Relationship between Acropora millepora juvenile fluorescence and composition of newly established Symbiodinium assemblage

Spawning and larval culturing

Eggs and sperm were acquired from eight gravid A. millepora colonies collected from Trunk Reef from about 3 metres depth (AIMS permit number: G12/35236.1) in November 2014 following spawning methods previously described in (Quigley, Willis & Bay, 2016). Briefly, positively-buoyant bundles were collected from the colonies, gently mixed with gametes from all 8 colonies, and then left to sit for 1.5 h to allow for fertilization to occur. After this time, fertilized embryos were washed three times by repeated transfer to new fresh-seawater containing bins. Bulk mixtures of gametes were maintained and larvae were raised in tanks with constant aeration and flow through set at 27 °C. Fully competent larvae were exposed to ground, autoclaved crustose coralline algae, a natural known settlement cue (Heyward & Negri, 1999), in sterile, plastic 6-well plates and allowed to naturally metamorphose for ∼24 h.

Fluorescence microscropy/ image analysis/ fluorescence quantification

After metamorphosis, fluorescent images of each individual juvenile were taken using a fluorescent stereo-microscope MZ FL-III (Leica, Bannockburn, IL, USA) equipped with a Canon G6 camera. Photos and image analysis was approached as in (Kenkel et al., 2011; Strader, Davies & Matz, 2015; Strader, Aglyamova & Matz, 2016). Juvenile fluorescence was imaged using the double-bandpass F/R filter (Chroma no. 51004v2), a filter that detects red fluorescence produced from the coral host while excluding any chlorophyll fluorescence. In addition, photographs were imaged prior to treatment exposure to Symbiodinium, therefore red fluorescence in coral hosts is not confounded by potential chlorophyll fluorescence. Image analysis was performed using ImageJ (Schneider, Rasband & Eliceiri, 2012). For each individual photograph, raw integrated RGB values were calculated across the area of the juvenile. A reference RGB value was calculated within a fixed circular area adjacent to the juvenile in each photograph and was subtracted from measured RGB values for each individual. Individual redness values were calculated as the normalized red value divided by the normalized green value plus the normalized red value.

Symbiodinium acquisition and community sequencing

Post-imaging, the six-well plates with settled juveniles were added to sediment treatments for natural uptake of Symbiodinium in the lab in the National Sea Simulator (Seasim) at the Australian Institute of Marine Sciences, resulting in 21 days of cumulative exposure to sediments. Six-well plates with juveniles were floated approximately 5 cm above the sediments using porcelain weights. Flow-through aquaria were fed with 27.2 °C, 0.4 µM filtered seawater that each contained 1L of sediments, with turn-over of the 45L tanks occurring once per hour. During this time, juveniles were exposed to natural light at 50 µmol photon illumination. Given the high level of water filtration and average Symbiodinium size of >4 µm (LaJeunesse et al., 2005), all Symbiodinium taken up by juveniles are assumed to be of sediment origins. Ten and twelve juveniles (n = 22) were sampled for green and red color morphs respectively from four tanks (tank 1: three red, three green; tank 2: three red, one green, tank 3: three red, three green, tank 4: three red, three green), and preserved in 100% ethanol and stored at −20 °C until sample processing. DNA extraction and sequence read analysis from individual juveniles followed (Quigley, Willis & Bay, 2016). Briefly, nucleic acid isolation included an initial chemical lysis, mechanical lysis using 1 mm silica beads (MPBio, Santa Ana, CA, USA), precipitation and clean-up (Wilson et al., 2002). DNA was sent to the University of Texas at Austin’s Genomic Sequencing and Analysis Facility (USA) for paired-end Miseq sequencing (Illumina, San Diego, CA, USA) of the ITS-2 region (Pochon et al., 2001). Raw reads were cleaned and analysed using the USEARCH and UPARSE pipeline (v.7) and identified to the clade/type level using a custom Symbiodinium database constructed from the NCBI database (see Supplementary Information) (Altschul et al., 1990; Camacho et al., 2009; Edgar, 2013).

