All corals were housed at Reef Systems Coral Farm (New Albany, Ohio). This facility consists of large ∼4,000 L raceways (8′ x 4′ x 4′) housed within a greenhouse facility (optically clear plastic roofing with no shade cloth). Smaller indoor tanks (∼1,100 L) contain additional corals under LED illumination. All tanks utilize the same artificial seawater (Reef Crystals in RO/DI filtered water) and receive frequent (10%–20%, 1–2 weeks⁻¹) water changes. All tanks contain submersible power heads (Tunze) which circulate water within each system between 50–100 times per hour. Additional overflow pumps exchange water (1–2 times volume of tank hr⁻¹) with external sumps equipped with mechanical (25-micron sieve) and foam fractionation filtration. All coral fragments are mounted on ceramic disks (3 cm) and attached using cyanoacrylate gel (coral glue). Peak irradiance (measured at the same tank depth for each tank) within the outdoor (greenhouse) tanks was 1,050 μmol m⁻² sec⁻¹ (walz, 4pi sensor). Indoor corals were illuminated using LED lighting (Radion Xr30 Pro) on a 10:14 h dark: light cycle with peak irradiance measured at 300 μmol m⁻² sec⁻¹. Indoor and outdoor corals thus differ both in max irradiance but also in light spectra (natural lighting vs. LED output).

Reef Systems Coral Farm (New Albany, Ohio) houses over 30 different species of coral, with thousands of individual fragments originating from coral colonies that were harvested in Fiji but have been captively grown at the facility for over 10 years. All individual fragments utilized in the study had been fragmented and mounted for at least 1 month prior to use. In order to maximize diversity, we collected three replicate fragments from twenty different coral colonies chosen to include a wide range of life strategies and histories, including both mound and branching species acclimated to greenhouse conditions; Psammacora contigua (1 genotype), A. yongei (1 genotype), A. millepora (1 genotype), A. humilis (1 genotype), A. valida (1 genotype), Acropora species (1 genotype), T. reniformis (2 genotypes), P. cactus (1 genotype), M. digitata (1 genotype), M. capricornis (3 genotypes), C. chalcidicum (1 genotype), P. damicornis (2 genotypes). In addition, indoor acclimated fragments representing species A. humilis (1 genotype), T. reniformis (1 genotype), and M. capricornis (2 genotypes) were also included in our analyses.

Prior to measurements, each individual coral fragment was dark acclimated for 20–25 min (Suggett et al., 2015). Photophysiological responses to changing light conditions were then measured using a prototype Chlorophyll-a fluorometer previously described in Hoadley et al., 2023. Briefly, fluorescence induction curves were produced through excitation with 1.3-μs single turnover flashlets spaced apart by 3.4-μs dark intervals (32 flashlets were utilized under 420, 442, and 458-nm excitation while 40 flashlets were utilized during 505 and 520-nm excitation). Each fluorescence induction curve was followed by a 300 ms relaxation phase consisting of 1.3-μs light flashes spaced apart with exponentially increasing dark periods (starting with 59-μs). Fluorescence induction and relaxation curves measured using each excitation wavelength were sequentially repeated five times per measurement timepoint. These photophysiological measurements were repeated 34 times during an 11-min variable actinic light protocol in which corals were exposed to an initial dark period (30s) followed by three different light intensities (300 μmol m⁻² sec⁻¹ for 3.5 min, then 50 μmol m⁻² sec⁻¹ for 1.5 min followed by 600 μmol m⁻² sec⁻¹ for 3.5 min), and a final dark recovery period (∼2 min) (Hoadley et al., 2023). Resulting fluorescence data was analyzed using custom R scripts (Hoadley et al., 2023) and according to equations set forth in Kolber et al., 1998 in order to calculate quantum yield of photosystem II (Φ_PSII), PSII functional absorption cross-section (σ_PSII), reaction center connectivity ρ), non-photochemical quenching (NPQ), antenna bed quenching (ABQ), photochemical quenching (qP), and two time constants for electron transport (τ₁ and τ₂) where τ₁ reflects acceptor-side changes of PSII and τ₂ reflects changes in plastoquinone pool reoxidation (Suggett et al., 2022; Hoadley et al., 2023). However, downstream PQ reoxidation kinetics (τ₂) are typically derived from a multiturnover induction and relaxation flash sequence (Suggett et al., 2022) whereas ours are derived from the second exponential component of a single turnover induction and relaxation flash sequence (τ₂ ^ST).

