Research Articles
Hybridization and the Evolution
of Reef Coral Diversity
Steven V. Vollmer* and Stephen R. Palumbi
Hundreds of coral species coexist sympatrically on reefs, reproducing in massspawning
events where hybridization appears common. In the Caribbean, DNA
sequence data from all three sympatric
does not erode species barriers. Species
distinct at two nuclear loci or share ancestral alleles. Morphotypes historically
given the name
showing morphologies that depend on which species provides the egg for
hybridization. Although selection limits the evolutionary potential of hybrids,
F
hybrids with unique morphologies.
Acropora corals show that mass spawningA. cervicornis and A. palmata areAcropora prolifera are entirely F1 hybrids of these two species,1 individuals can reproduce asexually and form long-lived, potentially immortalDiverse reef-building coral assemblages have
served as the foundation for complex reef
ecosystems with exceptional biodiversity and
productivity. Yet, the evolutionary genesis of
coral diversity remains mired in a paradox.
As many as 105 coral species from 36 genera
and 11 families reproduce in yearly, synchronous
mass-spawning events (
overwhelming opportunities for
hybridization among congenerics (
crosses from a number of mass-spawning
genera demonstrate that viable hybrids
occur among congenerics (
hybridization should blur coral species
boundaries and stifle species diversification,
yet many mass-spawning coral groups have
rapidly diversified. The juxtaposition of high
hybridization potential and high species diversity
in mass-spawning corals has confused
the picture of coral evolution and cast such
doubt on the cohesiveness of coral species
boundaries (
have been considered hybrid swarms (
1), thereby providing2). Laboratory2, 3). Interspecific4) that some species-rich genera3).Acropora
group (
the 115 species of
past 5 million years (My) (
are capable of hybridizing with sympatric
congenerics in laboratory crosses (
prominent hypothesis proposes that interspecific
hybridization promotes reticulate evolution
and morphological diversification in the
absence of genetically distinct species (
even though a genetic mechanism for this
, the world’s most speciose coral5), exemplify this view (2–4). Most ofAcropora arose over the6, 7), and many2, 8). One3),Department of Organismic and Evolutionary Biology,
Harvard University, 16 Divinity Avenue, Cambridge,
MA 02138, USA.
*To whom correspondence should be addressed. Email:
svollmer@oeb.harvard.edu
Fig. 1.
(
The Caribbean Acropora species: (A) A. cervicornis and (B) A. palmata, and (C) the bushy andD) palmate F1 hybrid A. prolifera morphs from Puerto Rico.R
E P O R T Swww.sciencemag.org SCIENCE VOL 296 14 JUNE 2002
2023Downloaded from
www.sciencemag.org on April 7, 2010hypothesis is lacking. Polyphyletic sequence
data for corals continue to be taken as direct
evidence of reticulate evolution (
due consideration to alternatives such as
incomplete lineage sorting.
To examine the potential role of hybridization
in coral speciation, we analyzed DNA sequence
variation at three loci in the three sympatric
species of Caribbean
8–11) withoutAcropora (Fig. 1).Acropora cervicornis
species with fossil records dating back at least 3
to 3.6 My (
and habitat preferences. The arborescent
“staghorn” coral
forereef and backreef habitats, whereas the
robust “elkhorn” coral
in high–wave energy reef-crest habitats
(
over a few nights each summer (
potentially hybridize. The third species,
prolifera
varies from being locally rare to occurring in
large patches (
intermediate between
causing many to consider it a species of
hybrid origin (
high heterozygosity support this possibility
(
high and yet surprisingly discrete. In Puerto
Rico, for example, there are two discrete
prolifera
form we term the “bushy” morph (Fig. 1C), and
a thicker formwith palmate, flattened branches
we call the “palmate” morph (Fig. 1D).
