and day 8 (P = 0.0552, 0.0932). In parallel, reduced O 2 level increased photochemical yield Based on the CHEMTAX analysis, the phytoplankton community composition changed with time under the different O 2 and CO 2 combination treatments (Fig. 4). The diverse phytoplankton community was originally dominated by diatoms, cryptophytes, and prasinophytes, but then shifted to have higher proportions of dinoflagellates and the pico-cyanobacterium Synechococcus (Fig. 4) when nutrients were depleted (Supplementary Fig. 2). While diatoms continued as one of the dominant groups throughout the incubation period, the proportion of dinoflagellates obviously increased in the LOAC treatment ( Fig. 4c-e, P = 0.0240, 0.0029, 0.0035, correspondingly). A similar trend was found in the LOHC treatment, although a significant increase was only observed at day 5 ( Fig. 4d, P = 0.0405).
Diatom culture experiment. Based on the field investigation and mesocosm experiment where diatoms were dominant, a diatom culture experiment was conducted to investigate photosynthetic performance, growth rate, and CCM efficiencies in the globally distributed coastal diatom Thalassiosira weissflogii. The cells were grown under four pO 2 :pCO 2 combinations for over nine generations in laboratory culture. DO, carbonate chemistry and cell numbers were maintained in a stable range (with~1000-5000 cells mL −1 , Supplementary Fig. 3) by diluting the medium every 24 h without using aeration. Levels of DO, pH T , and CO 2 in low O 2 (LO) and high CO 2 (HC) culture conditions differed from those in the ambient O 2 (AO) and ambient CO 2 (AC) treatments ( Supplementary Fig. 3e, f and Supplementary Table 6).  Supplementary  Table 3. (e) Nonphotochemical quenching (NPQ) measured during the noon period at day 4, 8, 10. Black symbols represent ambient O 2 (AO, 213 μM) and red symbols low O 2 (LO,~57 μM); Circles represents ambient CO 2 (AC,~13 μM); triangles represent high CO 2 (HC,~27 μM). Mesocosms were incubated under incident sunlight and natural levels of temperature ( Supplementary Fig. 1), and all data were obtained under growth conditions. Detailed information for the mesocosms experimental features are given in Supplementary Figs. 1 and 2. The values are the means with error bars indicating standard deviations of independent biological replicates (n = 3 mesocosms). Light-colored symbols are individual data corresponding to the treatments. Blue * and red * indicate significant differences (P < 0.05, LSD test) due to low O 2 under ambient (LOAC) and elevated CO 2 levels (LOHC), respectively, compared to the control treatment (AOAC).
Reduced O 2 levels significantly promoted net photosynthesis of the diatom by~14% under both AC and HC levels ( Fig. 5a, P = 0.0024, 0.0005). The absolute rates were higher by~31% in the AOHC and by~50% in the LOHC compared with the AOAC treatment ( Fig. 5a, P < 0.0001, 0.0001), respectively. Decreased O 2 concentration also increased the growth rate by~14% under AC and only by 9% under HC (Fig. 5b, P < 0.0001, 0.0001). This suggests that there was substantially less enhancement of growth by reduced O 2 under the influence of elevated CO 2 with lowered pH.
Decreased O 2 levels reduced mitochondrial respiration under the AC and HC levels by 41% and 68%, respectively ( Fig. 5c, P = 0.0054, P < 0.0001), suggesting that mitochondrial respiration was suppressed by reduced O 2 availability to a much greater extent under HC conditions. At the same time, LOAC-and LOHC-grown cells exhibited unchanged high values of photochemical efficiency compared with the cells grown under the AOAC treatment ( Supplementary Fig. 4a, P = 0.4717, 0.9663). This indicates that the cells were maintaining a healthy physiological state with high light use efficiency. NPQ decreased significantly in low O 2 treatments by 20% (P = 0.0016) and 32% (P < 0.0001) under AC and HC levels, respectively (Supplementary Fig. 4b), suggesting a more efficient energy transfer in LOgrown cells, which is consistent with the results from the mesocosm experiment using natural phytoplankton assemblages (Fig. 2e).