Once chlorophyll-a fluorescence-based measurements were complete, a portion of coral tissue was removed using the water-pick method (Johannes and Wiebe, 1970) and filtered seawater (artificial seawater vacuum-filtered, 0.2-micron filter). The resulting tissue-water slurry was homogenized using a hand-held tissue homogenizer (tissue tearer) and then measured using a graduated cylinder. Samples were then centrifuged (8,000 g, 10 min, room temperature). After centrifugation, the supernatant was removed, and the pellets were resuspended in sterile seawater, vortexed for 30 s and then centrifuged once more to wash the endosymbionts free of the host tissue. Resulting algal pellets were resuspended in 10 mL of sterile seawater.

One ml aliquots of resuspended algal samples were preserved with glutaraldehyde (0.1% final concentration), incubated in the dark for 20 min, flash frozen in LN₂, and then stored at −80°C for downstream flow cytometry. All samples were analyzed using a Attune NxT (Invitrogen, United States) equipped with a 488-nm (200-mW) laser and a 200-µm nozzle. Glutaraldehyde-fixed samples were re-pelleted using a centrifuge (12,000 g, 5 min), resuspended in 1x filtered PBS, and then filtered through a 50-micron mesh filter to remove residual coral tissue or large cell aggregates. Samples were then diluted 4x with 1x filtered PBS and spiked with 10-µL of 1:1,000 diluted fluorescent bead stock (Life Technologies 4.0-µm yellow-green 505/515 Fluospheres sulfate). 100-µL per culture was analyzed at a flow rate of 200-µL min⁻¹. For characterization of Symbiodiniaceae cells, data collection was triggered on forward light scatter (FSC), while red (695/40 nm bandpass filter) autofluorescence detected chlorophyll-a, and both were utilized to gate the cell and bead populations for bead-normalized FSC, side scatter (SSC) and chlorophyll fluorescence measurements.

For quantification of neutral lipid content, 500-µL of the diluted sample was first stained with 2-µL 5 mM Bodipy 493/503 (Invitrogen, 4,4-Difluoro-1,3,5,7,8-Pentamethyl-4-Bora-3a,4a-Diaza-s-Indacene) and then incubated in the dark at 37°C for 15-min. Stained samples were then run according to the same conditions described above. The Symbiodiniaceae cell population was identified using FSC-height and autofluorescence, and the gated population’s green fluorescence was quantified (bandpass filter 530/30).

For Carbon and Nitrogen analysis, 2-mL of each resuspended algal sample was filtered through a 13-mm ashed GF/F filter and dried in a 95°C oven for 24-h. Filters were packed into tin capsules, combusted (Costech Instruments 4010 Elemental Combustion System) and analyzed via Elemental Analyzer. Total carbon and nitrogen values were compared to an atropine standard. For particulate organic phosphorus (POP) analysis, 2-mL of each isolate was filtered through a 13-mm ashed GF/F filter and stabilized with a 0.17 M Na₂SO₄ rinse. Filters were placed in muffled scintillation vials with 2-mL aliquots of 0.017 M Na₂SO₄ and evaporated to dryness in a 95°C muffle oven for 24-h. Vials were covered in aluminum foil and baked at 450°C for 2-h, and then were baked with 5-mL 0.2 M HCl at 90°C for 30-min. Samples were diluted with 5-mL ultra-pure water and analyzed using the Skalar SAN + system and compared to an Adenosine Triphosphate standard.

To confirm that corals across different tanks were grown under similar nutrient conditions, 20-mL samples from each husbandry tank were collected and stored at −20°C until analysis, when they were thawed and loaded into the Skalar SAN + system’s autosampler. The samples were analyzed for µM nitrate, nitrite, ammonia, and phosphate via continuous flow analysis and according to EPA standard techniques (EPA, 1993a; EPA, 1993b).

