High genetic diversity in a remote island population system: sans sex


Author for correspondence:
Eric F. Karlin
Tel: +1 201 684 7743
Email: ekarlin@ramapo.edu


  • It has been proposed that long-distance dispersal of mosses to the Hawaiian Islands rarely occurs and that the Hawaiian population of the allopolyploid peat moss Sphagnum palustre probably resulted from a single dispersal event.
  • Here, we used microsatellites to investigate whether the Hawaiian population of the dioicous S. palustre had a single founder and to compare its genetic diversity to that found in populations of S. palustre in other regions.
  • The genetic diversity of the Hawaiian population is comparable to that of larger population systems. Several lines of evidence, including a lack of sporophytes and an apparently restricted natural distribution, suggest that sexual reproduction is absent in the Hawaiian plants. In addition, all samples of Hawaiian S. palustre share a genetic trait rare in other populations. Time to most recent ancestor (TMRCA) analysis indicates that the Hawaiian population was probably founded 49–51 kyr ago.
  • It appears that all Hawaiian plants of S. palustre descend from a single founder via vegetative propagation. The long-term viability of this clonal population coupled with the development of significant genetic diversity suggests that vegetative propagation in a moss does not necessarily preclude evolutionary success in the long term.


Long-distance dispersal of plants to the Hawaiian Islands, which are the most remote high volcanic islands in the world, is thought to be a rare occurrence. It is believed that there have been only 270–280 successful flowering plant colonists there, with subsequent adaptive radiation resulting in at least 956 species (Wagner et al., 1990). Hoe (1979) determined that there had been a similar number of successful Hawaiian moss colonists, with 225 original immigrants being sufficient to account for the 233 moss species native to Hawaii. Noting that the endemism of mosses in the Hawaiian Islands was the highest reported for a moss flora of any well-documented area and a notable absence of several moss taxa (at the family, genus, and species levels) which could reasonably be expected in Hawaii, Hoe (1979) concluded that long-distance dispersal of mosses to Hawaii was a rare event.

The dioicous Sphagnum palustre L. is a member of subgenus Sphagnum (Shaw et al., 2010) having a discontinuously circumboreal to temperate and subtropical distribution in the Northern Hemisphere. The plant is strongly associated with suboceanic regions, and its distribution is both amphi-Pacific and amphi-Atlantic, with an isolated occurrence in Hawaii (Karlin & Andrus, 1995; Karlin, 2001). The type species for the genus, S. palustre has recently been shown to be an allopolyploid with fixed, or nearly fixed, heterozygosity (Shaw et al., 2008; Karlin et al., 2010).

Sphagnum palustre is indigenous to Hawaii (Karlin, 2001) where its natural distribution appears to be limited to a small area (< 200 km2) in the summit region of Kohala Mountain, at elevations ranging from 900 to 1670 m (Bartram, 1933; Degener et al., 1969; Karlin & Andrus, 1995). This mountain is the oldest volcano on the Island of Hawaii (the Big Island), with vulcanism ending some 120 kyr ago (Sherrod et al., 2007). Annual precipitation about the summit ranges from 2500 to > 3500 mm (Giambelluca et al., 1986), and supports montane rain forest (Mueller-Dombois & Fosberg, 1998). Sphagnum palustre is a prominent species there, often forming a thick (80+ cm) blanket across the ground (Fig. 1). The occurrence of S. palustre in 23.9-kyr-old peat from Kohala Mountain (Karlin, 2001; S. C. Hotchkiss, unpublished data) suggests that colonization occurred during the last glacial, a time when the climate of the summit region was quite different from that of the present, being much cooler and drier (Hotchkiss, 2004). Given the remoteness of the Hawaiian Islands and the apparent lack of natural dispersal by S. palustre to other Hawaiian sites (Degener et al., 1969; Karlin & Andrus, 1995), Karlin (2001) concluded that the Hawaiian population probably resulted from a single dispersal event, with one diaspore serving as the founding agent.

Figure 1.

