Biogeochemical feedbacks to ocean acidification in a cohesive photosynthetic sediment

Ecosystem feedbacks in response to ocean acidification can amplify or diminish diel pH oscillations in productive coastal waters. Benthic microalgae generate such oscillations in sediment porewater and here we ask how CO2 enrichment (acidification) of the overlying seawater alters these in the absence and presence of biogenic calcite. We placed a 1-mm layer of ground oyster shells, mimicking the arrival of dead calcifying biota (+Calcite), or sand (Control) onto intact silt sediment cores, and then gradually increased the pCO2 in the seawater above half of +Calcite and Control cores from 472 to 1216 μatm (pH 8.0 to 7.6, CO2:HCO3− from 4.8 to 9.6 × 10−4). Porewater [O2] and [H+] microprofiles measured 16 d later showed that this enrichment had decreased the O2 penetration depth (O2-pd) in +Calcite and Control, indicating a metabolic response. In CO2-enriched seawater: (1) sediment biogeochemical processes respectively added and removed more H+ to and from the sediment porewater in darkness and light, than in ambient seawater increasing the amplitude of the diel porewater [H+] oscillations, and (2) in darkness, calcite dissolution in +Calcite sediment decreased the porewater [H+] below that in overlying seawater, reversing the sediment–seawater H+ flux and decreasing the amplitude of diel [H+] oscillations. This dissolution did not, however, counter the negative effect of CO2 enrichment on O2-pd. We now hypothesise that feedback to CO2 enrichment—an increase in the microbial reoxidation of reduced solutes with O2—decreased the sediment O2-pd and contributed to the enhanced porewater acidification.


Results
In the following, we report H + /O 2 concentrations ([H + ], [O 2 ]) and fluxes in the porewater of Control sediment under conditions of light and darkness, first in ambient seawater and then in CO 2 -enriched seawater. Following this, we assess how the addition of a surface calcite layer (+Calcite) altered porewater [H + ] and [O 2 ], again, first in ambient seawater and then in CO 2 -enriched seawater.
Our pH microprofiles describe depth gradients in proton concentrations [H + ] ([H + ] = 10 −pH ). Because the range of [H + ] observed in this experiment was small, we do not report these as pH but instead as nmol L −1 . We note that within the range of pH measured in our experiment, the [H + ] is directly proportional to the [CO 2 ] because CO 2  ] (µmol kg SW −1 ). The calcite and aragonite saturation states (Ω CA and Ω AR , respectively) were derived from the measured parameters for a temperature of 15 °C. www.nature.com/scientificreports/ which molecular diffusion is the dominant transport mechanism for solutes-indicated that, on average, the sediment removed 2.5 times more H + from the overlying seawater in light than it released in darkness ( Table 2, H + flux DBL ). Visual inspection of the sediment surfaces revealed that the motile pennate diatoms species that dominated the microphytobenthos of this sediment (see Vopel et al. 2018; genera Pleurosigma, Gyrosigma, Nitzschia, Thalassionema, and Bacillaria) relocated to the surface within 3 h of the sand and calcite layers being added (Fig. 1a). Photosynthesis lowered the [H + ] in the top millimetre of the sediment below that measured in the overlying seawater (Fig. 2a). This reversed the direction of the sediment-seawater H + exchange observed in darkness and increased the flux of H + from the bottom of the oxic zone to the surface sediment by a factor of 1.7 (H + flux sub in Table 2).