A 2-mL aliquot of isolated symbiont cells were first pelleted (centrifugation at 10,000 rcf for 5 min) and then stored in 1-mL of DMSO buffer solution (Pettay et al., 2015) at 4°C. DNA was extracted from each sample using the Wizard Genomic DNA Purification Kit (Promega). Quality of DNA samples were assessed on a 1.0 Nanodrop (Thermo Scientific) and all samples had 260/280 and 260/230 values above 1.5. For each sample, PCR was first performed targeting the internal transcribed spacer 2 (ITS2) region using the previously established primer pairs (ITSintfor2: 5′GAATTGCAGA ACTCCGTG-3′, ITS2-reverse: 5′GGGATCCATA TGCTTAAGTT CAGCGGGT-3′) (Sheffield et al., 1989; LaJeunesse, 2002). Resulting amplicons were subjected to a second round of PCR (only eight cycles) using the same primer pairs with adapter sequence (Forward- TCG​TCG​GCA​GCG​TCA​GAT​GTG​TAT​AAG​AGA​CAG​GAA​TTG​CAG​AAC​TCC​GTG; Reverse-GTCTCGTGGGCTCGGAGATGTGTATAAGAGACAGGGGATCCATATGCTTAAGTTCAGCGGGT). Adapter sequences are underlined in the above primer sets. All PCR was achieved using Hot Start Taq DNA Polymerase (New England BioLabs, Inc.) under the following settings: denaturation 94.0°C for 30 s, annealing 54.0°C for 35 s, extension 68.0°C for 3 min x 35 cycles (Eight cycles in round two), final extension 68.0°C for 5-min. Samples were then purified using the GeneJET PCR Purification Kit (Thermo Scientific) and visualized on agar gels to confirm results. Amplicons were submitted to the University of Delaware Sequencing and Genotyping Center for library preparation and metagenomic sequencing. Amplicons were dual indexed using the Nextera system and were run as a single library using a paired-end 300 base pair x 2 strategy on an Illumina MiSeq system. Demultiplexed FASTQ sequences were then uploaded to SymPortal for profiling (Hume et al., 2019). Given known high variability in rDNA copy numbers across species and genera (LaJeunesse and Thornhill, 2011) which may hinder translation of data into accurate relative abundances (Davies et al., 2022), absolute abundance ITS2-type profiles were then normalized to rDNA copy numbers according to Saad et al., 2020.

All statistical analyses were conducted in R (version 4.1.3). Algal genotypes for each coral sample were derived from SymPortal-generated ITS2 profiles normalized to reflect relative abundance of each symbiont genera. Importantly, known differences in Cladocopium and Durusdinium ITS2 copy numbers was addressed by normalizing to previously derived rDNA copy ratios of 2119:362 (Saad et al., 2020). The dominant (>70% of ITS2 sequences) symbiont type in each coral fragment (Figure 2B) was then utilized to screen for photophysiological metrics that were significantly different across algal species using either a one-way ANOVA or Kruskal–Wallis test if data did not meet the assumptions of normality. The resulting dataset was then transformed using Z-scores and then plotted as a heatmap (Figure 2A) with individual coral colonies represented by each column, and photophysiological metrics across rows. A clustering analysis (1,000 bootstrap iterations) was performed using the R packages pvclust (Suzuki and Shimodaira, 2006) and dendextend (Galili, 2015) to cluster individual coral samples by algal phenotype. Clustering analyses were also carried out on heatmap rows and resulting clusters were further analyzed using custom scripts to identify which photophysiological metrics were most important for separating our coral fragments into separate light-response phenotypes.

Full actinic light profiles for identified photophysiological metrics (Figure 2A) were plotted in Figure 3 and a repeated measures linear mixed model with a tukey posthoc (with Bonferroni correction) identified significant differences across algal light-response phenotypes (Supplementary Table S4) using the lmerTest (Kuznetsova et al., 2017) and multicomp (Hothorn et al., 2008) R packages. Spectrally-dependent differences within each photophysiological metric and algal phenotype were similarly assessed (Supplementary Table S3). Significant differences in cellular physiology (cell size, Granularity, Chl a, N:P, C:P, and C:N ratios) across phenotypes was also assessed. For each metric, normality was first determined (Shapiro-Wilks). If data were determined to be normal, a One Way ANOVA followed by a Tukey posthoc was performed. For data that did not meet the assumptions of normality, a Kruskal Wallace with Bonferroni correction was performed (Figure 4).