Sphagnum palustre hummocks in a Hawaiian montane rain forest. Photograph by S. C. Hotchkiss.

The plant’s abundance on Kohala Mountain has increased significantly within the past 100+ yr (Bartram, 1942; S. C. Hotchkiss & P. Vitousek, unpublished data). Anthropogenic activities have played a major role in both the plant’s increase on Kohala Mountain as well as its establishment at other Hawaiian sites, where it has flourished (Degener et al., 1969; Hoe, 1971; Karlin & Andrus, 1995; Waite, 2007; S. C. Hotchkiss et al., unpublished data; S. Joe & C. Smith, unpublished data). Two human activities in particular stand out. One is the harvesting of S. palustre on Kohala Mountain, which occurred as early 1919. The plants were harvested with minimal processing and used as packing for the roots of tree seedlings before planting at sites throughout the Big Island. This practice lasted at least into the early 1960s (Degener et al., 1969), being but one part of a large-scale forestry program which resulted in > 12 million trees (mostly non-native species) being planted in the Hawaiian Islands, primarily in the 1920s and 1930s (Bryan, 1947; Woodcock, 2003). In addition to providing an excellent mechanism for dispersal to a multitude of Big Island sites outside of Kohala Mountain, the harvest of live S. palustre for horticultural use would also have enhanced the plant’s dispersal on Kohala Mountain. Plant fragments from any given site likely adhered to shoes and boots and were dispersed to other Kohala locations. In addition, plants and plant fragments undoubtedly fell off transport vehicles as Sphagnum was moved from collection sites on Kohala Mountain to the government nursery located in Hilo (Degener et al., 1969). Thus, Sphagnum‘litter’ probably accumulated along the roads and trails on Kohala Mountain, and vegetative propagation in suitable habitats would result in the establishment of S. palustre along these pathways. Karlin (2006) discusses in depth the putative impacts harvesting and horticultural use have had on the distribution and ecological status of S. portoricense Hampe (subgenus Sphagnum) in Puerto Rico. The second notable human action was the deliberate transplantation of S. palustre from Kohala Mountain to the Mt Ka’ala Reserve on the island of Oahu in the late 1960s (Hoe, 1971). The growth of S. palustre at the latter site has been explosive, disrupting the ecological balance of a fragile habitat supporting a number of endangered species (Karlin & Andrus, 1995; S. Joe & C. Smith, unpublished data). In the 40+ yr since the plant’s introduction, a 1.25-ha Sphagnum carpet has developed at the site despite more than a decade of effort to arrest its rapid spread (Karlin & Andrus, 1995; S. Joe & C. Smith, unpublished data).

Some mosses are noted for their long-distance dispersal capacity, with recent studies indicating rapid and effective colonization over great distances within short time frames (Hassel & Söderström, 2005; Karlin et al., 2011). Microsatellite analysis has shown that multiple long-distance colonizations gave rise to the population of S. palustre in the Azores, which, like the Hawaiian Islands, is a highly isolated archipelago (H. K. Stenøien et al., unpublished data). Genetic study has also suggested that the moss Grimmia montana Bruch & Schimp. in the Canary Islands and Maderia (both part of Macaronesia) resulted from multiple long-distance colonization events (Vanderpoorten et al., 2008). These studies suggest that it is possible for a moss species indigenous to Hawaii to be represented by two or more independent colonizations resulting from long-distance dispersal.

We used microsatellites to investigate whether the Hawaiian population of S. palustre resulted from a single founding event and to compare the genetic diversity of this population to that of more extensive population systems of S. palustre and the closely related allopolyploid S. cristatum Hampe (Karlin et al., 2008) in other regions.

