The Genus Cerion (Gastropoda: Cerionidae) in the Florida Keys

The systematic relationships and phylogeography of Cerion incanum, the only species of Cerion native to the Florida Keys, are reviewed based on partial sequences of the mitochondrial COI and 16S genes derived from 18 populations spanning the range of this species and including the type localities of all four described subspecies. Our samples included specimens of Cerion casablancae, a species introduced to Indian Key in 1912, and a population of C. incanum x C. casablancae hybrids descended from a population of C. casablancae introduced onto Bahia Honda Key in the same year. Molecular data did not support the partition of C. incanum into subspecies, nor could populations be apportioned reliably into subspecies based on morphological features used to define the subspecies. Phylogenetic analyses affirmed the derived relationship of C. incanum relative to other cerionids, and indicated a Bahamian origin for the Cerion fauna of southern Florida. Relationships among the populations throughout the Keys indicate that the northernmost populations, closest to the Tomeu paleoislands that had been inhabited by Cerion petuchi during the Calabrian Pleistocene, are the oldest. The range of Cerion incanum expanded as the archipelago that is the Florida Keys was formed since the lower Tarantian Pleistocene by extension from the northeast to the southwest, with new islands populated as they were formed. The faunas of the High Coral Keys in the northeast and the Oölite Keys in the southwest, both with large islands that host multiple discontinuous populations of Cerion, are each composed of well supported clades that are characterized by distinctive haplotypes. In contrast, the fauna of the intervening Low Coral Keys consist of a heterogeneous series of populations, some with haplotypes derived from the High Coral Keys, others from the Oölite Keys. Individuals from the C. incanum x C. casablancae hybrid population inhabiting the southeastern coast of Bahia Honda Key were readily segregated based on their mitogenome lineage, grouping either with C. incanum or with C. casablancae from Indian Key. Hybrids with C. casablancae mitogenomes had haplotypes that were more divergent from their parent mitogenome than were hybrids with C. incanum mitogenomes.

Introduction around the salt ponds on Key West [20], and a neotype designated from this locality [19]. The name Cerion incanum fasciata (Binney, 1859) was proposed for an illustrated variety from Key Biscayne, Florida with irregular brownish longitudinal strips on a brownish base color [21]. Another phenotype with a somewhat larger shell with a tapering spire and frequently costate last whorl was given the name Cerion incanum saccharimeta by Pilsbry and Vanatta, 1899 [22], based on specimens from Sugarloaf Key. An additional thin-shelled form with a thin, narrow peristome and livid color from "Key Vaccas" was designated Cerion incanum var. vaccinum Pilsbry, 1902 [20]. All of these form names have subspecific rank (ICZN Article 45.6.4) [23].
Early authors [24,25,26] considered Cerion incanum to be of Cuban origin, transported to the Florida Keys on floating debris and subsequently dispersing throughout the archipelago from southwest to northeast. Later authors [27] suggested that the Cerion of the Florida Keys were of Bahamian origin. The recent discovery of the Cerion petuchi Harasewych, 2012 [13] from the Bermont Formation (~1.6 M YBP) of Palm Beach County, a species most closely related to C. agassizii Dall, 1894, from Pleistocene deposits of the Great Bahama Bank, indicates Bahamian affinities for the ancestor of Cerion incanum. A phylogeny of Cerionidae based on partial COI sequences [16] indicates a well-supported relationship between C. incanum and species of Cerion from Andros Island and Eleuthera, both on the Great Bahama Bank. This suggests that the populations of the High Coral Keys are the oldest, and have the oldest genotype. As the archipelago that is the Florida Keys was formed during the Pleistocene by extension from the northeast to the southwest, the range of Cerion incanum expanded in a southwesterly direction, populating the new islands as they were formed. The genetic relatedness among the populations of the various keys is expected to reflect this pattern.
In the present study, we use partial sequences of the mitochondrial cytochrome c oxidase I (COI) and 16S rDNA genes to investigate genetic variation among and within populations of Cerion inhabiting the Florida Keys in order to infer the relationships among them, and to gain insights into biogeographic patterns within the Keys as well as the interactions between native and introduced species.

