The effects of a variable temperature regime on the physiology of the reef-building coral Seriatopora hystrix: results from a laboratory-based reciprocal transplant

The effects of global climate change on coral reef ecosystems are becoming increasingly apparent (Hoegh-Guldberg, 1999) and have not only ecological (Hoegh-Guldberg et al., 2007), but also socio-economic implications (Hughes et al., 2003). As such, there is an urgent need to better understand the thermal biology of reef-building corals (Jokiel and Coles, 1990; Gates and Edmunds, 1999; van Oppen and Gates, 2006; Edmunds and Gates, 2008), as well as more fundamental components of their cellular (Chen et al., 2012) and sub-cellular behavior (Mayfield and Gates, 2007; Peng et al., 2011). Unfortunately, traditional temperature manipulation studies of corals are oftentimes characterized by shortcomings that prevent us from making strong claims about coral responses to elevated temperatures under environmentally relevant conditions. Importantly, experiments have typically challenged corals with a single elevated temperature for brief periods in an attempt to simulate an acute thermal stress event (e.g. DeSalvo et al., 2008; but see Dove, 2004; Putnam and Edmunds, 2009; Putnam et al., 2010; Oliver and Palumbi, 2011; Putnam and Edmunds, 2011). However, many coral reefs experience large variations in temperature (Craig et al., 2001; Leichter et al., 2005; Leichter et al., 2006; Sheppard, 2009), and some even undergo dramatic temperature changes over a time scale of minutes to hours. For instance, coral reefs of Nanwan Bay, the southernmost embayment of Taiwan, experience episodic upwelling (Lee et al., 1999) whereby the temperature may change up to 10°C over the course of a summer day during spring tides (Jan and Chen, 2008).

As thermal heterogeneity can be high within the seawater surrounding certain reefs, the use of static temperature manipulations alone may minimize the ability to interpret the physiological response of corals inhabiting such thermally variable regions. Reciprocal transplants in situ (e.g. Fan and Dai, 1999; Smith et al., 2007; Smith et al., 2008; Barshis et al., 2010; Bongaerts et al., 2011), in contrast, stand to greatly enhance our understanding of the phenotypic plasticity of corals (Coles and Brown, 2003), as they inherently avoid potentially unrealistic thermal regimes. In fact, thermal history has repeatedly been shown to influence the physiological response and acclimation capacity of reef corals (Coles, 1975; Coles and Jokiel, 1978; Warner et al., 1996; D'Croz and Mate, 2004; Castillo and Helmuth, 2005; Middlebrook et al., 2008; Howells et al., 2011; Oliver and Palumbi, 2011; Carilli et al., 2012; Guest et al., 2012), further promoting the utility of reciprocal transplants in developing our understanding of the coral response to changing temperatures. However, field-based studies are susceptible to impact by unexpected environmental variability, and results must always be interpreted conservatively.

With this in mind, by conducting a laboratory-based reciprocal transplant (LBRT) study, we took advantage of the thermally unique and dynamic environments of southern Taiwan in order to gain insight into how a common reef-building scleractinian, Seriatopora hystrix Dana 1846, acclimates to changes in seawater temperature. Coral colonies were collected from both Houbihu, a reef within Nanwan Bay that experiences episodic summer upwelling (Putnam et al., 2010), and Houwan, a reef on the western side of the Hengchun Peninsula that does not experience this phenomenon, and specimens from each site were incubated at either stable (26°C) or variable (23–29°C over a 12 h period) temperature for 7 days. This laboratory-based approach was also utilized in place of a reciprocal transplant in situ because of the multitude of other abiotic parameters (e.g. nutrient and dissolved oxygen levels) that are affected by upwelling events in Taiwan (Chen et al., 2004); while the physiological response to upwelling is indeed a worthy avenue for future research, only the temperature changes associated with such oceanographic events were of interest in this work. Collectively, it was hypothesized that corals exposed to upwelling conditions in nature would perform better under a variable temperature regime in the laboratory in comparison to those that experience relatively more stable annual temperatures. In other words, in the absence of acclimation, transplanted corals were expected to be physiologically compromised relative to non-transplanted controls, as corals have repeatedly been shown to be particularly sensitive to changes in temperature (e.g. Hoegh-Guldberg and Smith, 1989; Gates, 1990; Fitt and Warner, 1995; Hoegh-Guldberg and Jones, 1999).

The effects of temperature changes on reef coral physiology, and particularly photosynthesis, are relatively well documented (Jones et al., 1998; Warner et al., 1999; Fitt et al., 2001; Smith et al., 2005; Venn et al., 2008; Weis, 2008). Exposure to dramatically elevated temperatures results in photoinhibition of the endosymbiotic dinoflagellates (genus Symbiodinium) within corals (Jones et al., 2000). In extreme circumstances, Symbiodinium may be lost from the coral tissues and/or display reductions in chlorophyll a (chl a) concentration, phenomena known as bleaching (Glynn, 1983). While significant levels of bleaching were not anticipated for transplanted corals of this study due to the use of temperatures below the locally reported bleaching threshold of S. hystrix (Hung et al., 1998; Mayfield et al., 2011), decreased Symbiodinium densities and chl a concentration were hypothesized to be displayed by transplanted corals. In addition, it was anticipated that corals exposed to altered thermal regimes would demonstrate a decrease in maximum dark-adapted quantum yield of photosystem II (PSII) (F_v/F_m), as seen in previous studies (Putnam and Edmunds, 2009; Putnam and Edmunds, 2011). Given both the coral dependence on Symbiodinium photosynthesis for organic carbon (Muscatine and Cernichiari, 1969; Muscatine et al., 1981) and the strong connection between Symbiodinium photosynthesis and coral calcification (Gattuso et al., 1999), a reduction in both Symbiodinium density and F_v/F_m in transplanted corals would be expected to lead to decreased growth in these same samples.

