Marine Biology

, Volume 161, Issue 4, pp 899–910

Afro-Eurasia and the Americas present barriers to gene flow for the cosmopolitan neustonic nudibranch Glaucus atlanticus

Authors

  • Celia K. C. Churchill
    • Museum of Zoology and Department of Ecology and Evolutionary BiologyUniversity of Michigan
    • Marine Science InstituteUniversity of California, Santa Barbara
    • Department of Biological SciencesCalifornia State Polytechnic University
  • Diarmaid Ó Foighil
    • Museum of Zoology and Department of Ecology and Evolutionary BiologyUniversity of Michigan
Original Paper

DOI: 10.1007/s00227-014-2389-7

Cite this article as:
Churchill, C.K.C., Valdés, Á. & Ó Foighil, D. Mar Biol (2014) 161: 899. doi:10.1007/s00227-014-2389-7

Abstract

Pelagic species have been traditionally thought to occupy vast, genetically interconnected, geographic ranges in an essentially homogeneous environment. Although this view has been challenged recently for some mesopelagic planktonic taxa, the population structure of hyponeustonic (surface-drifting) species remains unknown. Here, we test the hypothesis of panmixis in Glaucus atlanticus, a cosmopolitan neustonic nudibranch, by assessing the genetic differentiation of multiple representatives from a global neustonic sampling effort. Specimens were collected from all subtropical oceanic gyre systems (North Atlantic, South Atlantic, North Pacific, South Pacific, and Indian Ocean). We sequenced a fragment of the mitochondrial cytochrome oxidase I gene for 98 individuals and performed population structure, differentiation (analysis of molecular variance, spatial analysis of molecular variance, FST, Jost’s D), and molecular clock analyses. Our results indicate that G. atlanticus is not globally panmictic, but that populations appear to be panmictic within ocean basins. We detected several topologically ectopic haplotypes in the Atlantic Ocean, but the molecular clock analysis indicates that these have diverged from closely related Indo-Pacific haplotypes over 1.2 MYA, coinciding with cooling in waters around in the southern tip of Africa and resulting oceanographic changes. These data and the fact that G. atlanticus is not known from polar latitudes suggest that gene flow between ocean basins is hindered by physical barriers (supercontinents) and water temperatures in the Arctic and Southern Oceans.

Introduction

After decades of treating the pelagic ocean as a vast, homogeneous environment and its constituent taxa as genetically interconnected (Day 1963; McGowan 1971; Finlay 2002), our understanding of open-ocean population structuring has recently become much more nuanced. Proposed speciation models for passive drifters contradict each other, ranging from the ubiquity hypothesis, in which huge population sizes eliminate dispersal as a limit to gene flow—thereby also eliminating endemism (Finlay 2002), to allopatric hypotheses of potential barriers to gene flow, such as persistent currents and continental land masses (Palumbi 1994; Dawson and Hamner 2008). Population genetic data for some pelagic microorganisms are broadly consistent with the former (de Vargas et al. 1999; Darling et al. 2000; Bucklin et al. 2003; Taniguchi et al. 2004; Ely et al. 2005; Goetze 2005). Yet, others show either subtle (Cowen et al. 2007) or pronounced (de Vargas et al. 1999; Darling et al. 2000; Goetze 2003, 2005; Selje et al. 2004; Dawson et al. 2011) genetic structuring within- and among-ocean basins and subtropical ocean gyres. Pelagic diversification processes are taxon-specific. Moreover, we now have empirical evidence that within- and among-basin water mass boundaries may be differentially permeable even to sister taxa (Goetze 2005), due to variation in details of ecology and life history, as predicted by early dispersal models (Gaylord and Gaines 2000).

A related obstacle to understanding the genesis of planktonic diversity is that not all drifters are the same. Most planktonic research focuses on species in the water column, but a subset of marine plankton, the neuston, is only associated with the air–water interface at the ocean surface (Marshall and Burchardt 2005). Neustonic taxa drift on wind-driven surface currents and have no vertical dispersal (e.g., diel migration, temperature-related mixing). In nonpolar latitudes, the neuston comprises a discrete community of species residing there temporarily (e.g., larval fish) and widespread, holoneustonic taxa (Marshall and Burchardt 2005). How are holoneustonic populations structured? Available data are almost completely restricted to the five epineustonic (above the water surface) insect species of Halobates, the oceanic sea skater, which show restricted gene flow among gyres within species, including prolonged reciprocal monophyly between the North Atlantic, North Pacific, and Indian subtropical gyres for the widespread species H. micans dated to 1–3 MY (Andersen et al. 2000). Because no adjacent subtropical gyres were sampled, it is impossible to determine whether equatorial surface currents versus land masses act as barriers to gene flow. The five oceanic Halobates species stem from two distinct colonizations by different coastal ancestral lineages (Damgaard et al. 2000), and there is little range overlap among recent sister taxa (Andersen et al. 2000). Inferred speciation patterns from Halobates gene tree topologies, and present day distributions, are most consistent with allopatric speciation. Given that the differential permeability of ocean basin and gyre boundaries to pelagic microorganisms is related to life history, however, Halobates spp. populations may be very differently structured than their hyponeustonic (below the water surface) counterparts. Halobates spp. lay eggs on driftwood, whereas hyponeustonic taxa have pelagic larval stages. Do hyponeuston also show trenchant genetic structuring by ocean basin or subtropical gyre?