Sample Collection
Neither Cerion incanum nor Cerion casablancae are endangered or protected species. Populations of Cerion were sampled from 18 locations between Key Biscayne and Key West (Table 1, Figs 1 and 2). Sampling sites were selected to span the known geographic range of the species based on historical museum records, and to include the type localities for each of the four named subspecies of Cerion incanum [C. incanum incanum (Leidy, 1851)-Key West; C. incanum fasciatum (Binney, 1859) [21]-Key Biscayne; C. incanum saccharimeta (Pilsbry, & Vanatta, 1899) [22]-Sugarloaf Key; C. incanum vaccinum (Pilsbry, 1902) [20]-Vacca Key] that had been proposed on the basis of shell morphology. Our sample from Indian Key is comprised of the descendants of 500 specimens of Cerion cassablancae Bartsch, 1920 [14], that were introduced to the Key on June 1, 1912, after a survey determined that no native Cerion were present on the island [28]. The following day, Bartsch introduced an additional 500 individuals of Cerion cassablancae onto the eastern portion of Bahia Honda Key, where they survived for multiple generations [14,28]. During a regular survey of this colony on September 23, 1931, Bartsch [29] was surprised to discover Cerion incanum on Bahia Honda Key, as well as the presence of hybrids between Cerion incanum and C. cassablancae near the site of his introduction. Our sample from the southwestern portion of Bahia Honda Key is comprised of Cerion incanum, while specimens from the southeastern shore of Bahia Honda are a sample of the hybrid population [Cerion incanum X C. cassablancae].
Samples were collected along the sides of public roads or along public beaches, not on private property. No specific permissions were required for these localities. Specimens were collected at each location, generally on and under low vegetation and rocks in close proximity to water, and transported living to the laboratory at the National Museum of Natural History, Smithsonian Institution. All specimens collected for this study are part of the collections of the National Museum of Natural History, Smithsonian Institution. Catalog numbers are provided in Table 1 [http://collections.nmnh.si.edu/search/iz (search by catalogue number in Table 1)].

DNA extraction, PCR amplification and sequencing
DNA was obtained from a subset of the individuals (n = 6-31) from each population by removing the dorsum of the shell using a Wizard Model 100 Saw with a fine diamond blade and dissecting half the buccal mass (~20 mg) from the animal. Genomic DNA was extracted using an Autogen GENE PREP965 (Autogen) employing the manufacturer's mouse-tail tissue protocol with a final elution volume of 100 μL. The shells were retained as voucher specimens and individually identified with a number suffix to the USNM catalog number assigned to each population. [Images of the sequenced specimens are appended to sample catalog records: http:// collections.nmnh.si.edu/search/iz/ (search by catalogue number in Table 1) and also appear on Table 1. Florida Keys populations of Cerion sampled for this study. The numbers of individuals sequenced from each population for the mitochondrial cytochrome c oxidase 1 and 16S rDNA genes, the numbers of different haplotypes within each population for each of the partial gene sequences, as well as for the concatenated sequences are provided. Samples from the type localities of named taxa are in bold. GenBank accession numbers for the various haplotypes of Cerion incanum, C. cassablancae and their hybrids are provided in Appendix 1.  Table 1]. Portions of two mitochondrial genes were amplified: a 655 bp region of the cytochrome c oxidase I gene using the primers HCO1490 AND LCO2198 [30] and a 510 bp region of the 16S ribosomal gene using the primers 16S-ar and 16S-br [31]. PCR amplifications were performed in 20 μL volumes that contained: 12.9 μL ddH 2 O, 2 μL 10x NH 4 reaction buffer (Bioline) 1.2μL of 50mM MgCl 2 (Bioline), 1 μL dNTPs, 0.5 μL of 100x BSA (BioLab), 0.6 μL of 10μM forward primer (Integrated DNA Technologies, Inc.), 0.6 μL of 10μM reverse primer (Integrated DNA Technologies, Inc.), 0.2 μL of Biolase DNA polymerase (Bioline) and 1 μL of extracted genomic DNA. For COI, the amplification protocol for the thermocycler (Applied Biosystems, Life Technologies) was as follows: initial denaturation at 95°C for 3 minutes, followed by 45 cycles of denaturation at 94°C for 30 seconds, annealing at 45°C for 45 seconds and extension at 72°C for 2 minutes, with a final extension at 72°C for 5 minutes. For the 16S rDNA gene: initial denaturation at 95°C for 3 minutes, followed by 40 cycles of denaturation at 95°C for 45 seconds, annealing at 48°C for 45 seconds, and extension at 72°C for 2 minutes, with a final extension at 72°C for 5 minutes. Resulting PCR products were visualized by agarose gel electrophoresis (1.5% agarose) and purified with ExoSAP-IT (Affymetrix) according to manufacturer's protocols, prior to sequencing.

