Experiments were conducted at the Smithsonian Marine Station (SMS) mesocosm facility between August-October 2022. The mesocosm system is described in detail in Johnson et al. (2021b). It consists of 12 independent, closed-circulation 40-liter tanks (Aqualogic) which were filled with filtered seawater (0.5 followed by 0.2 µm and CoralLife UV light). In each tank, seawater was circulated through a temperature control unit with an aquarium pump (Aqualogic) to maintain water temperature at 27 ± 0.5°C. Partial seawater changes (50%) were carried out twice weekly and salinity was maintained with the daily addition of approx. 500 ml reverse osmosis (RO) fresh water to each tank to counteract the effect of evaporation. Ambient lighting was maintained for 12 hours per day at PAR 173 ± 12 µmol photons m⁻² s⁻¹. Dissolved oxygen (DO, mg L^-1) concentration was measured continuously with an Oxyguard DO sensor in each tank, linked to a Neptune Apex control system. Oxyguard DO probe membranes were renewed at the start and calibrated each week throughout the experiment. Experimental DO concentrations were manipulated to create 2 deoxygenation treatments, 3.5 mg L^-1 and 1.5 mg L^-1, and a control treatment of 6 mg L^-1. Concentration of DO was controlled by bubbling nitrogen through the water causing oxygen to be expelled. Nitrogen was directly injected into a small circulation pump, so that it was mixed with water and dispersed throughout the tank. Care was taken to direct bubbles away from study organisms. An automated system of solenoid valves controlled the release of compressed nitrogen into each tank to maintain consistent DO concentrations at the target value. Control tanks were set up with an identical system but bubbled ambient air. In 6 of the tanks (2 per treatment), pH was logged each minute with Neptune pH sensors, which were calibrated 2 – 3 times per week following manufacturer protocol. Daily measurements of pH (NBS scale), DO, and temperature were collected using a YSI Pro 4 (pH was calibrated every 48 hours using YSI calibration buffers at pH 4, pH 7 and pH 10; DO was calibrated using 100% saturated air every 3 – 4 days) and temperature was cross-referenced daily with a handheld traceable thermometer (Fisherbrand) ( Table 1 ).

Larvae of three species of critically endangered, broadcast spawning corals were obtained from the Florida Aquarium Center for Conservation and Biscayne National Park in August 2022. The larvae were produced from gametes collected from spawning coral colonies at field sites (Orbicella faveolata) and from aquarium-reared corals (Colpophyllia natans, Pseudodiploria strigosa) ( Table 2 ). Larvae were produced via ex-situ assisted fertilization conducted following standard protocols (O’Neil et al., 2021; Banaszak et al., 2023) and swimming larvae were transported in plastic containers filled with filtered, oxygenated seawater to the SMS mesocosm facility where they were stored in plastic containers prior to settlement assays. Water changes using filtered (0.2 µm) seawater (27 ± 1°C) were carried out every 48 hours using 80 - 100 µm mesh to filter larvae. Larvae were placed in the experiment when they were competent to settle and metamorphose, between 3 to 9 days post-fertilization ( Table 2 for full timeline). The timeline of larval competency (the time during which the larvae are capable of settlement and metamorphosis) is generally capped at 30 days (Wilson and Harrison, 1998), and all the larvae used in this study were within that timeframe.

Experimental coral larvae were exposed to one of three DO concentrations (1.5, 3.5, or 6 mg L^-1) using settlement chambers ( Figure 1 ) placed in the mesocosm, with 4 replicate tanks per DO treatment (n=4). Fifty larvae from a given species were placed in each settlement chamber, and one settlement chamber for each species of coral larvae was placed in a given mesocosm tank, although different species started incubating at different times. Species-specific differences in spawning date and larval development resulted in a staggered start dates for each species. Since P. strigosa larvae were abundant they were divided into two batches (referred to as P. strigosa 1 and P. strigosa 2 herein), which were placed into the settlement chambers 6 days apart ( Table 2 ).

Settlement chambers consisted of a PVC tube (Length 10 cm, inner diameter 5.6 cm) fitted with mesh on both ends (100 µm pore size for O. faveolata and P. strigosa 2, or 180 µm pore size for C. natans and P. strigosa 1). Mesh was glued directly to the tube on one end of the chamber and to a removable lid on the other end ( Figure 1 ). For each settlement chamber, swimming larvae were selected and transferred using a pipette into transparent larval settlement chambers (described in Ritson-Williams et al., 2020) each holding a settlement tile (5 x 5 x 1 cm surface area) which had been pre-conditioned for approx. 7 weeks at a Florida Keys reef site (Wonderland Reef) and then maintained in flow through seawater at the SMS facility for approx. 3 months prior to the study. Settlement chambers were placed in mesocosms parallel to the flow of the seawater. Each mesocosm held up to four chambers, one containing each of the 3 species of coral larvae (and two for the P. strigosa larvae) ( Table 2 ).