To explore the mechanisms involved, we tested the CCM capacity of the diatom cells acclimated to different combinations of pO 2 and pCO 2 using direct comparisons under standard conditions (pH T = 8.00, 50-200 μM O 2 ). The LOAC-grown cells had a significantly lower half-saturation constant (K 0.5 ) for CO 2dependent photosynthesis (Supplementary Fig. 4c and Fig. 6a, P < 0.0001), indicating an increased photosynthetic affinity for CO 2 and an increase in CCM activity. Conversely, the HCacclimated cells grown under both AO and LO levels had lower CO 2 affinities and CCM activities, as revealed by their increased K 0.5 values compared to the AOAC-grown cells ( Supplementary   Fig. 3 Diurnal changes in photosystem II (PSII) quantum yield (Yield) and the effective functional absorption cross-section of PSII (σ PSII´, A 2 quanta -1 ) of phytoplankton assemblages grown under different O 2 and CO 2 treatments. Effective PSII quantum yield (a, c, e) and the effective functional absorption cross-section of PSII (b, d, f) at days 4, 8, and 10, respectively. Blue dots represent diel changes in photosynthetically active radiation (PAR, μmol photons m −2 s −1 ) during the experiment. Black symbols represent ambient O 2 (AO,~213 μM) and red symbols low O 2 (LO,~57 μM); circles represent ambient CO 2 (AC,~13 μM); triangles represent high CO 2 (HC,~27 μM). Detailed information for the mesocosms experimental features are given in Supplementary Figs. 1 and 2. The values are the means and the error bars represent standard deviations of independent biological replicates (n = 3 mesocosms). Light-colored symbols are individual data corresponding to the treatments. Blue * and red * indicate significant differences (P < 0.05, LSD test or Games-Howell test) caused by low O 2 under ambient (LOAC) and elevated CO 2 levels (LOHC), respectively, compared to the control treatment (AOAC). Fig. 6a, b, P = 0.0038, P < 0.0001). The efficiency of CO 2 acquisition, expressed here as the quotient of maximal photosynthetic rate (V max ) to K 0.5 , increased significantly with decreased O 2 by up to 187%, under the AC level (Fig. 6a, inset, P = 0.0001), but only by about 40% under the HC level (Fig. 6b, inset, P = 0.0481). This implies opposing effects of reduced O 2 and elevated CO 2 (lowered pH) on CO 2 acquisition efficiency.

Fig. 4c and
It appeared that the LO-acclimated cells increased their photorespiration when measured under the standard conditions (nearly ambient O 2 level) that at least partly repressed net photosynthetic O 2 evolution (Figs. 5a and 6a, b). To check if the divergences between conditions for physiological tests and for experimental cultures may potentially make the observed CCMrelated photosynthetic traits under the standard conditions inaccurately reflect those under growth conditions, we examined the activity of periplasmic carbonic anhydrase (eCA) involved in the extracellular conversion of bicarbonate to CO 2 using acetazolamide (AZ, as an inhibitor of eCA) . The inhibition of photosynthetic O 2 evolution by AZ measured under culture conditions was taken as a proxy of eCA-functional capacity and CCM activity (a greater inhibition of eCA relates to higher involvement of biophysical CCMs in photosynthesis). Inhibition was significantly greater in the LO-grown cells compared to AOgrown ones (Table 1, P = 0.0239). This indirectly supports the results showing that lowered O 2 concentration enhanced activity of the CCMs and CO 2 acquisition efficiency (Supplementary Fig. 4c and Fig. 6a, b). In addition, under the HC conditions AZ had an insignificant effect on the cells grown under both O 2 levels (Table 1), as revealed by unchanged net photosynthetic rates under both AO (P = 0.4982) and LO (P = 0.3838) conditions with AZ compared with that without AZ. This reflects that the elevated CO 2 alone was sufficient to cause downregulation of eCA and activity of CCMs, regardless of O 2 levels, leading to undetectable AZ impacts. Photorespiration of the diatom declined significantly (42%) in the LO-grown cells under AC levels (P = 0.0058), but decreased to a much lesser extent (20%) in cells grown under LO and HC levels (Fig. 6c, P = 0.0637). Once again, the effects of elevated CO 2 were opposite to the positive influence of reduced O 2 . Photorespiration correlated inversely with CO 2 acquisition efficiency (Fig. 6d, P = 0.0023, r = −0.7886), implying a shift from oxygenation to carboxylation catalyzed by ribulose-1,5bisphosphate due to low O 2 enhanced CCMs activity.

