In summer 2018 and 2020, a respective total of 800 and 240 adult blue mussels (Mytilus edulis) with a mean soft tissue dry weight (± SD) of 0.79 ± 0.28 g were sampled from a M. edulis longline in an offshore aquaculture pilot project approximately 10 km off the Belgian coast (51°11.02’N - 02°39.88’E). All samples were stored in aerated seawater and transported to the experimental facilities within 4 hours.

On both occasions, organisms were evenly and randomly distributed across four identical aquaria (100 × 5 × 0 cm). These were equipped with a continuous flow-through mechanism, allowing the homogenisation of approximately 400 L natural seawater in circulation per system. All aquaria were continuously aerated and pre-set at laboratory conditions mimicking the in situ seawater salinity, temperature and pH at the time of sampling (34 PSU, 20°C and pH 7.96; LifeWatch Belgium, 2015). After an initial acclimatisation period of minimum 48 h under ambient conditions, both seawater temperature and pH were manipulated individually across the aquaria. This resulted in a 2 × 2 factorial design with four experimental treatments: a control treatment [CTRL: in situ temperature and pH], an ocean acidification treatment [OA: in situ temperature and lowered pH], an ocean warming treatment [OW: elevated temperature and in situ pH] and a climate change treatment [CC: combined elevated temperature and lowered pH].

For each treatment, stepwise manipulations were imposed over the course of three days, with a daily seawater temperature increase of 1°C and/or pH decrease of 0.1 pH unit. Seawater temperature was regulated using TECO TK2000 heaters and pH was manipulated through the controlled bubbling of 100 % CO₂ in the OA and CC treatment tanks using the IKS AquaStar aquaristic computer system, simultaneously logging temperature and pH records every 15 minutes throughout the six-week experiments ( Table 1 ; Appendix A-Appendix B ). This resulted in environmental conditions of +3°C and/or -0.3 pH units in the corresponding treatments compared to the control settings, reflecting the IPCC RCP-SSP8.5 projections for ocean warming and acidification towards the end of this century (Hoegh-Guldberg et al., 2014; IPCC, 2021). These conditions were maintained for six weeks and organisms were fed three times a week by adding 5 mL Shellfish Diet 1800® (Instant Algae® mix by Reed Mariculture Inc.) to each aquarium. Glass pH electrodes were calibrated weekly using Hanna Instruments™ NIST Reference Buffer Solutions (4.01 and 7.01) and each aquarium was sampled weekly to determine Total Alkalinity (TA) using a CONTROS HydroFIA™TA alkalinity system. One third of each flow-through system’s water was renewed weekly. The carbonate chemistry of the seawater was calculated using CO₂SYS software (Pierrot et al., 2006) with the thermodynamic constants of Mehrbach et al. (1973).

The effect of temperature (current or elevated) and pH (current or lowered) on the production of N₂O in the WHOLE, SHELL, NaClO₃-inhibited and non-inhibited incubations was investigated using linear mixed effects models, with incubation core and/or experimental week added as a random factor in the models. If none of the random factors could explain left-over variance, a linear regression model was fitted. The effect of temperature, pH and experimental week (WK3 or WK6) on nutrient flux was analysed using linear regression models. Significance of the two-way interaction ‘temperature x pH’ and post-hoc pairwise comparison of the group means was used to identify possible combination effects of increased temperature and lowered pH. Normality of the residuals and model assumptions were checked using Shapiro-Wilk, Gaussian error distributions were used and model selection was based on the parametric bootstrap and Kenward Roger methods for mixed model comparison (Halekoh and Højsgaard, 2014) or on the backwards selection procedure for linear regression models. Statistical analyses were conducted using R v3.6.1 (R Core Team, 2019), linear mixed effects models were built using the R package lme4 (Bates et al., 2015) and linear regression models were fitted using the stats package (R Core Team, 2019).

To test for differences between oxygen microprofiles between experimental treatments in WK3 and WK6 (to keep a balanced design), a multivariate matrix was constructed with the oxygen concentrations at each distance above the shell (Widdicombe et al., 2013). Differences in the shell biofilm oxygen microprofiles were tested using permutational ANOVA (PERMANOVA) with experimental treatment and week as fully-crossed factors and with all profiles on all shells considered as replicates within the Euclidian distance matrix. For all significant PERMANOVA factors, post-hoc pairwise testing was used and homogeneity of multivariate dispersion was tested using PERMDISP. The SIMPER routine was used to identify which distances from the shell contributed most to any observed differences. Multivariate analyses were carried out in Primer v6.0 with PERMANOVA+ add-on software (Clarke and Gorley, 2006; Anderson et al., 2008).

