SEARCH

SEARCH BY CITATION

Keywords:

  • Australia;
  • dugong;
  • mitochondrial control region;
  • phylogeography;
  • population genetics;
  • seagrass

Abstract

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

We investigated phylogeography, demography, and population connectivity of the dugong (Dugong dugon) in Australian waters using mitochondrial control region DNA sequences from 177 Australian dugongs and 11 from elsewhere. The dugong is widespread in shallow Indo-West Pacific waters suitable for growth of its main food, seagrass. We hypothesized that the loss of habitat and creation of a land barrier (the Torres Strait landbridge) during low sea level stands associated with Pleistocene glacial cycles have left a persisting genetic signature in the dugong. The landbridge was most recently flooded about 7,000 yr ago. Individual dugongs are capable of traveling long distances, suggesting an alternative hypothesis that there might now be little genetic differentiation across the dugong's Australian range. We demonstrated that Australian dugongs fall into two distinct maternal lineages and exhibit a phylogeographic pattern reflecting Pleistocene sea-level fluctuations. Within each lineage, genetic structure exists, albeit at large spatial scales. We suggest that these lineages diverged following the last emergence of the Torres Strait landbridge (ca. 115 kya) and remained geographically separated until after 7 kya when passage through Torres Strait again became possible for marine animals. Evidence for population growth in the widespread lineage, especially after the last glacial maximum, was detected.

Dugongs are widespread in the tropical and subtropical Indo-West-Pacific (Fig. 1) where they generally feed on seagrasses in shallow waters (see Marsh et al. 2011 for references). Dugongs are long-lived (up to 70 yr) and slow-breeding animals (minimum breeding age 7–17 yr, with single calves produced at intervals of 3–6 yr) (Marsh et al. 2011). In Australian waters, dugongs occur around the northern coasts, from Moreton Bay in southeast Queensland to Shark Bay in Western Australia (Fig. 1). Currently there are no known barriers to movement within the Australian range.

image

Figure 1. Map showing geographical origins of Australian samples. From each locality, the number of sequences representing each lineage (widespread/restricted) is indicated. The shaded circles also provide a graphical representation of sequence numbers and lineages (dark segments represent the widespread lineage, light segments the restricted lineage). Each named locality was treated as a “population” for the gene flow analyses. Inset is a map of the Indo-West Pacific region showing the present-day range of the dugong (shaded).

Download figure to PowerPoint

This range is a Holocene phenomenon due to present-day high sea levels. Barriers existed in the past as a result of low sea level stands associated with Pleistocene glacial cycles and may have left persisting genetic signals in Australian dugongs. Lowered sea levels, in particular, likely impacted this species in two ways. The first, affecting many marine taxa, was the exposure of land barriers, fragmenting marine populations, influencing the distribution of species and potentially producing phylogeographic structure (e.g., Mirams et al. 2011). The most important land barrier lay between Cape York Peninsula (the northernmost part of mainland Australia) and New Guinea (Fig. 1). Today, these are separated by Torres Strait, which is only 12 m deep (Chivas et al. 2001). Despite substantial fluctuations, sea levels have rarely been at or above present-day levels during the last 2.5 million years (Shackleton 1987, Lisiecki and Raymo 2005, Raymo et al. 2006). Consequently, there have probably been few periods when marine organisms have been able to traverse Torres Strait as they can today. To illustrate this, Figure 2a shows historical sea levels over the most recent glacial cycle. A horizontal line at the −12 m level makes it clear that the Torres Strait landbridge was submerged for only a few thousand years after the penultimate glacial period (between about 125 and 115 kya) and then not again until ~7,000 yr ago, after the most recent glacial period.

image

Figure 2. (a) Plot of global sea-level changes during the last interglacial period modified after Chivas et al. (2001). A horizontal line at −12 m indicates the depth of the Torres Strait: the strait was open to transit by marine organisms only during periods when the sea level plot is above this line (shaded in gray). The horizontal line at −50 m indicates the depth of the Arafura sill. When sea level was higher than this, shallow seas were present in what is now the Gulf of Carpentaria. The inset shows sea-level changes over the last million years (after de Bruyn and Mather 2007). (b–d) Sea-level reconstructions produced by J. Guinotte (used with permission) showing the available area for dugong habitat at different sea-level stands, (b) representing the situation at the last glacial maximum and (d) the situation today. Areas shaded in black and dark gray are likely suitable for seagrass growth. Arrows show the present-day usual southern limits of the dugong's range in Australia.