Three days after the chambers were placed into the mesocosms, they were removed, and the tops, bottoms, and sides of the tiles were inspected for settled larvae under dissecting microscopes. The following categories of settlement were reported: (a) settled on the tile, (b) settled on the chamber or mesh, (c) swimmers, and (d) floating metamorphosed. Mortality was assumed for missing larvae, and sometimes decomposing larvae were observed during these counts indicating mortality was occurring. For purposes of analysis, categories a-d were considered surviving larvae, and categories a-b were considered settled larvae.

After the initial 3-day count, settlement tiles were removed from the chambers and returned to mesocosm tanks and maintained under the same experimental (or control) oxygen treatments for 42 more days (total of 45 days). Tiles were placed onto rigid plastic mesh with 1.5 cm openings in the tanks to minimize damage to corals settled on the underside of the tile. Corals were not fed during the experiment. Tiles were removed from the tanks for counts of surviving recruits after a further 42 days of exposure to the different oxygen treatments. Surviving coral recruits were counted on tops, bottoms, and sides of the tiles under dissecting microscopes.

All analyses were conducted in R Programming Language with RStudio version 4.2.2 (R Core Team, 2021) using the packages Tidyverse (Wickham, 2019) and Reshape2 (Wickham, 2007) for data wrangling, and ggplot2 (Wickham, 2016) and ggpubr for data visualization (Kassambara, 2023). The survival of larvae and settlement rate (on day 3), and recruit survival (on day 45) were measured as a proportion of the original number of swimming larvae placed into each chamber (n=50). Rates of larval survival, larval settlement, and recruit survival were all arcsine transformed prior to the analysis to support the use of parametric tests. Assumptions of normality were assessed using histograms, QQ plots and Shapiro-Wilkes tests, and variance was found to be equal between variables based on the results of Bartlett tests. Two-way ANOVAs were used to identify significant influence of species, deoxygenation treatment, and/or the interaction of these two factors on dependent variables. The data were subset and a priori one-way ANOVAs were conducted on individual species to test the effect of dissolved oxygen treatment for each species separately. Post-hoc tests were carried out with the emmeans (Lenth, 2016) and agricolae packages (Mendiburu and Yaseen, 2020) using Tukey HSD to determine pairwise significance when ANOVA p-values were ≤ 0.05.

Neither intermediate (3.5 mg L^-1) nor severe (1.5 mg L^-1) deoxygenation had a significant effect on larval survival ( Figure 2 ). Since coral larvae contain high lipid content and endogenous reserves to support many days or weeks of pre-settlement swimming, our findings imply that larvae are able to maintain sufficient metabolic rates such that oxygen availability at the concentrations we tested are not a strong factor for their survival. There were no prior studies quantifying the impact of deoxygenation on these species at the larval phase for comparison, but our results support previous findings with Acropora spp. that larval survival was not reduced in response to deoxygenation (Alderdice et al., 2022). Thermal stress has been shown to increase larval respiration in O. faveolata (Serrano et al., 2018), and there are no studies on respiration in C. natans larvae for comparison.

Similar to larval survival rates, we found that larval settlement rates were not significantly driven by deoxygenation conditions. This appears to run contrary to previous findings that coral larvae preferred to settle in water with a higher concentration of DO and actively avoided low-DO conditions for settlement (Jorissen and Nugues, 2021). It may be that when there is a choice, coral larvae prefer to settle in ambient DO concentrations but given only one option they will nevertheless settle under low DO conditions within a similar time frame.

Rates of larval survival and settlement were different between species, with the highest settlement and survival consistently observed in P. strigosa across all DO treatments. P. strigosa recruits have survived over longer timeframes than other spawning Caribbean coral species in recruitment experiments (Ritson-Williams et al., 2016), and it is one of the species predicted to increase in abundance as environmental conditions change (Edmunds, 2010). Our study identifies yet another stress (deoxygenation) to which they are more resilient, suggesting they are likely to play an increasingly important role on reefs of the future and should be a target of restoration activities. The second batch of P. strigosa had the highest rates of larval survival (>70% of larvae) and settlement (>50%) across treatments. The chambers used for P. strigosa 2 had a smaller gauge mesh than chambers used for P. strigosa 1 (100 µm compared to 180 µm, respectively), which could have impacted the results, however there was no evidence (no swimming or settled larval found outside of the mesh barrier) that larvae had escaped the larger mesh size used for P. strigosa 1. The differences between the two batches of P. strigosa were likely due to the difference in age of P. strigosa 2 compared to P. strigosa 1; the larvae of the two batches were produced at the same time and maintained under the same conditions, however there were 6 days between placing each batch into the settlement chambers, which may have influenced settlement behavior and sensitivity to environmental cues. Prior work has revealed increased mortality at early and late stages of the larval phase (Graham et al., 2008), and our results suggest likewise that coral larvae are more sensitive to hypoxic stress earlier in their development. This is also further evidence that depletion of energy reserves is not the primary factor limiting tolerance to deoxygenation as older corals with lower remaining reserves would be expected to be more sensitive to deoxygenation if that were the case.