COI 16S Concatenated
Sequencing reactions were performed in 10 μL volumes that contained: 6.25 μL ddH 2 O, 1.75 μL 5x sequencing buffer, 0.5 μL primer, 0.5 μL Big Dye, and 1 μL of purified PCR product, and run in the thermal cycler for 30 cycles of 95°C for 30 seconds, 50°C for 30 seconds, and 60°C for 4 minutes, then held at 12°C. Reaction products were purified using Millipore Sephadex plates (Millipore, Billerica, MA) using manufacturer's protocols and sequenced on an ABI 3730XL automated DNA Analyzer (Applied Biosystems and Hitachi, Ltd.). Geneious 7.1.2 (Biomatters, Ltd.) was used to visualize, trim, edit and assemble contigs from forward and reverse sequences. Individual sequences are identified by the Station letter designation ( Table 1) followed by the specimen number (e.g., KB3 refers to the sequence derived from specimen 3 from the Key Biscayne population and refers to specimen 3 of USNM 1201700). The program TCS v. 1.21 [32] was used to sort identical sequences for each gene into haplotypes, which are listed in Appendix 1 together with their corresponding specimens and GenBank (NCBI) accession numbers.

Species delimitation
The Automatic Barcode Gap Discovery (ABGD) method aggregates sequences into candidate species based on gaps in the distribution of pairwise differences that correspond to gaps between intraspecific and interspecific divergences [33]. It is one of several methods for species discovery and delimitation using molecular techniques recently evaluated [34] and found to be among the more conservative and objective.
Samples are parsed into a decreasing number of groups in a series of recursive partitions. The ABGD method (http://wwwabi.snv.jussieu.fr/public/abgd/; default parameters) was applied to a matrix of pairwise distances (Jukes-Cantor model, pairwise deletion of missing data) produced for the COI and 16S datasets using MEGA v. 5.05 [35] in order to evaluate hypotheses of species delimitation.

Phylogenetic analyses
Alignments of COI and 16S were obtained using MAFFT [36] for 16S and MUSCLE (Multiple Sequence Comparison by Log-Expectation) [37] for COI. The aligned sequences were concatenated using Geneious version 7.1.2 [38]. In order to expedite the analyses, the program TCS v. 1.21 [32] was used to identify and collapse individual concatenated sequences at each locality. In these analyses, collapsed sequences are identified by an exemplar sequence followed by an "x" and the number of individuals with that identical sequence (e.g., NNK1 x 15 indicates that there are 15 specimens with a concatenated sequence identical to sample NNK1).
A best-fit model of nucleotide sequence evolution (compatible with MrBayes) and partitioning arrangement for each locus was determined by PartitionFinder v.1.1.1 [39,40] using the "greedy" search and AICc scheme options. The GTR+G model was chosen for 16S, and models for each codon position (1 st : GTR+I, 2 nd : F81, and 3 rd : GTR+G) where chosen for COI.
Phylogenetic analyses were performed on a concatenated dataset (16S+COI) using Bayesian Inference (BI) performed with MrBayes 3.1.2 [41] and Maximum Likelihood (ML) with RAxML [42]. All analyses were run on the Topaz computing cluster, which is maintained by the Laboratories of Analytical Biology at the National Museum of Natural History. BI analysis was carried out for 10 million generations with two independent runs, each with four chains, and with trees sampled every 1000th generation. Model parameters (tratio, statefreq, shape, pinvar) were unlinked among partitions, and the rate prior (prset ratepr) was set to "variable". Convergence was determined when the average standard deviation of split frequencies was <0.01 and the potential scale reduction factor (PSRF) was 1.00. In addition, TRACER v1.5 [43] was used to verify that an adequate amount of trees were sampled from the posterior distribution, and to confirm the stationarity of the runs. To calculate posterior probabilities, a "burnin" of 25% of the total trees sampled per run adequately removed trees prior to convergence. ML options for RAxML included the GTRGAMMA model of nucleotide evolution (-m), rapid bootstrap analysis and search for best-scoring ML tree (-f m), and 1000 bootstrap replicates.