A network analysis was employed to look for significant correlation between photophysiological metrics and primary cellular traits using averaged values across coral species replicates (three species⁻⁻¹). Only correlations with a Pearson value above 0.6 were utilized.

Using the igraph package (Csardi and Nepusz, 2006), specific correlations (Pearson’s value of 0.55 or above) between symbiont cellular metrics and fluorescence-based phenotyping data were identified and are displayed in Figure 5.

Symportal analysis revealed that most corals were dominated by a single genotype (> 70% relative abundance) of Symbiodiniaceae. Acropora (including A. yongei, A. millepora, A. humilis, A. valida, and unknown species) predominantly hosted either C3 or C21 symbiont types, while Montipora (including M. capricornis and M. digitata) hosted primarily C15, but sometimes C26, variants (Figure 2B). The symbiont type D1 was primarily observed in Turbinaria, Psammacora, and Pocillopora.

Of the 1,360 algal biometrics derived from the fluorescence-based excitation profile, 987 were found to be significantly (p < 0.05) different across the dominant symbiont species (C1, C3, C15, C21, C26, and D1, Table 1). These identified algal biometrics were then used to organize samples according to trait-based (light-response) phenotypes (Figure 2A), with the resulting dendrogram organized into four distinct phenotypes according to the largest clustering groups (Figure 2A). Light-response phenotype 1 contains 14 of the 15 C15-dominated coral colonies (Montipora digitata and Montipora capricornis), with both high and low-light acclimated fragments clustering together (Figure 2). Two additional fragments of Acropora sp. (C3) were also found in phenotype 1. Light-response phenotype 2 was predominantly (12 of 16 fragments) comprised of Durusdinium trenchii (D1)-dominated corals (low-light T. reniformis, Pocillopora damicormis, and Psammacora contigua). The remaining four coral fragments in phenotype 2, were dominated by Cladocopium C1 (in Cyphastrea chalcidicum or Pavona cactus). Coral fragments belonging to light-response phenotype 3 were comprised of 3 A. millepora fragments (C3) and 5 fragments of high-light acclimated Durusdinium D1-dominated Turbinaria renifomis. Lastly, phenotype 4 was comprised of both high and low-light acclimated A. humilis (C3) fragments, along with all C21 dominated corals (A. yongei and Acropora sp.), all three fragments of M. digitata dominated by C26, two fragments of P. cactus (C1) and single C15-dominated (M. digitata) and D1-dominated (Turbinaria reniformis—high light acclimated) fragments. Of the 20 coral colonies represented in this study, only four contained fragments which did not all cluster within the same phenotype. Based on row clustering and custom scripts, quantum yield of PSII (Φ_PSII), functional absorption cross section of PSII (σ_PSII), non-photochemical quenching (NPQ), photochemical quenching (qP), along with the reoxidation kinetics (τ₁ ^ST and τ₂ ^ST) were determined to be the primary drivers for the observed phenotypic structure and were thus plotted in full detail (Figure 3) for each of the four light-response phenotypes described above.

A mixed linear model was used to identify spectrally dependent differences across light-response phenotypes for the photo physiological metrics, Φ_PSII, σ_PSII, NPQ, qP, τ₁ ^ST, and τ₂ ^ST reoxidation kinetics (Figure 3 and Supplementary Table S4). For all excitation wavelengths except 420 nm, phenotype 1 had significantly (p < 0.0159) lower Φ_PSII values compared to all other phenotypes. Under 420 nm excitation, Φ_PSII profiles for phenotype 1 were significantly (p < 0.0001) lower than phenotypes 2 and 4, but not phenotype 3. Additionally, Φ_PSII profiles for phenotype 3 were also significantly (p < 0.001) lower than phenotype 2 but only under 420 and 442 nm excitation whereas phenotype 3 different significantly (p < 0.043) from phenotype 4 under all excitation wavelengths except 525 nm. No differences across light-response phenotypes were observed for σ_PSII under any excitation wavelength and indicate that subtle differences may only exist when comparing across specific symbiont species (not phenotypes). Under all excitation wavelengths, nonphotochemical quenching profiles reached the highest values in phenotype 1 whereas relatively small changes were observed in phenotype 2. For phenotypes 3 and 4, NPQ values did not differ significantly from one another but represent a medium level that is significantly (p < 0.001) different from phenotypes 1 and 2. Photochemical quenching (qP) profiles observed in phenotypes 1 and 3 differed significantly (p < 0.003) from those in phenotype 2 and 4 under all excitation wavelengths. Differences in τ₁ ^ST were more sporadic across phenotypes as profiles under 420 nm excitation differed significantly (p = 0.001) between phenotypes 1 and 2. Under 442 nm excitation, τ₁ ^ST profiles for phenotype 1 were significantly (p < 0.011) different from those observed for phenotype 2 and 3. For τ₁ ^ST profiles measured under 505 nm excitation, phenotype 1 was significantly (p < 0.021) different from all others, while phenotypes 3 and 4 were also differed (p < 0.027) from one another. Lastly, τ₂ ^ST profiles for phenotype 1 had significantly (p < 0.002) slower (higher time constants) kinetics than those observed for all other phenotypes under all excitation wavelengths except 420 nm. Under 420 nm excitation, τ₂ ^ST profiles for phenotypes 1 and 3 were only significantly (p < 0.002) elevated over those found in phenotypes 2 and 4.