Discussion
We found that reduced levels of dissolved O 2 increased primary productivity of natural phytoplankton assemblages and stimulated growth and enhanced photosynthetic performance with increased activity of CCMs in a cultured diatom (Fig. 7). Mechanistically, low O 2 -enhancement of CCMs activity along with improved light use efficiency and the reduction in photorespiration allow low O 2 -grown phytoplankton to perform more efficient photosynthetic carbon fixation (Figs. 2 and 5) and result in faster growth in the diatom (Fig. 5). Reduced photorespiration from favored carboxylation may increase the demand for inorganic carbon, and the reduced mitochondrial respiration may result in decreased intracellular CO 2 supply through the respiratory pathway and thus enhance the CCM activity of cells grown under low O 2 levels. Although the antagonistic effects of increased CO 2 projected for the end of this century on CCMs partly canceled out the positive effects of decreased O 2 on the diatom ( Fig. 6 and Table 1), reduced levels of O 2 still significantly promoted their growth even under the elevated CO 2 conditions. Suppression of respiratory carbon loss (Fig. 5) might have also contributed to the enhanced growth rates of the low-O 2 grown diatom due to suppressed mitochondrial respiration, the rate of which depends on O 2 levels 13 . Especially under high CO 2 conditions, low-O 2 grown diatoms possessed higher growth rates with lower mitochondrial respiration, implying that the energy saved from the down-regulated CCMs could have supported the energetic demands for growth so that mitochondrial respiration diminished. These findings supported our hypothesis.
Whether the positive effects of reduced O 2 on phytoplankton assemblages observed in this work are true for dynamic in situ environments remains to be explored, in view of possible synergistic or antagonistic effects of multiple drivers. The sensitivity of phytoplankton to O 2 can be closely linked to their physiological conditions, types and/or efficiencies of CCMs and Rubisco 19,33,34 . Thus additional environmental stresses and diverse phytoplankton assemblage structures may complicate overall ecosystem responses 35 . For instance, changes in nutrient availabilities and phytoplankton communities in our mesocosms under fluctuating levels of PAR and temperature appeared to have affected the interactions of CO 2 and O 2 (Figs. 2-4). In addition, other components of the plankton communities, such as grazers, might have complicated the interactions within the mesocosm system. These factors may be at least partially responsible for observed differences in the magnitude of low-O 2 enhancement effects and high CO 2 dampening impacts on photosynthetic carbon fixation.
Most dinoflagellates are characterized by only moderately efficient CCMs and high O 2 affinity-form II Rubisco, and therefore may benefit more from reduced O 2 19 . This may account for the increased proportion of dinoflagellates in our low O 2 mesocosms after nutrients, especially after SiO 3 2− became exhausted. On the other hand, their complex nutritional modes, such as heterotrophic nutrition and phagotrophy, may give dinoflagellates more strategies to withstand low O 2 environments. As recently reported, Noctiluca scintillans, which relies on ingested endosymbionts, bloomed during a hypoxic event in the Arabian Sea 36 . The aforementioned positive effects of lowered O 2 and multiple nutritional modes might have increased the abundance of dinoflagellates encountering hypoxic waters. This implies that  (HC,~35 μM). The values are the means and the error bars represent standard deviations of independent biological replicates (n = 3 independent cultures). Light-colored symbols are individual data corresponding to the treatments. Different letters above the bars represent significant differences (P < 0.05, LSD test) among treatments. Detailed information for the experimental features and timing points for the above determinations are shown in Supplementary Fig. 3. hypoxic waters or ocean deoxygenation could enhance the development of harmful dinoflagellate blooms.