Download figure to PowerPoint

Barriers can also be produced as a result of loss of suitable habitat. The second effect of low sea level stands was the exposure of the Australian continental shelf. The shoreline at the last glacial maximum (LGM), about 18 kya, was on the very steep continental rise. This eliminated much of the shallow-water habitat suitable for growth of seagrasses, in particular along the east coast of Queensland (Hopley et al. 2007). In much of the Great Barrier Reef Province, conditions suitable for seagrass growth were probably rare at sea level stands deeper than about −50 m relative to modern levels (Hopley et al. 2007) (Fig. 2).

Here, we explore the hypothesis that Pleistocene sea-level fluctuations have strongly influenced the phylogeography and demography of the dugong in Australian waters. Alternatively, it is possible that any phylogeographical patterns have been obscured as a consequence of movement of dugongs leading to a degree of genetic homogeneity in the ~7,000 yr since the most recent flooding of Torres Strait. In support of this possibility, satellite-tagging studies have shown that individual dugongs are capable of long-distance movement covering hundreds of kilometers (Sheppard et al. 2006). Knowledge of the extent to which dugong populations are interconnected will inform the debate about management of the species in Australia. Of particular interest is the spatial scale at which it is legitimate to assess the eligibility of the species for listing under national and state legislation, which in turn determines the impact thresholds for government management action.

The time-scales are within the reach of mitochondrial markers (Avise 1994) and we therefore present inferences from mitochondrial control region sequences. Mitochondrial sequence data constitute a single, maternally inherited marker. Ideally, biparentally inherited nuclear markers, such as microsatellites, should also be employed in a study like this. However, the material available to us, while adequate for amplification of the mitochondrial locus, often did not provide template adequate for genotyping.

The work reported here extends that presented in two Ph.D. theses (Tikel 1997, McDonald 2005). These two authors each used DNA sequences from portions of the mitochondrial control region, as did we. Each found very strong evidence for the presence of two maternal lineages in Australia, as did we. One, the “widespread” lineage, occurs across the entire Australian range of the dugong, but is rare in southeastern Queensland. The “restricted” lineage was sampled primarily from the coast of Queensland.

Materials and Methods

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information
Generation of Molecular Data

Samples were obtained opportunistically from dugongs from the full extent of the species' range in Australia (Fig. 1, Table S1). Sources of material included dead stranded animals, animals taken by indigenous hunters, skin biopsies from live animals collected during satellite tagging experiments, and skin biopsies taken from free-ranging animals using a scraping device designed by Tikel (1997). Samples were also available from some dugongs from outside Australia (Table S1) to make a total of 188 (177 from Australia and 11 from elsewhere) for which a 411 bp portion of the control region was successfully sequenced (sequences with missing or ambiguous sites having been omitted).

DNA extraction followed van Oppen et al. (1999) or used an Epoch GenCatch tissue kit (Epoch Biolabs Pty. Ltd.) following the manufacturer's protocol. Initially, the “universal” forward primer L15926 (Kocher et al. 1989) and reverse primer A58 (5 CCTGAAGTARGAACCAGATGTC 3: Tikel et al. 1996, Tikel 1997) were used to amplify a 492 bp segment spanning the first hypervariable region of the mitochondrial control region. Subsequently, comparisons with published mitochondrial genomes for dugongs (AJ421723: Arnason et al. 2002 and AY075116) revealed six mismatches in the forward primer and 2–3 mismatches in the reverse primer. A new dugong-specific forward primer DLF (5 CATATTACAACGGTCTTGTAAACC 3) and reverse primer DLR (5 GTCATAAGTCCATCGAGATGTC 3) were designed, amplifying a fragment of 615 bp. The 5 primer is positioned in the tRNAPro and the 3 primer in the central conserved domain of the control region. Primers used for PCR were also used as sequencing primers. DNA amplification (PCR) was carried out in 25 μL reactions: 1 ×  PCR buffer, 2 mM MgCl2, 0.16 mM dNTPs, 1 ×  Q solution (Qiagen), and 1 unit of Taq DNA polymerase (Qiagen or Bioline Inc.), using the following amplification profile: 5 min at 96°C followed by 30 cycles of: 30 s at 96°C, 30 s at 50°C, 1 min at 72°C, with a final step of 10 min at 72°C.

PCR products were excised from a 1% agarose gel containing 40 mM Tris-acetate, 1 mM EDTA and purified using a QIAquick gel purification kit (Qiagen) following the manufacturer's instructions. Sequencing was done with ABI BigDye Terminator v3.1 chemistry (Applied Biosystems) and run on an ABI 377 sequencer, or ET chemistry (GE Biosciences) and run on a MegaBACE 1000 machine. Forward and reverse sequences for each sample were verified using Sequencher 3.1.1 (GeneCodes) and aligned in Se-Al v1.0a1 (Rambaut 1996) or BioEdit (Hall 1999).