Statistical analysis

Cleaned reads were variance normalized to account for differing sequence depth between samples using ‘DESeq2’ in R (R Core Team, 2013; Love, Huber & Anders, 2014). The functions ‘metaMDS’, ‘ordiplot’, ‘ordihull’, and ‘orditorp’ from the ‘vegan’ package and ‘ggplot2’ were used to construct NMDS plots using a Bray-Curtis distance matrix on variance-normalized OTU abundance data (Wickham, 2009; Oksanen et al., 2013). Gradient plots of the smooth response variable “redness” values over NMDS space were constructed using the ‘ordisurf’ function in ‘vegan.” Discriminant analysis of principle components (DAPC) was used to identify which OTUs may have attributed to juvenile groupings by fluorescence using the package ‘adegenet’ (Jombart, 2008).

Permutational multivariate analysis of variance using Bray-Curtis distance and Permutation test for homogeneity of multivariate dispersions was used to determine if Symbiodinium assemblages differed significantly between red and green juveniles using the ‘adonis’ and ‘betadisper’ functions in ‘vegan.’ To assess if there were significant differences in stabilized abundances between different Symbiodinium clades and types between red and green juveniles, independent 2-group Mann–Whitney U Tests (MWU) were performed (at alpha = 0.05). MWU Tests were run using the ‘wilcox.test’ function in the base R ‘stats’ package, which are able to account for the lack of homogeneity of variance across samples. To assess if there were significant differences in stabilized abundances between different OTUs found within red and green juveniles, negative binomial generalized linear models in ‘DESeq2’, using significant Benjamini–Hochberg p-values, were performed. The ‘ordisurf’ function and the ‘vegan’ package were used to generate Generalized Additive Models to determine if Symbiodinium assemblages differed significantly across redness values (Wood, 2006; Wood, 2008). To test for significant associations between variance normalized abundances and redness values for each OTU, Spearman’s rho correlation coefficients and associated p-values were calculated per OTU using the ‘cor.test’ function in the base R ‘stats’ package. Sample sizes per test for clades and types given in Tables S1 and S2. MWU and Spearman rho correlation tests were also randomized and re-run to assess if significant p-values were false-positives.

Fluorescent variation among juveniles

We found prominent variation in fluorescent phenotypes among young A. millepora juveniles (Fig. 1). Variation in fluorescence, quantified as ‘redness’, was continuous among individuals, however individuals were also binned into two categories and analysed as either ‘red’ or ‘green’ (Figs. 1; 2A, 2B). This fluorescent variation is due to variable expression of A. millepora GFP-like proteins, which vary in excitation/emission (Alieva et al., 2008; Beltran-Ramirez, 2010; Kenkel et al., 2011) and copy number (Gittens et al., 2015).

ITS-2 variation between color morphs

Eighty-nine OTUs were recovered between the two color morphs (62 in red and 71 in green). When fluorescence was treated continuously, there was a marginally significant relationship between the Symbiodinium community and redness of A. millepora juveniles (Generalized Additive Model: p = 0.053, R² = 0.33) (Fig. 2A). Of the 89 OTUs retrieved across juveniles, Spearman’s Rank Correlation tests indicated significant candidate OTUs as correlated with redness as a continuous variable (OTU79-C15, OTU985-A1; Fig. S1A). However, randomization tests (where redness values were shuffled among samples) indicated that these relationships might be false positives.

There was no significant relationship between Symbiodinium assemblage composition and fluorescence when color was treated as a categorical trait as found using either permutational manova (Adonis with Bray-Curtis distance: p = 0.58) (Fig. 2B) or discriminant analysis (Fig. 2C). OTU427 (C15) contributed the most to differentiating between the assemblages in the two color morphs as shown through loading weights of each OTU above discriminant analysis default thresholds (Figs. 2B, 2C). Although it appeared that green juveniles displayed a greater capacity to associate with a wider range of Symbiodinium OTUs and red juveniles displayed a more restricted assemblage (Figs. 2B, 2C), there was no significant relationship between the variances of the two color morphs (Permutation test for homogeneity of multivariate dispersions, df = 1, p = 0.5).