A mixed linear model was also used to compare spectrally dependent photo-physiological profiles within each light-response phenotype (Figure 3 and Supplementary Table S3). For Phenotype 1, Φ_PSII ⁴²⁰ Φ_PSII ⁴⁴² and Φ_PSII ⁵⁰⁵ profiles were significantly (p < 0.004) higher than Φ_PSII ⁴⁵⁸ and Φ_PSII ⁵²⁵. For phenotype 2, Φ_PSII ⁴²⁰, Φ_PSII ⁴⁵⁸, and Φ_PSII ⁵²⁵ profiles were lower (p < 0.005) than Φ_PSII ⁴⁴² and Φ_PSII ⁵⁰⁵. For phenotype 3, Φ_PSII ⁴²⁰ appeared to be significantly (p < 0.006) lower than all other profiles whereas Φ_PSII ⁵⁰⁵ was significantly (p < 0.007) higher than the rest. For phenotype 4, Φ_PSII ⁴²⁰ and Φ_PSII ⁴⁵⁸ profiles were on average lower (p < 0.029) than Φ_PSII ⁴⁴² and Φ_PSII ⁵⁰⁵ profiles. For all four phenotypes, spectrally dependent σ_PSII differed significantly (p < 0.021) from one another with σ_PSII ⁴²⁰ showing the highest and σ_PSII ⁵²⁵ the lowest values overall. Interestingly, and in contrast to that observed for σ_PSII, no spectrally dependent differences in qP were observed within any phenotype. NPQ⁴²⁰ and NPQ⁴⁴² displayed significantly (p < 0.0001) higher values as compared to NPQ⁴⁵⁸, NPQ⁵⁰⁵, and NPQ⁵²⁵ profiles within phenotype 1. For phenotype 2, NPQ⁴⁵⁸ profiles were similar to NPQ⁴⁴² and NPQ⁵⁰⁵ whereas all others differed significantly (p < 0.001) from one another, as NPQ⁴²⁰ values tended to be slightly higher than the rest. For phenotypes 3 and 4, the NPQ⁴²⁰ profile generated significantly (p < 0.001) higher values whereas NPQ⁵⁰⁵ values were significantly (p < 0.034) lower than all others. All τ₁ ^ST profiles for phenotypes 2 and 3 displayed significantly different responses from one another whereas τ₁ ⁵⁰⁵ and τ₁ ⁵²⁵ were similar to one another within phenotypes 1 and 4. Overall, τ₁ ⁴²⁰ and τ₁ ⁴⁴² produces slower reoxidation kinetics as compared to τ₁ ⁴⁵⁸, τ₁ ⁵⁰⁵, and τ₁ ⁵²⁵. Lastly, τ₂ ⁴²⁰ and τ₂ ⁴⁴² values were significantly lower than all others in phenotypes 1 and 2. For phenotype 3, τ₂ ⁴⁴² and τ₂ ⁵⁰⁵ produced significantly (p < 0.004) lower values than τ₂ ⁴²⁰, τ₂ ⁴⁵⁸, and τ₂ ⁴²⁵. In contrast, τ₂ ⁴²⁰, τ₂ ⁴⁴² and τ₂ ⁵⁰⁵ profiles produced significantly (p < 0.001) lower values than those observed for τ₂ ⁴⁵⁸, and τ₂ ⁵²⁵.