Photoinhibition initially manifests at the molecular level (Shapira et al., 1997), and it was hypothesized that Symbiodinium within transplanted corals would express decreased levels of two photosynthesis-targeted genes (PTGs) encoding proteins involved in photon capture and trafficking [photosystem I (psI, subunit III)] and the Calvin cycle [ribulose-1,5-bisphosphate carboxylase/oxygenase (rbcL)]. Although expression of Symbiodinium psI and rbcL have not been measured in reef corals, expression of the third PTG gene targeted herein, phosphoglycolate phosphatase (pgpase), which encodes a photorespiratory protein involved in removal of the photosynthetic by-product 2-phosphoglycolate (Husic and Tolbert, 1985; Mamedov et al., 2001), was previously documented to decrease in Symbiodinium within corals that had been exposed to elevated CO₂ partial pressure (Crawley et al., 2010), and the authors interpreted this to signify that those corals were photosynthetically compromised and hence unable to fix organic carbon at optimal rates. However, as photorespiration is also a mechanism of photochemical quenching (Heber et al., 1996), it is possible that corals may utilize this pathway as a mechanism to increase energy dissipation when experiencing stress-induced photoinhibition in order to avoid reactive oxygen species (ROS) production (Gorbunov et al., 2001), discussed in greater detail below. Therefore, unlike rbcL and psI, pgpase gene expression was hypothesized to increase in Symbiodinium in corals exposed to a temperature treatment (TT) different from that of their site of origin (SO).

Photoinhibition-derived ROS production has repeatedly been observed in corals exposed to elevated temperatures (e.g. Lesser, 1997; Downs et al., 2002), and it has been identified as a major factor eliciting the coral bleaching response (Franklin et al., 2004). As such, ascorbate peroxidase (APX1), which is a key enzyme required for the detoxification of ROS (Shigeoka et al., 2002) in algae and higher plants (Yoshimura et al., 2000), would potentially be induced in photosynthetically compromised Symbiodinium. Given the expectation for photoinhibition in transplanted corals, it was also hypothesized, then, that expression of the mRNA encoding APX1, apx1, would increase in transplanted corals. Finally, transplanted corals were also hypothesized to demonstrate elevated RNA/DNA and protein/DNA ratios, indicative of increased levels of gene and protein expression and turnover, respectively, which may be necessitated by the ROS (Lesser, 2006) and, more generally, cellular defense responses (Hochachka and Somero, 2002; Kültz, 2005). Collectively, it was hoped that by measuring both physiological and molecular parameters with methodologies that account for the endosymbiotic nature of the coral tissues (sensu Mayfield et al., 2011), a greater understanding of the coral response to altered temperature regimes would be obtained.

Five months prior to sampling (January 2010), HOBO Pendant data loggers (Onset, Pocasset, MA, USA) programmed to record temperature at hourly intervals were left at both Houwan (22°01′23.30″N, 120°41′18.29″E, non-upwelling site) and Houbihu (21°56′18.01″N, 120°44′45.54″E, upwelling site). The average temperature logged at Houwan between April and June 2010 (~26°C, Fig. 1A,B) was used as the ‘stable’ temperature treatment (Fig. 1C). Temperatures logged during spring-tide upwelling events occurring at Houbihu between May and June 2010 (Fig. 1D,E) were used to set the ‘variable’ temperature treatment used in the study (Fig. 1F), described below. Photosynthetically active radiation (PAR) was measured hourly on SCUBA with a cosine-corrected Li-Cor meter (LI-193 attached to a LI-1400 data logger via a 10 m cable; Li-Cor Biosciences, Lincoln, NE, USA) from 06:00 to 20:00 h (N=15 sampling times per day) at the approximate site of coral collection (7–8 m depth) on two non-overcast summer days prior to sampling and one non-overcast day following the sampling day at each of the two sites in order to estimate the average PAR level (μmol photons m⁻² s⁻¹) experienced by the S. hystrix colonies in situ. The mean (±s.d.) hourly values obtained across both sites (94.3±8.6 and 94.3±9.1 μmol photons m⁻² s⁻¹ for the non-upwelling and upwelling sites, respectively) were used to set the PAR level used in both the acclimation and experimental periods (~90–100 μmol photons m⁻² s⁻¹).

In May 2010, under Kenting National Park permit 0992900398, six S. hystrix colonies were removed on SCUBA from each site (7–8 m depth) and transported to Taiwan's National Museum of Marine Biology and Aquarium. The colonies were allowed to acclimate for 3 days at 26°C in an indoor, 50 kl, flow-through, sand-filtered seawater tank exposed to shaded natural light (mean hourly PAR=90±10 μmol photons m⁻² s⁻¹, maximum PAR=300–350 μmol photons m⁻² s⁻¹) under a clear ceiling. Twelve nubbins (~2 g nubbin⁻¹) were then created from each of the six colonies from each site with pliers for a total of 120 and 24 nubbins reserved for physiological and molecular analyses, respectively. The 72 nubbins from each site were strung on fishing line suspended ~10 cm below the surface, randomly allocated to a position within the acclimation tank and held at the conditions described above for 3 weeks (15 May–7 June) until tissue had covered the site of nubbin fracture.

