To evaluate the effects of ST and LI seasonal variation on FPs and PO, we used genets of A. cervicornis that we have been nursing in farming lines (donor farms) at water depths of 8 and 12m since 2016 at Punta Soldado, Culebra Puerto Rico. Five months before the start of the experiment, we selected 20 healthy-looking genets from the donor farming lines, ten genets per water depth. We divided them into three fragments (i.e., ramets), two of these between 15-20cm in length and the third one of 6cm ( Figure 1 ). Immediately afterward, the two large ramets were tagged with a unique number, and one of them was placed in a new farming line set at its original collection depth (hereafter shallow-shallow control [SS]) or deep-deep control [DD]; Figure 1 ). These two groups will be known as the control groups. The other large ramet was set from shallow to deep (shallow-deep [SD]) or deep to shallow (deep-shallow [DS], Figure 1 ). These fragments will be known as treatment groups. Farming lines where the ramets were set were constructed by hammering two-rod bars of 1.5m in length each, 0.5m into the sandy bottom, 2.5m apart, and connecting them with fish lines. The farms were visited monthly for the next five months to check for fragment bleaching and disease signs. Such signs were never observed during this period, suggesting that by the start of the experiment, all coral fragments were fully acclimated to the environmental conditions. The 6cm-long ramets were placed in numbered conical centrifuge tubes, frozen in liquid nitrogen, and transported to the laboratory, where they were stored in a -80°C freezer until further analyses. These 6cm fragments were subsequently analyzed as described below and understood to represent baseline levels of FP concentration and PO activity. Tissue fragments of 4-6cm in length from control (SS and DD) and treatment (SD and DS) colonies were subsequently collected on June 09, 2018 (Summer 2018), Nov 02, 2018 (Fall 2018), and May 18, 2019 (Spring 2019) following the same procedures as explained earlier for immune analyses. Farms were visited monthly for maintenance throughout the study, as described in Ruiz-Diaz et al. (2022).

Each frozen coral fragment was individually airbrushed to separate soft tissue from the skeleton using cold phosphate extraction buffer (50 mM phosphate buffer, pH 7.8 with 0.05 mM dithiothreitol) and separately collected into sterile tubes on ice. The crude extracts were then homogenized using a Fisher Scientific PowerGen 125 Tissue Homogenizer with a medium saw tooth tip for 60 seconds on the ice at low power. The homogenized tissue slurry was centrifuged at 4°C at 4,000 RPM for 5 minutes. The supernatant was aliquoted and used fresh or stored at -20°C. The coral fragments were processed using a modified Bradford assay to quantify the total protein concentration. To measure the total protein concentration, 5 μL of each extracted sample were plated in triplicate in 96-well plates along with a buffer-only control. Then 250 μL of Bradford reagent was added and left to incubate for 5 minutes at room temperature. Absorbance was recorded at 595 nm using a Scientific Multiskan GO spectrophotometer. A set of eight protein standards was prepared by serial dilution using bovine serum albumin at a 200 mg/mL concentration as a stock solution to prepare the calibration curve. The concentrations for each of the standard solutions were as indicated: 0.063, 0.125, 0.250, 0.500, 0.750, 1.000, 1.500, and 2.000 mg/mL.

Mean intrinsic protein fluorescence (RFU/mg of total proteins, hereafter mean intrinsic FP) in coral tissues was estimated through the spectral emission of these proteins following Palmer et al. (2009) with slight modifications. To measure the mean FP, 30 µL of the extracted samples was placed in a black 96-well plate, along with a control sample (buffer-only well). Triplicates of each sample were excited using an Infinite 200 Pro Multimode Microplate Reader at 280, 410, 470, and 520nm, and the emission spectra were measured from 450 nm to 750 nm at 5 nm intervals. Each sample’s relative fluorescence and absorbance were normalized to total protein concentration determined using Bradford Reagent and measuring absorbance at 595 nm, using Bovine Serum Albumin (BSA) to generate the mean fluorescence spectra per mg of total protein.

PO activity was quantified from fresh tissue extracts and processed as Palmer et al. (2011) described with slight modifications. PO activity was determined by adding 20 µL of the freshly extracted coral sample to a 96-well plate, in addition to 40 µL of 50 mM phosphate buffer, pH 7.8, and 25 µl of double-distilled water (ddH2O). After incubating at room temperature for 20 min, 30 µL of a 10mM dopamine hydrochloride solution was added as a substrate. The change in absorbance of the reaction was monitored every 5 min for 45 min using an Infinite 200 Pro Multimode Microplate Reader at 490 nm. The change in absorbance for the linear portion of the reaction curve was used to determine PO activity. Control reactions under the same condition, but with 70 µL of 50 mM phosphate buffer, pH 7.8, and no substrate were run in parallel as a control. PO activity was normalized to total protein concentration in each sample. The total protein concentration of the extracted coral sample was determined as described above.