We address this question by focusing on a member of the marine hyponeuston, Glaucus atlanticus, a nudibranch sea slug that floats upside down at the surface of subtropical gyre systems by storing gulped air inside its muscular stomach (Lalli and Gilmer 1989). A recent phylogenetic study of Glaucus revealed that the Indo-Pacific congener G. marginatus constitutes a species complex containing two cryptic pairs of sister species with overlapping distributions (Churchill et al. 2013). A parallel change in the reproductive morphology has occurred once in each cryptic species pair, suggesting that a nongeographic isolation mechanism—reproductive character differentiation—may be a primary driver of cladogenesis in this species complex. On the other hand, molecular phylogenetic analyses of the cosmopolitan G. atlanticus did not recover trenchant cladogenetic structuring nor evidence of cryptic species within- and/or among-ocean basins (Churchill et al. 2013).

Here, we revisit Churchill et al. (2013) G. atlanticus mitochondrial dataset using more sensitive population genetics methodologies and a number of new sequences representing all five subtropical gyres (including the South Atlantic for the first time) to test three general biogeographic hypotheses for this species: (1) Global panmixis (null hypothesis)—rates of global historical gene flow have been sufficient to result in the absence of genetic structuring among the five subtropical gyre systems (Fig. 1a). Allopatric speciation processes are therefore impeded. (2) Ocean basin panmixis—neustonic exchange occurs readily among adjacent subtropical gyre systems, but not among-ocean basins, producing three ocean basin-specific clades, each potentially capable of proceeding toward allopatric speciation (Fig. 1b). (3) Within-gyre panmixis—each subtropical gyre population is a discrete and isolated gene pool. Prolonged isolation is predicted to produce reciprocal monophyly, and eventually speciation, of gyre populations (Fig. 1c).
https://static-content.springer.com/image/art%3A10.1007%2Fs00227-014-2389-7/MediaObjects/227_2014_2389_Fig1_HTML.gif
Fig. 1

Hypothetical networks of global subtropical gyre genetic structuring. a Global panmixis. b Panmixis only within-ocean basins. c Panmixis only within-ocean gyres. NA North Atlantic; SA South Atlantic; NP North Pacific; SP South Pacific; IN Indian

Materials and methods

Sample collection, DNA extraction, amplification, and sequencing

Glaucus atlanticus specimens were collected as a part of a global sampling of neustonic invertebrates from 2006 to 2012 (Fig. 2). Specimens were either collected by hand as they washed up on beaches, or in the open ocean via neuston net tows. All specimens were preserved in 95 % ethanol and identified by external morphology. At least 15–20 individuals, when possible, were chosen arbitrarily from each ocean gyre for genetic analysis, but only five were available from the South Atlantic (Table 1).
https://static-content.springer.com/image/art%3A10.1007%2Fs00227-014-2389-7/MediaObjects/227_2014_2389_Fig2_HTML.gif
Fig. 2

Glaucus atlanticus and collecting localities. aG. atlanticus individual. Scale bar = 1.0 cm. b Map of collecting localities showing subtropical gyre boundaries. Collecting sites are indicated with red squares. Subtropical gyres are color-coded: blue North Atlantic; green South Atlantic; red North Pacific; black South Pacific; yellow Indian

Table 1

Material examined in this study, including isolate code, museum voucher number, coordinates (decimal degree), gyre system location, and GenBank accession numbers