Vertical [O 2 ] microprofiles measured in the DBL to compute the diffusive sediment-seawater O 2 exchange (DOE), and in the sediment porewater to compute the depth-integrated sediment O 2 production (R A ) showed that the Control released on average 1.9 and 1.4 times, respectively, more O 2 in light than it consumed O 2 in darkness ( Table 2). Note the large standard deviation of the average DOE measured in light indicating patchiness in the distribution of benthic microphytes (Fig. 1a). In darkness, O 2 diffused from the freely flowing seawater above the sediment to an average sediment depth of 7.6 mm (Fig. 3b). As expected, photosynthesis at the surface of the sediment supersaturated the porewater of the top 2 mm of the sediment with O 2 , reversing the dark O 2 flux and increasing the O 2 penetration by 0.9 mm (Fig. 3a). Note that when all the derived DOE and R A data (Table 2) were pooled, they were linearly correlated (R 2 = 0.96), but agreement between these two estimates gradually decreased with increasing O 2 production (Fig. S2). The three replicate Fourier Transform Infrared (FTIR) spectra of ground oyster shell granules in (b) indicate absorption peaks around 877 and 713 cm −1 . These are due to vibrations of the carbon-oxygen double bond in the carbonate ion of calcite, confirming that the ground oyster shells are primarily composed of calcite. The cores in (c, d) received a 1-mm surface layers of (c) sterile carbonate free sand (Control) or (d) calcite (ground oyster shells; +Calcite).     Table 2). In light, Control and +Calcite did not differ in any of the measured or derived sediment O 2 consumption proxies (Tables 2, 3, Fig. 3a). In darkness, the +Calcite cores removed on average less O 2 from the overlying seawater than the Control cores (DOE in Tables 2, 3, Fig. 3b). However, the estimates of R A and the volume-specific O 2 production, R V , did not confirm this difference (Tables 2, 3  In light, calcite deposition had no statistically clear effect on H + flux DBL ( Table 2, Fig. 2c). In darkness, however, this resulted in a H + flux DBL similar in size but opposite in direction of that in the Control (152 ± 98 vs. − 155 ± 111 mmol m −2 h −1 , Table 2, Fig. 2d). The effect of calcite deposition on the the measured or derived sediment O 2 consumption proxies was statistically not clear, in both light and darkness (Tables 2, 3).

Discussion
Our measurements in cores of a photosynthetic subtidal silt sediment with 1 mm of added sterile sand (Control) revealed that enrichment of the overlying seawater with CO 2 had increased the peak-to-peak amplitude of the diel porewater [H + ] oscillations (Fig. 4a, compare filled triangles vs. filled circles). Two effects likely contributed to this increase: (1) in darkness, sedimentary microbial reaction processes seem to have amplified the positive effect of seawater CO 2 enrichment on porewater [H + ] (Fig. 5a, filled symbols), and (2) replenishment of the CO 2 taken up during diatom photosynthesis by the bicarbonate pool may have consumed H + at a greater rate in CO 2 -enriched seawater than in ambient seawater (Fig. 5a, open symbols). Furthermore, we showed that seawater CO 2 enrichment decreased the sediment penetration of O 2 in both light and darkness (Tables 2, 3) (Fig. 2d). Consequently, the peak-to-peak amplitude of the diel porewater [H + ] oscillations decreased (Fig. 4a, open triangles). Such an effect was not observed in ambient seawater: the overlapping Δ[H + ] D-L profiles in Fig. 4a Tables 2, 3). These gradients may not correctly reflect the steady-state sediment-seawater O 2 exchange, but if so, then the +Calcite cores seemed to have removed less O 2 from their overlying seawater than the Control cores in both ambient (d ES = − 1.6) and CO 2 -enriched seawater (d ES = − 1.35). We note, however, that the proxies derived from the measured porewater [O 2 ] profiles (O 2 -pd, R A and R V ) did not return statistically clear differences.
We hypothesise that the observed increase in the flux of H + from the bottom of the oxic layer of the +Calcite sediment resulted from the dissolution of the sediment-facing boundary of the calcite deposit. Inspection of the [H + ] microprofiles shown in Fig. 2b reveals that this may have started at a porewater [H + ] of 15-20 nmol L −1 (pH 7.8-7.7)-the concentrations measured at 1 mm depth. We can infer from the Control that CO 2 enrichment of the sediment-overlying seawater then raised the [H + ] in the porewater of the calcite layer above ~ 25 nmol L −1 (below pH 7.6, Fig. 2d, Control) initiating dissolution, which lowered the porewater [H + ] to about 20 nmol L −1 (increased pH to ~ 7.7) as shown in Fig. 2d. If the dissolution-precipitation balance was to shift toward net dissolution at Ω CA < 1, and assuming that porewater total alkalinity, TA = 2.3 mmol kg −1 , then calcite should be stable at pH > 7.4, and the [H + ] profiles across the calcite layer should resemble the profiles measured in the Control. Our results suggest that dissolution started at a lower [H + ] (higher pH) confirming the evidence presented by others 24 who have observed that gross dissolution of whole-shell biogenic CaCO 3 occurred in treatments that were oversaturated (Ω > 1) with respect to calcite.