In all treatments, N₂O was produced in both the WHOLE and SHELL incubations, whereas no N₂O production was measured in the poisoned ZNCL incubations. In CTRL, whole animals produced on average (± SD) 4.55 ± 0.18 nmol N₂O ind^-1 h^-1 and throughout the experiment, the shell biofilm contribution ranged between 65-75% ( Figure 2 ). In OW, the average WHOLE production rate gradually increased from 5.47 ± 0.23 to 15.84 ± 0.48 nmol N₂O ind^-1 h^-1 over six weeks, while the shell biofilm contribution remained stable at 68% ( Figure 2 ). In both acidified treatments, the relative contribution of the shell biofilm to whole organism N₂O production decreased over time. In OA, the WHOLE N₂O production rate was stable throughout the experiment (mean 5.79 ± 0.33 nmol N₂O ind^-1 h^-1), but the contribution of the shell biofilm decreased from 63% to 23% over the course of six weeks ( Figure 2 ). Whole animals in CC produced on average 9.10 ± 1.82 nmol N₂O ind^-1 h^-1, while the associated shell biofilm contribution decreased from 60% to 12% over time, after an initial rise to 65% in the second week of experiments. Overall, temperature and pH had a significant, antagonistic effect on the parts of the animal that emitted N₂O, with an increasing production in warmer environments and a decreasing production in acidified or combined climate change conditions ( Table 4 ).

The current climate [CTRL] N₂O production of blue mussels (including their associated shell biofilm) was in line with other studies (e.g. Heisterkamp et al., 2010; Heisterkamp et al., 2013; Gárate et al., 2019) and exclusively biological, as ZnCl₂ poisoning fully inhibited N₂O production. The shell biofilm contribution to total N₂O production by M. edulis was important (± 70 %) and remained stable through time, as did the total N₂O production rate. Although still a considerable contribution, the partitioning measured in this study was lower than previously reported by Heisterkamp et al. (2013), where N₂O production originated almost exclusively from the shell biofilm (± 94 %). Possibly, methodological differences such as the use of natural seawater and a regular feeding regime in our experiments could have prompted the adept filter feeder M. edulis, with an established rich intestinal microbial system capable of considerable N₂O production, to continue to do so throughout the duration of this experiment.

Nitrifiers, a combination of ammonium-oxidising bacteria and archaea, produce N₂O during nitrification in oxic environments and during nitrifier denitrification in suboxic environments. Another set of bacteria, denitrifiers, produce N₂O during anaerobic denitrification in anoxic environments, comprising the final steps of CND. These three pathways are considered to be the main oceanic sources of N₂O production, while the main sink is the further reduction of N₂O to N₂ by (nitrifier) denitrification in sub- or anoxic environments (Bange et al., 2010). In this study, each of these three N₂O producing pathways were observed. Contrary to Heisterkamp et al. (2013), where only dissected shells were used in a similar experiment, nitrification apparently did not play as an important role when whole animals (incl. shell biofilm) were incubated. Most likely, the presence of the living and digesting animal itself caused the resulting total N₂O production to lean towards a more prominent inclusion of anoxic or suboxic pathways.

Denitrification of nitrite, targeted by the DNO2 precursor incubations, was the most dominant pathway of N₂O production by M. edulis in the current climate scenario, corroborating the findings of Heisterkamp et al. (2013). Production of N₂O in the presence of NaClO₃, actively preventing the oxidation of nitrite to nitrate, was on average an order of magnitude higher than without the inhibitor present. As such, we suggest that the availability of nitrite might be a limiting factor in the production of N₂O and that incomplete (nitrifier) denitrification of nitrite, where further reduction to N₂ has not (yet) happened, is indeed a major pathway of N₂O production by M. edulis and its biofilm. An overestimation of said denitrification due to anaerobic ammonium oxidation (anammox), where ammonium oxidises to N₂ using nitrite under anoxic conditions, was assumed negligible since no anammox activity has been associated with mussel biofilms before (Marzocchi et al., 2021).

Denitrification of nitrite can be associated with the presence of M. edulis in different ways, both in and on the organism. The incomplete denitrification of nitrite to N₂O by denitrifiers in the mussel’s anoxic digestive system could be the result of a high lysozyme activity, cutting short the denitrification pathway (Birkbeck and McHenery, 1982; Heisterkamp et al., 2013; Gárate et al., 2019). Additionally, this last step in the nitrifier denitrification pathway has only be attributed to a certain genus of nitrifying bacteria, Nitrosomonas sp., whose biofilm presence and/or (differential) activity is related to oxygen and nutrient availability (Schramm et al., 2000; Shrestha et al., 2002) and substrate affinity and availability (Foesel et al., 2008; Zhu et al., 2013). With the variable oxygenation of the shell biofilm measured in this study in mind, this could further explain why the produced N₂O is not always readily reduced to N₂ in a heterogeneous microbial biofilm on M. edulis’s outer shell. However, as the N₂O flux was several orders of magnitude lower than that of the other measured N species ( NH 4 + , NO 2 - and NO 3 - ), reduction of N₂O to N₂ must still have occurred to close the nitrogen budget in the experimental incubations. Research by Marzocchi et al. (2021) suggests that mussel-associated N₂ production, although currently often overlooked, might be substantial. The high CTRL nitrate consumption rates observed in this study comply with this assumption.