After the 3-week acclimation period, the 72 nubbins from each site were randomly assigned to either a stable temperature (26°C) treatment (N=3 tanks, 12 nubbins per tank) or a variable one programmed to fluctuate between 23 and 29°C over a 12 h period (N=3 tanks, 12 nubbins per tank). In the latter treatment, temperature was held at 23°C for 5 h, increased to 29°C over the course of 1 h, and then incubated at 29°C for 5 h prior to returning to 23°C over the course of 1 h. Nubbins from each site were incubated in separate tanks to maintain experimental independence, and the temperature in each of the twelve 150-l tanks was controlled by commercial heaters and chillers (AquaTech, Kaohsiung, Taiwan) as described previously (Putnam et al., 2010). All 144 nubbins were suspended on fishing lines 10 cm below the surface to receive similar PAR levels (mean=99 μmol photons m⁻² s⁻¹) generated by metal halide lamps (150 W), and each tank was characterized by a seawater turnover rate of ~150 l day⁻¹. During the 7 day variable temperature exposure period, which was initiated at 19:00 h on 7 June 2010, HOBO Pendant loggers were used to record the temperature at 10 min intervals in each of the 12 tanks, and a Li-Cor meter was used to document PAR during the light portion of the 12 h:12 h light:dark cycle (lights on 06:00 h, lights off 18:00 h) each day.

Prior to experimentation, all coral nubbins to be used for physiological analyses (N=120) were buoyantly weighed (sensu Spencer-Davies, 1989) in a chamber under an XP105 DeltaRange balance (Mettler-Toledo, Columbus, OH, USA). After 7 days of exposure (15 June 2010) to either the stable or variable temperature regime, F_v/F_m measurements were made on 2 h dark-adapted nubbins with a Diving-PAM underwater fluorometer (Walz, Effeltrich, Germany) as described in another work (Putnam et al., 2010). Nubbins were then re-weighed to calculate growth (normalized to surface area; described below) over the course of the experiment and frozen at −20°C. Upon thawing, 10 nubbins from each tank were blasted with a high-pressure water gun attached to a SCUBA tank to remove tissue, which was decanted into centrifuge tubes and frozen at −20°C after removing aliquots for chl a analysis and cell counts.

The tissueless skeleton was dipped twice in molten (65°C) paraffin wax incubated in 200 ml vats within a Thermo-Shandon histological incubation chamber (Waltham, MA, USA) in order to calculate surface area (SA) as described previously (Stimson and Kinzie, 1991). The chl a concentration within 1 ml of tissue slurry was measured in a spectrophotometer after overnight extraction in nine volumes of acetone as in another work (Jeffrey and Humphrey, 1975), normalized to SA and reported as μg cm⁻². From another aliquot (100 μl) of tissue slurry, the Symbiodinium density of formalin-fixed [4% in 0.1 μm filtered seawater (FSW)] cells was calculated with light microscopy and a hemocytometer (10 counts per sample) as previously described (Fitt et al., 2009). Symbiodinium density was also normalized to SA and is reported as cells cm⁻². Areal chl a was divided by the Symbiodinium density of the same sample to yield pg chl a cell⁻¹, which was analyzed and compared across treatments, as described below.

After 7 days of treatment exposure, small (~50 mg) biopsies were fragmented from the two replicates from each tank reserved for molecular analyses, submerged in 500 μl TRI-Reagent (Ambion, Austin, TX, USA) and immediately frozen at −80°C. It should be reiterated that sampling for molecular analyses was conducted on separate nubbins from those used for physiological measurements. This is due to the fact that the buoyant weighing process can require 5–10 min, a period of time in which molecular parameters, such as gene expression, could change. As buoyant weighing occurs at the temperature of treatment, however, it is unlikely that the Symbiodinium density and chl a content would change over this short period, and so the same nubbins can be used for growth, Symbiodinium cell counts and chl a analysis.

RNA, DNA and protein were extracted from each of the 24 S. hystrix biopsies as in a prior work (Mayfield et al., 2011) with the following exceptions. All centrifugation steps for RNA and DNA purification were performed at 13,000 g instead of 12,000 g, and nucleic acid pellets were always washed twice with 70% ethanol. Proteins were sonicated gently during the room temperature incubations between spins, as this was shown to better solubilize them in the guanidinium/ethanol wash buffer. RNA, DNA and protein quantity and quality were assessed as described previously (Mayfield et al., 2011). For each sample, total RNA (μg) was divided by total DNA (μg) in order to calculate an RNA:DNA ratio, a proxy for total gene expression. Similarly, a protein:DNA ratio was calculated for each sample as a proxy for total protein expression.

From the DNA phase, Symbiodinium hsp70 genome copy proportions (GCPs) were calculated with real-time PCR as in a prior work (Mayfield et al., 2011) for normalization of Symbiodinium gene expression. Briefly, this strategy is necessary when conducting macromolecular expression analyses with endosymbiotic samples, whose host:Symbiodinium biomass ratio may differ between samples, over time or in response to experimental treatments (Mayfield et al., 2009; Mayfield et al., 2010; Mayfield et al., 2011).