Data Analysis
Phylogeny and recognition of lineages

Because of the presence of multiple identical haplotypes and of haplotypes differing from each other by few substitutions, we regarded a median-joining network (Bandelt et al. 1999) as an excellent way to present the data. The network was constructed from pairwise sequence differences using the program Network v4.2.0.0 (http://www.fluxus-engineering.com/sharenet.htm). Epsilon was set to zero; “connection cost” was set as the median vector criterion; each character was weighted 10; transitions and transversions were equally weighted.

Demographic Analyses

Basic summary statistics, calculated using DnaSP v5.10 (Rozas et al. 2003, Librado and Rozas 2009), were haplotypic diversity (h) (Nei 1987) and nucleotide diversity (π) (Nei 1987). DnaSP v5.10 was also used to calculate neutrality indices and, by simulation (1,000 replicates, assuming no recombination), their associated expected distributions. Ramos-Onsins and Rozas (2002) and Ramírez-Soriano et al. (2008) suggested that the most robust neutrality indices for detecting the signature of population growth were Fu's FS (Fu 1997) and the R2 statistic (Ramos-Onsins and Rozas 2002). We did not estimate the widely used, but more conservative, Tajima's D (Tajima 1989) because of its low resolving power (Ramírez-Soriano et al. 2008, Lohse and Kelleher 2009). Nor did we use statistics associated with mismatch distributions (Harpending 1994). Despite their popularity, these have been shown to have little power to detect population increase (Ramos-Onsins and Rozas 2002, Lohse and Kelleher 2009).

Demographic histories can also be estimated from genealogies (phylogenies) in a Bayesian statistical framework using BEAST (versions 1.4.6 and 1.7.2, Drummond and Rambaut 2007; http://beast.bio.ed.ac.uk). MODELTEST (Posada and Crandall 1998), using the Akaike information criterion, indicated that the nucleotide substitution model of Hasegawa et al. (1985) was appropriate when additionally allowing for unequal substitution rates among sites and for a proportion of sites to be invariable. When using BEAST, runs were of sufficient length (typically 30 million or more) that effective sample sizes (ESSs) were always over 100, and usually very far over this value. Several runs were done for each input file to check for convergence. The program TRACER v1.4 (http://beast.bio.ed.ac.uk/) was used to analyze the output from BEAST. The first 10% of iterations in each run were discarded as burn-in.

Analyses Done Using Beast Are Outlined Below
Test for change in population size

A coalescent exponential expansion model was specified and a randomWalkOperator selected for the exponential.growthRate parameter (Supplementary data files 2 and 3). If the 95% highest posterior density (HPD) of the growth rate parameter includes zero, a hypothesis of constant population size cannot be rejected (Marino et al. 2011; https://groups.google.com/d/msg/beast-users/y-ppM_dB5UI/uPybHlRMYc4J).

Determination of mutation rate assuming lineages split during last closure of Torres Strait

Monophyly of Australian dugongs was forced, and a prior of 115,000 yr (normal distribution ± 5,000) (date of the closure of Torres Strait to transit by marine organisms at the start of the last glacial period inferred from sea-level estimates; Fig. 2) placed on the most recent common ancestor (MRCA) of all Australian dugongs (see Supplementary data file 6). Trees generated during this analysis were examined for the strength of support given to the individual lineages.

Bayesian skyline plots

Bayesian skyline plots (BSPs) (Drummond et al. 2005) show changes in effective population size (NE(FEMALE)) over time, along with credibility intervals. A major advantage of this approach is that it avoids problems associated with choosing a single demographic scenario such as “constant population size” or “exponential growth.” Sample input files are in Supplementary data files 4 and 5. The mutation rate prior was specified (following the analysis above) as normally distributed with a mean of 24.8% per million years and lower and upper bounds of 13.89% and 37.46% per million years, respectively.