Seawater temperature (ST) and light intensity (LI) were measured using Hobo Pendant temperature/light data logger 64k-UA-002-64 (Onset Company) attached to one of the rods of the experimental nursery lines at each depth. The devices were programmed to record one datum every hour for 30 -40 days, after which they were retrieved and substituted by new ones until the end of the study. For the statistical analysis, temperature records were averaged by summing all the temperature records between the collection of tissue divided by the number of days between the last and previous tissue collection. For LI, only the first ten days of records from each month between tissue collection were used, as algae and sediments covered the device sensor, potentially interfering with the detection of light after day 10 of the device being deployed. Therefore, LI records after the 10^th day were not reliable. These data were used to run the general linear mixed models.

To determine significant differences between ST and LI, between water depths (i.e., 8 and 12 m) two One-way repeated ANOVA were done with depth as the independent variable and ST, and SI measured at four distinct time intervals, (spring 2018, summer 2018, fall 2018 and spring 2019) as the dependent variable. Post-hoc analyses were performed to determine differences between periods.

Our statistical analysis also consists of repeated measures of FPs (including green and cyan FP) and PO activity taken in a seasonal-based fashion, i.e., June 09, 2018, Nov 02, 2018, and May 18, 2019. For these analyses, we used data from genets (control DD and SS) and ramets (experimental SD and DS) fragments of A. cervicornis, set at two depths, i.e., 8 & 12 m depths. Exploratory analyses of these data suggested a linear relationship between mean intrinsic FPs and PO activity across the monitoring periods, treatments (i.e., depths), ST, and LI. Therefore, to explore how these factors explain the observed variation in mean intrinsic FPs and PO activity across time, we developed a general linear mixed model (GLMM) with fixed and random effects described as follows. Fixed effects: (1) treatments with four levels, i.e., control shallow and deep and experimental shallow and deep groups; (2) ST, (3) LI, and (4) monitoring period with four levels, i.e., spring 2018, summer 2018, fall 2018 and spring 2019. The random effect only consisted of ramets fragments. To determine the effects of the natural variation (seasonal) of ST and LI on the immune-related proteins, we ran the model (one run per immune-related protein) considering only the results from control colonies, i.e., DD and SS colonies. We used data from control and experimental colonies to determine the effects of ST and LI between depths on the immune-related proteins. We performed three additional runs (one per immune-related protein), yet, in this case, we only used data from the summer of 2018 to the spring of 2019 since the experimental group had no data in the spring of 2018. The structure of the model was specified by backward stepwise analysis, excluding those non-significant effects. The final model components (i.e., most parsimonious)were selected by performing an Akaike Information Criterion (AIC) comparing nested models that only differed on the components of interest. These analyses involved choosing the more saturated model only if the reduction in the residual sum of squares was statistically significant compared to the simpler models.

ST at the shallow farms was statistically different from temperatures at the deep-water farms ( Figure 2 ; Table 1 ). However, the temperature at the two depths exhibited a similar temporal pattern of variation, increasing from spring to fall 2018, while decreasing from fall 2018 to spring 2019. Variations in temperature across time periods were also statistically significant ( Table 1 ).

Meanwhile, LI was significantly higher at the shallow farms than at the deep farms ( Figure 2 ). Likewise, LI also varied significantly across monitoring periods, yet in a slightly different manner than the temperature at both depths, decreasing from spring to summer 2018, reaching its highest values in fall 2018, and decreasing again in spring 2019.

Control groups (SS and DD) remained visually healthy throughout the study periods, yet detectable and varying levels of GFP and CyFP proteins and PO activity were observed in every colony during each monitoring period. However, levels of these immune-related proteins varied in magnitude across periods. For instance, GFP in both control groups steadily increased from spring to summer and further in fall 2018, while decreasing from fall 2018 to spring 2019 ( Figure 2 ). These differences among monitoring periods were statistically significant in both groups ( Table 2 ). Furthermore, seasonal fluctuations in ST and LI also showed a strong correspondence to the concentrations of GFP, for both control groups (F = 3.89; P = 0.013; Figure 2 ; Table 2 ). That is, increases in ST and LI from the spring to summer of 2018 corresponded to increases in GFP, with the highest peak of ST and LI mirroring the highest peak of GFP. Consistently, the observed decreases in ST and LI from Fall 2018 to Spring 2019 corresponded to a decrease in GFP detection. These effects were also statistically significant (ST: F = 7.531, P = 0.008; LI: F = 7.527, P = 0.008; Table 2 ) for both control groups.