Isolate

Voucher

Latitude

Longitude

Gyre system

GenBank #

01GaIWA

WAM S59354

−32.0333333

115.7500000

Indian

JQ699594

02GaIWA

WAM S59354

−32.0333333

115.7500000

Indian

JQ699595

03GaIWA

WAM S59354

−32.0333333

115.7500000

Indian

JQ699596

04GaIWA

WAM S59354

−32.0333333

115.7500000

Indian

JQ699597

05GaIWA

WAM S59354

−32.0333333

115.7500000

Indian

JQ699598

06GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961612

07GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961613

10GaIWA

WAM S59354

−32.0333333

115.7500000

Indian

JQ699599

11GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961614

12GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961615

14GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961616

15GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961617

16GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961618

17GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961619

18GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961620

19GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961621

20GaIWA

WAM S59356

−32.0000000

115.7500000

Indian

KF961622

01GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961623

02GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961624

03GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961625

04GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961626

05GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961627

06GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961628

07GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961629

08GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

JQ699574

10GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961630

11GNAGu

No voucher

27.4965000

−084.9965000

North Atlantic

KF961631

12GNAGu

No voucher

27.4965000

−084.9965000

North Atlantic

KF961632

13GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961633

14GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961634

15GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961635

16GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961636

17GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961637

18GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961638

19GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961639

20GNAGu

UMMZ 302975

27.4965000

−084.9965000

North Atlantic

KF961640

22GNADO

UMMZ 302975

27.4988000

−085.9983000

North Atlantic

KF961641

25GNADO

UMMZ 304380

26.4963000

−089.0013000

North Atlantic

KF961642

26GNADO

UMMZ 304381

27.5010000

−085.0115000

North Atlantic

KF961643

27GNADO

UMMZ 304381

27.5010000

−085.0115000

North Atlantic

KF961644

29GNADO

UMMZ 304382

28.4995000

−089.0005000

North Atlantic

KF961645

30GNADO

UMMZ 304382

28.4995000

−089.0005000

North Atlantic

KF961646

31GNADO

UMMZ 304383

28.4972000

−088.0063000

North Atlantic

KF961647

32GNAGu

UMMZ 304384

28.6627000

−085.4880000

North Atlantic

KF961648

33GNAGu

UMMZ 304384

28.6627000

−085.4880000

North Atlantic

KF961649

34GNAGu

UMMZ 304384

28.6627000

−085.4880000

North Atlantic

KF961650

35GNAGu

No voucher

28.6627000

−085.4880000

North Atlantic

KF961651

36GNAGu

UMMZ 304384

28.6627000

−085.4880000

North Atlantic

KF961652

37GNAGu

No voucher

28.6627000

−085.4880000

North Atlantic

KF961653

38GNAGu

UMMZ 302975

27.4965000

−084.9948000

North Atlantic

KF961654

39GNAGu

UMMZ 304385

26.9905000

−084.9982000

North Atlantic

KF961655

40GNAGu

UMMZ 302976

27.4997000

−086.0057000

North Atlantic

KF961656

41GNAGu

UMMZ 302977

28.0003833

−087.4846167

North Atlantic

KF961657

42GNAGu

UMMZ 302977

28.0003833

−087.4846167

North Atlantic

KF961658

43GNAGu

UMMZ 302977

28.0003833

−087.4846167

North Atlantic

KF961659

44GNAGu

UMMZ 302978

24.4932000

−083.5035000

North Atlantic

KF961660

45GNAGu

UMMZ 302978

24.4932000

−083.5035000

North Atlantic

JQ699582

46GNAGu

UMMZ 302978

24.4932000

−083.5035000

North Atlantic

KF961661

47GNAGu

UMMZ 304390

24.9853000

−084.9990000

North Atlantic

KF961662

48GNAGu

UMMZ 302979

26.0047000

−086.0083000

North Atlantic

JQ699583

49GNAGu

UMMZ 304386

25.9995000

−090.0108000

North Atlantic

KF961663

50GNAGu

UMMZ 304386

25.9995000

−090.0108000

North Atlantic

KF961664

51GNAGu

UMMZ 304386

25.9995000

−090.0108000

North Atlantic

KF961665

52GNAGu

UMMZ 304387

26.6057000

−088.5665000

North Atlantic

KF961666

03GaNPKo

UMMZ 302983

19.3550000

−156.7216667

North Pacific

JQ699588

05GaNPKo

UMMZ 302981

19.3550000

−156.7216667

North Pacific

JQ699585

07GaNPKo

UMMZ 304388

19.3550000

−156.7216667

North Pacific

KF961667

09GaNPKo

UMMZ 304388

19.3550000

−156.7216667

North Pacific

KF961668

10GaNPKo

UMMZ 304388

19.3550000

−156.7216667

North Pacific

KF961669

11GaNPKo

UMMZ 304388

19.3450000

−156.7366667

North Pacific

KF961670

12GaNPKo

UMMZ 304388

19.3550000

−156.7216667

North Pacific

KF961671

13GaNPKo

UMMZ 304388

19.3550000

−156.7216667

North Pacific

KF961672

15GaNPKo

UMMZ 304389

20.2205000

−157.0110000

North Pacific

KF961673

16GaNPWH

UMMZ 304391

19.1250000

−153.3616667

North Pacific

KF961674

01GaSA

UMMZ 303481

−34.1372000

018.4336000

South Atlantic

KF961675

02GaSA

UMMZ 303481

−34.1372000

018.4336000

South Atlantic

KF961676

03GaSA

UMMZ 303481

−34.1372000

018.4336000

South Atlantic

KF961677

04GasA

UMMZ 303481

−34.1372000

018.4336000

South Atlantic

KF961678

05GaSA

UMMZ 303481

−34.1372000

018.4336000

South Atlantic

KF961679

01GaSPWH

UMMZ 302980

−04.2650000

−141.7150000

South Pacific

JQ699584

02GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

JQ699587

03GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

JQ699590

04GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

JQ699591

05GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