The abundance of large motile pennate diatoms raises the possibility that the dissolution of the surface calcite was enhanced by the activity of extracellular enzymes. Diatoms use an extracellular carbonic anhydrase 25,26 , located in the periplasmatic space 27,28 , as part of a carbon-concentrating mechanism that increases the flux of CO 2 towards the carboxylating enzyme, ribulose-1,5-bisphosphate carboxylase-oxygenase (RubisCO) 29 . This enzyme catalyses the otherwise slow inter-conversion of CO 2 and HCO 3 − and in this case, facilitates CO 2 uptake by generating CO 2 from HCO 3 − at the cell surface 26 through a two-step process, a hydration-dehydration step followed by a rate-limiting transfer of protons, which is presumably buffered by the acidic polymerised silica of the diatoms' cell wall 30 . Evidence presented by Subhas et al. 31 suggests that carbonic anhydrase increases calcite dissolution for saturation states from 0.6 to near 1, the effect being most pronounced close to equilibrium (Ω CA = 1).

What amplified the effects of biogeochemical processes on porewater [H + ] in CO 2 -enriched seawater? In Fig. 5, we use vertical Δ[H + ] enr-amb profiles to assess how seawater CO 2 enrichment altered [H + ]
in both the sediment-overlying seawater and the sediment porewater. As mentioned above, for both Control and +Calcite, injection of CO 2 -enriched air will initially have decreased the porewater-seawater [H + ] gradient observed in darkness and so the dark H + efflux. If sedimentary reaction processes kept producing and consuming H + at unchanged rates, then the porewater [H + ] and sediment-seawater H + efflux must have gradually increased until net production and efflux were again in equilibrium. That is, the [H + ] at the sediment surface must have increased by as much as the [H + ] in the bottom seawater and Δ[H + ] enr-amb would gradually be attenuated with sediment depth. However, Fig. 5a shows that in darkness, Δ[H + ] enr-amb in the porewater of the upper 2 mm of the sediment exceeded that in the overlying seawater. This suggests that net-production of H + had increased leading to higher porewater concentrations and a higher H + efflux.
Following Middelburg et al. 32 , the instantaneous effect of a biogeochemical process on [H + ] is the product of the net charge exchanged during the process (Δcharge), the sensitivity factor of seawater (δpH/δCBA; CBA = charge balance alkalinity), and the process intensity (I process , mol m −3 s −1 ): Because the sensitivity factor and the net charge exchange are functions of pH, the influence of a biogeochemical process also depends on pH. In other words, the acidification of the sediment porewater can alter the positive or negative effect of biogeochemical reaction processes on porewater [H + ] even if the intensity of the reaction process remains unchanged. For aerobic mineralisation and the reoxidation of reduced solutes with O 2 , this pH dependency sees the production of [H + ] steeply increasing as porewater [H + ] increases above 10 nmol L −1 (pH decreases below 8.0 33 ). That is, an increase in porewater [H + ] will have increased the production of H + by porewater microbial reaction processes raising Δ[H + ] enr-amb above that observed in the overlying seawater, even if the reaction process intensity remained unchanged. It would follow then that the observed increase in Δ[H + ] enr-amb does not necessarily imply that an ecosystem process has responded to additional CO 2 .
While in darkness Δ[H + ] enr-amb in the porewater of the Control exceeded that in the overlying seawater, in light, it steeply decreased below that in the overlying seawater reaching a minimum just below the sediment surface (Fig. 5a). This suggests photosynthesis must have removed more H + in CO 2 -enriched seawater than in ambient seawater. Again, this may have followed from the pH dependencies of the seawater sensitivity factor and the net charge exchange (see above); Soetaert et al. 33 showed that the consumption of porewater H + by photosynthesis based on ammonium or nitrate increases steeply as the environmental [H + ] increases above 10 nmol L −1 . An ] enr-amb may have a common cause, a positive CO 2 response of microbial reaction processes that generate H + and consume O 2 : the microbial reoxidation of reduced solutes (iron, manganese and sulphur) with O 2 and aerobic mineralisation and nitrification. For example, an increase in the production and subsequent oxidation of Fe 2+ would have raised both the O 2 demand and the porewater [H + ] in the surface sediment adding to the effects of the pH-dependent sensitivity factor and net exchange of charge, as per Eq. (2).