We obtained sequence data for the Caribbean
and A. palmata are sister12, 13). Both have distinct morphologiesA. cervicornis occurs throughoutA. palmata occurs primarily14, 15). Both species spawn synchronously16) and canAcropora, occurs Caribbean-wide, where it7, 14, 15). It is morphologicallyA. cervicornis and A. palmata,7, 15). Pax-C intron data showing10). Morphological variation in A. prolifera isA.morphs—a thin, highly branchedAcropora
minicollagen and calmodulin genes, and at the
mitochondrial putative control region (
nuclear data indicate that the species
species at introns of the nuclear17). TheA. cervicornisand
and that the morphologically intermediate species
A. palmata are genetically distinctA. prolifera
(F
is actually a first-generation1) hybrid. Acropora cervicornis and A. palmatawere reciprocally monophyletic at minicollagen
(Fig. 2A). All of the
A. prolifera (n “22) were heterozygous at minicollagen, containing
one allele from each of the two species’
clades. The calmodulin data for
A. cervicornisand
B, and B
A. palmata formed three distinct alleles: A,‘ (Fig. 2B). Allele A was exclusive toA. cervicornis
palmata
species, making it either a shared ancestral allele
or an introgressed allele from recent or
historical hybridization. As with minicollagen,
all of the
at calmodulin (A/B
complete heterozygosity of
two nuclear loci strongly suggests that every
individual sampled was a F
Mitochondrial data show that the 45
unique haplotypes form a polytomy with
three clades (Fig. 2C), labeled as haplotypes
A, B, and C. The A and C haplotypes contained
only
The B haplotypes contained all three
taxa:
. B alleles were exclusive to A., but the variant B‘ was shared betweenA. prolifera (n ” 28) were heterozygous” 26; B/B‘ ” 2). TheA. prolifera at these1 hybrid.A. cervicornis and hybrid A. prolifera.A. palmata, A. cervicornis, and hybridA. prolifera
in
occur in both directions. Hybrids receive maternally
inherited mitochondrial DNAs from
either
. All three haplotypes were foundA. prolifera, indicating that hybrid crossesA. palmata (B haplotype) or A. cervicornis(A haplotype) “mothers.”
Although hybrid crosses occur in either direction,
mitochondrial DNA (mtDNA) introgression
appears unidirectional because
A. cervicorniscolonies possess all three haplotype
clades, but
indicate that “
passed to
of
B haplotypes in
common (
The presence of multiple B variants in
A. palmata colonies do not. The datapalmata” (B) haplotypes areA. cervicornis through backcrossingA. cervicornis with hybrid A. prolifera. IntrogressedA. cervicornis were(20%) and sampled at every site.A. cervicornisindicates the mtDNA introgression has
occurred more than once. Because nuclear loci
should sort more slowly than maternally inherited
mtDNAs (
the mitochondrial data but not the minicollagen
data are consistent with recent introgression
rather than incomplete lineage sorting.
In Puerto Rico, we sampled two distinct
morphs of
palmate morphs (Fig. 1, C and D). Although
all individuals, irrespective of morphology,
are F
donated its egg and mitochondrion to the
hybridization event. All bushy hybrids had a
18, 19), polyphyletic patterns inA. prolifera, i.e., the bushy and1 hybrids, they differ in which speciespalmata
whereas all of the palmate hybrids
had a
maternal and mitochondrial background,cervicornis background. This suggestsFig. 2.
putative control region. Likelihood searches were conducted in PAUP* 4.0b8 (
parameters and 25 random-addition heuristic searches with tree-bisection-reconnection branch swapping.