JQ699593

06GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961680

07GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961681

08GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961682

09GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961683

10GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961684

11GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961685

12GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961686

13GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961687

14GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961688

16GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961689

17GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961690

21GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961691

23GaSPPC

AMS C.462956

−33.7380000

151.3100000

South Pacific

KF961692

Approximately 30 mg of tissue was sampled from the left ceratal cluster or foot. Genomic DNA was extracted using the E.Z.N.A. Mollusc DNA Kit (Omega Bio-Tek) according to manufacturer’s protocols. Mitochondrial cytochrome oxidase I (COI) was amplified using the universal primers LCO1490 and HCO2198 (Folmer et al. 1994) using the general PCR protocol [2 min at 95 °C; 35 cycles of 30 s at 94 °C, 30 s at X °C, 1 min at 72 °C; 5 min at 72 °C], where X = 45 °C. PCR products were sequenced directly with PCR primers using an ABI 3730xl automated sequencer (Applied Biosystems, Inc.) by the University of Michigan DNA Sequencing Core. Forward and reverse primer chromatograms were aligned using the MUSCLE algorithm (Edgar 2004) implemented in CodonCode Aligner (CodonCode Corporation) and checked by eye. COI sequences were aligned according to amino acid translations using the mitochondrial code for Mytilus edulis (NCBI-GenBank).

Phylogenetic analyses

The sequence matrix (658 nucleotides of mitochondrial COI) was analyzed in jModelTest (Posada 2008) using likelihood calculations performed in PhyML (Guindon and Gascuel 2003) to determine the best-fit models of nucleotide substitution by Akaike information criterion (HKY + Γ). Bayesian Markov Chain Monte Carlo (MCMC) analyses were performed in BEAST 1.7.2 (Drummond et al. 2012) using the HKY model with empirical base frequencies, gamma site heterogeneity with four rate categories, and nucleotides were partitioned using the SRD06 model (Shapiro et al. 2006). All parameters were linked, and the clock model was strict with substitution rates estimated for nudibranch COI (1 %/MY; Shields 2009). Starting trees were generated randomly, and the tree priorly assumed the coalescent process of constant size. MCMC analyses were 10 million generations, logging parameters every 1,000 generations, and the first 25 % of trees were discarded as burn-in. Convergence was confirmed by eye using the “Trace” function in Tracer v. 1.5 (Rambaut and Drummond 2009) and by repeating all analyses three times. Three analyses were combined to generate summary statistics and maximum clade credibility trees (median node heights and >0.5 posterior probabilities).

Phylogeographic analyses

To visualize the genetic structure of G. atlanticus, a haplotype network was constructed using TCS 1.21 (Clement et al. 2000) with a 95 % connection limit. To identify genetic subdivisions between ocean basins and between subtropical gyres, haplotypes were assigned to five populations (North Atlantic, South Atlantic, North Pacific, South Pacific, and Indian Ocean) and grouped by ocean basin (Atlantic, Indian, and Pacific). These subdivisions were tested for genetic structure using the analysis of molecular variance (AMOVA) implemented in Arlequin 3.5 (Excoffier and Lischer 2010). Molecular diversity indices were calculated in Arlequin 3.5. Arlequin 3.5 was also used to calculate FST values as a measure of pairwise differences between all population groups. The significance of the pairwise FST values was estimated by performing 16,000 permutations, and the nominal p values were subsequently corrected for multiple test biases by the Bonferroni standard correction. Because F statistics and their relatives, such as GST, greatly underestimate differentiation when applied to highly polymorphic markers (Hedrick 2005; Jost 2008; Meirmans and Hedrick 2011), we also calculated the D statistic (Jost 2008), which is independent of within population diversity. GenAlEx 6.5 (Peakall and Smouse 2006, 2012) was used to calculate Jost’s D by treating the mtDNA haplotype data as genotype data (assuming homozygosity in all loci). The significance of the pairwise Jost’s D values was estimated by performing 9,999 permutations, and the nominal p values were subsequently corrected for multiple test biases by the Bonferroni standard correction. Additionally, a spatial analysis of molecular variance (SAMOVA) was performed using SAMOVA version 1.0 (Dupanloup et al. 2002), which aims to define groups of samples that are maximally differentiated from each other yet geographically homogenous. Because many sampling locations contained a single individual, locations that were geographically close (within 3° of longitude or latitude) and in the same gyre system were merged together and their coordinates averaged, resulting in five populations: North Atlantic, North Pacific, South Atlantic, South Pacific 1 (Australia), South Pacific 2 (Marquesas Islands), and Indian Ocean. A SAMOVA was performed for all five populations with 100 simulated annealing processes. The AMOVA and SAMOVA analysis results were compared for the a priori hypotheses of two groups (Atlantic, Indo-Pacific) and three groups (Atlantic, Indian, Pacific) to determine whether they produced similar clustering of genetic variation. Demographic histories were characterized using mismatch analyses in Arlequin 3.0 and by comparing Bayes factors for two demographic models in BEAST: coalescent with constant size, and coalescent exponential.