For restriction fragment length polymorphism (RFLP) analysis, DNAs were amplified with PCR using the Symbiodinium 18s rDNA primers (‘ss3z–ss5z’), reagent concentrations and thermocycling conditions of another work (Rowan and Powers, 1991b). The PCR products were purified with the AxyPrep PCR cleanup kit (Axygen, Union City, CA, USA) according to the manufacturer's recommendations, except with a 10 min incubation at 60°C prior to final elution in 20 μl nuclease-free water. The PCR amplicons were then digested with TaqI and Sau3 AI (New England Biolabs, Ipswich, MA, USA) in separate 20 μl reactions as in a prior work (Yang, 2001), and 10 μl of the digested DNAs were electrophoresed on 1.5% Tris acetate EDTA agarose gels, post-stained in an ethidium bromide bath for 20 min and visualized at 610 nm on a Typhoon Trio Scanner (GE Healthcare, Waukesha, WI, USA). The digestion patterns were then compared with those of a prior work (Rowan and Powers, 1991a) to verify Symbiodinium identity. Sub-cladal Symbiodinium diversity (sensu LaJeunesse, 2002; LaJeunesse et al., 2010; Bellantuono et al., 2011) was not assessed, as the markers used to infer such diversity, particularly the internal transcribed spacer region 2 (ITS2), have issues associated with intragenomic variation (Stat et al., 2009; Stat et al., 2010) that preclude the assignment of a DNA sequence to a single Symbiodinium cell (Pochon et al., 2012).

RNA (200 ng) was reverse-transcribed (20 μl reactions) with 1× Solaris RNA spike (Thermo-Scientific) and the High-Capacity cDNA synthesis kit (Applied Biosystems, Foster City, CA, USA), according to the respective manufacturer's recommendations. Assuming equal RNA loading, the Solaris spike controls for differences in reverse transcription (RT) efficiency caused by factors such as enzymatic inhibitors co-purified with the RNA and thereby circumvents the need for a housekeeping gene (Mayfield et al., 2009). Spike-inoculated cDNAs (2 μl reaction⁻¹) were used as the template in real-time PCRs (20 μl reactions) with the primers found in Table 1, 1× (10 μl) EZ-TIME SYBR Green mastermix with ROX passive reference dye (Yeastern Biotech Co., Taipei, Taiwan), and, in the case of the pgpase gene only, 0.5× (5 μg reaction⁻¹) bovine serum albumin. Real-time PCRs were conducted in an Applied Biosystems 7500 real-time PCR machine, and triplicate reactions of each sample (N=24) were run alongside three serial dilutions of random cDNA samples used to estimate PCR efficiency of each primer set on each 96-well plate (sensu Bower et al., 2007).

After a 10 min incubation at 95°C, thermocycling was performed at 95°C for 15 s followed by 60 s at the respective annealing temperatures for each gene (Table 1). Cycle numbers (Table 1) also varied for each assay, which was always terminated with a melt curve analysis from 65 to 95°C in 10 s increments. Unlike the Symbiodinium gene assays, the number of Solaris RNA transcripts reverse transcribed was quantified in each of the 24 cDNA samples using a TaqMan probe-based assay with a proprietary mastermix (including primers) according to the manufacturer's recommendations, and expression of each gene was normalized to the RNA spike recovery as recommended by the manufacturer. Then, the spike-normalized gene expression was divided by the Symbiodinium GCP as discussed previously (Mayfield et al., 2011) to control for differential ratios of host:Symbiodinium nucleic acids within a complex mixture of biological material. In short, this standardization for RT efficiency and Symbiodinium nucleic acid quantity differences between samples is required to generate accurate gene expression data for endosymbiotic, reef-building corals (Mayfield et al., 2009).

All statistical analyses were conducted with JMP (version 5.0, SAS Institute, Cary, NC, USA), and the Shapiro–Wilk test and Levene's test were used to determine whether data sets were normally distributed and of equal variance, respectively (Quinn and Keough, 2002). When either or both conditions were not met, log or square root transformations were conducted prior to statistical tests, and in such cases the back-transformed means were presented in the corresponding figures. Temperature and light data were compared and tested with Student's t-tests, and tank effects of physical data were compared with one-way ANOVAs with tank nested within SO and TT. To compare the temperature profiles of the six stable tanks against the six variable temperature tanks, means and standard deviations calculated on each day (N=7) were compared using Student's t-tests and one-way nested ANOVAs to determine TT and tank effects, respectively. For the variable TT, temperature data were also partitioned into ‘low’ (23°C), ‘transition’ (23–29°C) and ‘high’ (29°C) groups for a more detailed assessment of SO and tank effects. Two-way nested ANOVAs were used to test for the effects of SO, TT, their interaction and tank nested within TT × SO on both the physiological and molecular parameters. After verifying the absence of a tank effect for each response variable (data not shown), the tank factor was dropped from the model and samples were pooled across tanks, resulting in an N of 30 and 6 for each site and treatment for the physiological and molecular parameters, respectively. Tukey's honestly significant difference (HSD) tests were used to compare individual means. All means are presented ±s.d. unless otherwise indicated.

