Important questions in the study of evolution are whether the characteristics of organisms are mutually independent or behave and evolve as integrated systems or modules, and whether the integration of modules can be shifted, switched or canalized along lineages 1. In modular colonial organisms, which form by the reiteration of identical units, the emergent forms are usually complex networks 23. Colony form in these invertebrates is a consequence of modular (polyp) replication, and if there is tight integration (e.g. pleiotropy or linkage disequilibrium) among modular and supra-modular traits (e.g. polyp aperture, inter-polyp distance or spacing, branch thickness, internode and branch length), then changes at the module level may lead to changes in colony architecture. Alternatively, different groups of traits may change semi-independently (or conditionally independently) if pressures or developmental mechanisms differ. It has been found that all characters are mutually associated, integration being strongest among the colony level (network) characters, which suggests that branching characters within colonial organisms can be independent of module characters in interspecies comparisons 4. Nevertheless, the integration or decomposition of modular features at the intraspecific level is unknown in marine modular organisms.

A particular attribute of gorgonian coral colonies is the existence of different levels of nested modularity, such as various types of microscopic sclerites (length variation 0.1–1 mm), polyps that always have eight pinnate tentacles (1–20 mm), and modular branches placed at nearly fixed internodes (1–50 cm), which are all repetitive modules but morphologically fixed features throughout the colony (e.g., 567). Comparative analyses among several Caribbean gorgonians suggest decoupling of evolution at the polyp vs. branch levels 4. Consequently, evolutionary change in polyp form or size does not imply a change in colony form, or vice versa. These characteristics of marine modular invertebrates are consistent with the modular concept of phenotypic plasticity widely demonstrated in plants 8. In addition, developmental controls of colony form in gorgonian corals seem different from the traditional axial patterning of most unitary metazoans but are related to allometric changes in growth rate and reproductive onset or heterochrony 27. These features of astogenetic growth also accord with the view of developmental reaction norms as whole-organism allometric functions (e.g. 9), which reveals once again that phenotypic plasticity should be considered a feature of modular traits 8. Could the modular concept of phenotypic plasticity explain the morphological integration of modularity in marine invertebrates such as corals?

Pseudopterogorgia bipinnata (Verrill) (Octocorallia: Gorgoniidae) is a shallow water Caribbean gorgonian coral that can colonize most coral reef habitats. In a single location such as Carrie Bow Cay, Belize barrier reef, it is found continuously from wave-swept environments as shallow as 1 m to the deepest portions of the reef (~ 45 m) where there is very low water movement and less light (Fig. 1), as well as inshore lagoonal habitats such as sea grasses and mangrove roots in the Pelican Cays. In addition, the species ranges from purple and yellow coloration in lagoonal habitats to gray and beige in exposed fore-reef environments 9, and can adopt bushy, fan- and feather-like forms with over tenfold differences in branch length towards its environmental extremes. This explains why this species has been considered a taxonomic complex (e.g. 109). To integrate findings on the evolution of modularity in gorgonian corals at the intraspecific level, the aims of this study were: (1) to ascertain, using genetic information, whether the different P. bipinnata "complex" phenotypes belonged to the same species; (2) to quantify the variation of different modular traits in different environments; and (3) to determine the morphological integration and decomposition of the same traits in relation to the plasticity of the species.

The Pseudopterogorgia bipinnata complex provided a good model for exploring phenotypic plasticity and morphological integration at the intraspecific level. P. bipinnata did not show the same pattern of morphological integration found among species, which suggests tight morphological integration of all traits and decoupling of evolution at the levels of polyp (the colonial module sensu stricto) vs. branch/internode (an emergent modular feature produced by colonial growth) 3. Some features did not change at all despite 10-fold differences in other integrated features; more importantly, some colonial features showed dependence on modular features. The gorgonian corals Pseudopterogorgia bipinnata and P. kallos apparently belonged to the same species (at least phylogenetically), exhibiting remarkable plasticity, particularly in branch and internode lengths. Molecular evidence from nuclear and mitochondrial DNA sequences, as well as two microsatellite loci and ITS2 DGGE banding patterns, corroborated the view that the morphological differences corresponded to phenotypic plasticity. This finding differs from those in other corals (e.g. fire coral, Millepora dichotoma), where the different morphotypes found in different environments are associated with different fixed molecular patterns rather than phenotypic plasticity 11.