Models of sequence evolution were evaluated on distance-based topologies with hierarchical
likelihood ratio tests (
along major branches indicate substitutions. Sample sizes (alleles or haplotypes) are labeled in
parentheses (
(SC). Bootstrap values (
Maximum likelihood (ML) trees for (A) minicollagen, (B) calmodulin, and (C) mitochondrial31) with estimated model32) inMODELTEST 3.06 (33). Major allele/haplotype clades are labeled. Tick marksn). Site abbreviations: Yucatan ( Y); Panama (Pa); Jamaica ( Ja); Puerto Rico (PR); St. Croix)50%) from 300 replicates are labeled on relevant nodes. The Pacific congenerAcropora nasuta
AF507116 to AF507373). (A) Minicollagen ML tree constructed with a K80 model (ln score
(B) Calmodulin ML tree constructed with a HKY model (1 of 4 trees; ln score
Mitochondrial putative control region ML tree constructed with a F81
Palmate
was used as the outgroup. Sequences are available in GenBank (accession numbers” 654.81).” 592.86). (C)$*model (ln score ” 2014.96).A. prolifera hybrids are shown in blue; bushy hybrids are in red.R
E P O R T S2024
14 JUNE 2002 VOL 296 SCIENCE www.sciencemag.orgDownloaded from
www.sciencemag.org on April 7, 2010that maternal and/or cytoplasmic effects account
for the marked differences in these two
hybrid morphotypes. Thus, coral morphology
appears sensitive to not only nuclear genetic
effects, but also to nuclear-cytoplasmic interactions
within a hybrid nuclear genome.
Differential introgression of loci characterizes
many terrestrial hybridization systems
(
that the pattern is due to ancestral polymorphism.
We applied a two-population Bayesian
coalescent model (
20); however, a rarely explored alternative is21) to our data and the publishedPax
introgression [as migration (
product of effective population size (
migration (
introgression (
tests (LRTs) (
material (
data are consistent with low levels of
introgression (
one haplotype crossing the species boundary
every 5
size) generations. For the nuclear loci, the
-C data (10) to estimate the rate ofM) in units 2 + theNe) andm)] and test null hypotheses of noM ” 0) using likelihood ratio22). Results [Table 1 and supplemental23)] indicate that the mitochondrialM ” 0.20), roughly equivalent toNf (i.e., mtDNA effective populationPax
of introgression (
and calmodulin data were both consistent
with no introgression, suggesting that the
shared B
ancestral allele. Such differential cytoplasmic
and nuclear introgression is consistent with selection
against hybrid genotypes that is thought
to result from selection against nuclear genes in
foreign genetic backgrounds (
breakup of coadapted gene complexes in backcrossed
individuals (
The existence of hybrid
-C data were also consistent with low levelsM ” 0.30), whereas the minicollagen‘ allele at calmodulin is a retained24), and/or the25).A. proliferashows that complete barriers to hybridization
have not evolved between
cervicornis
prolifera
almost entirely of F
the reproductive potential of hybrid
A. palmata and A.. However, the observation that A.hybrid populations are composed1 individuals suggests thatA. proliferais severely limited or that hybrid breakdown
occurs in later generations. Some hybrid
A. prolifera
viable gametes, and are interfertile with
cervicornis
suggests that they are essentially sterile
“mules,” which have little genetic impact on
either parent species. Strict F
often ecologically rare in natural hybridization
systems (
selection manifest as hybrid infertility
or hybrid breakdown has been inferred, as
here (
only when hybridization is frequent or
F
(
clonally by fragmentation (
long-lived, potentially immortal hybrid genotypes.
These “immortal mules” may accumulate
over time, providing the opportunity for
rare backcrosses, and for the ecological persistence
of a diverse suite of
that is greater than the number of
species on reefs.
The Caribbean
corals diversify not only through conventional
species formation, but also through
the unprecedented formation of long-lived coral
hybrid morphotypes. In effect, hybridization,
through the formation of asexual coral hybrid
lines, generates new morphologies and potentially
new ecotypes without speciation. Similar
clonal niche partitioning is known for rare parthenogenetic
taxa (
for an ecosystem-defining group like
reef-building corals. Although it remains to be
seen how pervasive coral hybrid “mules” are,
the variety of intermediate morphologies in corals,
especially in regional endemics and putative
subspecies (
unique hybrids may be common. Because
of the potential for natural hybridization in
mass-spawning corals, the coral reticulate evolution
hypothesis suggested that genetic exchange
between “species” generates discrete
coral morphologies (
Instead, we suggest that reef-building coral
diversity is enhanced by hybridization through
the production of long-lived asexual hybrid
morphotypes, which have little evolutionary
potential.