Temperature was measured at each of the two sites from which the experimental corals were taken, Houwan (non-upwelling site, Fig. 1A,B) and Houbihu (upwelling site, Fig. 1D,E). The mean monthly temperature was similar at each site in 2010: 26.35±1.76°C (N=12) at Houbihu and 26.60±2.03°C at Houwan (Student's t-test of monthly means, t=0.32, P=0.75). Importantly, the mean temperature in the 3 months prior to collection (10 February–10 May 2010) was also similar between the two sites (t=0.91, P=0.37) and was 25.75±0.76 and 25.48±1.19°C at Houbihu and Houwan, respectively. In contrast, the temperature variation differed between the two sites (Student's t-test of monthly standard deviations, t=2.90, P<0.01); the mean monthly standard deviation over the course of the entire year was 0.99±0.32°C at Houbihu and 0.72±0.04°C (N=12) at Houwan. When looking at the mean monthly standard deviation in temperature in the 3 months prior to experimentation, there was also a significant difference (t=4.69, P=0.01), and the mean monthly standard deviation over this period was 0.93±0.06 and 0.73±0.03°C for Houbihu and Houwan, respectively. The mean monthly range over the course of the entire year was 6.33±2.03 and 3.19±0.61°C for Houbihu and Houwan, respectively, and this difference was also statistically significant (Student's t-test of monthly ranges, t=5.13, P<0.01). The mean monthly range in the 3 months prior to coral sampling was 6.15±1.44 and 3.46±0.27°C at Houbihu and Houwan, respectively, and this difference was statistically significant (t=3.90, P=0.02). The largest intra-month fluctuations in temperature at Houbihu were in June (8.4°C) and July (9.6°C), reflecting the periods in which upwelling events were most common. During spring-tide upwelling events at Houbihu in April and May 2010, the average temperature range was ~6°C (23–29°C) per day for up to seven consecutive days (Fig. 1D), and the variable TT used herein (Fig. 1F) aimed to mimic these events. Average diel PAR, based on hourly measurements made between 06:00 and 20:00 h at the approximate depth of coral colony collection (7–8 m) across 3 days, did not differ between sites (Wilcoxon rank-sum test, z=0.66, P=0.51).

The mean (±s.e.m.) daily temperature of the Houbihu and Houwan stable tanks (N=3) was 25.88±0.01 and 25.94±0.01°C, respectively, similar to the target of 26°C and the average temperature in the 3 months prior to coral collection and experimentation (~26°C) at each of these sites. Due to low variation between tanks, this small temperature difference was statistically significant (Student's t-test, t=7.82, P<0.01), and there was also a statistically significant tank effect (one-way ANOVA of tank nested within treatment, F=49.23, P<0.01), due predominantly to one tank of the Houwan stable temperature group having a significantly lower temperature than that of the other two tanks of that TT and SO (Tukey's HSD, P<0.05).

The variable temperature profile consisted of a 5 h incubation at 23°C, followed by an increase to 29°C over the course of 1 h. Then, samples were incubated at 29°C for 5 h before the temperature was reduced to 23°C over the course of 1 h (Fig. 1F). The mean temperatures of the Houbihu and Houwan variable temperature tanks were 26.40±0.05 and 26.26±0.05°C (N=1010 temperature measurements for each tank), respectively, and this difference was not statistically significant (Student's t-test, t=0.47, P=0.64). Furthermore, there were no tank effects (one-way ANOVA of tank nested within SO, F=0.61, P=0.73). However, when partitioning the recorded temperatures into three phases – ‘low’ (23°C), ‘transition’ (23–29 and 29–23°C) and ‘high’ (29°C) for each of the 7 days – statistically significant SO and tank effects were documented. First, the mean (±s.e.m.) low temperature was 23.27±0.02 and 23.11±0.02°C for the three tanks containing corals from Houbihu and Houwan, respectively, and this difference was statistically significant (Wilcoxon rank-sum test, z=3.92, P<0.01). Also, there was a statistically significant effect of tank nested within SO (F=537, P<0.01). Means of 26.66±0.04 and 26.51±0.04°C were recorded at the transition temperatures for the tanks containing corals from Houbihu and Houwan, respectively, and this difference was statistically significant (Student's t-test, t=2.47, P=0.02). There was also a statistically significant effect of tank nested within SO (F=3.80, P=0.01). Finally, the high incubation temperatures differed significantly between SO (Wilcoxon rank-sum test, z=2.91, P<0.01) and were 29.28±0.01 and 29.14±0.04°C for the Houbihu and Houwan tanks, respectively. There was a statistically significant effect of tank nested within SO (F=886, P<0.01).

The daily temperature standard deviations for the six stable and six variable tanks across the 7 days of experimentation were found to differ significantly between the two TTs (Wilcoxon rank-sum test, z=7.89, P<0.01), with means of 2.84±0.01 and 0.25±0.01°C for the variable and stable treatments, respectively. There was, however, a tank effect on the average diel standard deviation in temperature (one-way ANOVA of tank nested within treatment, F=28.40, P<0.01) due to one tank of the Houwan stable treatment demonstrating a diel temperature standard deviation that was approximately half that of the other two tanks of that SO and TT (Tukey's HSD, P<0.05). The mean (±s.e.m.) PAR of the 12 tanks (one measurement per tank across each of 7 days) during the light portion of the 12 h:12 h light:dark cycle was 99±1.81 μmol photons m⁻² s⁻¹, and PAR and did not vary significantly across tanks (one-way ANOVA of tank nested within TT × SO, F=0.24, P=0.86). Furthermore, the PAR was statistically similar (Wilcoxon rank-sum test, z=0.43, P=0.67) between the field sites (mean of both sites=94.3±4.8 μmol photons m⁻² s⁻¹) and the 12 tanks.

A series of physiological parameters were measured in each of the 120 nubbins dedicated for such analyses (Fig. 2). Despite small, though statistically significant, effects of tank on both mean and standard deviation of temperature, physiological and molecular data showed no such tank effects (data not shown), and so data were pooled across tanks, resulting in N=30 and 6 for physiological and molecular parameters, respectively. Growth (Fig. 2A) was not significantly affected by SO or TT but did vary in response to their interaction (Table 2); non-transplanted corals had higher growth rates, although there were no significant pairwise differences (Tukey's HSDs, P>0.05). However, Symbiodinium density (Fig. 2B) demonstrated a statistically significant SO response (Table 2), with ~15% higher endosymbiont densities in samples from the non-upwelling site, Houwan. Areal chl a (Fig. 2C) was responsive to both TT and the interaction of SO and TT (Table 2), and non-transplanted corals contained significantly higher (~20%) chl a cm⁻². Also, areal chl a was ~1.5-fold higher in corals from the upwelling site exposed to a variable temperature regime relative to samples from this site exposed to stable temperature (Fig. 2C). These changes also held when chl a was normalized per cell (Fig. 2C). Finally, both SO and TT significantly affected F_v/F_m (Fig. 2D, Table 2). Specifically, Symbiodinium from corals from the upwelling site exhibited ~1% higher F_v/F_m values than those of the non-upwelling site, and Symbiodinium from corals exposed to the variable temperature regime displayed ~1% higher values than those incubated at a stable temperature (Fig. 2D).