Here, for the first time, we presented evidence for a positive effect of seawater CO 2 enrichment on the peakto-peak amplitude of the dark-light oscillation in porewater [H + ] that characterise cohesive photosynthetic sediment. We discussed possible causes of this effect and showed that dissolution of carbonate, if added to the surface, will diminish these oscillations. The pH dependencies of both the seawater sensitivity factor and the net exchange of charge may explain why sediment biogeochemical processes in CO 2 -enriched seawater added and removed more H + to and from the porewater in darkness and light, respectively, than they did in ambient seawater. Ecosystem feedbacks in the form of CO 2 -induced changes in the intensity of photosynthesis in light and respiration in darkness may also explain an enhanced consumption and production, respectively, of porewater H + , but this is not consistent with the observed decrease in O 2 penetration in light and the similarity of the [O 2 ] microprofiles measured in ambient and CO 2 -enriched seawater. One possible process that explains this discrepancy is increased microbial reoxidation of reduced solutes with O 2 , and thus increased sediment O 2 demand, decreasing the O 2 penetration depth while increasing porewater [H + ]. The dissolution of the added calcite then effectively countered the increase in porewater [H + ] (Fig. 5b), but not the CO 2 -induced decrease in O 2 penetration. This dissolution should not be considered an ecosystem feedback because calcite was not naturally present at the sediment surface. It indicates, however, a challenge for recruits of calcifying fauna arriving at the sediment surface. A chemically more aggressive (acidified) surface layer may prevent macrofauna recruitment [20][21][22] and so alter the three-dimensional complexity of the sedimentary ecosystems 34 and associated carbon and nutrient remineralisation and sediment-seawater solute exchange processes.

Material and methods
Sediment collection and properties. On June 11th 2019, we collected 24 cores of subtidal silt with SCUBA at 10 m water depth in Man O'War Bay (S 36° 47′ 38″, E 175° 10′ 14″), Hauraki Gulf, New Zealand, as described in Vopel et al. 19,35 . The cores were kept below 15 °C, the in situ seawater temperature, during a 2.5 h trip to the laboratory. The salinity of the seawater was 34.5. Loss of weight of the silt's upper 10 mm layer after drying at 60 °C and combustions at 550 and 950 °C indicated that on average, porewater accounted for 68 ± 1.3% of the silt's wet weight, and 9.5 ± 0.4 and 4.1 ± 0.2% of the silt's dry weight were organic matter and the calcium carbonate, respectively (± 1 SD, n = 3). Chlorophyll a and phaeopigment contents determined following Jeffrey and Humphrey 36 were 12.5 ± 1.9 and 20.5 ± 2.4 μg (g dry weight) −1 (n = 3) giving a Chl a/phaeopigment ratio of 0.6. Particle sizes, determined using ~ 10 mL sample of the homogenized silt and a Malvern Mastersizer 2000, ranged from 0.24 to 350 μm with a volume weighted mean of 35.7 ± 0.9 μm (Kurtosis: 8.0 ± 1.3, Skewness: 2.6 ± 0.2).
Laboratory setup and seawater properties. We submerged 12 randomly selected sediment cores in two independent experimental units (EU) each circulating ~ 1120 L of natural seawater ( Table 2). Details of these units are described in Vopel et al. 19,35 . LED floodlights provided ~ 130 μmol quanta m −2 s −1 of photosynthetically active radiation (PAR) from 7 a.m. to 7 p.m. at the surface of the submerged sediment cores. This intensity was similar to that measured midday at the core collection site (K. Vopel unpublished observations). The seawater in each circulation unit was continuously sterilised with UV light and a sprinkler returning seawater from an inline particle filter ensured that the seawater was saturated with O 2 (K. Vopel, unpublished observations).
Starting on d 2 of the experiment, we manipulated the seawater circulating in one EU by automatic stepwise injection of CO 2 -enriched air (5% CO 2 , 21% O 2 in nitrogen) to decrease its pH by 0.04 units per day (for 10 d) until a pH of 7.60 was reached (Fig. S1). The seawater pH was then maintained at 7.60 for a further 9 d until the end of the experiment (d 21). The level of CO 2 enrichment was controlled by CapCtr software (Loligo Systems Aps), a SenTix HWD electrode connected to a pH 3310 m (WTW), and a solenoid valve. For additional details including the calibration of pH electrodes see Vopel et al. 35 .
