Heat tolerance and thermal preference of the copepod Tigriopus californicus are insensitive to ecologically relevant dissolved oxygen levels

Shifting climate patterns may impose novel combinations of abiotic conditions on animals, yet understanding of the present-day interactive effects of multiple stressors remains under-developed. We tested the oxygen and capacity limited thermal tolerance (OCLTT) hypothesis and quantified environmental preference of the copepod Tigriopus californicus, which inhabits rocky-shore splashpools where diel fluctuations of temperature and dissolved oxygen (DO) are substantial. Egg-mass bearing females were exposed to a 5 h heat ramp to peak temperatures of 34.1–38.0 °C crossed with each of four oxygen levels: 22, 30, 100 and 250% saturation (4.7–5.3, 5.3–6.4, 21.2–21.3, and 50.7–53.3 kPa). Survival decreased at higher temperatures but was independent of DO. The behavioral preference of females was quantified in seven combinations of gradients of both temperature (11–37 °C) and oxygen saturation (17–206% or 3.6–43.6 kPa). Females avoided high temperatures regardless of DO levels. This pattern was more pronounced when low DO coincided with high temperature. In uniform temperature treatments, the distribution shifted toward high DO levels, especially in uniform high temperature, confirming that Tigriopus can sense environmental pO2. These results question the ecological relevance of OCLTT for Tigriopus and raise the possibility of microhabitat selection being used within splashpool environments to avoid physiologically stressful combinations of conditions.


Results
Thermal tolerance. As expected, there was a significant main effect of temperature on survival of T. californicus (χ 2 1 = 51.20, P < 0.001). Specifically, survival decreased rapidly with increasing temperatures, from an average of approximately 90% at 34.1-34.9 °C to < 10% at 36.3 °C. No females survived at 37.2 and 38.0 °C (Fig. 1). Survival did not differ among low, normal and high DO levels (χ 2 1 = 0.72, P = 0.40). The interaction of temperature and DO also was not statistically significant (χ 2 1 = 0.69, P = 0.41). Thus, the temperature-induced mortality of T. californicus was independent of oxygen at the levels tested.
Thermal and oxygen preferences. Overall, females avoided high temperatures regardless of the oxygen levels in the water ( Fig. 2A-D). The distribution of females tended to shift more toward lower temperatures when the oxygen levels were high in this part of the chamber (Fig. 2B,C, pairwise comparisons between treatments 2, 3 and treatment 1, P = 0.075 and 0.051, respectively); this pattern was less clear when low DO occurred across the thermal gradient (Fig. 2D, pairwise comparisons of treatment 4 and treatments 1-3, all P-values < 0.05). Female distributions were skewed toward high oxygen levels when there was no thermal gradient (Fig. 2E,F, pairwise comparisons between treatments 5 or 6 and 1-4, all P-values < 0.001), and this pattern was stronger at 36 °C (Fig. 2F) than at 12 °C (Fig. 2E, pairwise comparison between treatment 5 and treatment 6, P < 0.001). When there were no temperature and oxygen gradients, females were distributed evenly in the chambers, and no preferred positions were observed (Fig. 2G, pairwise comparisons between treatment 7 and treatments 1-6, all P-values < 0.001). A full list of pairwise P values is provided in Table S1 in Supplementary information S2.