Molecular clock analyses

Date estimates for selected uncalibrated nodes in the BEAST Bayesian tree were derived by using r8s 1.7 (Sanderson 2003) using the finalization of the closure of the Isthmus of Panama as the calibration point, estimated in 3.5 MYA (Coates et al. 1992; Coates and Obando 1996). The nodes tested represent the divergence between several haplotypes of Indo-Pacific affinity found in the Atlantic from their Indo-Pacific closest relatives. Analyses using the Langley–Fitch and penalized likelihood methods were run. For the Langley–Fitch method, analyses were run using the Powell and Truncated Newton as well as the local molecular clock procedure. The penalized likelihood analysis was run with the Powell algorithm.

Results

Phylogenetic reconstruction

Phylogenetic analysis of within-species G. atlanticus mitochondrial haplotype diversity produced a poorly resolved tree (Appendix S1) lacking significant support for most branches.

Phylogeographic analyses

Figure 3 shows the haplotype network of relationships in G. atlanticus, which is largely consistent with the results of the phylogenetic analyses, but incorporates non-bifurcating genealogical information derived from population-level divergences. Most Atlantic haplotypes, north and south, clustered together; in fact, all five South Atlantic specimens shared a haplotype with a North Atlantic conspecific. At least 10 individuals were haplotyped from each ocean gyre except for the South Atlantic (N = 5). Haplotype diversity (h) was maximal (1.00 within each ocean gyre population), and so an additional 30 individuals were genotyped from one ocean gyre (North Atlantic) to investigate whether the addition of more individuals might reach saturation, but North Atlantic haplotype diversity (h) remained at 1.00 (Table 2). This is almost entirely due to synonymous substitutions in the third codon position, as evidenced by the much higher rate of evolution (2.54 vs. 0.23) of the third codon position generated in the Bayesian phylogenetic reconstruction. Average pairwise diversity (π) was 6.37–7.96 for all gyre populations except the North Pacific, which was higher (π = 10.42). Tajima’s D values are negative for all gyre populations, significantly so for all except South Atlantic (Table 2); this further indicates an excess of rare haplotypes.
https://static-content.springer.com/image/art%3A10.1007%2Fs00227-014-2389-7/MediaObjects/227_2014_2389_Fig3_HTML.gif
Fig. 3

Haplotype network of G. atlanticus mtCOI sequences. Subtropical gyres are color-coded: blue North Atlantic; green South Atlantic; red North Pacific; black South Pacific; yellow Indian

Table 2

Molecular diversity indices for ocean gyre populations of G. atlanticus

Population

Sample size (N)

Haplotype diversity (h)

Nucleotide diversity (π)

Tajima’s D

p value

North Atlantic

47

1.0000 ± 0.0044

6.799260 ± 3.260492

−2.23902

0.00094

South Atlantic

5

1.0000 ± 0.1265

7.000000 ± 3.963246

−0.20309

0.49844

North Pacific

10

1.0000 ± 0.0447

10.42222 ± 5.197645

−1.68011

0.02831

South Pacific

18

1.0000 ± 0.0185

6.366013 ± 3.164781

−2.00876

0.00881

Indian

17

1.0000 ± 0.0202

7.955882 ± 3.891067

−1.92848

0.01394

Number of specimens haplotyped (N), nucleotide diversity—mean number of pairwise differences (π), haplotype diversity (h), and Tajima’s D with associated p values for all populations examined

Hierarchical AMOVA analyses comparing variance between ocean gyres and ocean basins show that φST (between gyre) and φCT (between ocean basin) values are similar and somewhat low: 0.26 and 0.25, respectively (Table 3; φST is significant). In one additional nonhierarchical AMOVA (Table 3), samples were grouped by Atlantic basin versus Pacific + Indian basins, (φST = 0.29, p = 0.00), and in this case, φST is higher than in the original AMOVA. In both cases, the great majority of the variation (70–75 %) is explained by variation within populations (oceanic gyres) and almost 0 % is explained by variation among gyres within groups (ocean basins, i.e., North and South Atlantic, North and South Pacific). These results probably reflect the presence of six haplotypes of Indo-Pacific affinity in the Atlantic Ocean and one haplotype (13GaNPKo) of Atlantic affinity in the Pacific Ocean (Fig. 3).
Table 3

AMOVA tests of regional (ocean basin) and population (ocean gyre) genetic structure in G. atlanticus with two different data partitions

Partitioning into three groups (ocean basins): Atlantic, Pacific, Indian

df

Sum of squares

Variance components

% Of variation

F statistics

Among groups (ocean basins)

2

79.145

1.21792 Va

24.93

φCT = 0.24930, p = 0.00000

Among populations (gyres) within groups (ocean basins)

2

7.845

0.02564 Vb

0.52

φSC = 0.00699, p = 0.30594

Within populations (gyres)

92

335.041

3.64175 Vc

74.54

φST = 0.25455, p = 0.00000

Partitioning into two groups (ocean basins): Atlantic, Pacific + Indian

df

Sum of squares

Variance components

% Of variation

F statistics

Among groups (ocean basins)

1

74.147

1.43171 Va

27.94

φCT = 0.27943, p = 0.10238

Among populations (gyres) within groups (ocean basins)