Genetic Structure and Gene Flow

Given that most samples came from distinct localities where sampling was possible, we simply used those localities as the basis for assigning individuals to “populations.” With some clustering of localities by geographic region if samples were few in number, we could define 11 populations on this basis (each represented by a pie chart in Fig. 1). There are no clear criteria for a priori grouping of these populations for a hierarchical analysis such as AMOVA (Excoffier et al. 1992). Analyses using methods designed to identify maximally differentiated groups (e.g., SAMOVA, Dupanloup et al. 2002) yielded equivocal results (data not shown). However, we did AMOVA analyses in Arlequin v3.5 (Excoffier et al. 2005), treating the lineages separately and placing populations into geographically defined regions that maximized “among region” variation (Table 1, and see Fig. 1 for localities). Thus, populations in the widespread lineage were placed into five regions as follow: (Shark Bay; Exmouth, etc.; Beagle Bay, etc.), (Darwin), (Bluemud Bay, etc.; Torres Strait), (Cooktown, etc.; Townsville, etc.; Shoalwater Bay), (Hervey Bay including Moreton Bay). For the restricted lineage, populations were placed into two regions as follow: (Torres Strait including Bluemud Bay), (Shoalwater including Townsville, etc.; Hervey Bay; Moreton Bay).

Table 1. Results of the AMOVA tests for differentiation among populations grouped into geographical regions and among all populations. See text for details of assignment of populations into regions
Lineage (analysis #)Source of variationVariation (%)Fixation index P
  1. a

    Excludes haplotype h44 from New Caledonia.

  2. b

    Therefore is related to the degree of differentiation among all populations, as measured by FST.

Widespread lineage onlyaAmong regions21FCT = 0.2070.000
Among populations within regions4FSC = 0.0560.072
Within populationsb75FST = 0.2510.000
Restricted lineage onlyAmong regions31FCT = 0.3140.000
Among populations within regions0FSC = −0.0150.633
Within populationsb69FST = 0.3030.000

We also calculated FST values by lineage between all pairs of populations (those shown in Fig. 1), and performed an exact test of population differentiation based on haplotype frequencies using Arlequin v3.5. Significance values of relevant fixation indices were estimated using 10,100 permutations.

Isolation by Distance

We explored the relationship between genetic and geographic distance using Mantel tests, which have the null hypothesis that genetic difference is not correlated with geographical distance (Mantel 1967). Analyses were done using GenAlEx v6 (Peakall and Smouse 2006). Permutations (9,999) were used to test for significance. One test was done for each lineage using pairwise sequence differences among all pairs of individuals as the genetic distance matrix. Another was done using a matrix of genetic distances, FST/(1 − FST), between the specified populations. This matrix was calculated in Arlequin following Slatkin (1995) and then manually entered into a spreadsheet for analysis using GenAlEx. Geographic distances between individual sample sites were calculated in several ways, primarily headland-to-headland around the coast, and also closely following the shoreline. For Torres Strait and very large embayments where exact sampling localities were not always recorded, distances were measured from the geographic center of that region. For the analyses using populations, some samples were grouped in order to increase sample size. Specifically, in the restricted lineage, the one sample from Blue Mud Bay in the Northern Territory was included with the 41 samples from Torres Strait. Similarly, the single Townsville representative of this lineage was included with the 18 from Shoalwater Bay. In the widespread lineage, samples from Darwin, Blue Mud Bay, and Mornington Island were grouped together, as were samples from between Cooktown/Starcke and Townsville (NE Queensland), and samples from Shoalwater Bay to Moreton Bay (SE Queensland). For such groupings, geographic distances were measured from the midpoint of the range of samples.

Results

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information
Phylogeny and Recognition of Lineages

Sequences reported here are available from GenBank with the accession numbers EU835761–EU835816. The sequence alignment of 411 sites from 188 dugongs comprised 60 variable sites (including one single-site indel) and 56 haplotypes. Fewer than half of the haplotypes (25) occurred in more than one individual (Table S1, Fig. 3). The most common haplotype (h1) was within the restricted lineage and occurred in 31 individuals. Thirteen haplotypes belong in the restricted lineage (95 individuals) and 34 haplotypes in the widespread lineage (81 individuals including one from New Caledonia). Trees generated by BEAST provided posterior support probabilities of 1 for the widespread lineage and 0.9988 for the restricted one (data not shown). Figure 1 shows the numbers of representatives of each lineage from each sampled locality in Australia and Table S1 gives details on distribution of each haplotype.

image

Figure 3. Median-joining network showing the relationships between haplotypes, the number of sequences representing each haplotype (given in parentheses if greater than one) and the geographical origin of each. “Central E coast Queensland” includes all populations from Hervey Bay to Cooktown/Starcke (see Fig. 1 for locations). Widespread and restricted lineages are indicated. Numbers of mutations inferred as occurring along each branch are indicated by slash marks.

