Fecund colonies of Acropora tenuis were collected from Magnetic Island, Australia (19°10′S, 146°50′E) and spawned at 7:00 pm on the 20th of October 2008, 5 days after the full moon. The colonies were kept in individual 60 L tanks with flow-through 1 μm filtered seawater (FSW). Gametes were collected from spawning colonies and fertilized in separate containers filled with 1 μm FSW. After fertilization, the embryos were transferred to 600 L tanks with flow-through FSW, where they were raised until they developed into planula larvae. Four days after spawning, the larvae were transferred to a temperature-controlled laboratory (27 °C) at the Australian Institute of Marine Science for allocation into experimental treatments. Planula larvae were eight days old (12 days post fertilisation) when the experiment commenced.

Fresh extracts of Symbiodinium ITS-1 types D and C1 were obtained from fragments of Acropora millepora and A. tenuis colonies, respectively, collected from Magnetic Island (19°10′S, 146°50′E). Coral tissue containing Symbiodinium cells was isolated by high-pressure airbrushing and collected in a bag containing 0.2 μm FSW. Slurries were homogenized and filtered through a 60 μm mesh, centrifuged (3000 g for 4 min) and washed 3 times, then filtered through a 10 μm membrane (Millepore) to remove most coral cells and skeletal debris. Symbiodinium extracts were kept in 0.2 μm FSW for 26 h, after which cell densities were determined in five replicate samples per type using a haemocytometer. DNA was extracted from donor coral branches and Symbiodinium cell extracts, and the dominant Symbiodinium type was confirmed using Single Strand Conformation Polymorphism (SSCP) [52]. To examine differences in cell integrity of Symbiodinium C1 and D, the proportion of dead cells was quantified in fresh Symbiodinium isolates in Oct 2010. Symbiodinium C1 cells were extracted from A. tenuis and D cells from Acropora pulchra and genotyped as above. A. millepora could not be used as the D-type donor because of significant mortality at the only known location where this species associates almost exclusively with D-type Symbiodinium [36]. Dead cells were stained with Evans blue [53] immediately following extraction, and at 24, 48 and 72 h after extraction. Four replicate counts of the number of stained dead cells vs. unstained live cells were conducted using a haemocytometer and light microscope.

Five symbiont cell surface modifications were undertaken using two enzymes [α-Amylase (Sigma A6255) and trypsin (Sigma T6567)] and three lectins [(Concanavilin A (ConA) from Jack bean (Sigma C7275), LPA from Limulus polyphemus (Sigma L2263) and WGA from Triticum vulgaris (Sigma L9640)]. The two enzyme treatments modified molecules on the surface of Symbiodinium cells: α-Amylase hydrolyses α-(1,4) glycan linkages, whereas trypsin hydrolyses any exposed peptide bonds on the carboxyl side of arginine and lysine residues. Specifically, trypsin digests the 40–100 kD range of glyco-protein exudates of Symbiodinium microadriaticum [20]. The lectin treatments blocked specific glycans: ConA blocked α-mannose and α-glycose residues; LPA blocked N-acetyl neuraminic acid, glycuronic acid and phosphorylcholine analog residues; and WGA blocked N-acetyl-β-D-glycosaminyl and N-acetyl-β-D-glycosamine oligomers. The lectin and enzyme treatments were selected to overlap with those used in previous studies (as outlined in [30,31]) and to reflect the likely surface glycans of Symbiodinium strains used [29]. Symbiont cells were diluted to 1 × 10⁵ cells/mL and digested with final concentrations of either 5 mg/mL α-Amylase (resuspended in 25 mM Tris–HCl pH 7.5, 100 mM KCl), 6 μg/mL trypsin (resuspended in 1mM HCl) or 0.1 mg/mL of lectin (resuspended in PBS) (as outlined in [30,31]). Cells in all treatments were incubated at room temperature for 2 h in the dark, mixing gently every 20 min. Following incubation, algae were centrifuged at 3000 g for 5 min, washed 3 times and re-suspended in fresh 0.2 μm FSW before infection.