Discussion
There is extensive evidence that the splashpool copepods of the genus Tigriopus can tolerate extremely highamplitude fluctuations of environmental conditions 36,37,45,46 . Our results showed strong mortality at temperatures higher than 34.9 °C and only less than 10% of females survived after being exposed to 36.3 °C; no surviving females were observed after exposure to 37.2 and 38.0 °C. These results are comparable to published estimates of the thermal tolerance of T. californicus from the same climatic zone 37,38,47 . The thermal tolerance of T. californicus populations have been physiologically linked to the ATP synthesis capacity in the mitochondria; ATP synthesis declines at temperatures close to the knockdown temperatures 38 . The level of heat shock protein upregulation also has a positive correlation to thermal tolerance of T. californicus 41 . Suppression of a specific heat shock protein (HSPB1) has been shown to reduce the thermal tolerance of T. californicus to acute heat stress 17,48 . Importantly, the overall survival of T. californicus did not decrease at two low, but ecologically relevant DO conditions relative to normal DO (~ 100% saturation), regardless of peak exposure temperatures, nor did high DO level mitigate the effects of acute thermal stress. These results appear to contradict predictions of OCLTT 19,21 . A role for OCLTT, particularly during acute heat stress such as imposed in our experiments, is supported by evidence from a range of taxa in both aquatic and terrestrial ecosystems (reviewed in Table 1 in Ref. 21 ). However, it is likely that insensitivity of thermal tolerance to low DO levels occurs in species with high capacity to regulate oxygen intake and delivery (e.g., in the snail Planorbis planorbis) or when DO levels are not lower than the critical levels (e.g., P crit of 1.1-1.3 kPa for T. californicus 50 ). For example, the snail Planorbis planorbis has 2-4 times higher hemoglobin levels than its congeners 51 ; under hypoxic conditions the thermal tolerance of P. planorbis was not reduced, but the CTmax of P. carinatus was lowered by 1.2-2.1 °C 49 . Tigriopus species do not have gills, lack respiratory pigments 50 , and appear to have lost both the transcription factor HIF-1α and oxygen sensing prolyl hydroxylase repressor, EGLN, from their genome 43 . It has been suggested that a high surface-area-to-volume   50 . This may allow them to maintain their oxygen consumption rate independent of P O2 in the environment until the critical P O2 of 1.1-1.3 kPa 50 , which is approximately 4-5 times lower than the lowest P O2 of ca. 4.7-5.3 kPa in our experiment. Therefore, the insensitivity of thermal tolerance of T. californicus to low, but ecologically relevant DO levels observed in our experiment was in line with previous studies; reduced thermal tolerance of water-breathing species has been observed only at extremely low DO levels 4,26 . As noted above, in the splashpool system, DO levels below P crit of T. californicus are exceedingly unlikely to coincide with high temperatures. Therefore, the oxygen level may not be the limiting factor for thermal tolerance in ecologically relevant scenarios for Tigriopus and in similar cases. Instead, the capacity to physiologically cope with other consequences of temperature extremes may be the primary determinant of their thermal tolerance. Mortality under extreme temperatures may be the result of the dysfunction of a number of physiological processes such as the collapse of ATP synthesis 38 , membrane and protein structure instability 52,53 , and suppressed expression of heat shock protein (HSPB1) 17,48 . In light of our whole-organism survival data, we hypothesize that macromolecular disruption, rather than any direct effect of oxygen availability, explains recently published patterns of decreased ATP production capacity in Tigriopus mitochondria 38 , particularly because those in vitro experiments appear to have been run under normoxic conditions.
Irrespective of the mechanism(s) for the insensitivity of the thermal tolerance of T. californicus to ecologically relevant DO levels, our results join a growing literature suggesting that the OCLTT hypothesis may not be a universal principle for predicting the survivability of species in ecologically relevant conditions in nature 4,26 . Whether the exceptional environmental covariation of temperatures and DO levels in splashpools has selected for high thermal tolerance regardless of DO levels remains to be seen; it is equally plausible that selection for surviving periods of extreme night-time low DO may have coincidentally increased the ability to survive bouts of high temperature using anaerobic ATP production pathways. Recent work reveals that T. californicus can survive several days of anoxia 43 . It also is clear that the elevated DO levels that tend to naturally coincide with high temperatures in splashpools 30 do little to alleviate the effects of high temperature stress on these animals. Although beyond the focus of this study, it will be interesting to explore whether low DO levels may also alter critical thermal minimum (CT min ) of T. californicus and other species when low DO levels in their habitats occur during cold nights or the winter period 54 . In terrestrial ecosystems, the CT min values of some insects such as false codling moth Thaumatotibia leucotreta 55 , the beetle Tenebrio molitor 56 and crustaceans such as Porcellio scaber 56 are independent of oxygen availability.
Overall, female T. californicus demonstrate a strong avoidance of elevated temperatures, with oxygen playing a secondary role in influencing behavior. For example, their high distribution in the low DO region in treatment 1 ( Fig. 2A) was likely just to avoid the potentially lethal effects of extreme temperatures in the oxygen-rich water. The distribution of T. californicus was considerably more concentrated in oxygen-rich water only when low temperatures coincided with high oxygen (Fig. 2B,C) or there was no thermal gradient in the chambers (Fig. 2E,F). These results suggest that T. californicus can avoid low DO conditions.