Underlying differences in cellular physiology were also compared across the four fluorescence-based light-response phenotypes (Figure 4). Cell size (FSC) was significantly (p < 0.002) higher in phenotype 2 as compared with phenotypes 1 and 4 (Figure 4A). Granularity (SSC) was significantly (p < 0.008) higher in phenotypes 2 and 4 as compared to phenotypes 1 and 3 and may indicate differences in light scattering abilities across groups (Figure 4B). Fluorescence-based chlorophyll-a measurements were significantly (p < 0.025) lower in phenotype 2 as compared with phenotypes 1 and 4 (Figure 4C). N:P and C:P ratios were significantly (p < 0.004) higher in phenotype 1 as compared with phenotypes 2 and 4 (Figures 4D, E).

In order to look for broad connections between primary (cellular) and secondary (photophysiological) traits, a network analysis (Figure 5) was used to search for significant correlation between each of the 1,360 fluorescence-based measurements and traditional cellular characteristics (Carbon per cell, Nitrogen per cell, Phosphate per cell, C:N ratio, N:P ratio, Cell Size, Chlorophyll-a (FSC), Granularity (SSC), and neutral lipids). Our analysis identified 415 correlations having a significant Pearson value of 0.6 or above. The cellular metrics N:P, C:P, and SSC displayed the greatest number of significant correlations (269, 63, and 70 respectively). For N:P ratios, the majority of positive correlations were with τ_s ^ST measurements, while most negative correlations were with Φ_PSII values. For C:P ratios, most correlations occurred with NPQ or qP values while SSC correlated more broadly with various metrics including Φ_PSII, τ_s ^ST, qP and connectivity. A select number of these cellular to photo-physiological correlations are displayed in full detail in Figure 6.

Light acclimation state can mask species-specific differences in certain physiological metrics, as higher irradiances often lead to upregulation of stress-mitigating pathways (Ragni et al., 2010) and a reduction in photopigment production (Hoadley and Warner, 2017). This has traditionally made it difficult to capture species-specific trait-based differences without first accounting for light acclimation state. However, differences in the degree of impact that light acclimation state has on photophysiology may be largely species dependent, as high- and low-light acclimated fragments of Cladocopium C15 and C1 (in A. humilis) all clustered within light-response phenotypes 1 and 4 respectively, indicating minimal impact of light acclimation state on the overall photophysiological phenotype derived through our ‘high content’ chlorophyll-a fluorescence protocol (Figure 2). In contrast, high- and low-light acclimated T. reniformis coral fragments containing Durusdinium D1 symbionts clustered into separate phenotypes, as did two of the low-light acclimated C3-dominated Acropora humilis coral fragments (Figure 2). Suggett et al., 2022 also noted that the variance in light acclimation state (light niche plasticity) differed across three different species of coral found along the same reef system and at a similar depth. Understanding how various environmental factors constrain individual coral and symbiont species combinations and their underlying phenotypes has not been feasible with industry standard fluorometers such as the diving PAM (Krämer et al., 2022). However, single-turnover fluorometers capable of measuring additional photochemical parameters are becoming increasingly popular, especially as trait-based approaches are further applied toward coral restoration and conservation practices (Voolstra et al., 2021a; Suggett et al., 2022). Still, careful consideration towards the limits of fluorescence-based measurements must still be accounted for as their utility is further explored in applied settings.