Download figure to PowerPoint

The only sequences from Australian waters that did not belong to the widespread or restricted lineages were from two dugongs from Ashmore Reef (Fig. 1, 3), which lies on the edge of the Australian continental shelf almost 400 km off Western Australia and ~120 km from Timor from which it is separated by a deepwater trench. A third dugong from Ashmore Reef carried a sequence representative of the widespread lineage. The Australian lineages were represented outside Australia by a single sequence (h44, widespread lineage) from New Caledonia, about 1,500 km east of the closest part of Queensland. All remaining sequences from outside Australia form a loose cluster in Figure 3, but given the diversity they exhibit and the very limited sampling, this cluster may not represent a single lineage (hence we did not present neutrality indices and some other analyses for this lineage alone).

Demographic Analyses

Overall, dugongs exhibit high haplotypic diversity (0.946) and rather low nucleotide diversity (0.026) (Table 2). The restricted lineage, despite including a larger number of samples, displays much lower haplotypic diversity and nucleotide diversity than the widespread lineage (Table 2).

Table 2. Summary statistics and neutrality indices for the mitochondrial control region sequences from dugongs
Lineage# segregating (variable) sites# haplotypesHaplotypic diversity hNucleotide diversity πAverage pairwise nt differencesFu's FSa R 2 b
  1. Note: *indicates 0.05 >  > 0.01, **indicates < 0.001.

  2. a

    Calculated in Arlequin. Significance tested by simulation (10,000 replicates). A significant value implies population growth.

  3. b

    Calculated in DnaSP v4.10. Significance tested by simulation (10,000 replicates). A significant value implies population growth.

  4. c

    Including haplotype h44 from New Caledonia.

  5. d

    Excluding haplotype h44 from New Caledonia and including a single-site indel in a member of the widespread lineage.

Widespreadc (81)26340.9450.0083.4−25.08 **0.065
Restricted (95)15130.8280.0062.36−2.1370.078
Australiad43460.9380.0249.845  
All othersd (12)2490.9550.0198.09  
Total (188)60560.9460.02610.64  

When the data were explored for evidence of population growth, strikingly different results were obtained for each lineage. Runs in Beast rejected the hypothesis of constant population size for the widespread lineage but not for the restricted one. The neutrality indices (Fu's FS and R2; Table 2) did not support population growth for the restricted lineage. However, the highly significant value for Fu's FS statistic indicates that the widespread lineage has experienced growth. Values for the R2 statistic did not reject the null hypothesis of constant population size in either lineage and was our only evidence against growth in the widespread lineage. The Bayesian skyline plot (Fig. 4a) suggests recent expansion for the widespread lineage after a period of near-stasis. Both median and mean values for effective population size (NE(FEMALE)) through time are shown in Figure 4a, b because they differ from one another more than we had expected and it is not clear which should be preferred. The present-day mean estimate of NE(FEMALE) in the widespread lineage is 16,646 (but with a very wide 95% HPD of 1,121–73,358). The present-day median value is 9,571. For the restricted lineage, the Bayesian skyline plot (Fig. 4b) suggests that there has been little change in population size. However, several runs, totaling hundreds of millions of generations, had to be combined to bring the ESS for some parameters close to the recommended minimum value of 100 for this lineage. This suggests that the data are inadequate to recover a strong signal for this lineage.

image

Figure 4. Bayesian skyline plots: (a) widespread lineage and (b) restricted lineage. Population size change over time is indicated (mean is solid thick line and median is thick dashed line), with 95% HPD intervals (thin dashed lines). The X-axis shows time in years back to a specified cut-off. The Y-axis (logarithmic scale) indicates NE(FEMALE) estimates multiplied by generation time (25 yr: Marsh et al. 2011).

Download figure to PowerPoint

When the tMRCA for the two Australian lineages was set to 115 kya, a mutation rate of around 25% per million years (95% HPD ~14%–37% per million years) was inferred.

Genetic Structure within Lineages and Isolation by Distance

All analyses indicate that there is genetic structure within each lineage. Values for FCT and, consequently, of FST, calculated using AMOVA, were always significant (Table 1), indicating that there is significant differentiation among regions and among populations across regions. Values for FSC were never significant, implying little differentiation between populations within a given region. However, the number of samples varied substantially across populations and many populations were small, limiting statistical power. Population pairwise FST values for each lineage are shown in Table S2, S3. Many pairs of populations are significantly differentiated, but rarely those within any regional grouping used in the AMOVA analyses. The spatial sampling of individual haplotypes, especially in the widespread lineage, must have had a strong influence on this analysis. For example, all three haplotypes present in Blue Mud Bay were also found in Shoalwater Bay (over 4,000 km away along the coast), which explains the apparent lack of differentiation between these localities (Table S2). Other examples of widely distributed haplotypes can be found in Table S1. It is striking that representatives of the restricted lineage in Torres Strait (including one representative from Blue Mud Bay in the Northern Territory) form a population strongly differentiated from dugongs of the same lineage in southern and central Queensland (Table S3).