Materials and Methods


All genetic analyses were based on gametophytes. A total of 54 gametophytes of S. palustre from eight locations in Hawaii were screened for 16 microsatellite markers (Supporting Information Table S1). Thirty-four specimens were from six sites on Kohala Mountain, six were from Wailuku, a site on Mauna Kea, and 14 were from the Mt Ka’ala Reserve on Oahu (Table 1). The specimens covered the range of morphological variation observed in Hawaiian S. palustre. Described by Karlin & Andrus (1995) and Karlin (2001), this morphological variation includes differences in pigmentation, the porosity of branch leaf hyaline cells, the porosity of stem cortical cells, and the shape of branch leaf chlorophyll cells. Collections were typically made from different patches with a minimum interpatch distance ≥ 1 m. Most specimens were < 10 yr old at the time of DNA extraction; the oldest was collected in 1976. To compare the genetic diversity of Hawaiian population systems of S. palustre to that of other regions, an extended microsatellite data set of 107 additional specimens of S. palustre and S. cristatum Hampe from other regions was used: 22 samples from 15 sites in eastern North America, 14 samples from four sites in western North America, 33 samples from four sites in eastern Asia, and 38 samples of S. cristatum from 13 sites on South Island, New Zealand. Data from European plants were not available. The majority of these data came from previously published studies (Karlin et al., 2008, 2010; Shaw et al., 2008), and were augmented by microsatellite analysis of 14 specimens of S. palustre from western North America and three specimens of S. palustre from eastern Asia (Table S1).

Table 1.   Collection sites for the Hawaiian specimens of Sphagnum palustre used in this study, including the number of samples (n), number of multilocus haplotypes (MLHs), and the multilocus haplotype clusters (MHLCs) for each site
SitenMLHsMLHCsIslandLocationLatitudeLongitudeElevation (m)
  1. MLHCs in parentheses probably belong to the indicated MLHC, but are ambiguous because of null alleles. MLHCs with no prefix belong to subgroup A, those with the prefix ‘B’ belong to subgroup B, and the one with the prefix ‘C’ belongs to subgroup C.

HI-1323, 6HawaiiKohala Mt20.0464°N155.6238°W903
HI-221133,4,5,6,7,8,9, (B1)HawaiiKohala Mt20.0698°N155.6700°W1180
HI-353(6), 7, C1HawaiiKohala Mt20.0568°N155.6890°W1221
HI-422B1HawaiiKohala Mt20.1111°N155.7306°W1271
HI-5221, B1HawaiiKohala Mt20.1056°N155.7028°W1300
HI-611(2)HawaiiKohala Mt20.0889°N155.6875°W1315
HI-765B2, B3HawaiiWailuku19.7057°N155.2708°W1107
HI-81431, 2OahuMt Ka’ala21.5024°N158.1486°W1160

All available Hawaiian specimens of S. palustre at the following herbaria were examined: SUNY Binghamton (BING), Bishop Museum (BISH), Duke University (DUKE), Harvard University (FH), University of Michigan (MICH), Missouri Botanical Garden (MO), New York Botanical Garden (NY), and the Smithsonian Institution (US).

Microsatellite genotyping

DNA extractions were accomplished according to the protocols described by Shaw et al. (2003). Primer sequences and microsatellite characteristics for the markers analyzed in this study were described by Shaw et al. (2008). The following 16 microsatellite markers, numbered as in Shaw et al. (2008), were assayed for the present study: 1, 4, 5, 7, 9, 10, 12, 14, 17, 18, 19, 20, 22, 28, 29 and 30. Microsatellites were amplified in 8 μl of multiplexed reactions, each targeting a set of three loci. Primer sets were arrayed for multiplexing according to expected fragment sizes (for nonoverlapping amplification products) and alternating fluorophores. Each primer pair included a forward primer fluorescently labeled with the dyes HEX or 6-FAM (Integrated DNA Technologies, Coralville, IA, USA). Multiplexing was accomplished using a Qiagen Multiplex PCR kit (Qiagen, Valencia, CA, USA), scaled for smaller reactions, but otherwise used according to the manufacturer’s recommendations. Five to 20 ng of genomic DNA in 3 μl of distilled H2O served as the template in each reaction. A standard thermocycling regime was implemented for all primer sets, with no additional optimization. This consisted of an initial denaturation and hot-start activation at 95°C for 15 min, then 30 cycles of 94°C for 30 s, 54°C for 90 s, and 72°C for 60 s, with a final extension at 60°C for 30 min. PCR products were diluted in sterile water, and 1.2 μl of the dilution was mixed with GS500 size standard and Hi-Di Formamide (Applied Biosystems, Foster City, CA, USA) for electrophoresis on an ABI 3730 sequencer (Applied Biosystems, Foster City, CA, USA). Size determinations and genotype assignments were made using GeneMarker 1.75 software (Softgenetics, State College, PA, USA).