are reproductive, produceA.. Yet, the limited introgression1 hybrids are26). Where F1 hybrids dominate,27). Such F1 hybrids should be common1 offspring are long-lived. Like many corals28), hybrid A. prolifera can propagate29), allowing forAcropora morphotypesAcropora show that reefbuilding30), but has never been postulated5), suggests that morphologically3) without genetic isolation.References and Notes
1. P. L. Harrison
2. B. L. Willis, R. C. Babcock, P. L. Harrison, C. C. Wallace,
et al., Science 223, 1186 (1984).Coral Reefs
3. J. E. N. Veron,
and Evolution of the Scleractinia
Univ. Press, Ithaca, NY, 1995).
4. C. C. Wallace, B. L. Willis,
237 (1994).
5. J. E. N. Veron,
of Marine Science, Townsville, Australia, 2000).
6.
16, s53 (1997).Corals in Space and Time: The Biogeography(CornellAnnu. Rev. Ecol. Syst. 25,Corals of the World (Australian Institute!!!!,R. Kelley, Assoc. Australas. Palaeontol. Mem.6
7. C. C. Wallace,
of the Coral Genus Acropora
Acroporidae
Morphology, Phylogeny and Biogeography
Scientific and Industrial Research Organisation,
Melbourne, Australia, 2000).
8. M. Hatta
9. D. M. Odorico, D. J. Miller,
(1997).
10. M. J. H. van Oppen, B. L. Willis, H. W. J. A. van Vugt,
D. J. Miller,
11. M. J. H. van Oppen, B. J. McDonald, B. Willis, D. J.
Miller,
12. A. F. Budd, T. A. Stemann, K. G. Johnson,
, 1 (1988).Staghorn Corals of the World: A Revision(Scleractinia; Astrocoeniina;) Worldwide, with Emphasis on(Commonwealthet al., Mol. Biol. Evol. 16, 1607 (1999).Mol. Biol. Evol. 14, 465Mol. Ecol. 9, 1363 (2000).Mol. Biol. Evol. 18, 1315 (2001).J. Paleontol.68
13. A. F. Budd, K. G. Johnson,
Symp.
14. T. F. Goreau,
15. S. D. Cairns, Ed.,
Scleractinia
Institution Press, Washington, DC, 1982).
16. M. de Graaf, G. J. Geertjes, J. J. Videler,
, 951 (1994).Proc. 8th Int. Coral Reef1, 423 (1997).Ecology 40, 67 (1959).Stony Corals (Cnidaria: Anthozoa:) of Carrie Bow Cay, Belize (SmithsonianBull. Mar. Sci.64
17. A total of 131 individuals were sampled across five sites
in the Caribbean; samples per site and species (
A. palmata
as follows: Yucatan
Jamaica
9 palmate); St. Croix
a CTAB (hexadecyltrimethylammonium bromide) buffer,
proteinase K (100
extraction methods. Amplifications were obtained
with GeneAmp XL PCR kits under normal polymerase
chain reaction (PCR) conditions, 30 to 35 cycles, and
annealing temperatures of 51° to 54°C. A 374–base
pair (bp) fragment of minicollagen, including the second
intron, was amplified with published primers. A calmodulin
intron (343 bp) was amplified with coral-specific
primers CalMf (5
and CalMr2 (5
mitochondrial putative control region (933
83 bp of cytochrome oxidase III was amplified with
primers CRf (5
3
3
CATAGTGAGGGTGAGGGAACTGGC-3
(5
were sequenced directly; heterozygous nuclear
alleles were observed as double peaks confirmed in
samples sequenced in both directions.
18. P. Pamilo, M. Nei,
19. S. R. Palumbi, F. Cipriano, M. Hare,
(2001).