Haplotype networks
The program TCS v. 1.21 [32] was used to estimate haplotype networks for the COI and for the 16S genes (with outgroups removed) under default settings (gaps treated as a fifth base, 95% confidence interval for statistical parsimony). Correlations between haplotypes and the specimens at the various localities are provided in Appendix 1.

Results
Specimens from each of the sampled populations (Table 1, Figs 1 and 2) were sorted to the named subspecies of Cerion incanum (Table 2) based on a comparison of their shells to typical specimens of the respective subspecies (Fig 1A-1D). In many cases, these attributions were provisional, as the phenotypes exemplified by the type specimens are extreme. Morphological features on which the subspecies are based are less pronounced in the majority of specimens. Many individuals have phenotypes that are intermediate between subspecies. Some populations have significant proportions of three (BHSW) or all four phenotypes (BPSW).
Sequences for portions of the COI and 16S rDNA genes were obtained for 450 individuals of Cerion incanum spanning the entire geographic range of this species, 30 individuals from a C. incanum x C. casablancae hybrid population along the southeastern coast of Bahia Honda Key, and 6 individuals of Cerion casablancae from the introduced colony on Indian Key (Table 1).
Only the population from the southwestern portion of Bahia Honda Key (BHSW) was monomorphic for the COI gene, and only the Missouri Key population (MK) was monomorphic for the 16S gene (Table 1, Fig 3). For all Cerion incanum populations spanning the Florida Keys, the maximum sequence diversity for COI was 2.14% within populations, and 2.90% between populations. For the 16S gene, the values were 2.55% within populations and 2.75% between populations. Maximum sequence diversity within the introduced population of Cerion casablancae on Indian Key (IK) was 1.53% for the COI gene and 1.56% for the 16S gene. Differences between Cerion incanum and the Indian Key population of C. casablancae ranged between 6.56% and 8.55% for COI gene and between 4.51% and 6.08% for the 16S gene. Levels of sequence diversity within the C. incanum x C. casablancae hybrid population in the southeastern region of Bahia Honda Key (BHSE) reached 8.09% for COI and 5.88% for 16S, but were clearly bimodal [ranges for COI = 0.00-3.97% and 7.33-8.09% / for 16S = 0.00-2.16% and 4.90-5.88%], corresponding to comparisons between homologous and heterologous mitogenomes respectively.

Species delimitation
Results of Automatic Barcode Gap Discovery (ABGD) analyses of the COI and 16S datasets are shown in Table 3. In both the COI and 16S analyses, the first two partitions segregated nearly all haplotypes as separate groups. All Cerion incanum haplotypes were combined into a single group in the third and all subsequent partitions in both analyses. Indian Key samples of C. casablancae were progressively grouped with some to all C. incanum x C. casablancae hybrids with mitogenomes derived from C. casablancae in partitions 3-6, and with C. incanum in partitions 7-10, in analyses of both genes. The outgroup taxa remained distinct in all partitions in all analyses.