To examine the binding of lectins to the surface of Symbiodinium cells, Alexa488 fluoro-labelled ConA and WGA lectins (Invitrogen: C11252 and W11261) were hybridised to fresh extractions of Symbiodinium C1 and D obtained in Oct 2010, as described above. Alexa488 labelled LPA was not available and therefore could not be tested. For each cell type, approx 1 × 10⁷ cells were preserved in 4% paraformaldehyde 24 h after extraction and stored at 4 °C. After 4 days, fixed cells were spun at 3000 g for 6 min and washed twice in PBS (pH 7.2). Separate aliquots of 500 μL (1 × 10⁶ Symbiodinium cells/mL) were incubated with the two fluorescently labelled lectins in a final concentration of 50 μg/mL for one hour in the dark, alongside a non-modified control. In two separate 500 μL aliquots, C1- and D-type cells were incubated with ConA and WGA lectins that had been pre-exposed to lectin inhibitors (WGA with N-acetyl-D-glucosamine [Sigma A8625]; ConA with Methyl alpha-D-mannopyranoside [Sigma: M6882]) at a concentration of 1 M for 1 h. Following incubation, cells were washed twice in PBS and resuspended in 100 μL PBS. Ten microlitres of each cell suspension was placed into 4 wells of a teflon coated eight-well microscope slide (ProSciTech G350805-BK) and mounted with fluorescent mounting media (VectaSheild H-1400) for imaging using confocal microscopy. Unstained control cell suspensions were also prepared for both C1 and D type symbionts to allow for spectral profiling of the innate fluorescence. Symbiont cells were imaged using a Zeiss Meta 710 confocal microscope, excited using a 488 nm (argon) laser. Emissions were collected using 34 spectral detection channels within 488 to 690 nm emission. Following spectral mapping of both innate fluorescence and labelled cell fluorescence, online profiling was used to image 800–1100 individual cells of both labelled and control cell suspensions. The use of online fingerprinting and confocal microscopy to determine the proportion of cells labelled allows for the resolution of multiple distinct fluorescent emissions associated with a single cell within a pixel resolution [54]. The auto-fluorescence of the 5 μm diameter accumulation body within the 10 μm Symbiodinium cell results in a similar 590 nm emission peak as the Alexa488 surface labelled lectins. Therefore this method allows for high-resolution spectral separation of labelled and unlabelled cells, which is otherwise unachievable using flow cytometry.

Approximately 200 larvae were placed in experimental tubs containing 200 mL of 0.2 μm FSW. After 24 h, 3 × 10⁶ untreated Symbiodinium cells (positive control, POS) or similar numbers of Symbiodinium from one of five cell surface modification treatments were separately added to three replicate containers per treatment, resulting in a concentration of 1.5 × 10⁴ cells/mL. A negative control was also employed, i.e., larvae in containers without Symbiodinium cells. Fifteen coral larvae were sampled randomly from each of the three containers per treatments at 12, 24, 48, 72 and 96 h post infection (PI) and preserved in 1 mL 2.5% glutaraldehyde in FSW. Two hours after each sampling, the fixative was removed and the coral larvae were washed 3 times then stored in 1 mL 0.1 M phosphate buffer (0.2 M NaH₂PO₄, 0.2 M Na₂HPO₄ and pH = 7.2). The proportion of larvae infected with Symbiodinium and the cell abundance within larvae were quantified by counting the number of Symbiodinium cells in 15 larvae per tub per treatment (total n per treatment = 45 larvae). Larvae were squashed under a cover slip and symbiont cells were counted at 20 × magnification under fluorescent light (Axioskop 2 Zeiss). Samples of 20 larvae from all surface treatments were preserved in 100% EtOH at the end of the experiment. These samples were analysed using SSCP of ITS-1 to confirm that only a single Symbiodinium type (C1 or D) had been taken up [52].