Allometric change during growth is a characteristic of gorgonian coral colonies. It is known that gorgonian colonies tend to spread their branches against unidirectional flow, forming a 2-dimensional network even after transplantation 1213. In essence, most species vary from 3-D bush-like colonies in multidirectional flows to fan-like 2-D networks in environments with unidirectional flow (e.g. 14). Other colonial traits such as axial core composition and stiffness differ considerably within species owing to the preferred level of water movement 15. This is related to gorgonian community structure, where many species overlap totally or partially along an environmental continuum in wave exposure and depth 9. This change in growth forms and preferences of species may be interpreted as allometric changes that enable resources to be exploited optimally 16. The observed change in P. bipinnata, in contrast, did not seem to follow a simple continuum allometric pattern but seems to support the modular concept of phenotypic plasticity 7. An integrated allometric change that produced the nested modularity of gorgonian corals would have to include changes in most traits; but it was clear that some traits changed less than others and one trait, inter-polyp distance, did not change at all across environments.

Traits may differ in their evolutionary constraints, but the rate of change of any one trait may or may not depend on others. For gorgonians, the pattern of phenotypic integration at the intraspecific level showed the same interdependence of branches and internodes as found at the interspecific level, but also showed an integrated loop among branch length and minute "polypar" traits such as branch thickness and polyp aperture. This was not wholly inconsistent with the interspecific pattern of decoupled evolution between the "colonial" vs. "polypar" traits 3 because despite a more than tenfold change in branch and internode lengths across habitats, interpolyp distance did not correlate (e.g., zero-order PCC) with any other feature, suggesting a decoupled mechanism of change for this trait. The link between micro- and macro-evolutionary changes in the shape of modules and colonies in gorgonian corals may possibly depend on which traits are independent and which are integrated.

When a species trait is invariant in different environments while others show different phenotypic responses, there may be canalization of this particular phenotype. However, it is difficult to determine the level of canalization since different factors may interact from gene expression to morphology, and life history traits to behavior 17. If the species Pseudopterogorgia bipinnata is divided in two lineages by a geographic barrier, for instance, their descendants could include short and long branched colonies that all have the same inter-polyp distance. On the other hand, inter-polyp distance might undergo a process of decanalization in which plasticity starts to occur under new environmental challenges. It is suggested that canalized traits can attain variability depending on the degree and conditions driving this event 17. In addition, phenotypic plasticity for some traits can be an adaptive response to new selective pressures (e.g. 1819). The interplay between canalization and morphological integration might be the link from phenotypic plasticity to what we perceive as macroevolution.

One indication of the link between micro- and macro-evolutionary changes could be the interplay among plasticity and integration with genetic variability and assimilation (e.g., 20, but see 21). The link between micro- and macro-evolutionary changes in the branch and polypar features of P. bipinnata seems to be related to constraints on change in some already-fixed features, but implies integration of changes in branch length with other unconstrained traits: inter-polyp distance, which varies widely among species 3, was constant at the intraspecific level. Moreover, other modular traits, apparently not integrated at the macroevolutionary level, became correlated with branch length at the intraspecific level (Fig. 4).

The three main phenotypes did not exhibit clines and were found in overlapping zones. Although the change of phenotype was apparently independent of genotype, two phenotypes were seldom seen side-to-side sharing the same environment (Fig. 1). This could mean that although this species has characteristic phenotypes at certain depths, there were exceptions where the ancestral, derived or intermediate phenotype could co-exist with a phenotype that was not expected to occur in these conditions. Speculatively, there could be maternal-specific developmental "signals" determining phenotype in marginal environments; or, alternatively, it could be an incipient case of genetic assimilation. It is also important to mention that phenotypic plasticity can show variations in which the population mean and variance change 22. Therefore, the presence of two morphotypes at the same depth may just be some type of variation occurring in some locations. It would be interesting to have more estimates of these overlapping zones so that more precise conclusions could be drawn. It would be even more desirable to place experimental colonies at in-between depths to see whether there is plasticity, as seems to be the case, or whether a phenotype becomes fixed. Nonetheless, more refined molecular and experimental approaches are needed to evaluate genotypic vs. phenotypic variance.