Interestingly, the genome of T. californicus appears to lack prolyl hydroxylase and HIF-1α 43 , but our results clearly illustrate that these animals can sense the relative abundance of oxygen in the water. These results imply the existance of an alternative, extracellular (and perhaps superficial) mechanism(s) for T. californicus to sense pO 2 . Under uniformly extreme high temperatures, a shift in the distribution of females toward oxygen-rich water was even stronger; none were found in the low DO regions (Fig. 2F). This may be the result of higher basal metabolic demand, indicated by a general higher oxygen consumption rate at higher temperatures in a congener 50 . Finally, T. californicus did not show any preferred region within the chambers when they were in the control treatment at room temperature of 19 °C and the DO was maintained at 100% saturation. They also traveled a longer distance in this control treatment than in all other treatments, an indication that our observations of their distributions were consistent throughout the trial and uninfluenced by other confounding factors such as light.
Lastly, the behavioral results complement the physiological results to suggest a potentially novel explanation for how T. californicus can thrive in splashpools, where both temperature and DO are highly fluctuating and often extreme. Specifically, the results for thermal tolerance suggest that T. californicus may occasionally not be able to physiologically cope with extreme temperature. To survive in splashpools with extremely high temperatures during the day, there must be an alternative mechanism. Interestingly, field observations indicate that splashpools may be highly stratified over their small spatial scales of a few 10 s of centimeters or less (Fig. S1 in Supplementary information S1, Ref. 30 ). Indeed, our behavioral preference test showed that females avoided near-lethal temperatures even if by doing so they had to deal with lower oxygen levels at lower temperatures. This behavioral preference of T. californicus, which remains to be demonstrated in a natural setting, supports a recent prediction that behavioral responses of natural populations may enable them to exploit microclimatic variations in heterogeneous habitats as an important mechanism to rescue species from rapidly changing environments 17,18 .
Our study provides empirical evidence for the insensitivity of both heat tolerance and thermal preference of the splashpool copepod T. californicus to the ambient DO level. These patterns apply in ecologically relevant low and high DO conditions. It is likely that warming may be more stressful for T. californicus than low DO levels. Our results are among a small but growing collection of studies showing that the OCLTT hypothesis may not be a universal tool for predicting the thermal tolerance of species, particularly in ecologically relevant scenarios where environmental conditions fluctuate dynamically across a small scale of space and time. The strong behavioral preference of T. californicus suggests that together with physiological adjustments [39][40][41] , microhabitat selection might be used as an alternative mechanism for T. californicus to survive in highly fluctuating and often extreme conditions in splashpools 30  To create different DO levels in the PCR tubes during the heat ramp, we prefilled tubes with seawater adjusted to one of the desired levels. Dissolved oxygen levels were manipulated in a 20 L water bath using a custom-built, Arduino microcontroller system that regulated DO (while maintaining a constant pH of ~ 8.05) by coordinating the opening/closing of solenoid valves connected to oxygen, carbon dioxide, and nitrogen gas cylinders. The systems included calibrated temperature, DO (Honeywell DL5000), and pH (Honeywell Durafet III) sensors connected to a Honeywell UDA1282 Universal Dual Analyzer. The milliamp outputs of this analyzer provided feedback to the Arduino on the current conditions in a header tank; after comparing the current conditions to the desired setpoints, the Arduino triggered brief (10 s of milliseconds) pulsed openings of the solenoid valves to regulate gas flow. This cycle continued on a continuous loop, constantly monitoring and maintaining the DO level. Target DO levels (in mm Hg) were confirmed in the PCR tubes using a fiberoptic oxygen sensor (Neofox, Ocean Optics) prior to the tubes being capped and sealed with parafilm. To sustain these DO levels during the heat ramp, the entire thermocycler used for the thermal tolerance assay was housed in a sealed incubator, in which we manipulated the atmospheric oxygen levels in parallel with the desired DO in the seawater by pumping nitrogen or oxygen into the incubator. This arrangement was necessary because in preliminary trials all DO levels equilibrated with the atmosphere by the mid-way point of the 5-h heat ramp. Due to safety concerns around high-amperage electrical equipment, we could only increase oxygen levels in the incubator to 150% saturation. Using the incubator, DO levels within the PCR tubes were 4.7-5.3, 5.3-6.4, 21.2-21.3, and 50.7-53.3 kPa (22, 30, 100 and 250% saturation), respectively, at the start of the thermal tolerance assay. We confirmed in preliminary trials that these target DO levels in PCR tubes (each containing one female copepod) were maintained during the heating phase until reaching the peak temperature, but they drifted by the end of the assay. At the completion of the heat ramp, DO levels in the PCR tubes were 12.7-13.3, 14.0-14.4, 21.2-21.3, and 40-42.7 kPa (60-62, 66-67.5, 100, 188-200% saturation, respectively). Thus, although DO conditions did not remain constant for the entire duration of the heat ramp, they remained different from each other. Both low DO treatments (22.5 and 30% saturation) remained within the ecologically relevant low DO ranges found in splashpools during the heating phase 30 . Experimental assays were run using identical methods as these preliminary trials, but we did not measure DO levels in PCR tubes for experimental copepods.