Phylogenetic analyses
Both Bayesian and Maximum Likelihood analyses of the concatenated (16S + COI) dataset resulted in a tree that resolved the relationships among the outgroup taxa and the Cerion of the Florida Keys with high support and partitioned all Cerion samples from the Florida Keys into two clades: the casablancae clade, which included all specimens of C. casablancae from Indian Key, and a subset of C. incanum x C. casablancae hybrids from the Bahia Honda SE population with mitochondrial genomes derived from C. casablancae; and the incanum clade, which included all specimens of Cerion incanum, including all C. incanum x C. casablancae hybrids from the Bahia Honda SE population with mitochondrial genomes inherited from C. incanum. Five major clades were resolved within the incanum clade (Fig 5). Clade 1 is the most basal, and was composed of two subclades. Clade 1A, consists of 8 haplotypes and 47 individuals, all from the High Coral Keys. It includes all the specimens from Key Biscayne and representatives of both populations from Key Largo, but none from Plantation Key. Clade 1B is larger (19 haplotypes, 76 individuals) and includes specimens from the High Coral Keys (Key Largo to

Haplotype networks
The program TCS sorted 486 COI sequences from Cerion incanum, C. casablancae and their hybrids into 114 haplotypes. Of these, 48 (42%) haplotypes were represented by single individuals (singletons) and 8 (7%) were present in more than one population. The corresponding 16S sequences included 78 haplotypes, of which 30 (38%) were singletons and 15 (19%) present in multiple populations (Appendix 1). The haplotype network for the COI gene (95% confidence interval, connection limit = 11 steps) is shown in Fig 6, the network for the 16S gene (95% confidence interval, connection limit = 9 steps) in Fig 7. Haplotypes from the High Coral Keys, Low Coral Keys and Oölite Keys are shaded in different tones. Haplotypes present in more than one of these areas are shaded proportionally. In both analyses, haplotypes occurring in Cerion casablancae and C. incanum x C. casablancae hybrids with mitogenomes derived from C. casablancae form a network that is separate from the network of haplotypes occurring in C. incanum and C. incanum x C. casablancae hybrids with mitogenomes derived from C. incanum, indicating that it would    require more than 11 steps to join the COI networks, and more than 9 steps to join the 16S networks. In one case, the COI haplotype (Haplotype 106) was more than 11 steps from either network.
With the exception of the Plantation Key population, COI haplotypes tended to be geographically restricted. All COI haplotypes occurring in samples from the northern High Coral Keys (KB, KLN, KL) were shared by and confined to these three populations, and were not represented elsewhere in the Keys (Fig 8). Similarly, COI haplotypes occurring in the Oölite Keys were restricted to these Keys. A notable and anomalous exception was Plantation Key, which shared a single wide-ranging COI haplotype only with four of the Oölite Keys. Haplotypes of COI occurring in the Low Coral Keys were limited to this group of islands. Excluding the special case of the C. casablancae haplotype shared by Indian Key and the Bahia Honda SE population, which was due to well-documented introductions, only Windley Key, Long Key and Crawl Key had COI haplotypes in common. Except for Key Biscayne (KB) and the population from southern Key Largo (KL) the majority (64%) of COI haplotypes occurring in each of the sampled populations were endemic to that population. Other than the wide-ranging Haplotype  As with the COI haplotypes, most of the 16S haplotypes occurring on the High Coral Keys are endemic to the High Coral Keys (Fig 9), and differ by 1-10 mutational steps (Fig 7). The exceptions are Haplotype 26, which occurs on Plantation Key in addition to several Oölite Keys, as well as Haplotypes 35, 60 and 61 which are within 2 mutational steps of Haplotype 26 and endemic to Plantation Key. Most of the 16S haplotypes occurring on the Oölite Keys are interconnected, often through complex linkages that may include one or more haplotypes (eg., Haplotypes 20,26,54,63) with ranges that extend beyond the Oölite Keys. 16S haplotypes represented on the Low Coral Keys are more widely distributed, and few are interconnected. Unlike the COI haplotypes, a much smaller number of populations sampled had a majority of 16S haplotypes that were endemic to a Key.