The number of larvae infected and the level of infection (density of Symbiodinium cells per larva) were analysed using bootstrap-based ANOVA [55]. The assumption of homoscedastic variances required for ANOVA was violated and transformations did not rectify this problem. The observed F-statistics were therefore tested against the distribution of F-statistics resulting from 4999 randomisations of the data. The negative control treatment never resulted in infection, and was therefore not included in the analyses. To test whether symbiont abundance per larva were independent at the experimental container level, we first conducted a random bootstrap-based one-way ANOVA. Because infection levels did not vary significantly among larvae within an experimental container (F = 1.69; p = 0.19), this level was omitted from further analyses. However, the number of infected larvae did vary among experimental containers within a treatment (F = 0.71; p = 0.007), and therefore the experimental container was used as the unit of replication.

Three-way random bootstrap-based ANOVAs were used to test whether the abundance of Symbiodinium cells per larva and the number of larvae infected per sample differed between Symbiodinium types (n = 2), or among surface treatments (n = 6) or sample times (n = 5). This analysis is appropriate for proportional data when test distributions are generated via bootstrapping [55,56]. The model used contained all interaction terms among the following main effects:S = μ + α + β + γ + ε, where S is the number of Symbiodinium cells per larva or the number of larvae infected per sample (α, β and γ are the main effects of Symbiodinium type, time and treatment, and ε the error). Post-hoc comparisons were conducted using Matlab's multcompare function [57]. This function performs multiple comparisons while correcting for repeated tests. Tukey's HSD criterion was used to provide the correction [58,59]. To reveal within-type differences potentially obscured by large overall differences between Symbiodinium C1 and D, post-hoc analyses were conducted separately for each symbiont type. Differences in the proportion of dead cells between C1 and D type Symbiodinium extractions were tested using a two-way bootstrap-based ANOVA (Time = 0, 24, 48, 72 h; Type = C1, D) and post-hoc analyses as described above.

The mean abundance of Symbiodinium cells taken up by coral larvae in the first 96 h following symbiont addition and the mean proportion of larvae infected per treatment differed significantly between Symbiodinium types (Table 1 and 2). The proportion of larvae infected by C1 was almost 100%, whereas only 29–71% of larvae exposed to D type Symbiodinium were infected at the end of the experiment. Post-hoc comparisons revealed a more than five-fold greater mean (± SE) abundance of symbionts in larvae offered C1 (45.64 ± 2.36) than in those offered D Symbiodinium (1.05 ± 0.17). The process of freshly extracting cells from coral hosts affected Symbiodinium types differently, resulting in significantly higher, albeit still relatively low, mean (± SE) proportions of dead cells for D (5.8 ± 2.3% dead cells) compared to C1 Symbiodinium (1 ± 0.41% dead cells) (Figure 1; Table 3).

Treatments that modified Symbiodinium cell surface molecules significantly affected the number of cells in larvae (Table 1) but not the proportions of larvae infected (Table 2). Variation in C1 symbiont abundance within larvae was high among treatments, and notably, significant differences were found between enzyme-treated versus both untreated and lectin-treated cell surfaces for this symbiont type (Figure 2a). In general, larvae contained significantly more algal cells when they were treated with the enzymes α-Amylase [AA] and Trypsin, which modified glyco-protein surface exudates, compared to unmodified Symbiodinium cells (Figure 2a). In our larval study, the numbers of Symbiodinium C1 cells taken up when cells were treated with the lectins ConA and WGA were similar to those of unmodified cells, but significantly lower than numbers of enzyme-treated cells (Figure 2a). In contrast to patterns of infection found for Symbiodinium C1, larvae supplied with Symbiodinium D had greater mean levels of infection at 96 h when supplied with untreated and lectin-treated cells compared to enzyme-treated cells, although post hoc tests did not indicate that differences were statistically significant overall (Figure 2b).