While the degree of thermal tolerance can vary across specific host/symbiont combinations (Suggett et al., 2017), comparatively high bleaching resistance in Durusdinium D1 and Cladocopium C15 has been a focal point of coral research (Voolstra et al., 2021b). Whether bleaching resistance is derived through similar functional traits or if mechanisms of thermal tolerance differ across the two species is currently unknown. Coral endosymbiont thermal tolerance is often linked to photochemistry (Warner et al., 1999; Fitt et al., 2001; Wang et al., 2012), yet light-response phenotypes differed across Cladocopium C15 and Durusdinium D1 dominated coral fragments in this study (Figures 2, 3), suggesting differences in their photosynthetic characteristics. For example, higher reliance on non-photochemical quenching (NPQ) in response to rapid changes in light is noted for the Cladocopium C15-dominated light-response phenotype (phenotype 1—Figure 3I) whereas Durusdinium D1 symbionts from phenotype 2 and 3 relied more heavily on photochemical quenching to mitigate excess excitation energy as a larger proportion of PSII reaction centers remain closed throughout the actinic light protocol (Figures 3N–O). Reoxidation kinetics between the C15 and D1 also differed as light-response phenotype 1 had lower τ₁ ^ST values than phenotype 2 (under 420, 442, and 505 nm excitation) or 3 (under 442, 458, 505, and 525 nm excitation) and indicate faster rates of electron transport between the Q_a and Q_b sites within the PSII reaction centers of Cladocopium C15 symbionts (Figures 3Q–S). Interestingly, faster τ₁ ^ST kinetics for C15 were coupled with much slower τ₂ ^ST rates as compared to other phenotypes (Figures 3U–W). Importantly, τ₂ ^ST values are derived from a 2 exponential equation fit model and thus do not necessarily reflect a specific rate constant (Hoadley et al., 2023), the higher values likely indicate slower rates of electron transport within and downstream of the plastoquinone pool. Alternative electron sinks or cyclic electron transport can play an important role in coping with excess excitation energy (Roberty et al., 2014; Vega de Luna et al., 2020) and differences in their utility across symbiont species may help drive the different reoxidation profiles observed here. Overall, stark contrasts in how each species copes with light energy during rapid changes in light are perhaps not surprising given the >100 million years of evolutionary history that separate the two species (LaJeunesse et al., 2018). How and if these different functional traits drive unique thermal acclimation strategies will need to be the focus of a future study.

Use of multiple LED colors to excite chlorophyll-a fluorescence allows for potential differences in photopigment utilization to be incorporated into our phenomic analysis. For example, spectrally-dependent variance in NPQ responses may point towards differences in how photopigments are utilized to cope with excess excitation energy within light-response phenotypes 1 and 3 (Figures 3I, K). In contrast, phenotype 2 displayed much lower levels of NPQ in response to changes in actinic light, and little spectral variance in its profile (Figure 3J). Non-photochemical quenching broadly encompasses various mechanisms utilized by photosynthetic organisms to dissipate harmful excess excitation energy absorbed by light harvesting antennae (Lacour et al., 2020). For many eukaryotic photoautotrophic taxa, the xanthophyll cycle (XC) is a major energy dissipation mechanism regulating observed changes in NPQ. While XC is not always involved in NPQ regulation, and its role in the family Symbiodiniaceae is not fully resolved, excess light energy is dissipated as heat through the inter-conversion of the photopigments (zeaxanthin to violaxanthin) or (diadinoxanthin to diatoxanthin). The sum of these various photopigments are collectively known as the xanthophyll pool, and higher concentrations are often associated with acclimation to high light (Lacour et al., 2020; Schuback et al., 2021). Importantly, these different photopigments have unique absorption spectra which may be preferentially excited by our multispectral analysis. For example, absorption spectra for extracted zeaxanthin and violaxanthin pigments indicate that both absorb light from 420, 442, and 458 nm excitation, but longer wavelength excitation (505 and 525 nm) may not be as readily absorbed by violaxanthin (Ruban et al., 2001). The spectrally-dependent variance in NPQ response observed for phenotypes 1 and 3 (Figures 3I–K) may potentially reflect differences in the relative abundance and utilization of various xanthophyll pigments. Similarly, differences in the absorption spectra between diadinoxanthin and diatoxanthin may also be present and lead to observed differences in spectrally dependent photosynthetic metrics. While additional research is needed, such a connection between XC pigment pool/utilization and excitation wavelength could provide an additional dimension for understanding NPQ responses and how they might differ across species and environmental conditions.

Significant spectrally-dependent variability is also notable within the τ₁ ^ST and τ₂ ^ST reoxidation kinetics. These time constants reflect the rate of electron transport between the Qa and Qb site of the PSII reaction center (τ₁ ^ST) and further downstream kinetics involving the PQ pool (τ₂ ^ST) and further downstream electron transport. Previous work has indeed demonstrated the utility of τ₁ ^ST for characterizing light or thermal acclimation state in reef corals (Hoadley et al., 2015; Suggett et al., 2022; Hoadley et al., 2023) or productivity rates in marine algae (Gorbunov and Falkowski, 2021). Values from τ₂ ^ST are less well understood yet the clear structure observed in our profiles suggest this metric is indeed useful for assessing trait-based differences across species and/or environmental conditions.