A variety of RNA, DNA and protein-based parameters were measured in each of two nubbins within each of the 12 tanks. From the DNA phase, a Symbiodinium hsp70 GCP (Fig. 3A) was calculated with real-time PCR and shown to be statistically similar across sites and treatments (Table 3). Similarly, the RNA/DNA ratio (Fig. 3B) was unresponsive to SO, TT or their interaction (Table 3). In contrast, the protein/DNA ratio (Fig. 3C) demonstrated a significant SO effect (Table 3) and was ~25% higher in samples of the non-upwelling site, Houwan (Fig. 4C). Finally, RFLP-based genotyping (supplementary material Fig. S1) revealed that all 24 samples possessed Symbiodiniumof clade C based on results from both TaqI and Sau3 AI digests of the 18s rDNA gene. Previous work in Southern Taiwan (R. Gates, unpublished data) has found that the same Symbiodinium ITS2 types (C1, C3 and C59) were associated with S. hystrix at both Houbihu and Houwan. While the intragenomic variation associated with the ITS2 marker does not allow for a current assessment of the degree to which Symbiodinium diversity may have driven the changes in physiological response documented herein, it is hoped that new molecular markers for Symbiodinium (Pochon et al., 2012) and functional work across a variety of Symbiodinium types (sensu Sampayo et al., 2007; Stat et al., 2008) will allow for future elucidation of this critically important topic in our field.

High-quality holobiont RNAs inoculated with exogenous spikes were reverse transcribed to cDNA and used as a template in real-time PCRs for Symbiodinium gene expression analyses. There was a statistically significant interaction effect on the recovery of the Solaris RNA spike (Table 3), suggesting that the RT reaction was not equally efficient across the 24 samples. After normalizing mRNA expression of the Symbiodinium candidate genes to both Solaris spike recovery and the Symbiodinium hsp70 GCP, it was found that rbcL expression (Fig. 4A) differed significantly in response to TT, with ~2-fold more transcripts in specimens of the variable TT (Fig. 4A). psI (subunit III) expression (Fig. 4B) also differed in response to TT, with ~1.9-fold more transcripts measured in samples of the variable TT. Unlike rbcL, psI expression also was significantly influenced by SO (Table 3), with 1.5-fold higher expression in Symbiodinium from corals from the upwelling site. TT also significantly affected expression of pgpase (Fig. 4C, Table 3), which was expressed at 1.6-fold-higher levels in corals exposed to variable temperature. However, none of the factors or their interaction had a significant effect on the expression of apx1 (Fig. 4D, Table 3).

Thermal history has been shown to influence the capacity of reef-building corals to acclimate to elevated temperatures (e.g. D'Croz and Mate, 2004; Castillo and Helmuth, 2005; Middlebrook et al., 2008; Edmunds, 2009), though this topic is still quite poorly understood, particularly from a molecular perspective. Previous studies on this topic have had few consistent conclusions, though some researchers found that prior exposure to stressful conditions could lead to enhanced ability to tolerate later stressors (Brown et al., 2002a; Brown et al., 2002b; Oliver and Palumbi, 2011; Thompson and van Woesik, 2009; Guest et al., 2012). These works led us to hypothesize that corals would show an interaction effect in response to the LBRT, whereby corals from an upwelling environment would demonstrate enhanced physiological performance upon exposure to a variable temperature regime, while corals from a more stable thermal environment would perform relatively better under the stable temperature regime. In fact, this observation was only supported by growth and chl a content, both of which were higher in non-transplanted controls (Fig. 2). Specifically, corals from the upwelling site possessed higher chl a content and exhibited higher growth rates when exposed to the fluctuating temperature regime, while those specimens from the non-upwelling site exhibited higher chl a content and growth rates at the stable temperature regime.

The authors of a similar study (Smith et al., 2007) conducted a reciprocal transplant in situ, whereby corals were transplanted between a low-energy backreef and a high-energy forereef, and observed faster linear extension in non-transplanted controls. As such, both the coral specimens of this prior work and those sampled herein demonstrated phenotypic plasticity, though they were ultimately well adapted to the environment from which they were initially sampled. However, as discussed in greater detail below, while growth is arguably the most easily interpretable measure of coral fitness (Hughes, 1987), data derived from cellular and sub-cellular responses are required to develop the physiological mechanisms by which the coral acclimation to altered temperature regimes documented herein occurred.