As global warming and eutrophication have perturbed the O 2 budget of the ocean, degradation of habitat fitness for aerobic marine organisms has occurred both regionally and globally 3,4,8 . Importantly, recently reported time-series data suggest the occurrence of upwelling-induced continuous hypoxia events (~1-2 weeks) in shallower layers 37 . In our study, however, natural phytoplankton assemblages and the diatom T. weissflogii benefited from reduced O 2 concentrations that were low enough to be detrimental for most marine animals 15,38 . Accordingly, even under elevated CO 2 conditions, low O 2 -enhanced photosynthesis can accelerate "re-oxygenation" in illuminated waters bỹ 193-250% (based on the net photosynthetic values of day 5 in Fig. 2, and an assumption that the photosynthetic quotient is 1.0), and thus may progressively alleviate the impacts of diminished oxygen on animals (Fig. 7). Considering that open ocean diatoms are more sensitive to rising CO 2 than coastal ones 39 , the combined impacts of reduced O 2 and increased CO 2 levels on coastal and pelagic phytoplankton taxa are expected to differ in extent. Thus, the present result, showing that lowered availability of O 2 enhanced primary production of phytoplankton, indicates a possible negative feedback effect on ocean deoxygenation.
Marine primary producers are exposed to multiple stressors along with progressive ocean acidification (OA) and warming 40 , being affected by enhanced nutrient limitation in pelagic waters and by deteriorating eutrophication in coastal areas, along with ocean-warming-induced decrease in oxygen solubility. Ocean deoxygenation has been predicted to cause a further 1-7% decline in the global ocean O 2 inventory over this century, due to global warming 8 . Moreover, increasing discharges of nitrogen and phosphorus to coastal waters 41 and strengthening upwellingfavorable winds 42 may make invasions of hypoxic waters into the   (Fig. 7). Other key biological responses under multiple drivers along with long-term selection and evolution of dominant phytoplankton to life under low O 2 / high CO 2 conditions are unknown, but should be a priority for further research. Future studies on the ocean deoxygenation effects are also encouraged to include more drivers, to better reflect the real complexities of future ocean environments.

Methods
Field studies. Photosynthetic carbon fixation was investigated at eight different stations in the Pearl River estuary of the South China Sea (Fig. 1a and Supplementary Table 1), where the phytoplankton assemblages were dominated by diatoms 45 during the time of our investigation (June 2015). Samples were collected from 10 to 20 m depths and transferred immediately into 50 mL quartz tubes and sealed to prevent gas exchange. The samples were inoculated with 100 μL of 5 μCi (0.185 MBq) NaH 14 CO 3 solution for 2.15 h. All the incubations were carried out under incident solar radiation, attenuated with neutral density filters to simulate light intensities at the sampling depths, and the temperature was controlled with flow-through surface seawater. After incubation, the cells were filtered onto glass-fiber filters (25 mm, Whatman GF/F, USA) and stored at −20°C until measurement, during which the filters were exposed to HCl fumes overnight and dried (20°C, 6 h) to remove unincorporated NaH 14 CO 3 as CO 2 . The incorporated radioactivity was measured by liquid scintillation counting (LS 6500, Beckman Coulter, USA), and photosynthetic carbon fixation rates were estimated as previously reported 46 . Since the measurements were carried out under varying and low light levels similar to in situ levels at depths of 10 and 20 m, we normalized the photosynthetic rates to light intensity (μmol C (μg Chl a) −1 h −1 (μmol photons m −2 s −1 ) −1 ) to obtain the light use efficiency of photosynthesis (PLUE). This was done to allow for a meaningful comparison among different stations according to the linear relationship of photosynthetic carbon fixation under low solar irradiance levels 46 , which lies within the range of sunlight levels used in the present fieldwork (<100 μmol photons m −2 s −1 ).