The Mantel tests comparing pairwise population genetic distances with geographical distances (Fig. S1) suggested a degree of isolation by distance, but none was significantly different from null expectations regardless of the approach used to estimate geographical distances. When Mantel tests were done using individuals rather than populations and a genetic distance matrix based on pairwise numbers of differences between sequences, significant isolation by distance was implied for each lineage ( 0.001 in each case) (Fig. S1).

Discussion

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

Australian dugongs, with the exception of two individuals from Ashmore Reef, fell into two distinct maternal lineages. The widespread lineage occurs throughout the dugong's Australian distribution but is rare in Moreton Bay. The restricted lineage occurs along the east coast of Queensland and in Torres Strait, but rarely further west (Fig. 1). The data indicate a persisting genetic signature of past events. The emergence of the Torres Strait landbridge at about 115 kya is the most likely cause of geographical separation and we used this event as a calibration point. Population growth was detected in the widespread lineage, but there is no evidence for growth in the restricted lineage. Genetic structure can be detected within each lineage.

Comparisons with Other Sirenians

Australian dugong populations exhibit high levels of genetic variation in the mitochondrial control region (Table S4) in comparison with other sirenians. Overall haplotypic diversity is greater than in any species of manatee (genus Trichechus) (Table S4). The West Indian manatee, Trichechus manatus, has quite high overall values of h and π. However, this species, like the dugong, has several (three recognized by Vianna et al. 2006) mitochondrial lineages with varying degrees of geographic overlap. The considerable divergence between these lineages means that the overall values for genetic diversity are far greater than that seen in any single lineage. Diversity indices for individual lineages of T. manatus are particularly low (Table S4).

Genetic Patterns: Scale of Population-genetic Structure Reflecting Movement of Animals

Our analyses suggest population-genetic structure exists, but at large geographic scales (i.e., hundreds to thousands of km). The AMOVAs indicate significant differentiation among geographically defined regional groups of populations and (consequently) among all populations. Many population-pairwise values of FST were significant, but rarely those within any regional grouping used in the AMOVA analyses. Isolation by distance was not found when the genetic distance matrix was calculated for populations. However, a significant result was obtained when Mantel tests were conducted using individuals rather than populations. Only a single individual of the restricted lineage has been found west of Torres Strait to date, implying an historical range on the east coast of Queensland. This lineage also exhibits significant differentiation of populations either side of the Townsville-Cooktown-Starcke coastal tract from which it is hardly represented in our samples. We have no samples as yet from the region between Cape Melville (just north of Starcke) and Torres Strait—a straight-line distance of about 530 km that is known to support a population of several thousand dugongs (Marsh et al. 2002). It remains unclear whether the scarcity of members of the restricted lineage between Townsville and Starcke is merely a sampling artefact.

Dugongs are capable of long distance movement (Sheppard et al. 2006). They can also transit deep water occasionally, as demonstrated by the presence of typically Asian haplotypes at Ashmore Reef and the presence of a member of the widespread lineage in New Caledonia. Other evidence of deepwater crossings has been reported by Marsh et al. (2002, 2011) and Hobbs et al. (2007). Given the continuous shallow-water habitat now present throughout the dugong's Australian range, why is there not more complete geographical mixing of mitochondrial lineages? We have no clear answer to this question: it is possible that barriers or filters to movement, of which we are unaware, occur at some points around northern Australia. These barriers could be behavioral as well as physical. Marsh et al. (2011) discussed maternally transmitted learned behavior in sirenians including intriguing observations that suggest that the use of space may follow matrilines. However, sex-biased dispersal, as has been noted for the Florida manatee (Bengston 1982, cited in Garcia-Rodriguez et al. 1998), is not likely to provide an explanation. Both male and female dugongs have been recorded as traveling long distances, but seasonal or other patterns have not been detected, possibly due to inadequate sample sizes (Sheppard et al. 2006). Addition of nuclear markers will help to clarify the situation.