Had only growth and chl a content been measured, the overarching hypothesis proposed herein, as well as those proposed by researchers in coral biology (e.g. Brown et al., 2002b) and other fields (Cunningham and Read, 2003; Stillman, 2003; Somero, 2010), would have been substantiated; corals perform better at a familiar temperature regime. However, as a wider range of biological scales was examined herein, a closer look at the additional parameters measured (Table 4) reveals that, in fact, the physiological response of corals to variable temperature may be quite complex. For instance, certain traits, such as the protein/DNA ratio (Fig. 3C) and Symbiodinium density (Fig. 2B), showed adherence to SO and were unaffected by TT or the SO × TT interaction; specifically, corals from the non-upwelling site were shown to possess higher Symbiodinium densities and protein:DNA ratios at the termination of the experiment. Interestingly, another study (D'Croz and Mate, 2004) also documented higher Symbiodinium densities and total protein in corals sampled from a non-upwelling reef relative to those from an upwelling reef. It is likely that corals with higher Symbiodinium densities would have higher protein content, as the Symbiodinium occupy nearly the entire volume of the coral gastrodermal tissues (Chen et al., 2012) and are highly proteinaceous (Weston et al., 2012). In fact, our prior work (Mayfield et al., 2011) has shown that the protein:DNA ratio was variable over time in S. hystrix, and we suggested that such variation could indeed be due to differences in Symbiodinium density. Both the spatial differences in Symbiodinium density and holobiont protein documented herein and previously (D'Croz and Mate, 2004) and the temporal differences in these parameters documented in prior studies (Mayfield et al., 2010; Mayfield et al., 2011) highlight the importance of normalizing Symbiodinium macromolecular expression in a way that accommodates variation in the host:Symbiodinium biomass ratio.

It is unclear why corals from reefs characterized by decreased thermal heterogeneity would possess higher densities of Symbiodinium. It is possible that corals from upwelling reefs possess lower Symbiodinium densities due to the potential for Symbiodinium to generate ROS in response to rapidly increasing temperatures (Lesser, 1997). mRNA levels of apx1, whose respective protein degrades certain ROS species into less harmful intermediates, was detected at similar levels between SOs (Fig. 4D). Although this may indicate that ROS levels were similar between corals of the two sites, a direct analysis of the concentration of ROS species, as well as expression of additional molecules involved in ROS detoxification [e.g. catalase (Lesser and Shick, 1990)], is warranted to conclusively determine the role of ROS in driving Symbiodinium density differences. As such, it remains to be demonstrated whether relatively lower densities of Symbiodinium are characteristic of upwelling reefs and whether or not such an adaptation is due to a correlation between Symbiodinium density and ROS production under thermally variable temperatures. Finally, as a second explanation for these findings, Houwan is characterized by significantly higher nutrient levels than Houbihu (Liu et al., 2012), and these relatively higher nitrogen levels, in particular, may allow for a higher standing stock of Symbiodinium, which are thought to be nitrogen-limited in many corals and sea anemones (Wang and Douglas, 1998).

In addition to Symbiodinium density and the protein:DNA ratio, both F_v/F_m and Symbiodinium psI (subunit III) mRNA expression were also characterized by significant SO effects (Table 4), and were, specifically, higher in samples from the upwelling site, Houbihu. F_v/F_m is a common index used by coral biologists to infer the efficiency of the initial stages of electron transport in PSII within dark-adapted Symbiodinium (e.g. Warner et al., 1996). Although a statistically significant 1% increase in F_v/F_m was found in corals from the upwelling site relative to those of the non-upwelling site, it is unclear whether such a small change would have implications for coral photosynthesis and health. It is tempting, though, to speculate whether there is a biologically significant correlation between photosystem gene expression and F_v/F_m; it may be that cells with more efficient photosystems could accommodate higher electron loads and thus require higher levels of photosystem gene and protein expression.

F_v/F_m and psI mRNA expression were also affected by TT, and, in addition to expression of pgpase and rbcL, were found to be at higher levels in corals exposed to a temperature regime that varied from 23 to 29°C over a 12 h period. In other words, both F_v/F_m and expression of the three PTGs were higher in samples exposed to a variable temperature regime for 7 days. Interestingly, our prior work (Putnam et al., 2010) also documented higher F_v/F_m in samples exposed to fluctuating temperatures in comparison to those incubated at stable temperatures, though in that study, only corals from the upwelling site, Houbihu, were used, and so this result was not unexpected. However, we also documented negative effects on coral physiology when the fluctuating treatment encompassed temperatures of 30–32°C (Putnam and Edmunds, 2009; Putnam and Edmunds, 2011), which have long been known to elicit the coral thermal stress response (Brown, 1997; Fitt et al., 2001). As temperatures used in the present study were below levels known to evoke stress in this species (Hung et al., 1998; Mayfield et al., 2011), it does not seem as if the coral specimens exposed to the variable temperature regime were physiologically compromised. In fact, an argument will be made below that they had an elevated photosynthetic capacity based on a variety of indices.

Unlike psI gene expression and F_v/F_m, which varied in response to both SO and TT, but not their interaction, the rbcL and pgpase genes appeared to be more strongly influenced by temperature directly and not the thermal history of the corals. Symbiodinium rbcL has been molecularly (Rowan et al., 1996) and biochemically characterized (Lilley et al., 2010), and a recent study found RBCL protein expression to be decreased in response to elevated temperature (Doo et al., 2012); although the latter was conducted with a foram–diatom symbiosis, it is possible that the strong temperature effect on rbcL/RBCL expression is a ubiquitous phenomenon of marine microalgae (Leggat et al., 2004). Importantly, these substantial changes in both rbcL and pgpase gene expression between temperature regimes, regardless of SO, point to a capacity for plasticity at the molecular level in this widely distributed Indo-Pacific reef coral that could have implications for carbon fixation.