Field DO, chlorophyll a (Chl a) concentration and nutrients were measured as described previously 5,47 . Briefly, field DO was manually measured on board using the Winkler titration method 48 . The Chl a content was measured with a Turner Designs Model 10 Fluorometer. The nitrogen (NO X , NO 3 − + NO 2 − ), NH 4 + , and SiO 3 2− concentrations were measured with a nutrient-autoanalyzer (Quickchem 8500, Lachat Instruments, USA) following the description of Kirkwood et al. 49 . This equipment has detection limits of 0.014 and 0.075 μM for NO X and SiO 3 2− , respectively.
Dissolved inorganic carbon (DIC) concentrations at investigated stations were estimated based on measured salinity and the relationship between salinity and DIC concentrations in the published literature 50 in the same area of the Pearl River estuary during the same season. CO 2 concentration and pH T were calculated using CO 2 SYS software 51 , using the equilibrium constants K 1 and K 2 for carbonic acid dissociation 52 .
Mesocosm studies. Surface seawater (0-1 m) with natural plankton assemblages was sampled from a harbor near the Dongshan Swire Marine Station of Xiamen University (23.65 o N, 117.49 o E) with an acid-cleaned plastic bucket, filtered (180 μm) to remove large grazers, and transported to the station within 1 h. The incubation system used 30-liter cylindrical polymethyl methacrylate tanks (n = 3), which allowed 91% PAR transmission and were water-jacketed for temperature control with a re-circulating cooler (running water). We set two O 2 and two CO 2 levels with three pO 2 :pCO 2 combinations: (1) ambient O 2 (AO,~213 μM) & ambient CO 2 (AC,~13 μM), AOAC; (2) low O 2 (LO,~57 μM) & ambient CO 2 , LOAC; (3) low O 2 & high CO 2 (HC,~27 μM), LOHC. The presented O 2 and CO 2 concentrations are average values across the entire experiment. N 2 , CO 2 , and air were mixed proportionally to create different and stable pO 2 :pCO 2 combinations in the gas stream. The incubation tanks were continuously aerated (0.5 L min −1 ) under incident solar radiation. The O 2 concentration was measured (20:00) with a precise single-channel fiber optic oxygen sensor (Microx 4, PreSence, Germany) every day. CO 2 concentrations of seawater were calculated from daily measured pH NBS (20:00) and TA measured every other day using CO 2 SYS software. The pH was determined according to Dickson (2010) 53 with a high-quality pH meter (Orion StarA211, Thermo, USA) which was calibrated with standard National Bureau of Standards (NBS) buffer solutions (Hanna). The pH NBS values were converted to pH Total (pH T ) using the CO 2 SYS software as described above.
For nutrient measurements, water samples were stored in 80-mL polycarbonate bottles, instantly frozen, and stored at −20°C until analysis. Samples for silicate determination were fixed with 1‰ chloroform and preserved at 4°C. Nutrients were measured with an AA3 Auto-Analyzer (Bran-Luebbe, GmbH, Germany) with  Fig. 2 assuming that the photosynthetic quotient is 1.0). In natural environments, low O 2 -enhanced production of phytoplankton biomass makes them a more effective O 2 source that may help to counteract the negative effects of hypoxia on heterotrophs. Black, red, and blue arrows indicate directions, increase and decrease, respectively. detection limits of 0.08, 0.08, and 0.16 μM for NO X , PO 4 3− , and SiO 3 2− , respectively.
Samples for analysis of Chl a and other pigments were filtered onto glass-fiber filters (25 mm, Whatman GF/F, USA) which were immediately preserved in liquid nitrogen until analysis. Measurement was conducted with a high-performance liquid chromatography system (UltiMate 3000, ThermoFisher Scientific, USA) after filters were submerged in N, N-dimethylformamide and then mixed 1:1 (V:V) with 1-M ammonium acetate 54 . Chlorophyll a and other pigments were identified by their retention times and quantified using peak areas and standard curves. Quantification was performed with standards purchased from DHI Water & Environment, Hørsholm, Denmark. Chemotaxonomic analysis was carried out using CHEMTAX software 55,56 .
