The corals Montastraea annularis and M. faveolata each host three distantly related taxa7 of the dinoflagellate genus Symbiodinium, denoted A, B and C, that are identified by restriction-fragment length polymorphisms (RFLPs) in genes encoding small ribosomal RNA (srRNA)7. A and B are common in shallow-water corals (high-irradiance habitats), whereas C predominates in deeper corals (low-irradiance habitats). Mixed samples A + C and B + C, common at intermediate locations7, suggest that symbionts may actually exist as complex communities that track differences in irradiance within a colony2.

To test this hypothesis we sampled four locations in each of 46 colonies (Fig. 1). All M. faveolata, and all but one colony of M. annularis, yielded two or three types of symbionts. As predicted, Symbiodinium A and B dominated locations with higher, downwelling irradiance (communities 1 and 2, unshaded colony tops), and C dominated locations of lower irradiance (communities 3 and 4, colony sides and shaded colony tops) (P < 0.001; χ2 test). These patterns of intra-colony zonation largely disappear at slightly greater depths (8–11m in M. annularis, and 6–12m in M. faveolata), where Symbiodinium C is predominant overall7. As before7, Symbiodinium A was more common in M. faveolata than in M. annularis.

Figure 1: Symbiont communities in M. annularis
figure 1

(a) and M. faveolata (b). Each symbol represents one core sample that contained Symbiodinium A, B, C or mixtures of taxa summarized according to the code shown below. Columns in the data matrices represent individual coral colonies (depth increases from left to right), and rows represent locations of higher (rows 1 and 2) and lower (rows 3 and 4) irradiance, as defined in the diagrams to the left. Samples were collected in January 1995.

Analyses at a finer scale (1cm) confirmed that symbionts occupy distinct but overlapping habitats (Fig. 2). Unshaded columns of M. annularis create a localized gradient of low (on the side, no downwelling) to high (on the top, full downwelling) irradiance, which we sampled along transects. At intermediate depths (3–7m) this gradient coincides with the transition from Symbiodinium C to B, B + C, or A (Fig. 2a–d). Analyses of shallower (1–2m) and deeper (9–12m) corals (Fig. 2d) show that depth7 and intracolony zonation of the symbionts occur in parallel. These consistent patterns strongly argue that zonation is controlled by ambient irradiance. Furthermore, experimentally toppled columns, which experienced immediate and severe changes in irradiance gradients, largely re-established expected patterns of symbiont zonation during a six-month period (Fig. 2e, f). This response shows that the patterns are maintained dynamically.

Figure 2: Symbiont zonation in columns of M. annularis.
figure 2

a, b, TaqI digests of srRNA genes. a, Transect from column side to top (lanes 1–11; positions shown in c) and standards for Symbiodinium A, B and C. b, Standards (see Methods). Lane 1, C alone; 2–10, B + C mixtures (numbers below are C:B ratios); 11, B alone. c,Column sampled in a; symbols 1–11 (dark, Symbiodinium C; white, Symbiodinium B) represent the data in lanes 1–11, respectively in a. d, Zonation in 14 columns from 14 colonies sampled as in c (differences in shading are for clarity only); x-axis, location relative to lower side and top centre; y-axis, genotype ranging from all C to all B or all A (A in one column only); z-axis, depth. e, Zonation in a column transplanted to 90° from vertical for 6 months, presented as in c. f, Zonation in 7 columns (all at a depth of 6m) transplanted and sampled along two transects (white and black) as in e, presented as in d; zonation in natural columns at a depth of 6m (from d) are plotted for comparison (grey).

Symbioses between corals and dinoflagellates are stable mutualisms, with the notable exception of coral bleaching, which involves the loss of symbionts and/or photosynthetic pigments3,8. Bleaching is an ecologically important but poorly understood response to environmental stress3,8,9. Many bleaching events exhibit intra-specific variation distributed within and among habitats in ways that are difficult to explain9,10,11,12. Because irradiance and temperature act synergistically to induce bleaching12,13,14, and the symbionts of M.annularis and M. faveolata exhibit different associations with irradiance (Figs 1 and 2), we hypothesized that symbiont polymorphism underlies this variation.