By nature, adult blue mussels have a wide thermal tolerance window (e.g. Seuront et al., 2019; Kamermans and Saurel, 2022). Since this species has adapted to survive in such variable conditions, this study opted to push towards the boundaries of its natural tolerances in order to get the most realistic stress response to variable climate scenarios. In the North Sea, the highest mean temperatures occur over the summer period (when organisms were sampled) when the mean seawater pH is fairly stable (LifeWatch Belgium, 2015), meaning the adopted climate manipulations should elicit a relatively accurate physiological response outside M. edulis’ natural tolerances. Results show that the effects of seawater temperature rise and ocean acidification were generally of an antagonistic nature. N₂O production rates of whole animals, as well as the production rate of dissected shells with an intact microbial biofilm, significantly increased with a higher temperature and significantly decreased in an acidified environment (whole animals) or a combined high temperature/low pH climate change environment (biofilm).

These experiments were set up to mimic North Sea summer conditions in CTRL, with additional manipulations in terms of temperature and/or pH levels according to the experimental treatments. Although this summer setting is the most relevant in terms of studying potential stress responses due to climate change, the temperature-dependent nature of physiological and microbial processes suggest that the absolute values of these responses would differ between the seasons (Lesser et al., 2010; Boulêtreau et al., 2012; Múgica et al., 2015). However, whether the observed patterns and nature of the interaction between ocean warming and acidification varies seasonally remains to be confirmed.

The increasing effect of elevated temperature on N₂O production rates by M. edulis and its associated shell biofilm are most likely linked to the mussel’s concomitantly increasing filtration rates in combination with temperature-induced increased biofilm activity. As bivalve filter feeding increases with temperature (Kittner and Riisgård, 2005; Ong et al., 2017; Voet et al., 2021), more denitrifying microorganisms might be ingested and/or the concentration of available N precursors in the digestive system might increase. In addition, general microbial activity increases with warmer temperatures (Kroeze and Seitzinger, 1998; Boulêtreau et al., 2012). This study showed that even though the N₂O production by whole animals increased with temperature, the relative contribution of the microbial shell biofilm remained stable over time in the ocean warming [OW] scenario. Indeed, this would suggest a similar positive effect of warmer environments on both the animal’s and the shell biofilm’s N₂O producing capacities. Moreover, increased feeding rates will result in higher biodeposition rates and in turn, a higher nutrient load, potentially further stimulating (de)nitrification activities in the microbial shell biofilm.

On the other hand, the relative contribution of the shell biofilm to N₂O production evidently decreased over time in both acidified scenarios, on top of the overall decrease in whole animal N₂O production rates. A negative effect of ocean acidification on the marine nitrogen cycle was previously observed by Beman et al. (2011), reporting reduced oceanic nitrification at even relatively small experimental pH reductions (0.05-0.14). Together with a profound effect of reduced pH on the community composition of marine biofilms (Espinel-Velasco et al., 2021), this could potentially explain the observed lower N₂O production rates under acidified conditions, as well as the decreasing relative contribution of the microbial biofilm. The latter is most likely also tied to a thinner, patchier or even disappearing shell biofilm, as is suggested by this study’s significantly affected oxygen microprofiles in both acidified conditions. The microprofile analysis indeed confirmed an adverse effect of pH on the metabolic activity (respiration) and composition of the shell biofilm. In both acidified treatments, perceived biofilm thickness considerably decreased, as well as the net biofilm oxygen consumption. Furthermore, net oxygen consumption at the shell surface sometimes dropped to (near) zero under acidified conditions, indicating either an incomplete biofilm cover or a profound shift in functional metabolism and/or the biofilm’s microbial makeup. A more in-depth functional analysis of M. edulis’ shell biofilm under climate change conditions is described in more detail by Dairain et al. (manuscript in preparation) ² . Nitrifier denitrification, shown to be the dominant N₂O producing pathway for M. edulis (e.g. this study and Heisterkamp et al., 2013), gained even more importance in the OA treatment. Affirmingly, this could (in part) be a result of the increasing prevalence of suboxic conditions provided within a thinning and/or receding shell biofilm.

The blue mussel is by far the most abundant mollusc species colonising North Sea artificial hard substrates (Coolen et al., 2020) and these numbers will multiply even further by planned co-location strategies involving offshore wind farms and bivalve mariculture (Schupp et al., 2019; MSP, 2020; Steins et al., 2021). Considering the prevalence of these organisms and the potential biofilm-substrate they provide are on the rise, their role in the local emission of the potent GHG nitrous oxide could be noteworthy and a valuable research topic. On the other hand, this study also suggested that reduction of this GHG to inert N₂ could be prevalent in or around these mussels, as was the case in other recent mussel microbiome research (Marzocchi et al., 2021). This would suggest that colonising and aquaculture mussel communities could play a role in the removal of nitrogen from the water column, actively counteracting marine eutrophication. Contrary to climate change-induced changes in (i.a.) survival or ecophysiology of blue mussels (Voet et al., 2021), the antagonistic effects of temperature and pH on N₂O production described in this study would most likely not amount to large-scale enhancements of future mussel-associated N₂O emissions. With the expansion of the offshore renewable energy effort in a necessary attempt to abate GHG emissions causing global climate change, insights such as described in this study are paramount. Complementary, marine climate science would benefit greatly from further research into the link between impacts of future climate effects on current marine communities and the potential effects of adaptation and ecophysiological acclimatisation over longer time periods.