A similar trend was observed when considering CyFP in the DD and SS colonies, with detection levels steadily increasing from spring to summer and then to fall 2018 while decreasing from fall 2018 to spring 2019 ( Figure 2 ). Differences in CyFP among monitoring periods were statistically significant (F = 3.048, P = 0.036; Figure 2 ; Table 2 ). Seasonal fluctuation of ST and LI also fluctuate across time in a similar way as CyFP fluctuates ( Figure 2 ; Table 2 ), with the highest peak of ST and LI coinciding with the highest of CyFP and lowest peaks of ST and LI also coinciding with the lowest peak of CyFP. ST and LI had a significant effect on CyFP in both experimental groups (ST: F = 5.995, P = 0.018; LI: F = 5.416, P = 0.024; Table 2 ).

The activity of the phenoloxidase enzyme (PO) exhibited a slightly different pattern of fluctuation than FPs in the control groups. PO activity from DD colonies increased from spring 2018 to summer and fall 2018 and then decreased from fall to spring 2019 ( Figure 3 ). However, PO activity in SS colonies decreased from spring 2018 and then increased from summer to fall 2018, and decreased from fall 2018 to spring 2019 ( Figure 3 ). These differences among monitoring periods were statistical differences for both control groups (F = 22.854, P =0.000; Table 2 ). Meanwhile, seasonal fluctuations of ST and LI also showed a similar fluctuation as the PO activity ( Table 2 ); increasing from spring to summer 2018, with the highest peak coinciding with the highest peak of ST and LI ( Figure 3 ), and as ST and LI decreased from Fall 2018 to Spring 2019, so does PO activity. ST and LI had a significant effect on PO activity in both control groups (ST: F = 5.098, P = 0.028, LI: F =4.474, P = 0.039; Table 2 ).

GFP concentration from DD colonies showed significant differences compared to SS colonies (F = 7.323, P = 0.009, Table 2 ). Specifically, DD colonies exhibited significantly higher GFP in the spring, and summer and SS exhibited significantly higher GFP in the fall of 2018 and spring of 2019 ( Figure 2A ).

A depth-related impact in GFP was also observed between the control and experimental groups. For instance, GFP was higher in DS colonies compared to DD colonies in the summer of 2018. By the fall of 2018, detection levels in DS increased threefold with respect to DD colonies. However, in spring 2019, levels of GFP in DD colonies almost tripled compared to DS colonies ( Figure 4 ). These differences were statistically significant ( Table 3 ). A similar detection pattern was also observed when comparing DD and SD colonies. However, differences in GFP detection levels show no statistical differences.

GFP levels between SS and SD/DS colonies exhibited a slightly different pattern. For instance, GFP levels from SD colonies double that of SS during the summer of 2018. However, by the fall of 2018, both groups increased. Nonetheless, SS doubled the concentration of SD. Whereas by spring 2019, both groups decreased in GFP levels. However, SD exhibited higher detection levels than SS colonies ( Figure 4 ). Differences between SS and SD were statistically significant ( Table 3 ). A similar detection pattern was also observed between SS and DS colonies, although differences in GFP levels show no statistical differences.

CyFP also varied between depths in the control and experimental colonies, consistent with previous analyses. For example, the data revealed statistical differences in CyFP between DD and SS colonies (F = 4.688, P = 0.034), with DD colonies exhibiting lower mean CyFP than SS colonies in all monitoring periods except from Spring 2019 ( Figure 4 ; Table 3 ). Meanwhile, CyFP in DS colonies almost doubled that of DD colonies in the summer of 2018. However, by the fall of 2018, CyFP from DD colonies nearly quadrupled compared to the previous monitoring period, whereas DS colonies decreased by half. By contrast, in spring 2019, CyFP concentration in DD colonies decreased by half, while DS colonies increased by half with respect to fall 2018. CyFP concentration between DD and DS colonies was statically significant (F = 2.98, P = 0.042; Table 3 ). A similar pattern was also observed between DD and SD colonies, although differences in CyFP concentration show no statistical differences. CyFP concentration between SS and SD/DS colonies behaved similarly. In the summer of 2018, SD colonies doubled the concentration of CyFP compared to SS colonies. By the fall of 2018, CyFP in both groups increased; however, CyFP concentration in SS colonies increased by order of five, while SD colonies increased by two. By spring 2019, both groups showed CyFP, similar to spring 2018. Differences across time between SS and SD were statistically significant (F = 2.98, P = 0.042; Table 3 ). Variation in CyFP between SS and DS showed similar increase/decrease patterns. Nonetheless, these variations were not statically significant.

Meanwhile, the effect of depth on the PO activity was less than clear, as no statistical differences were observed between control colonies, and between control and experimental colonies ( Figures 3 , 5A, B ; Tables 2 , 3 ). Instead, the observed fluctuations of PO activities were related to seasonality and WT and LI variation.