Mangrove communities are critically important ecosystems that are high in aquatic and terrestrial biological diversity in tropical and subtropical ecosystems worldwide
[2, 5]. Today they are threatened by high rates of anthropogenic disturbance, including habitat destruction, pollution, fragmentation, and changes in oceanic and estuarine environments due to climate change
[9, 10, 34]. The goal of this study was to document spatial genetic structure of two dominant Neotropical mangrove species at three spatial levels (1) among four estuaries in Panama (2) Between two estuaries from same coastal line and (3) within each one of the four estuaries. These data provide critical information to understanding how genetic diversity is structured and maintained within mangrove species and communities.
Both mangrove species showed a strong genetic break across the CAI. However, the patterns of diversity observed in this study were the opposite of what we had expected for both species. We found significant differences between estuaries from same coastline and also IBD within estuaries in both mangrove species. In addition, both mangrove species showed comparable outcrossing rates, contrary to other reports on their mating systems. Further, A. germinans showed much higher levels of genetic diversity, especially in plastid genomes, than R. mangle, in spite of the fact that R. mangle populations have a more continuous distribution. Finally, although there is documented evidence for extreme LDD in both mangrove species, our evidence mostly indicates restricted gene dispersal overall and largely equivalent rates of seed and pollen dispersal. Below, we interpret the observed population genetic differences of these species as well as the combined effects of species’ life histories, gene dispersal limitation, and biogeographic history.
The ecological imprint on genetic structure: mating system and Pollen vs. seed movement
The two mangrove species showed strong genetic structure across four estuaries analyzed, including evidence of substructure and IBD within estuaries. Nevertheless, the patterns of genetic structure were very different between species. Microsatellites revealed lower gene diversity and lower genetic structure in R. mangle than A. germinans. This contradicts the predictions based upon an outcrossing mating system where assumed outcrossing species, such as A. germinans, are expected to have lower genetic structure than the mixed-mating R. mangle[33, 35].
The protandry reported in A. germinans is the main evidence that supports an outcrossing mating system in this species because the flower-developing mechanism makes autogamy unlikely (i.e. within-flower pollination)
. However, absence of self-pollination (autogamy) is not equivalent to self-incompatibility because pollination could ocurr between different flowers on the same plant (geitonogamy). Other Avicennia species from Indo-West Pacific region including A. marina and A. officinalis show this pattern
[36, 37]. There is no direct evidence of self-incompatibility in A. germinans that we know of. It is possible that the patchy spatial distribution of A. germinans populations observed in Panama has an effect on the levels of outcrossing and therefore on the genetic structure in this species. In isolated individuals or low density populations geitonogamy could be an advantageous breeding system
. Moreover, the r=−0.64 estimates of pollen vs. seed flow suggests that although seed dispersal is likely equivalent to pollen dispersal in A. germinans, in a patchy matrix, this seed dispersal still is usually very local, leading to greater biparental inbreeding, and increasing population genetic differentiation and spatial genetic structure in nuclear genomes
Our data also suggest that historically ambophilous pollen dispersal mechanism of R. mangle has been more efficient promoting outcrossing and long-distance gene flow than entomophilous pollination system of A. germinans. Based on previous genetic studies, it is predicted to have higher dispersal potential and therefore low genetic structure in species with wind pollination system over species with insect pollinator system
. However, reproductive biology studies in other ambophilous species suggest a completely opposite trend indicating that in self-compatible species as is the case of R. mangle, wind is actually the mechanism that promotes selfing and that out-crossing is associated with insect pollen distribution. The reason is that abiotic mechisms as wind do not target distant receptive flowers as efficiently as insects
Although we do not know the rates of wind-to-insect pollen dispersal in R. mangle, the ratios of pollen-to-seed dispersal of r=7.7 estimated in this study are lower than those estimated in exclusively wind-dispersed plants (r=17
 and r=200
). In addition, based on the Hamilton and Miller’s method calculations, pollen vs. seed was not significantly different from each other meaning that ambophilous pollen dispersal is less efficient than exclusively anemophilous pollen dispersal and/or that seed dispersal in this species is comparatively higher than in other exclusively wind pollinated species.