3

12.844

0.05023 Vb

0.98

φSC = 0.01361, p = 0.15446

Within populations (gyres)

92

335.041

3.6415 Vc

71.08

φST = 0.28923, p = 0.00000

Group 1, Atlantic Ocean (2 populations); Group 2, Pacific Ocean (2 populations); Group 3, Indian Ocean (1 population)

The SAMOVA analysis of two groups confirmed the a priori hypothesis of an Atlantic Indo-Pacific differentiation (Table 4). The results of the analysis are similar to the AMOVA results (Table 3), with approximately 70 % of the haplotype diversity found within populations, but an appreciable amount of haplotype diversity (28.09 %) separates groups. The differences among populations within regions are very small (0.52 %). As in the AMOVA analysis, φST (within populations) and φCT (among groups) values are similar and somewhat low: 0.28 and 0.28, respectively (φST is significant). The SAMOVA analysis of three groups failed to separate the remaining populations by ocean basins (Atlantic, Pacific, Indian) as originally hypothesized, but instead, it separated them into the following three groups: Atlantic, (Indian, North Pacific, South Pacific 1), and South Pacific 2. In this case, the haplotype diversity among groups is slightly smaller than in the two-group analysis (27.52 %).
Table 4

SAMOVA tests of genetic structure in G. atlanticus with group partitions into two and three groups

Partitioning into three groups: NA + SA, SP2, NP + SP1 + IN

df

Sum of squares

Variance components

% of variation

F statistics

Among groups

2

77.548

1.40182 Va

27.52

φCT = 0.27518, p = 0.02248

Among populations within groups

2

7.845

0.02564 Vb

0.52

φSC = 0.01246, p = 0.00000

Within populations

92

335.041

3.64175 Vc

74.54

φST = 0.28422, p = 0.00000

Partitioning into two groups: NA + SA, SP1 + SP2 + NP + IN

df

Sum of squares

Variance components

% of variation

F statistics

Among groups

1

74.146

1.43903 Va

28.09

φCT = 0.28087, p = 0.06647

Among populations within groups

4

16.073

0.03819 Vb

0.75

φSC = 0.01037, p = 0.00000

Within populations

91

331.031

3.64629 Vc

71.17

φST = 0.28087, p = 0.00000

IN Indian Ocean; NA North Atlantic; NP North Pacific; SA South Atlantic; SP1 South Pacific 1 (Australia); SP2 South Pacific 2 (North of Marquesas)

The analysis of mismatch distributions generated in Arlequin resulted in θ1 upper limit values of 99,999 for all populations (ocean gyres). This indicates that the population sizes are so large that there are no recent coalescent events and that demographic history cannot be estimated precisely by this method with these data (Schneider and Excoffier 1999); thus, those results are not shown.

The genetic structure of G. atlanticus recovered in the haplotype network was tested using FST and Jost’s D pairwise analyses (Tables 5, 6). The results indicate that North Atlantic and North Pacific populations are not significantly different from each other. Also, North Pacific, South Pacific, and Indian Ocean populations are not significantly different. However, all Atlantic populations are significantly different from Indo-Pacific populations, suggesting limited gene flow between ocean basins. Our G. atlanticus data therefore reject hypotheses of exclusive within-gyre panmixis, and of global panmixis, but provide a mixed outcome for among-ocean basin panmixis: rejection for Atlantic/Indo-Pacific, but not for Indian/Pacific Ocean populations.
Table 5

Matrix of the population group comparisons results, with FST values (lower triangular) and associated p values (upper triangular)

 

NA

SA

NP

SP

IN

NA

0.47312 ± 0.0045

0.00000 ± 0.0000*

0.00000 ± 0.0000*

0.00000 ± 0.0000*

SA

−0.00635

0.00020 ± 0.0001*

0.00000 ± 0.0000*

0.00000 ± 0.0000*

NP

0.24921

0.23789

0.14048 ± 0.0035

0.15840 ± 0.0034

SP

0.32328

0.40346

0.01473

0.06871 ± 0.0025

IN

0.28405

0.31773

0.01302

0.01780

After Bonferroni correction (10 comparisons), significant values are p < 0.005. Significantly distinct FST values in bold, significant p values marked with an asterisk

NA North Atlantic; SA South Atlantic; NP North Pacific; SP South Pacific; IN Indian Ocean

Table 6

Matrix of the population group comparisons results, with Jost’s D values (lower triangular) and associated p values (upper triangular)

 

NA

SA

NP

SP

IN

NA

0.492

0.001*

0.001*

0.001*

SA

0.000

0.001*

0.001*

0.001*

NP

0.018

0.022

0.214

0.191

SP

0.023

0.032

0.000

0.060

IN

0.021

0.027

0.001

0.001

After Bonferroni correction (10 comparisons), significant values are p < 0.005. Significantly distinct Jost’s D values in bold, significant p values marked with an asterisk

NA North Atlantic; SA South Atlantic; NP North Pacific; SP South Pacific; IN Indian Ocean