We observed ‘paling’ in several colonies of M. annularis and M. faveolata on 18 September 1995, and bleaching was extensive by the second week in October, both in Panama and elsewhere15. At our study site, this event was ‘typical’: like a similar event there in 1983 (ref. 16), it followed a prolonged excursion above the mean summer maximum of temperature (Fig. 3e); it also coincided with atypically high water clarity (data in ref. 17), which implies increased irradiance at depth2. We also observed complex9,10,18 bleaching patterns in both M. annularis and M. faveolata. Bleaching was rare or slight at both shallow (<2m) and deep (>15m) sites; in between, however, both species displayed a curious pattern, with shallower colonies bleached preferentially in shaded places (Fig.3a, b) and deeper colonies bleached preferentially in unshaded places (Fig. 3c, d). Among M. annularis partitioned as in Fig. 1a (communities 1 and 2 versus 3 and 4) and by depth (above 8m versus below 8m), this difference was significant (n = 76 colonies, 64 bleached; P < 0.05, χ2 test). Some colonies exhibited a ‘ring’ of bleaching at the boundary between column top and side (Fig. 3a). M. faveolata colonies are not easily partitioned into two distinct irradiance microhabitats, but they clearly showed the same reciprocal pattern (Fig. 3b, d), with a shallower (6m) centre. Such observations have previously been hard to explain12 because the environment is isothermal, and the associations with irradiance and colony morphology are inconsistent.

Figure 3: Bleaching in M. annularis (a, c) and M. faveolata (b, d) at the study site on 28 October 1995 showing ‘shallow’ (a, b) and ‘deep’ (c, d) patterns. e, Sea surface temperatures (three-week running means, from satellite data30) at the San Blas Islands, Panama.
figure 3

Temperatures above 29°C in 1983 and 1995 were associated with coral bleaching16 (this study). Records from our study site (at 7-m depth) since 1993 (Marine Environmental Sciences Program, Smithsonian Tropical Research Institute) corroborate satellite data.

Symbiont zonation provides a simple hypothesis to explain these bleaching patterns. Bleaching was disproportionately common in what seems to be the upper limit of Symbiodinium C's ‘adaptive zone’: low-irradiance parts of corals in shallower water, and high-irradiance parts of corals in deeper water. Slight increases in temperature and irradiance might push these symbioses, but not others, beyond some physiological limit, resulting in bleaching. This hypothesis accounts for our bleaching observations, including areas of slight bleaching9,15,19 (see Fig. 3a, b), if Symbiodinium C were expelled selectively from mixed symbiont communities.

An analysis of symbionts collected in late October supported this interpretation of events. Post-bleaching samples were obtained <1cm from many sites sampled the previous January (Fig. 1). All available samples from communities that had previously contained mixtures of Symbiodinium C plus either A or B (or both) were identified (Fig. 1) and analysed. We reasoned apriori (Fig. 2) that these sites accurately defined the limit of Symbiodinium C in corals under non-bleaching conditions. Such mixtures also allow the fates of different symbionts to be compared directly. We also tested archived samples taken at the same time as, and <1cm away from, the original (pre-bleaching event) samples. In every case, Symbiodinium srRNA RFLPs in these replicate, pre-bleaching pairs were equivalent (data not shown), indicating that the small distance between pre- and post-bleaching samples was unlikely to be significant.

As predicted, Symbiodinium C had decreased in relative abundance in all 18 communities tested (see Fig. 4a–c). Absolute responses of different symbionts within a mixed community were compared by using estimates of relative abundances (from RFLP data; see Fig. 4a–c) to partition direct counts of symbionts into each genotype (Fig. 4d). Losses of Symbiodinium C were typically close to 100%, whereas B underwent a median decrease of 14%, and A more than doubled in 3 of 5 instances. The single sample that contained all three symbionts exhibited these same trends (Fig. 4c, d). Changes in colony chlorophyll contents and subjective assessments of bleaching paralleled changes in symbiont numbers (Fig. 4e). From these data we can tentatively rank the ‘fitness’ of the different Symbiodinium taxa as symbionts under ‘bleaching conditions’. The ranking obtained in this manner is: A > B C.

Figure 4: Symbiont communities before (January 1995) and during (October 1995) an episode of coral bleaching.
figure 4

ac, Lanes contain TaqI (a, b) or DpnII (c) digests of srRNA genes. a, Standards for B, C and B:C ratios of 8:1 (lane 1) and 1:1 (lane 2); lane pairs compare symbionts before (left) and during (right) bleaching in M. annularis (lanes 3–6) and M. faveolata (lanes 7–10). b, Standards for A, C and A:C ratios of 2:1 (lane 1) and 1:8 (lane 2); lane pairs 3 (M. annularis) and 4–7 (M. faveolata) compare symbionts as in a. c, Standards for A, B, C and equal amounts of A, B and C (lane 1) and A and B (lane 2); lane pair 3 compares symbionts in M. faveolata, as in a and b. The vertical bracket identifies bands that identify each symbiont. d, Densities of A (grey), B (white) and C (black) before and during bleaching (left and right bars of each pair, respectively) in samples reported in a (B + C, communities 3–10), b (A + C, communities 3–7) and c (ABC, community 3). e, Chlorophyll contents of the samples reported above, presented as in d. Samples were scored as ‘normal’ (not marked) or ‘slightly pale’, ‘pale’, or ‘bleached’ (marked by asterisks) when collected.