To measure gross and net primary productivity, respectively, seawater samples were inoculated with 200 μL of 10 μCi (0.37 MBq) NaH 14 CO 3 solution (ICN Radiochemicals, USA) for 2 h (gross) and with 100 μL of 5 μCi (0.185 MBq) NaH 14 CO 3 solution for 24 h (net). All the incubations were carried out under incident solar radiation in a flow-through water bath to obtain a uniform temperature. Photosynthetic carbon fixation rates in the mesocosm experiment were estimated as described above.
Photosynthetic fluorescence parameters were measured with a fluorescence induction and relaxation system (In-Situ FIRe, Satlantic, NS Canada). NPQ was estimated by the equation of Genty et al. 57 : where F md is the maximal fluorescence measured before sunrise and F m ' is the effective yield at 11:00 a.m. under incident sunlight.
Diatom culture studies. The diatom Thalassiosira weissflogii (CCMP 1336) was incubated in artificial seawater prepared according to the Aquil* medium recipe 58 Table 6). The dissolved O 2 and pH of seawater were measured before and after diluting the culture medium ( Supplementary Fig. 3 and Supplementary Table 6). The dissolved O 2 was measured with a Clark-type oxygen electrode (Hansatech, UK). Parameters of the seawater carbonate system (Supplemental Table 6) were calculated from pH and TA with CO 2 SYS software, and the pH NBS values were converted to pH Total (pH T ) using the CO 2 SYS software as described above. Photosynthesis vs CO 2 curves (n = 3) and other parameters (n = 3) were obtained from two separate experiments under the same experimental conditions after the cells had acclimated for at least nine generations (see Supplemental Fig. 3 for detail).
Cell concentrations were measured with a Counter Particle Count and Size Analyzer (Z2, Beckman Coulter, USA) before and after the dilutions every 24 h. The cells had acclimated for at least nine generations before the growth rate was measured. The specific growth rate (μ, d −1 ) was calculated as where N 1 and N 0 represent cell concentrations at t 1 (before the dilution) and t 0 (initial or just after the dilution), respectively. A Clark-type oxygen electrode was used to measure mitochondrial respiration (after acclimation for~13 generations) under the conditions of pH, O 2 levels, and temperature used for growth, and the oxygen consumption rates were monitored in the dark (~10 min). About 6-8 × 10 5 cells were harvested by gentle vacuum filtration (<0.01 MPa) onto polycarbonate membrane filters (1.2 μm, Millipore, Germany). These cells were then re-suspended in seawater (2 mL) buffered with 20 mM Tris (without introducing additional DIC into media, pH T = 8.00 for AC of 14 μM and pH T = 7.70 for HC of 34 μM) to maintain stable pH in the media. Trisbuffered seawaters were flushed with pure nitrogen and ambient air to achieve the culture O 2 levels.
During the measurements of photosynthetic O 2 evolution and photorespiration, 5-6 × 10 5 cells were harvested after acclimation for~18 generations and resuspended as above. Photosynthetic O 2 evolution was tested under growth O 2 levels (~255 μM for AO and~57 μM for LO), and photorespiration ( Supplementary Fig. 5) was estimated as the difference in photosynthetic O 2 evolution of the cells under reduced (~25 μM) and culture (~255 μM for AO and 57 μM for LO) O 2 conditions, an approach which has been used widely 26,59 . However, this method might have overestimated the absolute value of photorespiration to some extent because of the ignored mitochondrial respiration rates at different O 2 levels. Therefore, we re-estimated the photorespiration (Fig. 6c) using the differences of dark-respiration rates between the samples measured under~25 μM O 2 and growth O 2 conditions (~255 μM O 2 for AO and 57 μM O 2 for LO), assuming that the mitochondrial respiration rates for the cells grown under the treatments were the same under light and darkness. To obtain the reduced or ambient levels of O 2 , pure nitrogen gas or ambient air were bubbled into Tris-buffered seawater (20 mM, pH T = 8.00 for AC of about 14 μM and pH T = 7.70 for HC of about 34 μM). Light intensity and temperature were the same as in the growth experiment.