Clade and type abundances varied between color morphs (Fig. S1B), although after randomization, only clade A was robustly found in significantly greater abundances in red juveniles (Mann–Whitney U Tests, p = 0.03). The same OTU (427-C15) predicted as the OTU driving the greatest divergence of the two communities in the discriminant analysis was found at 15 fold lower abundances in red compared to green juveniles and OTU985 (S. microadriaticum) occurred in 74 fold higher abundance in red compared to green juveniles (DESeq2-Bejamini-Hochberg p-adjusted values both = 0.03).

Variable Symbiodinium types between color morphs

We detected significant changes in the Symbiodinium community as redness values shifted, potentially driven by significantly higher abundance of clade A in redder individuals and clade C in greener juveniles. Clade A is a generalist clade able to tolerate low and high light levels (Iglesias-Prieto & Trench, 1997a) and exhibits decreased growth in red compared to green light (Kinzie, Jokiel & York, 1984). C15 Symbiodinium often inhabit extreme shallow environments of high irradiance in species such as Montipora digitata, but can also dominate across substantial geography and depth in Porites spp. (Veron, 2000; LaJeunesse et al., 2003; LaJeunesse et al., 2004). It is possible that different abundances of specific Symbiodinium types in red and green fluorescent juveniles and significant changes to assembly composition with redness was observed because variation in the host internal light environment, or the availability of photosynthetically usable light, affects the proliferation or physiology of various Symbiodinium types differently. Spectral properties in coral tissues are regulated by FPs, with different FP types significantly changing the available wavelengths of light inside coral tissues (Salih et al., 2000). Green FP absorbs blue light and therefore diminishes the light available for photosynthesis and can suppress the cell-cycle of Symbiodinium (Kinzie, Jokiel & York, 1984; Alieva et al., 2008; Wang et al., 2008; Suggett et al., 2015). Therefore, GFP might act to regulate the amount of photosynthetically available light or proliferation of Symbiodinium, thus allowing uptake of an assemblage with specific physiologies. Alternatively, more red individuals are unable to regulate photosynthetically available light for Symbiodinium, which may promote the higher abundance of generalist clade A Symbiodinium and might reflect uptake of a more neutral Symbiodinium assemblage.

In addition, it is possible that greener host individuals are better at attracting more evolutionary derived C15 Symbiodinium as a way to avoid more opportunistic clade A Symbiodinium. Furthermore, opportunistic clade A Symbiodinium could avoid greener individuals as a means of increasing a competitive advantage with other Symbiodinium types. However, the actual dynamics of host/symbiont control regarding uptake and winnowing needs to be determined with follow-up studies. Regardless of what mechanisms may be at play, our results suggest subtle differences in Symbiodinium assemblages along a range of recruit fluorescence and drivers of these differences have yet to be investigated. It is also possible that the differences we observe may reflect variable irradiance preferences between Symbiodinium for specific light environments, although this is highly speculative and necessitates further study.

Differences in irradiance preferences may vary across specific Symbiodinium strains (Rowan et al., 1997), which could reflect differences in photo-acclimation potential. It is unclear how difference in irradiance preferences could impact coral performance, although perhaps this could occur through variability in carbon/nitrogen acquisition across strains, which occurs along depth gradients in Stylophora pistillata (Ezzat et al., 2017). Intra-clade level differences are extensive across a range of Symbiodinium traits and therefore the clade level may not be a good predictor of physiology. For example, in a comparative study, Symbiodinium microadriaticum and Symbiodinium pilosum (A2) had distinct photophysiological characteristics (Iglesias-Prieto & Trench, 1997b), and type A3 was found living in high irradiance environments through the production of MAA mycosporine-glycine (Banaszak et al., 2006). Photophysiological traits such as electron transport rates, cellular RCII concentrations, and light harvesting also differed across distinct types and did not cluster by clade; again suggesting that types within clades A and F are physiologically distinct (Iglesias-Prieto & Trench, 1997a; Suggett et al., 2015). Finally, the unique OTUs for both color morphs suggests that these are distinct subtypes and that variations between these types exist (for example: C3-u vs. C3-z prevalence between light environments (LaJeunesse et al., 2010).