Underlying Symbiodiniaceae cellular physiology differed significantly across the four light-response phenotypes derived from chlorophyll-a fluorescence-based measurements. Linking underlying cellular physiology with more easily measured secondary traits such as photo physiology is critical for broadening the utility of multispectral and single-turnover chlorophyll-a fluorometers. These non-invasive, optical tools could serve as highly informative platforms for monitoring health and resilience of photosynthetic organisms, including reef corals (Suggett et al., 2022). Cellular traits, such as granularity which broadly measures the light scattering properties of a cell, were significantly higher in phenotypes 1 and 3 and may serve to deflect excess excitation energy. Reductions in photochemical quenching are more quickly relaxed in phenotypes 1 and 3 and higher granularity may serve to mitigate rapid shifts in light, functioning to reflect excess excitation energy away from the cell and reducing reliance on downstream processes such as closing PSII reaction centers in response to high light (qP, Figures 3N, P). In contrast, large cell size and lower chlorophyll content cell^-1 for phenotype 2 could reduce light capture per cell as compared to phenotypes 1 and 2. From the cellular perspective, lower light capture by phenotype 2 maybe be beneficial for optimal performance within high-light environments, without the need to rely on secondary trait characteristics such as NPQ to mitigate excess light (Figure 3J). Indeed, cellular characteristics may help explain the photo physiological strategies employed by each phenotype. However, host skeletal and reflectance properties and their impact on the photo-symbiont’s light environment and acclimation strategy need also be considered (Enriquez et al., 2005; Enríquez et al., 2017) as potential sources of variability within the chlorophyll-a fluorescent light-response phenotypes.

As chlorophyll-a fluorescence-based measurements are increasingly utilized for understanding photosynthetic characteristics and the utilization of stress response mechanisms by coral photosymbionts, identifying direct linkages between photo physiology and ecosystem services or underlying cellular traits are needed (Suggett et al., 2017). Our network analysis identified key correlations between basic cellular traits and photo physiological parameters across 20 different coral/symbiont combinations (Figure 5). Certain chlorophyll-a fluorescence-derived photophysiological parameters may serve as useful biomarkers for some primary cellular traits, especially when properly contextualized within specific environmental and/or acclimatory conditions. For example, N:P ratios are inversely correlated to quantum yield of PSII measurements and directly correlated with τ₂ ^ST reoxidation kinetics suggesting that phosphorous limited cells downregulate photochemical activity. Specifically, reductions in the quantum yield of PSII indicate reduced efficiency of light utilization for photochemistry whereas increases in τ₂ ^ST reflect slower rates of electron transport, both of which appear to occur as N:P ratios rise. From a genotype perspective, both cell size and C:P differ significantly across the predominantly Cladocopium C15 and predominantly Durusdinium D1 phenotypes (phenotypes 1 and 2 respectively, Figures 4A, E; Figures 6A, C). These genotype level differences in cellular physiology may also be reflected in the reoxidation kinetics values where Durusdinium D1 (phenotype 2) appears to have slower (higher rate constant) light acclimated τ₁ ^ST (Figures 3Q, R) yet faster (lower rate constant) τ₂ ^ST (Figures 3U, V) reoxidation rates than those for Cladocopium C15 (Figure 6D). Linking both primary and secondary trait differences, especially across species known for their thermal tolerance, can be valuable for understanding what function traits are important for establishing resilient coral symbioses.

Carbon to phosphorous ratios also appear to be correlated with various photochemical metrics, most notably photochemical and non-photochemical quenching mechanisms which function to balance light utilization within the cell. Phosphorous limitation thus appears well linked to photochemical metrics, potentially regulating gene expression along with cell ultrastructure (Ferrier-Pagès et al., 2016; Rosset et al., 2017; Lin et al., 2019). Granularity is also linked with many different photochemical metrics which is perhaps not surprising given the strong phenotype differences observed for this cellular trait (Figure 4). Overall, the strong linkages observed in our network analysis help strengthen our understanding of how differences in cellular traits across Symbiodiniaceae species regulate chlorophyll-a fluorescence-based phenotypes.