Survival of C. natans and O. faveolata recruits exposed to 45 days of severe deoxygenation treatment (1.5 mg L^-1) was reduced in comparison to control and intermediate deoxygenation, indicating higher sensitivity to deoxygenation in these species than P. strigosa, which exhibited no difference in survivorship between treatments. Since this sensitivity was not observable at the larval stage, it is possible that the acquisition of symbionts impacted tolerance to deoxygenation. The larvae used in this study were asymbiotic, therefore their energetic needs were met entirely by maternal reserves and through respiration until symbionts were acquired after recruitment had taken place. Recruits with symbionts likely acquired their symbionts from conditioned tiles since no adult corals were placed in the mesocosms. Since respiration relies upon the availability of O₂, it is possible that C. natans and O. faveolata recruits have higher respiratory needs than P. strigosa, which did not exhibit any reduction in settlement nor recruit survival in response to deoxygenation. In the current study, almost all O. faveolata recruits (97.5 ± 2.5% of original larvae) had died after 45 days of exposure to deoxygenation. This aligns with previous findings that respiration rates in O. faveolata colonies were reduced in response to severe deoxygenation (Gravinese et al., 2022). One potential mechanism driving variable hypoxia response could be the micro-scale dynamics of oxygen exchange, which take place inside the DBL. The speed of molecular oxygen transfer is controlled by coral surface area, DBL thickness, and water flow rate (Shashar et al., 1993; Shashar et al., 1996; Mass et al., 2011). Metabolism increases with flow rate (Patterson et al., 1991; Mass et al., 2010), and it is possible that reduced water velocity inside settlement chambers may have slowed oxygen transfer for respiratory processes of asymbiotic larvae in this study. However, measuring water flow inside the settlement chambers was not possible within the constraints of our study and instrument capacity. Future research should target such knowledge gaps to better understand the micro-scale fluctuations in DBL oxygen dynamics and their role in hypoxia response.

In some cases, apparent increases in the number of recruits between initial settlement and 42-days post-settlement were observed, likely due to difficulty with finding the smallest recruits on tiles immediately after settlement. After 42 days of development, the coral recruits were slightly larger and had acquired symbionts based on observations of the color of the recruits, which made them more visible. Additionally, crevices are one of the preferred settlement micro-habitats for coral larvae, and recruits settled within a crevice on the tile may have been missed on the first count on day 3 but became visible as tentacles were extended by day 45. For this reason, we did not analyze change over time for given replicates, instead we compared differences among species and treatments in larval survival, settlement, and recruit survivorship for a given time point. We assume that the rate of error was the same within each species, although we interpret the data with the caveat that our findings may indicate an underestimation of initial settlement rates, and therefore the real rate of reduction in recruits over time.

Our study complements prior research on coral tolerance to deoxygenation by providing additional life history stages to species that were previously examined. The reduced survival of 45-day-old O. faveolata recruits exposed to 1.5 mg L^-1 DO (2.5 ± 2.5%) in comparison with control and intermediate deoxygenation treatments (22.5 ± 4.5 and 26 ± 7.55%, respectively) contributes to previous results investigating the impact of low-DO on adult O. faveolata corals. In Johnson et al. (2021b), adult colonies of O. faveolata were tolerant to severe deoxygenation over 11 days with no signs of bleaching or mortality, indicating that exposure time may play a critical role in hypoxia response, and that this may also vary at different life history stages within coral species. Since intermediate deoxygenation (3.5 mg L^-1 DO) did not significantly impact recruit survival in our study, we infer that the threshold for C. natans and O. faveolata at the recruit stage lies somewhere between intermediate and severe deoxygenation categories (1.5 and 3.5 mg L^-1 DO). More research is needed to explore the interaction between coral life history and hypoxia response.

As mass coral bleaching and mortality events have increased over recent decades, the importance of recruitment is critical to coral population recovery and resilience (Hughes et al., 2019). The current study is one of the first to look at this critical demographic bottleneck with both larval survival and post-settlement early recruit survival under low-DO conditions. Our results indicate that studies looking at larval survival and settlement over limited timescales of hours to days may underestimate the effects of longer-term exposure deoxygenation events (for example, 45 days in the current study) on the survival of coral recruits, and so longer-term studies are needed to better understand the potential for deoxygenation to impede ecosystem recovery (Adelson et al., 2022). With the current focus on enhancing and upscaling reef restoration efforts and increasing the yield of assisted fertilization efforts for endangered coral species, there is a critical need to improve understanding of the oxygen requirements for different species of coral larvae and early juvenile recruits. Rearing of larvae and recruits for reef restoration currently lacks information on the optimum DO conditions for settlement and survivorship, and the findings of such studies on deoxygenation will be important to guide restoration protocols. For example, our study would suggest that juvenile coral rearing tanks should be designed to meet species-specific oxygen requirements to reduce the risk of recruit mortality.