Genetic Patterns: Demography and Historical Population Size

Runs in BEAST (including BSPs) and values of Fu's FS indicate growth in the widespread lineage. BSPs indicate this has occurred primarily since the LGM. Only the R2 statistic failed to find evidence for population growth in this lineage. This statistic is held to be sensitive to population growth over a broad range of conditions, but generally performed less well than Fu's FS when sample sizes are as large as in this study (Ramos-Onsins and Rozas 2002). Beyond this, we are unable to say why the R2 statistic gave this result with our data. Methods that take into consideration underlying genealogy are much better at detecting demographic change than those that do not (Lohse and Kelleher 2009). Analyses implemented in Beast do consider genealogy (Ho and Shapiro 2011) (and require phylogenetic signal to be present in the data), whereas R2 and FS do not. In contrast to the widespread lineage, no analyses provided evidence of growth in the restricted lineage. We suspect that our data are not very informative for this lineage because of the small number of haplotypes represented (13) and their high level of similarity to each other. Heller et al. (2008) have discussed a similar problem in relation to their study on African buffalo (Syncerus caffer). Runs in BEAST for the restricted lineage, including those used to generate BSPs, did not mix well and required very large numbers of generations to reach an acceptable ESS. This was in contrast with runs for the widespread lineage, which mixed well and yielded high values for ESS. The form of the genealogy inferred for the restricted lineage in Figure 3 (common and similar central haplotypes from which a few new haplotypes are separated by only one or two mutations) suggests a population just starting to recover from a bottleneck during which haplotypic and nucleotide diversity had diminished (see e.g., Korsten et al. 2009).

The available means for estimating effective population size are dependent on the value chosen for the mutation rate: higher mutation rates imply smaller values for NE. By setting the divergence time of the two Australian lineages at 115 kya, a mutation rate of about 25% per million years was inferred. Short-term mutation rates in the control region at the level of population or species comparable with this rate have been noted for other mammals (e.g., Shapiro et al. 2004, Ho et al. 2007, Saarma et al. 2007, de Bruyn et al. 2009, Korsten et al. 2009, Phillips et al. 2009). Tikel (1997), based on scant fossil evidence, estimated a mutation rate for the control region of dugongs of 2% per million years. If this rate is used, then all estimates of NE will be ~12 times greater. This or similar rates have been used in studies on other sirenians yielding values for NE that are very, perhaps unrealistically, high. For example, Cantanhede et al. (2005) estimated NE(FEMALE) of 454,600 for the Amazonian manatee and values of around 90,000 for each of the T. manatus lineages.

The relationship between NE and census population size is not simple (Charlesworth 2009). Baleen whales have life histories comparable to that of dugongs: age at first parturition is at least several years with single calves produced at intervals of one to several years; longevity is many decades. Roman and Palumbi (2003) and Alter et al. (2012) suggested that total population sizes of baleen whales should be about six times the value for NE(FEMALE), although other studies suggest the multiplier should be larger (e.g., Frankham 1995). Using a multiplier of 6 and the NE(FEMALE) values from BSPs, the current mean census population size of the restricted lineage is estimated to be 15,403 (95% HPD 238–84,555) and that of the widespread lineage 96,000 (95% HPD 6,726–440,148), summing to an Australian mean total of ~111,500.

Aerial survey estimates of dugongs in Australian waters sum to ~85,000–100,000 animals (Marsh et al. 2002). This is slightly lower than our mean census estimates based on the mutation rate of nearly 25% per million years, but not all the dugong habitat in Australia has been surveyed from the air and such surveys underestimate absolute population size (Marsh et al. 2011). In addition, values derived from mitochondrial sequence data represent long-term estimates that will not reflect recent anthropogenic population declines (Roman and Palumbi 2003).

Phylogeographic Interpretations

We suggest that the penultimate flooding of Torres Strait was the key event producing the genetic patterns that we have reported. That the two Australian lineages are nearest sisters to each other, required under this scenario, is consistent with our data. In this scenario, dugongs were probably not present in what is now the Great Barrier Reef (GBR) region immediately prior to ~125 kya, presumably because of limited suitable habitat (Fig. 2). Following the last interglacial warm period (peaked at about 120 kya), during which a continuous population of dugongs of a single lineage probably spanned the present-day range of the species in Australia, falling sea levels at about 115 kya (Fig. 2) separated eastern and western populations. East coast populations were subsequently likely small, effectively bottlenecked and restricted to limited areas in what is now the southern GBR region,1 and perhaps the Coral Sea plateaus (Wörheide et al. 2002), and would have been restricted to these refugia until shortly before the final inundation of the Torres Strait land bridge about 7,000 yr ago (Fig. 2). In contrast, the areas of occupancy and sizes of the populations west of Torres Strait must have been larger after 115 kya (Fig. 2). Consequently, the western (“widespread”) lineage contains many more haplotypes and exhibits greater haplotypic and nucleotide diversity (Table 2) and has a longer history of population growth. The finding of identical haplotypes on either side of Torres Strait at widely separated localities (Table S1) suggests that the east coast representatives of the widespread lineage are descended from individuals migrating there since the flooding of the land bridge some 7,000 yr ago and the development of suitable habitat, which might not have occurred until about 4,000 yr ago in Torres Strait (Crouch et al. 2007).