Confocal microscopy revealed that the ConA and WGA lectins used here bind to the cell surface of both Symbiodinium cell types, either in localised areas or as a smooth layer surrounding Symbiodinium cells (Figure 3b, d, g, i). The absence of corresponding fluorescent patterns in control cells (Figure 3a, f) and reductions in localised areas of fluorescence when inhibitors were added to lectin treatments (Figure 3c, e, h, j) confirm that the fluorescent patterns observed are attributable to the lectins. These lectin binding patterns are consistent with those detected by Logan et al. [29] on cultured material and support lectin specificity for algal fractions rather than host debris in fresh isolates. Overall, lectin labelling was high (76–85%) in three treatments, but was only 25% for C1-type cells labelled with ConA (Figure 3b). In comparison, 10–32% of cells were labelled with lectins pre-exposed to an inhibitor (Figure 3c, e, h, j), suggesting incomplete inhibition and/or non-specific labelling at levels similar to those detected by flow cytometry by Wood-Charlson et al. [30].

We detected significant changes in densities of both Symbiodinium types and the proportion of larvae infected in the first 96 h (Table 1 and 2). The proportion of larvae infected with C1 Symbiodinium differed significantly among earlier but not later sampling times for type-C1, and almost all larvae were infected in all treatments after 48 h (Figure 4a). The proportions of larvae infected with Symbiodinium D were lower and more variable across all treatments, and post-hoc comparisons revealed a significant increase in the proportions of D infected larvae at 48 h compared to earlier and later time points (Figure 4b). While almost 100% of larvae were infected with Symbiodinium C1 at 48 h (Figure 4a), post hoc analyses of cell densities revealed significant overall increases at 48 h and further changes at 72 and 96 h (Figure 2a). In contrast, overall cell densities in larvae exposed to Symbiodinium D did not increase significantly until 96 h (Figure 2b). Interestingly, Symbiodinium C1 cell densities declined for larvae exposed to untreated Symbiodinium (POS: p = 0.009) and two lectin-blocked cell extracts (ConA: p = 0.02 and WGA: p = 0.03) between 72 and 96 h but not in the remaining cell surface treatments (AA: p = 0.92; LPA: p = 0.08; Trypsin: p = 0.74). Conversely, Symbiodinium D cells increased in abundance within larvae between 72 and 96 h in ConA treated and unmodified cells (ConA: p < 0.001; POS: p = 0.02) but not in the remaining treatments (AA: p = 0.96; LPA: p = 0.13; WGA: p = 0.44; Trypsin: p = 0.66).

This study found significant differences in the degree to which A. tenuis larvae procured C1 compared to D Symbiodinium in the early establishment stages of symbiosis, leading to a higher proportion of larvae infected and up to a five-fold greater abundance of C1 Symbiodinium compared to D Symbiodinium after four days of symbiont exposure. In contrast to our finding of greater initial procurement of C1 Symbiodinium by larvae, one-month-old A. tenuis juveniles, naturally inoculated and grown at the parental location (Magnetic Island), host mostly Symbiodinium D for up to 2 years [7,47]. Differences in symbiont patterns between larvae and juveniles could arise from differences in infectivity of symbiont types and/or post-infection growth rate, both potentially affected by host “winnowing” mechanisms and local environmental conditions [2,12,23,24,46,60]. The greater larval uptake of the Symbiodinium type dominating adult populations of A. tenuis (i.e., Symbiodinium C1) found here accords with results found for other coral species. In Fungia scutaria, for example, the most robust symbiosis occurs when larvae were exposed to the Symbiodinium type dominating adult F. scutaria populations (i.e., C1f) [46,60,61]. Conversely, differences in infection levels may also exist because of symbiont dynamics. For example, Symbiodinium A from Montastraea annularis and Tridacnea crocera has been described as “weedy” because of its higher high stress tolerance, superior ability to infect larvae, and greater post-infection growth rates compared to the more variable C type [10,43,62]. Prevailing environmental conditions also affect rates of uptake and proliferation of different Symbiodinium types [12]. Symbiodinium C1 is more infective and proliferates more readily within Acropora larvae at temperatures between 25–28 °C, with infection rates decreasing significantly at temperatures above 29 °C [63]. In contrast, uptake of Symbiodinium D is unaffected by temperature, with low initial uptake increasing over time at temperatures between 26–32 °C [63]. Our study was conducted at 27 °C, which promotes uptake and proliferation of Symbiodinium C1, potentially explaining the higher proportion of larvae infected by Symbiodinium C1 and its high abundances within A. tenuis larvae.