Inhibition of photosynthetic O 2 evolution by acetazolamide (AZ) 60 , an inhibitor of periplasmic carbonic anhydrase (eCA), was determined with a Clark-type oxygen electrode under culture conditions. We added the AZ dissolved in 0.05 mM NaOH at a final concentration of 100 μM; an equal amount of 0.05 mM NaOH was added as a control treatment. The cells used for this test had been acclimated to the growth O 2 and CO 2 levels for about ten generations, and~5 × 10 5 cells were harvested and re-suspended in 2 mL seawater buffered with 20 mM Tris to maintain the CO 2 partial pressures as mentioned above. O 2 levels were achieved and controlled as above.
The photosynthesis vs CO 2 curves was determined with a Clark-type oxygen electrode under standard conditions commonly used for CCM studies 19 . Approximately 4-10 × 10 5 cells were harvested as above after acclimation for approximately nine generations and were re-suspended in DIC-free seawater (2 mL) medium buffered with 20 mM Tris (pH T = 8.00). The concentrations of DIC in the seawater were then adjusted by adding sodium bicarbonate solution, and the final DIC concentration reached to 8 mM. DIC (μM) values were converted to CO 2 (μM) with CO 2 SYS software. All the cells from different treatments were measured under the same standard conditions (pH T = 8.00, light intensity = 400 μmol photons m −2 s −1 , O 2 was in the range of 50-200 μM, and the temperature was controlled at 20 ± 0.1°C). CO 2 acquisition efficiency was calculated as CO 2 acquisition efficiency ¼ V max =K 0:5 ðCO 2 Þ; ð3Þ where V max and K 0.5 were calculated by fitting the photosynthetic O 2 evolution rates at various CO 2 concentrations with the Michaelis-Menten formula. Measurements of chlorophyll fluorescence parameters were carried out with a pulse amplitude modulated (PAM) fluorometer (XE-PAM, Walz, Effelrich, Germany) after the cells had acclimated for~12 generations. Effective photosystem II (PSII) quantum yield of photosystem (Yield) was measured with an actinic light level of 226 μmol photons m −2 s −1 (similar to that of the culture level). Nonphotochemical quenching (NPQ) was also measured at this actinic light intensity.
Approximately 5-8 × 10 5 cells were harvested (~18 generations) for measuring elemental composition. Particulate organic carbon (POC) and particulate organic nitrogen (PON) were determined by filtering cells on the pre-combusted (450°C for 6 h) GF/F filters (25 mm, Whatman), storing at −80°C before measuring. Filters were treated with HCl fumes to remove inorganic carbon and dried before analysis on a CHNS elemental analysizer (vario EL cube, Elementar, Germany). Biogenic silica (BSi) was determined by the spectrophotometric method 61 , and the cells were harvested onto Polycarbonate filters (1.2 μm, Millipore, Germany). Production of POC, PON, and BSi was calculated by multiplying the cellular content by specific growth rate.
Statistics and reproducibility. The data are expressed in raw form, or presented as means ± standard deviation (SD) with n = 3 (triplicate cultures or mesocosms). We used one-way ANOVA to assess significant differences among the treatments. Prior to analyses, data were checked for homoscedasticity. If required, data were Ln transformed, and then LSD test was used for post hoc investigation. If the data, even after transformation, did not meet the assumption for equal variance, Games-Howell tests were chosen for post hoc investigation. Linear fitting analysis was conducted with Pearson correlation analysis (two-tailed). Partial Correlation Analysis was employed to explore the net correlation between DO and photosynthetic light use efficiency in the Pearl River estuary investigation. Parameters including pH T , cultured temperature, DIN, SiO 3 2− , DIC, and CO 2 were under control. A 95% confidence level was used in all analyses.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The source data that underlying the main charts are provided as Supplementary Data 1. ARTICLE COMMUNICATIONS BIOLOGY | https://doi.org/10.1038/s42003-022-03006-7