The hybrid population on Bahia Honda Key
Thirty individuals from the C. incanum x C. casablancae hybrid population inhabiting the southeastern coast of Bahia Honda Key were sequenced. Of these, 19 had mitogenomes derived from Cerion incanum and 11 from C. casablancae. The COI haplotypes derived from C. incanum mitogenomes were all endemic to the BHSE population and were within 3 mutational steps of haplotypes present on MK, NNK or from a wide ranging haplotype present on the Oölite Keys and Plantations Key.
Of the specimens with the C. casablancae derived mitogenomes, one shared a COI haplotype present in a specimen of C. casablancae from Indian Key, the rest had COI haplotypes that were endemic to the BHSE population. Most differed by 1-8 bases from Indian Key specimens of C. casablancae, but one could not be connected to either of the mitogenome networks within the 95% confidence interval (11 steps). The 16S haplotypes present in specimens with C. casablancae mitogenomes were either C. casablancae or endemic to the BHSE population and within 6 mutational steps from C. casablancae haplotypes. Specimens with mitogenomes inherited through C. incanum maternal lineages had more diverse 16S haplotypes. Four were endemic to the BHSE population, 7 also occurred on neighboring MK, 2 on NNK, 5 on BPN and one was shared with populations ranging from PK to SKN.

Discussion
Cerion incanum is endemic to the Florida Keys, ranging from Key Biscayne to Key West, and the only species of Cerion native to the archipelago. Like most Cerion, it forms dense local colonies that are not widespread, and does not occur on every Key. A major factor affecting dispersal among islands is the transport by hurricanes of vegetation to which these animals are attached [1,22,44]. Prior to his initial series of experiments, Bartsch [28,29] surveyed many of the Keys and introduced Bahamian species of Cerion only onto islands on which C. incanum was not found. Indian Key and Bahia Honda Key were among the uninhabited Keys on which he introduced a large propagule [500 adult specimens] of C. casablancae (Fig 1E) in June, 1912. Descendants of these introductions survive little changed on Indian Key after more than a century ( Fig 1F). The Bahia Honda population of C. casablancae remained isolated for more than a decade, but on returning to study this colony in 1931, Bartsch [30] reported the appearance of C. incanum at the site, as well as the presence of C. incanum x C. casablancae hybrids that persist to the present day. Bartsch [14,29,45] monitored the populations he introduced for decades, documenting instances of Cerion colonies being extirpated or decimated by fires and hurricanes and of propagules being transported among the Keys by hurricanes.
Our efforts to apportion the specimens in our study to subspecies on the basis of shell characters used to define them met with limited success, as a significant proportion of the shells had phenotypes that were intermediate between those of the named subspecies. With the exception of the nominotypical subspecies Cerion incanum incanum, phenotypes corresponding to the type specimens were either entirely absent or a minor component of the populations sampled from their type localities and frequently had broad and discontinuous distributions ( Table 2) rather than representing geographically isolated races that are usually associated with subspecies. In his monographic treatment of the terrestrial gastropods of North America, Pilsbry [46], who earlier named two of the subspecies of Cerion incanum, treated all of the subspecies as forms, noting that "they have little significance", and listed them as synonyms of Cerion incanum.
When COI and 16S sequences were evaluated using the Automatic Barcode Gap Discovery (ABGD) method, all Florida Keys individuals (except for C. casablancae from Indian Key and individuals from the hybrid population on Bahia Honda Key with mitochondrial genomes inherited from C. casablancae) were grouped together in the third and all subsequent partitions, indicating the absence of discernible gaps in the distribution of pairwise differences within our sample.