For each heat ramp, PCR tubes were placed in an Eppendorf Mastercycer gradient thermocycler, which was custom-programmed to generate a gradual rise and fall of temperature over a five-hour period. The start temperature for the heat ramp was 20 °C. The thermocyler was programmed to a new setpoint every 10 min. Over the first 25 min all columns increased by 1 °C. Subsequent setpoints were programmed to increase 1 °C every 10 min to a preset peak temperature of 37 ± 3 °C at 185 min. Using the gradient feature, each column reached a unique peak temperature of 34.1, 34.4, 34.9, 35.6, 36.3, 37.2 or 38.0 °C. Upon reaching the peak, the temperature was maintained for 1 h. Following the 1-h exposure, temperature was decreased to 20 °C over the course of one hour. This protocol created a thermal profile more similar to the environmental temperature variation experienced by T. californicus in the wild 39 . Preliminary trials revealed that females exposed to low DO conditions and a peak temperature of 36 °C suffered 100% mortality. In order to reduce unnecessary use of animals, it was determined that low DO levels, 22.5 and 30% saturation, combined with peak temperatures of 37.2 and 38.0 °C would not be tested as part of this experiment. Therefore, a total of 40 females were tested at each of the low DO levels of 22% and 30%, and 56 females were tested at the normal and high levels. This heat-ramp procedure was repeated twice at each dissolved oxygen level. DO levels were randomized and only one was examined per day. The survival of females was checked immediately after the ramp and daily for the following 4 consecutive days. Mortality was determined when females were unresponsive to mild shaking of the vial, changed colors to bright red, and the urosome was bent sharply at a right angle to the cephalothorax 57 . Statistical analyses were conducted using survival at day 4.
Behavioral preference assay in the presence of temperature and oxygen gradients. To test whether DO may impact thermal preference of T. californicus, we determined the positions and distance traveled by egg-mass bearing females (n = 14-15 individuals per treatment) in each of 7 treatment conditions (Table 1) in preference chambers (Fig. 4). Preference chambers were in-house designed based on the system for Daph-Scientific Reports | (2020) 10:18885 | https://doi.org/10.1038/s41598-020-75635-z www.nature.com/scientificreports/ nia magna described in Zeis et al. 58 , with extensive modifications (Fig. 4). Specifically, 5 identical chambers were constructed of acrylic (L × W × H = 22.86 × 1.27 × 1.27 cm, volume = 36.87 ml). The chamber was sealed with a rubber gasket and acrylic lid. It was divided into 18 equal intervals, 1.27 cm each, by marking the outside of the chamber. Copepods could swim freely throughout the chamber without any physical barriers. There were inlets at each end, which connect to two different header tanks (size: L × W × H = 28.58 × 31.12 × 40 cm; volume = 20.8 L) where the water temperature and DO were controlled by two programmable, recirculating water baths and the N 2 and O 2 gas systems. The pH in the water was controlled at around 8.05 ± 0.05 by the CO 2 gas system. Gas flows were automatically controlled by an in-house Arduino microcontroller system as described for the thermal tolerance assays above. Each preference chamber has 4 outlet ports distributed equally along one side (the distance between two adjacent outlet ports is 4.57 cm); inlet and outlet ports were fitted with a 50 µm mesh screen to prevent copepods swimming out of the chamber. The outflow rates were 0.8-0.9 mL/min for outlets. This resulted in a total of roughly 3.25 mL water outflow per min, approximately 9% of the volume of the preference chamber. This flow rate is comparable to an assay for Daphnia magna 58 . Copepods swam freely throughout the preference chambers, therefore the flow did not impede the swimming or behavioral preferences of T. californicus. Two outlet ports from each side of the chamber were connected to one sump tank and two others were connected to a second sump tank. The sump tanks have the same size and volume as the two header tanks. The water in each of the sump tanks was pumped up to the corresponding header tank to create two closed and recirculating water systems, which could be regulated independently for temperature and DO. Prior to an experiment, the desired temperature (11 to 37 °C) and DO levels (0-200% saturation) were established in the two header tanks using recirculating water baths and the Arduino system, respectively (see Table 1). In the chambers, we confirmed the temperature and DO gradients at 6 positions; temperature fluctuated 34.5-35.1 °C at the high end; 28. 