Molecular clock analyses

The molecular clock analyses revealed divergence times varying between 4.25–3.74 MYA and 1.47–1.2 MYA among Atlantic haplotypes of Indo-Pacific affinity and their closest Indian Ocean haplotypes (Table 7). Assuming that the temporal calibration point is correct (see “Discussion”), this means that one North Atlantic haplotype (37GNAGu) diverged from its co-clustering Indo-Pacific haplotypes (Fig. 3) before the closure of the Isthmus of Panama, about 3.5 million years ago (Coates and Obando 1996), whereas the rest of the topologically ectopic North Atlantic haplotypes (25GNAGu, 49GNAGu, 12GNAGu, 43GNAGu) diverged more recently, between 1.51 and 1.2 MYA. Our dataset lacks evidence of Atlantic/Indo-Pacific gene flow within the last 1.2 million years.
Table 7

Divergence time ranges (in millions of years) between Atlantic haplotypes of Indo-Pacific affinity (37GNAGu, 25GNADO, 49GNAGu, 12GNAGu, and 43GNAGu) and their closest Indo-Pacific haplotypes, estimated with r8s using different methods, algorithms, and calibration dates

 

37GNAGu

25GNADO

49GNAGu

12GNAGu

43GNAGu

Mean rate variation

Standard deviation

Langley–Fitch/TN

3.74

1.28

1.20

1.55

0.002602

1.93e−09

Langley–Fitch/Powell

3.73

1.28

1.20

1.52

0.00261

2.212e−09

Langley–Fitch/local

3.73

1.28

1.20

1.52

0.00261

2.212e−09

PL/Powell

4.25

1.51

1.47

1.87

0.00204

0.0005069

TN truncated Newton algorithm; PL penalized likelihood method

Discussion

Glaucus atlanticus is not globally panmictic

Our data were inconsistent with all three hypotheses outlined in Fig. 1, but best matched with Fig. 1b: panmixis only within-ocean basins. This is only partially consistent with Halobates micans (Andersen et al. 2000), but provides additional resolution that surface currents at gyre boundaries alone do not act as barriers to hyponeustonic gene flow. Although the lack of resolution in the phylogenetic tree and the mosaic topology of the haplotype network are consistent with global panmixis, the haplotype network recovered a group of mainly Atlantic haplotypes and a group of mainly Indo-Pacific haplotypes with some limited topological intermixing, suggesting a certain level of ancestral gene flow between ocean basins. Limited gene flow also probably accounts for the low φCT values in the AMOVA analysis. However, the FST and Jost’s D analyses clearly show that populations in one ocean basin (Atlantic) are genetically distinct from those in the other two (Indian and Pacific) and do not support the hypothesis of global panmixis. Furthermore, the FST and Jost’s D analyses indicate that individual gyre populations in the Atlantic and Indo-Pacific are not genetically distinct, rejecting the hypothesis of exclusive within-ocean gyre panmixis and supporting the hypothesis of within-ocean basin panmixis. On the contrary, the within-ocean basin panmixis, hypothesis is not supported by the SAMOVA analysis that maintained Indian, North Pacific, and South Pacific 1 populations in the same group, while separating South Pacific 2 into a group on its own. This, however, could be an artifact, because the South Pacific 2 population was represented by a single sequence. Both SAMOVA and AMOVA analyses indicate that most of the haplotype diversity is found within each population with an appreciable amount of diversity separating ocean basis (among groups). These results combined seem to indicate that G. atlanticus is not a panmictic species, although total or partial within-ocean basin panmixis cannot be rejected.

The nonpolar distribution of Glaucus likely reflects its inability to survive the cold waters of the Arctic and Southern Oceans, thereby limiting its capacity to maintain gene flow between the Atlantic and Indo-Pacific basins. However, available mtDNA evidence failed to reveal structure within-ocean basins or gyres in G. atlanticus, suggesting that this species is a capable disperser throughout physically connected tropical and subtropical oceans. Intriguingly, its sister group, the Glaucus marginatus species complex, is composed of species with much more limited geographic ranges (Churchill et al. 2013) even though both groups share similar life histories and behaviors. Although sexual selection seems to be an important force resulting in speciation in the G. marginatus species complex, speciation events are most consistent with sympatry and do not explain their smaller ranges or absence of this species complex from the Atlantic Ocean. Further research on the feeding behavior or larval development of Glaucus may explain the different dispersal abilities of these species.

Cape of Good Hope disjunction

One unexpected pattern found in our G. atlanticus data is the clustering of South Atlantic specimens (sampled near Cape Town, South Africa), exclusively in the Atlantic group instead of the Indo-Pacific group, which are found only about 1,850 km eastward, in Durban. Above the subtropical convergence zone, the Agulhas current flows southwest down the eastern side of South Africa and enters a retroflection region off the Cape of Good Hope. On the western side, the Benguela current flows northeast from the retroflection region (Walker 1989). Either the South Atlantic sample size was too low (N = 5) to recover expected global clade lineages, our sequence data are not sufficiently variable to provide needed resolution to separate the South African population from the rest of the Atlantic, or this further supports that physical models, in this case of ocean surface currents, are inadequate in accurately predicting neustonic gene flow.