The microsatellite data were treated as ‘diploid’ for the calculation of heterozygosity but as haploid (Karlin et al., 2009) for all other calculations. Aside from the determination of heterozygosity, in which the presence of one allele at a marker was considered to represent a ‘homozygous’ state, a sample was considered to have a null allele in one of the two component genomes (i.e. it was not treated as being homozygous) when a marker typically having two alleles per sample had just one (Karlin et al., 2009). Because of differences in allele sizes, it was possible to treat the data as haploid for markers having two alleles amplified per sample. Such markers were thus considered as representing two loci (e.g. locus 10a and 10b). Although usually not an issue, when it was not possible to unambiguously assign an allele to a locus we arbitrarily assigned that allele to a locus based on achieving the minimum genetic diversity. Markers typically having one allele per sample were treated as one locus. We use the term ‘marker’ (e.g. ‘marker 10’) when we consider all of the alleles amplified per individual at a given microsatellite marker to be a collective unit (‘diploid’ data), and ‘locus’ when the data were treated as being haploid. This was deemed a suitable approach given that gametophytes were sampled and that S. palustre is an allopolyploid having fixed heterozygosity. It was also required to treat the data as being haploid in order to calculate the time to most recent ancestor (TMRCA). If the data were treated as diploid one would be determining the time to the last common ancestor of the two parent species of S. palustre and not the TMRCA for the population being studied. A locus was considered to be polymorphic when the frequency of its most common allele was ≤ 95%.

A data set of 27 multilocus haplotypes (MLHs) with 8.0% missing data across 24 loci was obtained for the Hawaiian plants (excluding copies and ambiguous MLHs – those having the same alleles as another MLH but with one or more null alleles). Eighteen of these MLHs had no missing data across the eight polymorphic loci detected in Hawaiian S. palustre (see the Results section), and this eight-locus data set was used for analyses of population structure. Fifteen of the 27 Hawaiian MLHs were represented by samples having no missing data across all 24 loci.

Microsatellite data were analyzed as fragment sizes; alleles were coded as numbers of nucleotides rather than repeat numbers. Most statistical calculations were made with GenAlEx 6.4b5 (Peakall & Smouse, 2006). These included Shannon’s mutual information index (MI; Sherwin et al., 2006; Sherwin, 2010), Nei’s genetic distance (Nei, 1972), and PCO (Orloci, 1978).

Population structure was tested for using PCO and the program Structure 2.3.3 (Pritchard et al., 2000; Falush et al., 2003). PCO was based on a distance matrix generated by the ‘Haploid’ genetic distance option in GenAlEx (Schneider et al., 1997). This distance measurement generates a pairwise, individual-by-individual genetic distance matrix for multilocus haplotypes based on haploid data. In Structure, four models were tested: ‘admixture with correlated allele frequencies’, ‘admixture with independent allele frequencies’, ‘no admixture with correlated allele frequencies’, and ‘no admixture with independent allele frequencies’. Two runs were performed for each K (estimated number of populations) from 1 to 10, with 100 000 steps of burn-in followed by 200 000 replications. The K having the highest ‘estimated log probability of data’ (logeP(D)) was determined to be the best fit for the data (Kbest). A total of 10 runs were then performed on Kbest−1, Kbest, and Kbest+1 to further test this conclusion. Genetic divergence between groups was measured using Nei’s genetic distance and MI, which was calculated using log base = 2. MI ranges from 0.0 (all genetic diversity within groups) to 1.0 (all genetic diversity between groups). For statistical comparisons (PCO, Structure and MI) with populations from other regions marker 30 was treated as being homozygous in the Hawaiian plants (i.e. loci 30a and 30b were assigned the same allele).