It is possible that other physiological differences among juveniles besides fluorescence may be influencing Symbiodinium uptake. The higher abundances of specific Symbiodinium types may be due to different by-products produced by these two juvenile color morphs. The sugar trehalose has been found to be a strong chemical attractant for coral larvae (Hagedorn et al., 2015) and different Symbiodinium types may be similarly attracted to trehalose or other sugar derived chemicals. Potentially, different color morphs could produce variable attractive compounds or quantities. Indeed, the variable production of N-acetyl glucosamine and C6 sugars in coral mucus has been found to significantly alter bacterial assemblages (Lee et al., 2016). Finally, since our experiment was performed on a bulk culture containing multiple genotypes, it is also possible that the variation we see in regards to Symbiodinium could be a result of genotypic variation influencing uptake as there are heritable components to both fluorescence and Symbiodinium acquisition (Kenkel et al., 2011; Quigley, Willis & Bay, 2017). Further work is needed to determine if other traits aside from variable fluorescence are associated with juvenile physiology that may explain variable uptake of symbionts.

Although we found a significant correlation between the abundance of two specific OTUs representing two Symbiodinium types and redness fluorescence, a previous report found no significant relationship between symbiont abundance and green fluorescence in Seriatopora hystrix (Roth, Fan & Deheyn, 2013). This may be due to three reasons: (1) that study did not differentiate between different Symbiodinium clades or types, (2) all Seriatopora hystrix larvae measured were only of green fluorescence and did not exhibit red to green variability as measured in A. millepora, and (3) differing methods in Symbiodinium quantification. In addition, it is possible that fluorescence shifts during juvenile ontogeny, making comparisons between larval and juvenile fluorescence incongruent.

Outlook for future research

Despite the caveats in our results, this study highlights the potential of juvenile fluorescence to modulate the host light environment impacting the physiology of specific Symbiodinium types or function as an attractant for specific types of ecologically diverse Symbiodinium. Any studies going forward would need to incorporate larger sample sizes, as it is possible that the low number of sequenced animals in this study limited our power to distinguish between significant changes and noise in Symbiodinium types and OTUs variable between fluorescent morphs. The first step forward would be to disentangle potential parental effects of fluorescence and Symbiodinium assemblage uptake (Kenkel et al., 2011; Quigley, Willis & Bay, 2017). Despite potentially confounding parental effects, our study provides the first evidence that juvenile fluorescence is a trait potentially linked to mechanisms of Symbiodinium acquisition and warrants further investigation. In addition, it would be ideal to monitor the time course of symbiont winnowing among fluorescent color morphs, as the study presented here is merely a snapshot in time during a dynamic period of restructuring of Symbiodinium assemblages (Quigley, Bay & Willis, 2017). It is possible that the differences in Symbiodinium OTUs and types we observe in this study are due to differences in the timing of winnowing between color morphs (Abrego, Van Oppen & Willis, 2009b). If true, this would support a role of juvenile fluorescence in the winnowing process. It would also be ideal to disentangle if Symbiodinium acquisition is dependent on the light transforming properties of FPs by performing manipulative light experiments using specific filters to modulate the host fluorescence available to Symbiodinium prior to infection, as in (Strader, Davies & Matz, 2015). This would test the hypothesis of host fluorescence acting as an attractant, a hypothesis that has also been proposed in (Hollingsworth et al., 2005). It is imperative to investigate the spectral and behavioural properties of specific symbiont types, specifically characterizing the wavelengths of light that can be detected as well as irradiance preferences between specific Symbiodinium types (Suggett et al., 2015; Innis et al., 2018). For example, our data suggest that clade A could be less attracted to high abundances of GFP in juveniles, although the physiological response of clade A Symbiodinium to different light spectra and intensity need further substantiation. To take it a step further, it would be interesting to identify if different irradiance preferences were related to differential nutrient acquisition or other aspects of coral performance (Ezzat et al., 2017). Finally, our findings provide a framework for further experimental studies exploring the ecological function of fluorescence during coral ontogeny and its potential influence on the establishment of symbiosis.