Historic gene flow between ocean basins

The molecular clock analysis indicates that all Atlantic haplotypes of Indo-Pacific affinity are long diverged from their closest Indo-Pacific haplotypes: There is no evidence of gene flow between ocean basins for the last 1.2 MY. The main mechanism by which Indian Ocean marine organisms have migrated into the Atlantic during the Pleistocene is the so-called Agulhas Leakage (Vermeij 2012), which is the exchange of fauna between the Indo-Pacific and the southwest Atlantic. This faunal exchange is promoted by an enhanced Agulhas current around the southern tip of Africa, which oscillates in intensity with global temperatures (Peeters et al. 2004; Biastoch et al. 2009) and allows some dispersal of Indian Ocean species into the Atlantic during interglacial periods (Vermeij 2012). Fish stocks are thought to have colonized the North Atlantic and Mediterranean during these episodic oceanic interchanges (Alvarado Bremer et al. 2005), including top predators such as the great white shark (Gubili et al. 2011) and the (marine mammal) orca (Foote et al. 2011). However, the sedimentary record in the southern tip of Africa suggests that the pattern of glacial-interglacial contrasts was established 1.2 MYA (Diekmann and Kuhn 2002), making the Agulhas Leakage more episodic after a global cooling event during the mid-Pleistocene. In the case of Glaucus, it appears that gene flow between the Indo-Pacific and the Atlantic may have been ceased during this mid-Pleistocene cooling period suggesting that the late Pleistocene Agulhas Leakage may be less permissive for passive drifters than for active swimmers. This may also be a factor in the absence of the Indo-Pacific G. marginatus species complex from the Atlantic Basin.

However, the results of the molecular analysis must be interpreted cautiously. Without external calibration points for neustonic taxa (e.g., given global surface current connectivity, it is impossible to say with certainty that the closure of the Panamanian Isthmus completely separated North Atlantic and North Pacific neuston populations), the divergence times are only as accurate as the molecular clock estimates. Shields (2009) reviews the support for using 1 %/MY in nudibranchs from the Ross Sea—the rate is based upon a fossil-calibrated bivalve phylogeny of trans-isthmanian species pairs (Marko 2002), and it has also been recommended for circumtropical gastropods in a fossil-calibrated phylogeny (Frey and Vermeij 2008) and for invertebrate mitochrondrial COI in general (Dawson et al. 2011).

The biogeographic relationship of Glaucus spp. to other hyponeustonic taxa depends on critical genetic data from their prey species, the neustonic hydrozoans Physalia, Velella, and Porpita, and a co-occurring neustonic mollusk lineage, the bubble-rafting snails Janthina janthina. Unfortunately, there are almost no data available on the population genetics of these organisms, but a recent study suggests that there is substantial cryptic diversity among Physalia present in New Zealand coastal waters (Pontin and Cruickshank 2012). Considering the morphological variation in Physalia between larger regions, there is the possibility that this group contains a substantial amount of cryptic variation.

Global panmixis as a model of planktonic population structure

Few studies address planktonic population structure at the global scale, but the overwhelming majority that do have found no support for true panmixis. Macrogeographic IBD signals have been recovered in populations of a nearshore rotifer (Mills et al. 2007), two sister species of oceanic copepods (Goetze 2005), a superspecies of high-dispersal diatoms (Casteleyn et al. 2010), and eight mophospecies of planktonic foraminiferans (reviewed in Darling and Wade 2008), for example. In general, hypotheses of planktonic gene flow are conceived as large-scale versions of studies of smaller areas (e.g., seas, offshore regions) where panmixis is common. Our study adds to the small, but growing body of evidence that global planktonic panmixis occurs chiefly in theoretical frameworks.

Acknowledgments

We thank the following for their assistance: USA: J. Lyczkowski-Shultz (Southeast Area Monitoring and Assessment Program) and R. Humphreys (Pacific Islands Fisheries Science Center) of the National Oceanographic and Atmospheric Administration, the students and crew of SEA Semester (www.sea.edu), T. Lee (University of Michigan Museum of Zoology); Australia: P. Colman (Australian Museum), L. Beckley (Murdoch University), S. Slack-Smith and C. Whisson (Western Australian Museum); South Africa: R. van der Elst (Oceanographic Research Institute), D. Herbert (Natal Museum), Mark Gibbons (University of the Western Cape), W. Florence and E. Hoenson (Iziko Museums of Cape Town). D. Riek photographed live glaucinins. Funding for this research comes from NSF award OCE 0850625 and National Geographic Society award 8601-09 to D.ÓF.

Supplementary material

227_2014_2389_MOESM1_ESM.docx (343 kb)
Appendix S1 Bayesian maximum credibility tree (DOCX 343 kb)

Copyright information

© Springer-Verlag Berlin Heidelberg 2014