The contrast of seed viability between the two species may also have a strong influence in levels of genetic structure. Rhizophora mangle species have the highest longevity seeds of any mangrove genus
[15, 29]. Direct experiments regarding establishment of seeds after long periods of floating exposure in sea water have showed a 60% success rate after 247 days floating for R. mangle, higher than any other Rhizophoracea species
. There is no similar quantitative data on establishment success after dispersal for A. germinans; however, its seed longevity is shorter than R. mangle. The local genetic structure observed in Panama largely corroborates this pattern, especially in the Caribbean, where two estuaries separated by 300 km showed identical chloroplast and nuclear genetic diversity in R. mangle but strong structure in A. germinans. Although A. germinans showed evidence of shared cpDNA haplotypes among estuaries on the same coast, there are also some cpDNA haplotypes and nuclear alleles that are restricted to each estuary, generating strong structure, even within estuaries. Thus, in a variable estuarine environment where seed movement is stochastic
, our data suggests that higher propagule longevity leads to a greater chance of successful establishing at long distances, increasing gene flow and decreasing population structure
The historical imprint on genetic structure
The Isthmus of Panama represents a 20 My to three My old barrier to seed gene flow between the Atlantic and Pacific oceans, and is the narrowest terrestrial area in the New World that separates mangrove populations in each ocean
[47, 48]. Based on microsatellite and chloroplast genetic diversity observed across the Isthmus, this land mass has created high levels of genetic structure and a strong barrier for contemporary seed gene flow. Moreover, our results indicate that the Isthmus also represents a strong barrier to pollen flow in each species. This is likely due to the absence of a continuous terrestrial population that spans the entire distance between coasts. Thus, the strong isolating effect that the rise of the CAI has had on population differentiation for these two and other species highlights the role of restricted seed dispersal in creating spatial genetic structure in hydrochorious marine species
Population differentiation observed between different oceans is exceptionally high compared to other tropical tree species. For example, Dick and Huertz
 report an average F
= 0.14 for microsatellites variation from the Neotropical tree Symphonia globulifera. Within Panama, S. globulifera averaged F
= 0.11 for samples taken across the Isthmus of Panama, which is within the range of observations that we see when comparisons are made within oceans for the mangrove species here. However, even trans-Andean F
between Mesoamerican populations of bird pollinated, mammal dispersal S. globulifera and those in the Amazon separated by > 3,000 km showed maximum F
= 0.27, still lower than lowest pairwise F
for A. germinans made between trans-Isthmian populations of Montijo Gulf and Costa Arriba (F
= 0.31) separated by < 100 km. Furthermore, populations of insect pollinated and wind dispersed mahogany (Sweitenia macrophyla) across Mesoamerica also show a lower overall F
= 0.10, with the largest pairwise differences (F
= 0.238) observed between Panamanian and Guatemalan populations at a distance of > 1,600 km
In spite of similar effects of CAI in the genetic structure of these two mangroves, we found unexpectedly high levels of cpDNA diversity and structure in A. germinans that suggests a level of diversity that could be more influenced by population history and demography than current gene flow
. The cpDNA diversity observed in A. germinans is remarkable given the small sample size and short geographic distances separating the populations both within and between coasts. This high level of diversity could be indicative of historical processes determining spatial genetic structure of populations. Although the fossil record of mangroves in Pleistocene is scarce and therefore the reconstruction of current mangrove distribution is very speculative, several lines of evidence suggest that mangrove ecosystem were under episodic crises during the Quaternary, specially associated to sea-level and temperature/humidity fluctuations
[55–57]. In particular, it is possible that current populations of A. germinans represent remnants of refugial populations created during the Pleistocene
. In fact, Panamanian populations are genetically diverse compared to other regions, and it has been suggested that both ancient introgressive hybridization and secondary contact between A. germinans and its sister species A. bicolor has occurred in Panama (especially on the Pacific side), generating a hotspot of genetic diversity