We measured the seawater salinity daily and kept it between 34.4 and 34.7 by adding ultrapure water to account for evaporation. Seawater analyses in previous experiments (K. Vopel unpublished observations) revealed that this addition had no measurable effect on the seawater carbonate chemistry. One litre of seawater collected from each EU weekly was analysed for dissolve inorganic carbon (DIC) with a SOMMA (Single Operator Multiparameter Metabolic Analyzer) coulometer system, and total alkalinity (TA) with a closed-cell potentiometric titration system following the SOP's 2 and 3a procedures 37 . D. Pierrot's adaptation of the CO 2 Sys.BAS program 38 computed the seawater pCO 2 and pH (total scale, mol kg-SW −1 ). The dissociation constant for HSO 4 − was taken from Dickson 39 ; the values of K 1 and K 2 of carbonic acid were from Mehrbach et al. 40  www.nature.com/scientificreports/ Millero 41 . The CO 2 Sys.BAS computations confirmed that the stepwise increase in the injection of CO 2 -enriched air increased the seawater pCO 2 by a factor of ~ 2.6 and decreased seawater pH from 8.0 to 7.6 ( Table 1).

Sediment treatment.
On d 1 of the experiment, we added a 1 mm thick surface layer of biogenic calcite ( Fig. 1) to six cores in each of the two experimental units. To do so, we briefly ground (Omni Ruptor 4000 Ultrasonic Homogenizer) bleach sterilised and rinsed oyster shells and sieved the material to exclude particles > 125 µm. Fourier transform infrared spectroscopy of this material confirmed the identity of the CaCO 3 mineral (Fig. 1). We weighed 2 g into each of 12 seawater-filled 100-mm diameter petri-dishes to create a consistent, 1 mm thick layer at the bottom of each petri dish. The content of each petri dish was then frozen at − 80 °C and one of the resulting solid seawater/calcite disks was placed into the headwater space of each core. As the disks thawed, an even layer of calcite was distributed onto the surfaces of the sediment cores. To avoid disturbing the settling particles, the cores were isolated from the surrounding seawater until the surface layers of calcite had formed (~ 20 min , and integrated volume-specific O 2 production (R V = R A /O 2 -pd, nmol cm −3 h −1 ), with the model PROFILE 42 neglecting the sediment biodiffusivity and irrigation. Estimates of the porosity of the top 10 mm sediment, based on microelectrode measurements of apparent gas diffusivity in intact sediment cores (for technical details, see Revsbech et al. 43 and Vopel et al. 44 ), revealed an average value of ϕ = 0.85 (data not shown). The molecular diffusion coefficients of O 2 and H + (D 0 ) were from Broecker and Peng 45 and Cussler 46 , respectively, and corrected for temperature and salinity as described by Li and Gregory 47 . The diffusivities corrected for tortuosity were calculated as D S = D 0 × ϕ, following Ullman and Aller 48 . The difference in molecular diffusivity between the silt and the overlying seawater caused a distinct change in the slope of the measured [O 2 ] profile at a position that marked the silt surface 49,50 .
Carbonate species distributions were determined in PHREEQC 3.0 using the seawater composition presented in Nordstom et al. 51 to calculate the activity of associated inorganic complexes. Calculations were completed for each microprofile depth. While the total alkalinity in pore waters was not known, the activity ratio of CO 2 :HCO 3 is independent of [DIC] and can be computed using the relevant equilibrium constants and the concentration of protons in solution.
Statistical data analyses. The vertical profiles of Δ[H + ] D-L and Δ[H + ] enr-amb shown in Figs. 4 and 5 were calculated using averages of [H + ] measured in replicate cores under conditions of light (L) and darkness (D) or in acidified (enr) and ambient (amb) seawater, respectively. We used the language proposed by Dushoff et al. 52 when describing the conclusion from our statistical tests in terms of statistical 'clarity' rather than 'significance' . (These authors argue against the over-reliance on a single (somewhat arbitrary) p value for determining the 'significance' of a test result). We used a two way analysis of variance (ANOVA) to test the effects of seawater pCO 2 (ambient, enriched) and sediment surface deposit (sand, calcite) and their interaction on O 2 -pd, DOE, R A , H + flux DBL and H + flux sub under light and dark conditions. When a clear pCO 2 × Deposit interaction was detected, separate Tukey HSD post-hoc analyses were undertaken for Deposit under pCO 2 conditions and vice versa. All statistical tests were conducted using Statistica (StatSoft GmbH). Prior to analysis, data were checked for normality and homogeneity of variance (visual inspection of residuals); no transformations were required. The d Effect Size (d ES) and 95% Confidence Interval were computed using the Excel routine created by Jared DeFife, Emory University, 2009 (http:// web. cs. dal. ca/ ~anwar/ ds/ Excel4. xlsx).

Data availability
The datasets are available from the corresponding author on reasonable request.