An alternative explanation for differences in initial infection and/or post-infection growth found between Symbiodinium cells could arise as a consequence of processes used to freshly extract Symbiodinium cells from donor coral tissues. A follow-up test found that extraction processes compromised Symbiodinium D significantly more than Symbiodinium C1 cells, although mortality of Symbiodinium D was only 6% after 72 h (i.e., a loss of 900 cells). Although we cannot discount the possibility that differences in infection dynamics between D and C1 symbionts were a consequence of compromised health of D cells, the magnitude of the differences found, both in the proportion of larvae infected and in cell densities between the two Symbiodinium types, suggest that it is unlikely that they were due to Symbiodinium health alone. The difficulty in culturing many Symbiodinium types justifies the use of freshly extracted Symbiodinium in experimental research on corals, however extraction effects must be (but rarely are) quantified because of their potential to affect infection dynamics. For example, Weis et al. [46] found that repeated washing of freshly isolated Symbiodinium cells from four coral sources reduced their ability to establish symbiosis with Fungia scutaria. To correctly assign differences in infection and post-infection growth among Symbiodinium types, cultures of ecologically relevant, similarly acclimated types must be used. If fresh extraction cannot be avoided, their health and viability must be accounted for using techniques such as PAM fluorometry and vital stains.

Treatments that modified Symbiodinium cell surface molecules significantly affected the number of cells in larvae, but not the proportion of larvae infected. This suggests that Symbiodinium cell surface molecules are involved in post-phagocytosis, rather than pre-phagocytosis recognition [12,36,42,63]. Coral larvae are promiscuous and initially associate with multiple types of Symbiodinium, suggesting that their pre-phagocytosis recognition system is non-specific. Little is known about post-phagocytosis processes, but the high abundance of symbionts found in larvae exposed to enzymatically treated C1 Symbiodinium cells suggests that the 40–100 kD glycome, including glycoproteins and long chain starch residues, plays a role in controlling post-phagocytic survival and growth of C1 symbionts within coral larvae. Our findings contrast with previous studies, which have uniformly found lower (or similar) infection levels with enzyme-treated compared to untreated Symbiodinium cells [30,31]. Differences in ontogenetic patterns of symbiosis among corals may explain these contrasting results. Coral species examined in previous studies were dominated by the same symbiont type in both the juvenile and adult life stages [30,31], whereas two symbiont types occur in A. tenuis and vary in dominance between juvenile and adult corals. We hypothesise that cell surface recognition molecules may differ among Symbiodinium types and vary in their capacity to maintain symbiosis in different life stages of the coral. However, the hypothesis that surface recognition molecules can limit uptake or post-infection cell proliferation in flexible coral-Symbiodinium symbioses requires further testing in a range of species and environmental conditions.