Our results in A. germinans contrast strongly with R. mangle, where very little cpDNA diversity was observed. These two species differ in their density and distribution, with R. mangle forming extremely dense, continuous forests near to shore, and A. germinans forming patchily distributed, lower-density stands at low and middle intertidal zones. One hypothesis that could explain the current distribution of cpDNA diversity is that mangrove populations represent relicts of much larger ancestral populations but that after Pleistocene-Holocene sea-level fluctuations, the mangrove composition shifted to a R. mangle -dominated community
[55, 57, 59]. Under this scenario, R. mangle could have resulted in a more efficient colonizer than A. germinans follow a stepping-stone dispersion pattern. This process combined with self-fertilization observed in R. mangle could be the responsible of the current continous and dense populations and the low cpDNA and nuclear diversity observed in this species
. Alternatively, it is also possible that only certain older lineages of R. mangle that are well adapted to current conditions survived in the Pleistocene-Holocene sea-level fluctiations and that only those exclusive lineages recolonized available habitats during the Holocene or that longer time of R. mangle presence in neotropics generated more lost of diversity via genetic drift than younger A. germinans.
Regardless of the explanation, A. germinans joins the ranks of tropical tree species in Panama that show complex biogeographic history with disproportionately high levels of cpDNA diversity relative to other parts of their range
. The historical complexity of Panama was also evident when we analyzed the geographical variation in structure with greater population differentiation in Pacific versus Caribbean estuaries for both species in nuclear and plastid genetic markers. In the case of R. mangle, introgressive hybridization between R. mangle and its sister species, R. racemosa, is a novel source of genetic variation exclusive to the Pacific
. Thus, the levels of genetic structures and patterns of biodiversity in R. mangle are completely different between Caribbean and Pacific due to independent historical processes in the occurrence and sympatry of R. mangle and R. racemosa.
The influence of coastal morphology
Hydrochory per se is assumed to be one of the most efficient mechanisms for LDD in plants. Therefore, it is expected that hydrochorious plant species should have high gene flow among populations and low genetic structure, especially in local geographic areas
[61–63]. Our data contradict this hypothesis because the two mangrove species resulted genetically structured among and within estuaries, indicating local restrictions to both pollen and seed dispersal, especially in A. germinans species. Currently, one of the major threats in mangroves is fragmentation and sea-level changes associated with climate changes
[10, 11]. Historically, both mangrove species have experimented sea-level fluctuations at several times and climate changes
[3, 4, 55, 56], thus, our results suggest that historically R. mangle have maintained more gene dispersal of both pollen and seed dispersal than A. germinans. However, our results also suggested that geographic location is important in predicting levels of genetic structure in mangroves. Pacific populations proved to be more structured than Caribbean populations in both mangrove species. One possible explanation is biological. The Pacific coast has been characterized by ancient hybridization between A. germinans and A. bicolor, but there is also the current scenario of introgressive hybridization between R. mangle and R. racemosa. Both hybridization processes are apparently complicating current levels of structure compared with Caribbean populations
[13, 31]. The other possible explanation is abiotic, including basin geomorphology and connectivity due to ocean currents. Our results suggest that current or historical landscape characteristics of Pacific estuaries are in some way enhancing pollen and seed dispersal limitations, generating more structure compared to Caribbean estuaries. Similarly, the density of A. germinans is very variable and in some places, for example in French Guiana A. germinans could be more dense and extended than R. mangle. In consequence, patterns of genetic structure among and within estuaries in that region could be completely different to observed in Panama. Thus, although life-history traits are important to predict expected genetic structure, landscape settings are generating a variety of situations on local scales that complicate any prediction in terms of expected levels of genetic structure
. Therefore, although long distance gene flow between South America and West Africa has been observed in both A. germinans and R. mangle[12, 13], estuarine geomorphology and ocean currents in Panama, especially on the Pacific side, seem to be more complex, preventing pollen and seed dispersal.