3-30.6, 25.5-27.1, 18.5-22.7, 15.7-18.4 °C at the four outlet ports; and 13-13.4 °C at the low end. Similarly, observed DO ranges from the high to low DO ends of the apparatus were 51.7-62.5, 26.4-53.3, 21.9-26.4, 9.2-14.7, 4.8-8.0 and 0.7-3.2 kPa, respectively. These are also ecologically relevant for the temperature and DO levels observed in splashpools occupied by Tigriopus (Fig. S1-Supplementary   Figure 4. Experimental system for the behavioral preference assay, in the presence of oxygen and/or temperature gradients. H1 and H2 are header tanks where both temperatures and dissolved oxygen can be independently controlled within the ranges of 1-100 °C and 0-200% saturation, respectively. WB1 and WB2 are water baths where temperatures were set up to control temperatures in the H1 and H2 tanks, respectively. Gases (O 2 , N 2 , and CO 2 ) were fed from cylinders to each header tank via airstones, and an Arduino-controlled system pulsed gas flows independently to each header tank through solenoid valves to maintain pH and pO 2 at desired levels. AS = Airstones and P = pump. Arrows indicate the directions of water flow.
Scientific Reports | (2020) 10:18885 | https://doi.org/10.1038/s41598-020-75635-z www.nature.com/scientificreports/ information S1, Ref. 30 ). For each experimental run, we randomly collected five egg-mass carrying females from the lab culture (temperature of 17-18 °C) and assigned one individual per preference chamber. In preliminary trials, we observed that copepods tended to modify their behavior in the presence of a con-specific, so our protocol isolated the effects of environment from any social factors. Females were allowed 10 min to explore the chambers before beginning data collection. Our preliminary observations showed that females swam freely within the first three to five minutes, and subsequently showed a more stable position in the chamber. The female positions in each chamber were observed once per min for 29 min (30 observations). All preference assays were conducted in diffuse light to avoid stressing the animals, and the orientation of the gradients relative to the room was randomly reversed for some chambers to avoid systematic influence of the surroundings on copepod behavior.
The distance traveled by a female during the observation period was calculated by summing the distances from one observation to the next. This cumulative measure of distance traveled may not be accurate for two reasons. First, it captures movements in increments of 1.27 cm; second, some females may move forward and back several times between consecutive observations. Nonetheless, this method provides a rough estimate of how active each female was within the observation chambers. The distance traveled was standardized to units of cm per minute, as not all females had all 30 observations due to the difficulty of observing them in the chambers under dim light. Specifically, 96/101 (95%) behavioral assays had 27-30 observations. The other five behavioral assays (5%), each had 10, 18, 21, 24 and 26 observations.

Statistical analyses.
To test for the effects of peak extreme temperatures, DO levels and their interaction on the thermal tolerance of T. californicus females, we ran a generalized linear model in R, using a binomial link function (0 = dead, 1 = alive). Temperature and DO were included as fixed factors. For thermal and oxygen preferences, the number of times that a female was observed in a specific position in the gradient (1-18 in the chamber) was counted to calculate the percentage of time in each position. The distribution of T. californicus in each position in the preference chambers is the main indicator of thermal and oxygen preferences. A Chi-square test was employed to test for pairwise differences in the distribution of females among treatments. For the distance traveled by females per minute, data were initially checked for normality using a Shapiro-Wilk test and the homogeneity of variance using Levene's test; both assumptions for ANOVA were not met (P < 0.05). Therefore, we used a non-parametric Kruskal-Wallis test to examine differences in the distance that copepods traveled per minute. P values < 0.05 are considered statistically significant. All analyses were run in R (v. 3