Interestingly, the LPA lectin treatment resulted in significantly lower numbers of Symbiodinium C1 cells compared to all other treatments but the ConA treatment, indicating that N-acetyl neuraminic acid, glycuronic acid and/or phosphorylcholine sugar moieties on Symbiodinium C1 cell surfaces are involved in early recognition of compatible Symbiodinium cells by the coral host. Conversely, the similarity in infection levels among WGA-treated, ConA-treated and untreated cells suggests that α-mannose, α-glycose, N-acetyl-β-D-glycosaminyl and/or N-acetyl-β-D-glycosamine sugar residues on Symbiodinium cell surfaces do not play an important role in determining levels of early uptake under our laboratory conditions. This result is surprising, because these sugars commonly occur on the surfaces of a range of micro-algae [28-30,34], and have been implicated in cellular recognition [28-32,34]. The presence of a range of sugar molecules can vary among cell stages, species and geographical locations [28,64], potentially explaining the results obtained here. We exposed coral larvae to Symbiodinium cells for the duration of the experiment, during which time cell surface molecules may have changed. For example, Aguilera and Gonzáles-Gil [28] found that while α-mannose, α-glycose and N-acetyl residues were common and relatively stable on the surface of four dinoflagellate species (but not Symbiodinium spp), they were reduced or absent at certain sampling times during the 52-hour experimental period. In contrast, Logan et al. [29] found that all eight Symbiodinium cultures tested displayed significant binding to ConA at all sampling times (3.5 and 5.5 weeks after sub-culturing, repeated over four culture cycles) indicating the widespread and stable presence of mannose residues on these cell types, assuming specificity of the ConA probe. Likewise, 50% of cell types tested bound equally well to WGA at all sampling times, demonstrating the common presence of N-acetyl residues on the cell surfaces of cultured Symbiodinium types. Our results confirm the presence of both mannose and N-acetyl groups on both the C1 and D-type Symbiodinium used here. If these molecules are temporally stable in the C1 and D Symbiodinium types studied here, then our results suggest that these residues are less important in the early onset of this symbiosis compared to long chain starch and glyco-proteins. It is also possible that other glycans, not tested here, may be important for cell recognition in A. tenuis. For example, lectins specific for galactose-residues (termed galectins) bound to most Symbiodinium types tested by Logan et al. [29] and may be important in the maintenance of successful symbioses by keeping Symbiodinium cells in a non-motile phase [65]. Lin et al. [31] and Wood-Charlson et al. [30] detected a reduced uptake of Symbiodinium cells whose galactose residues had been obscured. Detailed characterisations of the glycomes of Symbiodinium populations and types, for both freshly extracted and cultured Symbiodinium material (as demonstrated by [29]), combined with examinations of the ability of respective Symbiodinium cells to establish symbiosis with coral species will be required to further explore the role of cellular recognition in onset and maintenance of coral-Symbiodinium symbiosis.

The temporal scale over which infection by Symbiodinium cells occurs varies among coral species. In Fungia scutaria, larvae are infected within a few h, after which densities of cells remain relatively constant [46,60]. In contrast, we detected significant variation in the temporal infection dynamics of C1 and D type Symbiodinium used here. While almost 100% of larvae were infected with Symbiodinium C1 after two days overall cell densities in larvae exposed to Symbiodinium D did not increase significantly until four days after symbiont exposure. Interestingly, C1 cell densities declined between days three and four in larvae exposed to untreated Symbiodinium and to two out of three lectin-blocked cell extracts. In contrast, Symbiodinium D cells increased in abundance over the same time scale within larvae exposed to ConA treated and unmodified cells, possibly suggesting the initiation of a winnowing process involving specific glycan-lectin interactions. The time frame over which these declines in C1 and increases in D occurred is similar to delays in cell “sorting” processes reported for Hydra endosymbiosis, cell sorting being the process by which Hydra modifies their Chlorella complement through disintegration or ejection of cells [66]. Similarly, Davy et al. [23] found that both homologous and heterologous symbiont densities declined four days after uptake, but ultimately increased and stable symbioses were formed with both types 36 weeks after infection. Our results suggest that glycan-lectin interactions, in particular those involving N-acetyl residues, may be associated with the initiation of a winnowing mechanism after only four days and, if continued, may explain the dominance of the D type in juveniles of A. tenuis [7,36]. To elucidate the complex recognition mechanisms that corals may use to shape their symbiont communities during the establishment of symbiosis and subsequent winnowing process, future studies should examine infection dynamics over an extended temporal scale.