Phylogenetic analyses of the concatenated (16S + COI) sequences affirmed the previously reported derived relationship of Cerion incanum to cerionid outgroups, and particularly its sister group relationship to species from the Great Bahama Bank, adding further evidence for a Bahamian origin for the Cerion fauna of southeastern Florida and the Florida Keys [26]. The highly supported clade containing all samples of Cerion incanum, including all hybrids with mitochondrial genomes derived from C. incanum, was partitioned into five major groups that formed a ladder-like arrangement (Figs 4 and 5). With the exception of a single specimen from No Name Key, the most basal of these five groups (Clade 1) was composed only of specimens from the older Key Largo Formation (Key Biscayne to Bahia Honda Key). The intermediate clades (Clades 2-4) consisted of specimens that ranged from Plantation Key, the southwesternmost of the High Coral Keys, to Sugarloaf Key, in the central Oölite Keys, while the most derived clade (Clade 5) contained only specimens from the Oölite Keys of the younger Miami Formation (Fig 5). The composition and relationships among the clades of C. incanum in this phylogeny indicate that the northernmost of the keys, closest to the Tomeu paleoislands that had been inhabited by Cerion petuchi, the presumed ancestor of C. incanum, have the oldest populations, and that this species expanded in a southwesterly direction, colonizing newly emergent Keys as they were formed, with the populations of the Oölite Keys being the youngest.
Phylogenetic analyses as well as haplotype distribution and networks indicate that the faunas of the northern High Coral Keys and the Oölite Keys comprise well supported clades with distinctive haplotypes within each region that are directly interconnected. In contrast, the faunas of the intervening Low Coral Keys consist of heterogeneous groups that include some haplotypes derived from High Coral Keys and others from Oölite Keys. Key Largo of the High Coral Keys and many of the Oölite Keys (eg., Big Pine Key, Sugaloaf Key) are comparatively large islands, each with multiple discontinuous populations of Cerion, while Plantation Key (High Coral Keys) and the Low Coral Keys include many small islands, some with a single population of Cerion, others without living colonies of Cerion.
Upon reaching the Oölite Keys, Cerion incanum would have greatly expanded its range across the large land areas cumulatively provided by these Keys. These are topographically low islands, many separated by shallow, narrow channels whose positions and courses could be altered by hurricanes. Stochastic events such as fires and hurricanes, as well as sea level changes during the Holocene would have amplified genetic diversity in Cerion incanum inhabiting the Oölite Keys by recombining and isolating the numerous, geographically proximal populations, resulting in the high number of haplotypes occurring on this group of islands [ Table 1]. Effects of such stochastic events would be more extreme on Cerion populations inhabiting smaller, more isolated Keys separated by greater distances and deeper channels (including, but not limited to the Low Coral Keys). Negative effects could include local extinctions or decimations resulting in genetic bottlenecks, while the likelihood of introductions of new propagules would be inversely proportional to the distance of a Key to a neighboring Cerion population. The larger populations of the Oölite Keys and Key Largo were likely more stable over time and served as a source of genetic material to reestablish colonies in the Low Coral Keys or periodically contribute haplotypes to existing small populations.
Discordant patterns in distribution among two mitochondrial genes conflict with the general model that mitochondrial genomes are uniparentally transmitted, homoplastic and nonrecombining, yet similar patterns have been reported in the Cerion uva complex in the southernwestern Caribbean [47]. Explanations for this phenomenon range from convergent sequence evolution in haplotypes with broad geographic ranges to heteroplasmy and recombination within mitochondrial genomes. More recent studies have shown that the mitochondrial genome is not always clonal, neutrally evolving or clock-like [48,49], and have brought to light examples of departures from the uniparental, homoplastic and non-recombining transmission especially in interspecific or interstrain hybrids [50,51].
In addition to these natural processes by which Cerion is transported among islands, railroad bridges connecting Miami to Key West were built between 1905 and 1912, while the Overseas Highway extending to Key West was completed in 1938. Quarries on Windley Key and later on Plantation Key were sources for both fill dirt and slab rock used to build these structures, raising the possibility of more recent anthropogenic transport of Cerion propagules in the direction of Key West, perhaps accounting for the anomalous distributions of COI Haplotype 30 and 16S Haplotypes 20 and 26.