Module integration within gorgonian coral species can be shifted, switched or canalized along lineages. Marine modular organisms such as corals are variations on a single theme: their modules can couple or decouple, allowing them to adapt to all marine benthic environments. Adaptation to a certain environment does not always mean that the trait is fixed 18. Ecotypes can make some traits exhibit phenotypic plasticity but are not so extreme as to fix these traits in the species genome (see discussion in 21). Nonetheless, genetic assimilation 2023 can be a viable event for octocorals, where habitat seems to be a conditioning factor for niche separation. Species that are geographically widely distributed, such as P. bipinnata, also tend to be widely distributed across reef habitats 9. This trend has been found in many groups of terrestrial organisms (e.g. 24). Even so, there is no clear evidence of such a pattern in marine organisms. Phenotypic plasticity costs have been studied and it has been shown that the type of response by the organism determines its viability (e.g. the response of enzymes to the environment has little energetic cost in Stellaria longipes plants: 25; see 26 for plasticity cost models). It remains unclear whether is it better for P. bipinnata to fix a certain phenotype or express it depending on the environment; the costs and benefits of this plasticity have yet to be defined.

Genetic differences among Pseudopterogorgia bipinnata morphotypes and congeners were initially estimated using mitochondrial and nuclear DNA sequences as well as two microsatellite loci allelic variation. Collected specimens were prepared for sclerites observations with compound microscope and a few sclerites were observed with Scanning Electron Microscopy (see methods in 27). Mitochondrial sequences comprised the 5' end of msh1 gene using primers by 2829 and the nuclear sequences comprised a portion of the tandem array of ribosomal genes including complete ITS1, 5.8S and ITS2 sequences using primers by 30. DNA extractions were made according to 31 protocol using CTAB from alcohol preserved material (Ethanol 95%). PCR conditions were followed 4 and sequencing as in 32 using the Edge Biosystems kit and sequencing using BigDye 3.1 (AB 3100, capillary electrophoresis automated sequencer). The matrix editing, translation, and alignment for the diverse genes were accomplished using Bioedit 34 and ClustalW 35. Phylogenetic analyses were carried out in PAUP* 36 coupled with Modeltest for maximum likelihood analyses 3637. New sequences were deposited in GENBANK (accession nos. EU04312–EU043127). Predicted RNA secondary structures were modeled for the portion corresponding to the ITS2 following the methods from 33.

Two heterologous microsatellite loci, Pel74 and Pel1, developed for the sister species Pseudopterogorgia elisabethae Bayer by 38 were used to assess finer genetic differences among morphotypes. For that purpose, 4 colonies from the Bahamas, 8 from Bocas del Toro (Panama), and 30 from different habitats around Carrie Bow Cay, Belize were screened. DNA extractions we made as above and PCR conditions as suggested by 38. Microsatellite alleles were visualized using an AB3100, capillary electrophoresis automated sequencer, with fluorescently labeled primers.

In addition, a study of the ITS2 variation among morphotypes and locations was done for 32 colonies from Belize (14), Panama (12), and Colombia (6). The ITS2 region was amplified using primers and protocols from 33. PCR reactions were separated in a Denaturing Gradient Gel Electrophoresis-DGGE (ABS) containing 8% polyacrilamide, 1× TAE Buffer and linear denaturing gradient from 40% to 80% of Urea and Formamide. Before electrophoresis the gels were pre-run at 60°C and 150 V for 20 min, followed by the electrophoresis at 60°C and 150 V for 9 h. The gels were stained with Ethidium Bromide during 15 min and analyzed using a BIORAD chemidoc system. All the reactions were conducted without CG-clamp.

Plasticity and morphologic integration in Pseudopterogorgia bipinnata were studied using the same traits and methods as in 3 for 30 colonies collected in the vicinity of Carrie Bow Cay, Belize, between 0.2 and 45 m of water depth (same colonies used for microsatellite loci above). Measurements (10 per colony, 30 colonies, 5 variables), included branch thickness, polyp aperture, inter polyp distance, internode and branch length. Patterns of morphological integration were examined as proposed by 1 using third-order Partial Correlation Coefficients (PCCs) that were determined among the studied variables 3.

JAS collected all the data, did the analysis and wrote most of the manuscript. CA contributed with part of the writing and discussion as well as with the phylogenetic analyses. DD and NM did all the DGGE analyses and contributed with part of the writing. All authors read and approved the revised manuscript.