The fact that psI, rbcL and pgpase gene expression were relatively elevated in samples of the variable TT suggests that an increase in carbon fixation could indeed have occurred in these samples. First, elevated psI expression could suggest an enhanced capacity for electron transport and capture (Varotto et al., 2000), which would increase the number of ATP and NADPH molecules available for Calvin cycle enzymes. Such a circumstance might have driven the relative increase in rbcL expression in these same samples. rbcL encodes the rate-limiting enzyme of the Calvin cycle, RBCL, a protein known to be particularly inefficient (Whitney and Yellowlees, 1995). As such, increased levels of RBCL expression could allow for carbon fixation to occur at a faster rate and increase the potential for autotrophy in the coral holobiont. Secondly, increased expression of pgpase/PGPase could be necessary when both the light and dark reactions are occurring at high rates due to the enzyme's role in metabolizing negative regulators of the carbon fixation pathways (Kaplan et al., 1991; Suzuki, 1995). Thus, in contrast to the hypothesis that elevated pgpase expression would be indicative of photorespiration stemming from thermal-stress-derived photoinhibition, it is now argued that its upregulation instead could lead to higher capacity for Calvin cycle enzyme function. Ultimately, to determine whether such increases in Symbiodinium PTG expression actually lead to increases in carbon fixation, as has been documented in higher plants (Mayfield et al., 1995; Murchie et al., 2005) and phytoplankton (Paul and Pichard, 1998), and presumed to occur in corals (Crawley et al., 2010), future studies should simultaneously measure both PTG expression as was conducted herein and the degree of Symbiodinium carbon fixation and translocation to the host with radiolabeling-based approaches (sensu Furla et al., 2000; Cantin et al., 2009).

Numerous studies have attempted to use molecular tools to document the coral response to environmental changes (e.g. DeSalvo et al., 2008; Bay et al., 2009; Császár et al., 2009; Reyes-Bermudez et al., 2009; Voolstra et al., 2009; DeSalvo et al., 2010; Portune et al., 2010; Starcevic et al., 2010; Hoogenboom et al., 2011; Kenkel et al., 2011; Leggat et al., 2011; Levy et al., 2011), generating a wealth of potentially interesting findings with regard to the molecular capacity of corals to acclimate to changes they may face over the coming century. However, only one prior study has documented the sub-cellular response of corals to environmental changes with molecular methods that accommodate the endosymbiotic nature of coral tissues (Mayfield et al., 2011). In fact, in order to generate biologically meaningful macromolecular expression data for endosymbiotic organisms, such as reef-building corals, it is essential to employ not only commonplace methodological controls, such as exogenous spikes to ensure that RT efficiency, for instance, is similar between samples, but also biological composition controls (i.e. Symbiodinium GCPs) to ensure that differences in the host:Symbiodinium biomass ratio do not bias gene expression data. For instance, decreased levels of Symbiodinium gene expression may be measured in a bleached coral relative to a healthy coral if appropriate biological composition controls, such as those utilized herein, are not taken, simply because of the former possessing lower densities of Symbiodinium relative to the latter. Therefore, we further advocate the use of biological composition controls for all works seeking to document macromolecular expression in endosymbiotic organisms.

This represents the first reciprocal transplant study of corals that employs techniques that allow for the generation of biologically meaningful data with respect to the molecular biology of these environmentally sensitive organisms, and it is hoped that future studies will utilize similar approaches to investigate the sub-cellular changes of reef corals exposed to, for instance, global climate change simulations. We used real-time PCR-based gene expression, as well as other physiological and molecular approaches hypothesized to gauge the coral physiological response, and the ensuing data set suggests that corals may indeed perform well when incubated under fluctuating temperatures, possibly via an enhanced capacity for photosynthesis and carbon fixation. Ultimately, though, corals grew faster at conditions consistent with their thermal history (Podrabsky and Somero, 2004), as hypothesized, and growth is arguably the most indicative sign of organismal fitness. Collectively, then, it appears that both environmental history and temperature regime are important for predicting the physiological response of reef corals, and future transcriptome (cDNA) sequencing work of the samples used herein with Illumina Tru-Seq (San Diego, CA, USA) technology and a Genome Analyzer IIx will help to better uncover the molecular mechanisms through which such acclimation occurred.

Steve Doo is acknowledged for his assistance with sample processing and comments on earlier versions of the manuscript. Dr Peter Edmunds is also thanked for valuable discussions on experimental design. Dr Yi-Yuong Hsiao is graciously thanked for sharing of laboratory space, within which the majority of the molecular analyses were conducted. Finally, we thank Dr Ruth D. Gates for her contribution of unpublished Symbiodinium nucleic acid sequence data. This is Hawaii Institute of Marine Biology contribution no. 1524 and the School of Ocean and Earth Science and Technology (SOEST) manuscript no. 8775.

 
• apx1/APX1

  ascorbate peroxidase

 
• chl a

  chlorophyll a

 
• FSW

  filtered seawater

 
• F_v/F_m

  maximum dark-adapted quantum yield

 
• GCP

  genome copy proportion

 
• HSD

  honestly significant difference

 
• hsp70/HSP70

  heat shock protein-70

 
• ITS2

  internal transcribed spacer region 2

 
• LBRT

  laboratory-based reciprocal transplant

 
• PAR

  photosynthetically active radiation

 
• pgpase/PGPASE

  phosphoglycolate phosphatase

 
• psI/PSI

  photosystem I

 
• PSII

  photosystem II

 
• PTG

  photosynthesis-targeted gene

 
• rbcL/RBCL

  ribulose-1,5-bisphosphate carboxylase/oxygenase large subunit

 
• RFLP

  restriction fragment length polymorphism

 
• ROS

  reactive oxygen species

 
• RT

  reverse transcription

 
• SA

  surface area

 
• SO

  site of origin

 
• TT

  temperature treatment