20. J. C. Avise,
Evolution
21. R. Nielson, J. Wakeley,
22.
separate populations. Models were run independently
for each gene. The mode of the integrated posterior
probability distribution for the migration parameter
(
introgression. LRTs compared probabilities of
, 189 (1999).A. cervicornis,, and A. prolifera, respectively) were” 3, 3, 3; Panama ” 7, 5, 0;” 3, 4, 0; Puerto Rico ” 41, 22, 19 (10 bushy,” 0, 12, 6. DNA extractions used,g), and standard phenol-chloroform‘-GAGGTTGATGCTGATGGTGAG-3‘)‘-CAGGGAAGTCTATTGTGCC-3‘). The$ bp) plus‘-GCTTAGACAGGTTGGTTGATTGCCC-‘) and CO3r (5‘-CTCCCAAATACATAATTTGAACTAA-‘), and two internal sequencing primers, CRseqf (5‘-‘) and CRseqr‘-ATAACCCAACAAAGTCTAATTCCC-3‘). AmplificationsMol. Biol. Evol. 5, 568 (1988).Evolution 55, 859Molecular Markers, Natural History and(Chapman & Hall, New York, 1994).Genetics 158, 885 (2001).Acropora cervicornis and A. palmata were treated asM in units of 2Nem) was used to estimate the rate ofM versusM
polymorphism.
Multiple simulations confirm model convergence
[supplemental material (
the best values of
10 for both parameters.
23. Supplementary material is available on
at www.sciencemag.org/cgi/content/full/296/
5575/2023/DC1.
24. N. Takahata, M. Slatkin,
(1990).
25. R. S. Burton,
26. M. L. Arnold,
" 0 to test the null hypotheses of ancestralP values were divided by 2 after (21)23)]. The model searched forM and T within the bounds of 0 toScience OnlineTheor. Popul. Biol. 38, 331Evolution 47, 1814 (1990).Natural Hybridization and Evolution(Oxford Univ. Press, Oxford, 1997).
27. J. D. Nason, N. C. Ellstrand, M. L. Arnold,
Am. J. Bot.79
28. R. C. Highsmith,
29. A. Bowden-Kerby,
2063 (1997).
30. R. C. Vrijenhoek, in
, 101 (1992).Mar. Ecol. Prog. Ser. 7, 207 (1982).Proc. 8th Int. Coral Reef Symp. 2,Population Biology and EvolutionK. Wohrmann, V. Loeschecke, Eds. (Springer-Verlag,
Berlin, 1984), pp. 175–197.
31. D. L. Swofford,
Parsimony
Associates, Sunderland, MA, 1996).
32. N. Goldman,
33. D. Posada, K. A. Crandall,
(1998).
34. Thanks to those who assisted with field collections;
to R. Nielson and J. Wakeley for assistance with the
coalescent modeling; and to P. Barber, S. Belliveau, E.
Weil, and four anonymous reviewers for valuable
comments. Photographs were provided by H. Ruiz.
Supported by National Science Foundation grants
(S.R.P.) and a NIH Genetics Training Grant Fellowship
(S.V.V.).
3 January 2002; accepted 3 May 2002
PAUP*: Phylogenetic Analysis Using(and Other Methods), version 4.0. (SinauerJ. Mol. Evol. 36 (1993).Bioinformatics 14, 817Table 1.
of the Bayesian coalescent modeling for each gene
showing the estimated rates of introgression (
2
tests (LRTs). NS, not significant;*
Estimated genetic introgression. ResultsM inNem units) and the results of the likelihood ratioP ” 0.05; **P “0.01.
Gene 2
Nem LRT PMinicollagen 0.00 0.00 1.000 (NS)
Calmodulin 0.08 2.17 0.071 (NS)
Pax
MtDNA control region 0.20 4.31 0.019*
-C 0.30 6.02 0.007**R
E P O R T Swww.sciencemag.org SCIENCE VOL 296 14 JUNE 2002
2025