The hybrid population on Bahia Honda Key
Woodruff and Gould [11] provided a succinct summary of the history of Bartsch's introduction of C. casablancae onto Bahia Honda Key in 1912, including the reduction of the population size to about 55 individuals following a fire in 1915, and the subsequent appearance of C. incanum at the site by 1931 that resulted in the C. incanum x C. casablancae hybrid population present along the southeastern coast of this Key. In a detailed study, these authors analyzed samples of C. incanum, C. casablancae, and hybrid populations using multivariate morphometric techniques as well as studies of variation in 17 allozyme loci. They concluded that the hybrids are gradually approaching C. incanum in shell morphology based on comparisons of samples collected in 1933 and 1977. The distribution of allozymes, which are coded by nuclear genes and subject to Mendelian inheritance, indicated that by 1977 no pure C. casablancae remained on Bahia Honda and that this introduced genome was diminished by 30% or more in the various localities they sampled. With multiple caveats, they projected that "we may be unable to detect C. casablancae genes on Bahia Honda Key 350 years from now." [11].
Our sample from southeastern Bahia Honda Key was collected on the coast, south of Woodruff and Gould's station 551, an area that in 1977 had individuals with, on average, only 28% of the C. casablancae nuclear genome [11]. Our analyses, however, deal only with two mitochondrial genes, which are maternally inherited in species that are panmictic, outcrossing and hermaphroditic. Of the 30 individuals analyzed, 19 (63%) had mitochondrial genomes derived from C. incanum, and 11 (37%) from C. casablancae. Changes in the proportions of C. incanum vs. C. casablancae mitogenomes in this population over time would require assortive mating, differential mortality or immigration of individuals, or that these mitogenomes have different effects on fecundity, survivorship or generation time. Although we did not conduct multivariate morphometric analyses of shell form, there did not appear to be a conspicuous correlation between shell morphology and mitochondrial lineage.
Eight COI haplotypes and six 16S haplotypes were present among the 19 individuals with the C. incanum mitogenome. All of the COI haplotypes in this group were endemic to BHSE and within 3 mutational steps of populations on MK, NNK, PK or a haplotype ranging from PK to BTN. Ten individuals were within 4 mutational steps of the population on the southwestern coast of Bahia Honda. In contrast, most of the 16S haplotypes were shared with populations from MK, NNK, BPN or a haplotype ranging from PK to SKN. Endemic haplotypes were within 2 steps of populations on MK, WK or a haplotype present on BPN and LK. Eight specimens were within 2 steps of BHSW. Thus, there is no clear indication as to the geographic origin of the C. incanum that Bartsch discovered at BHSE in 1931. There were six COI and six 16S haplotypes among the 11 individuals with the C. casablancae mitogenome. One of these hybrids shared a COI haplotype, and two shared a 16S haplotype with specimens of C. casablancae from Indian Key. The remaining COI haplotypes differed from those of C. casablancae by 1 to more than 11 bases, while the remaining 16S haplotypes differed by 1 to 6 bases.
Differences between hybrids with the C. casablancae mitogenome and C. casablancae are greater than those between hybrids with the C. incanum mitogenome and C. incanum.
The vast majority of proteins present in the mitochondria are encoded by nuclear genes, synthesized in the cytoplasm and imported into the mitochondria. Cytochrome c oxidase is a large, transmembrane protein composed of multiple subunits, of which 11 are nuclear in origin and 3 are encoded by the mitochondrial genome. Although both ribosomal RNA subunits are encoded by mitochondrial genes, there are more than 50 protein components for mitochondrial ribosomes. Proteins coded by nuclear genes are subject to Mendelian inheritance, while proteins coded by mitochondrial genes are generally maternally inherited. In cases of hybridization, subunits derived from the mitochondrial genome of one parent species must combine with subunits encoded by nuclear genes of both parent species in order to produce functional composite molecules, potentially accelerating the evolution of novel variants of mitochondrial genes by processes similar to those that result in hybrizymes [10,51]. As calculated by Woodruff and Gould [11], the proportion of C. casablancae genes in the nuclear genome in the hybrid population closest to our sample was 28% in 1977, and was projected to reach zero in less than 350 years. Thus, hybrids with mitogenomes inherited from C. casablancae have more than twice the number of heterologous nuclear proteins to incorporate into functional multisubunit molecules within the mitochondria than do hybrids with C. incanum mitogenomes, with the proportion increasing over time. As a consequence, surviving hybrids with C. casablancae mitogenomes would be expected to exhibit haplotypes with a greater and increasing number of sequence differences than would hybrids with C. incanum mitogenomes. The proportion of individuals with C. casablancae derived mitogenomes within hybrid populations may also decrease over time due to selection against reduced mitochondrial function brought about by an increasing load of heterologous protein subunits. The data from our study support both these projections.
Although such large scale transplantation experiments as conducted by Bartsch over a century ago would not be permitted today, the results of his experiments, surviving in the eastern portion of Bahia Honda Key for more than 30 generations, provide an invaluable resource for studies of the long term effects hybridization and especially the interactions of the nuclear and mitochondrial genomes.