Our study provides a fuller understanding of M. annularis and M. faveolata, which are dominant members of western Atlantic reefs20 and are widely used as model systems in reef biology and geology11,13,18,19,21,22. Each coral ‘species’ encompasses one animal and dynamic, multi-species communities of symbiotic dinoflagellates. This strongly contradicts the ‘one host, one symbiont’ view of reef corals1, in which host taxa alone are adequate units of biodiversity, environmental variability is accommodated largely by physiological acclimatization2,3,4, and bleaching variability is often not understood12. We conclude that polymorphic symbiont communities underlie the broad distributions20 and bleaching ecology of these corals. Directed shifts in symbiont populations following extreme environmental change (Figs 2e, f and 4) suggest that similar shifts may also occur over annual cycles of environmental variation19. For these corals, and for mutualisms in general, the ability to cope with environmental change through changes in symbiont community composition reflects the selective advantage of hosting several distinct symbionts, despite the potential for destabilizing competition among them5,6.

How typical and important are the patterns documented here? M. annularis and M. faveolata in the Bahamas also host Symbiodinium A, B and C (data not presented), and published photographs18 and descriptions9,10 of bleaching elsewhere strongly resemble our own (Fig. 3a–d). With respect to Caribbean corals in general, bleaching is often predominant at intermediate depths9. We can attribute this pattern (and its within-colony correlate) in M. annularis and M.faveolata to symbiont polymorphism and zonation. Moreover, at least three other species of Caribbean corals host (at least) both Symbiodinium A and C (ref. 23; our unpublished data). For other species, which might host multiple but not so distantly related symbionts, refinements of Symbiondinium taxonomy would be essential. However, symbiont polymorphism does not exclude the significance of other attributes that are important features of coral biology, such as physiological acclimatization of hosts and symbionts2,3,4 and genetic differences among hosts11,14.

It has been hypothesized that a global warming trend, with increased frequencies of world wide coral bleaching induced by increasing temperature or ultraviolet irradiation, could have catastrophic consequences for many living coral reefs3,8,21. Alternatively, coral communities may adjust to climate change by recombining their existing host and symbiont genetic diversities24,25,26. Our findings supply a precedent for this idea: that one species of coral can flexibly host more than one taxon of Symbiodinium to produce symbioses with distinct ecological properties. For example, M. annularis and M. faveolata might adjust to a warmer Atlantic ocean by hosting more Symbiodinium A and less Symbiodinium B and C. However, long-term consequences of such replacements would depend on how they affect rates of coral growth and reproduction.


Field collections and manipulations. Coral samples were collected at Aguadargana reef in San Blas, Panama27 by coring (1.1cm2 surface area) and freezing immediately in liquid nitrogen (data in Figs 1 and 4). Other colonies (data in Fig. 2) were sampled by removing a defined circular area (0.12cm2) of living tissue from freshly collected colonies with an airbrush. In transplant experiments (Fig. 2e, f), columns of M. annularis were broken off 15cm below the living tissue, turned on their sides, and cemented (at the non-living base) back to the colony at a comparable location; this increased (new top), decreased (new side), or did not change (side) local irradiance. All 28 transplants at a depth of 6m seemed to be normal after 6 months. Analyses of non-transplanted (control) columns showed that natural zonation patterns were stable over this period (data not shown).

Laboratory analyses. Symbionts and symbiont DNA were isolated from frozen7 and from fresh28 samples. Nuclear srRNA genes were amplified using the ‘universal’ PCR primers ss5 and ss3 (all data in Fig. 1) or a combination of ss5 and the ‘Symbiodinium-biased’ primer ss3Z (all data in Figs 2 and 4), and analysed with TaqI and DpnII (data were consistent in every case)7. The biased primer (ss3Z) does not discriminate (within this study) against unknown, specific Symbiodinium genotypes (discussed in ref. 28), as confirmed by sequencing7 and by comparing results from ‘universal’ and ‘biased’ amplifications of 30 samples that contained two genotypes (A + C or B + C) in various proportions.

Cloned srRNA genes from Symbiodinium A, B and C were used as standards7. They were amplified singly and from defined mixtures of two (Fig. 2b) or three (Fig. 4c) types7 to assign field samples to classes of symbiont relative abundance by visual comparison (Fig. 2a, b). To validate this procedure, approximately equal numbers of A, B and C cells, from three natural isolates of each type, were mixed in pairwise combinations and analysed. The results implied that Symbiodinium B and C yield (on a per-cell basis) equal signals, whereas A yields about twice that amount. Standard mixtures of cloned genes were adjusted accordingly.

Symbiont densities and chlorophyll contents (Fig. 4d, e) were determined from haemacytometer counts (8 replicate grids per sample) and spectrophotometrically from methanol extracts29, respectively. These symbionts were isolated quickly (with minimal washing) from frozen samples at 4°C under dim light. Symbiont genotypes, numbers and chlorophyll contents were obtained from subsamples of each isolate.