Other scenarios must also be considered. A mutation rate for the mitochondrial genome of around 2% per million years (Brown et al. 1979) has long been used as a rule-of-thumb when exploring divergence times of mammal species. The lower mutation rate yields figures for dugong NE that are more than an order of magnitude greater than present-day census estimates. Furthermore, such rates imply that the Australian mitochondrial lineages coalesce over a million years ago. There have been many glacial-interglacial cycles since that time and no clear reason why events so far back in time would have produced a still-detectable signal whereas much more recent cycles have not. For these reasons, we do not favor this scenario as an explanation for the genetic structure reported here.

Conclusions

Data from the mitochondrial control region have demonstrated the presence of two maternal lineages in Australian dugongs. Analyses of these data show a phylogeographic pattern consistent with Pleistocene sea-level fluctuations. This pattern can still be discerned despite the potential for geographic mixing of dugong populations to either side of Torres Strait for about the last 7,000 yr. Within each lineage, genetic structure exists albeit at large scales, but demonstrating that gene flow remains restricted. These results strengthen the arguments by Marsh et al. (2011) for the need to assess the eligibility of the dugong for listing under national and state legislation in Australia at regional scales and to customize the management approach to the regionally diverse impacts. Further research using nuclear markers is required to identify the appropriate management units.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

This work was supported by funding provided through an Australian Research Council—Strategic Partnerships with Industry–Research and Training grant (Grant number C00002084), Funding was also provided by the Australian Fisheries Management Authority, James Cook University and an anonymous donor. Material in-kind support was provided by the Great Barrier Reef Marine Park Authority and Queensland Environmental Protection Agency. This research was conducted while BM was the recipient of an Australian Postgraduate Award (Industry). We thank the following people and their organizations for samples: Kanjana Adulyanukosol, Lem Aragones, Potchana Boonyanate, John Bowen, Hans de Iongh, Nick Gales, Claire Garrigue, Caroline Gaus, Bruce Hill, Donna Kwan, Ivan Lawler, Col Limpus, David Parry, Robert Prince, Mark VanderWal, and Scott Whiting; David Savage and others at QPWS, Drs. Rachel Bowater and Steve Johnson, and others at the Queensland Department of Primary Industries Oonomba Veterinary Laboratory; Marcus Barber, Dave Holley, Duncan Limpus, James Sheppard, and members of the Mabiaug, Badu, and Boigu communities in Torres Strait. We also thank Drs. David Hopley and Scott Smithers for advice on sea levels around Australia during the Pleistocene, Dr. John Guinotte for the sea level maps, Adella Edwards for help with figures, and Alana Grech for calculating the distances between sampling locations. Thanks also to Rod Peakall, Alexei Drummond, and Simon Ho for advice on portions of the population-genetic analyses and to Vimoksalehi Lukoschek and anonymous referees for comments on the manuscript. The High Performance Computing cluster at James Cook University made analysis in BEAST possible.

  1. 1

    Personal communication from D. Hopley, Adjunct Professor, James Cook University, Townsville, Queensland, Australia, November 2012.

Literature Cited

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Materials and Methods
  4. Results
  5. Discussion
  6. Acknowledgments
  7. Literature Cited
  8. Supporting Information
FilenameFormatSizeDescription
mms12022-sup-0001-Supplementaryfile1.docxWord document140K

Supplementary File 1 contains:

Table S1. Sample numbers, localities and haplotypes found.

Table S2. Pairwise population FST values for the widespread lineage.

Table S3. Pairwise population FST values for the restricted lineage.

Table S4. Comparisons with other sirenians.

Figure S1. Representative graphs generated from Mantel tests.

mms12022-sup-0002-Supplementaryfile2-6.docxWord document10KSupplementary files 2–6 BEAST files. Note that names for sequences in these files are the original working names and that the user will need to enter chain length, logging frequency and output file names in the MCMC section (indicated in the files by “xxxx” or user_specify”) before any of the files will run.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.