Genetic diversity within each of the five regional populations was assessed using four indices: mean number of alleles over loci, mean effective number of alleles over loci, number of polymorphic loci, and the maximum number of pairwise differences. The allelic information index (I; Brown & Weir, 1983) was expressed as its effective numbers equivalent, which is 2I (Jost, 2006). As the data were treated as being haploid, the diversity analyses focused on within-genome variability; between-genome variation was ignored. Thus the genetic diversity values were lower than they would have been if the data had been analyzed as diploid.


Upper and lower bounds of mutation rates were estimated based on the minimum and maximum possible ages of the clone (Ally et al., 2008, 2010). The upper bound was set to the oldest fossil date for Hawaiian S. palustre (23.9 kyr; Karlin, 2001; S. C. Hotchkiss, unpublished data) and the lower bound was set to the date of cessation of vulcanism for Kohala Mountain (120 kyr; Sherrod et al., 2007). Because we found that mutations accumulated in a manner consistent with a star-like phylogeny (Calafell et al., 2002) in Hawaiian S. palustre, the probability that a mutation had accumulated at a locus in either of two haplophasic ramet lineages is expected to equal 2μTCA, where μ is the mutation rate per haplophasic ramet per locus per year and TCA represents the clone age (Ally et al., 2008, 2010). By setting the age of the most divergent pair of MLHs detected among the Hawaiian samples (πmax = pairwise differences averaged over all loci) to the lower and upper bounds established above, we used πmax = 2μTCA to obtain estimates for the lower (μlower) and upper (μupper) bound mutation rates across loci (Ally et al., 2008, 2010). For TMRCA analyses, the mean of these two rates (μmean) was used as the mutation rate per year across loci. Assuming that all population growth was a result of vegetative propagation, the TMRCA was calculated using the method of Walsh (2001) using the infinite alleles model. The TMRCA was calculated for the most divergent pairs of MLHs detected both between and within the two subgroups (see the Results section) occurring in Hawaii.


In the Hawaiian plants, observed heterozygosity ranged from 0.92 to 1 at eight markers and appeared to be fixed, or nearly so. The other eight markers in the Hawaiian plants had one allele per sample. Plants from all regions showed fixed, or nearly fixed, heterozygosity at the same eight microsatellite markers (Fig. 2). However, while all of the Hawaiian samples (except two which had no amplicons at this marker) had just one allele per individual at marker 30, 99.0% (103/104 samples) of plants from other regions had two alleles per individual at this marker (Fig. 2). Thus, 24 loci were detected among the 16 microsatellite markers in the Hawaiian plants and 25 loci were detected among the plants from other regions. For the Hawaiian samples, the number of alleles per locus ranged from one to 12, with 15 loci being monomorphic (having one allele). Four loci were fixed across all samples of S. palustre and 11 additional loci were fixed in the Hawaiian plants but polymorphic (most common allele having ≤ 95% frequency) for all samples of S. palustre combined. Nine loci in the Hawaiian plants each had two or more alleles, with the most hypervariable being locus 14. Eight of the nine loci in Hawaiian plants having ≥ 2 alleles were polymorphic (Table S2).

Figure 2.

Observed heterozygosity across nine microsatellite markers in Sphagnum palustre from Hawaii (green columns) compared with S. palustre (from eastern Asia, eastern North America and western North America) and S. cristatum (from South Island, New Zealand) from other regions (blue columns). Note that observed heterozygosity in Hawaiian plants at marker 30 was 0.00 while it was 0.99 in the plants from other regions.

Analysis with both PCO (Fig. 3a) and Structure using only unique MLHs detected two subgroups in the Hawaiian plants. The primary axis of the PCO-based ordination (with 41.5% of the genetic variation) resolved subgroup A on the left side of the ordination from subgroup B on the right side of the ordination (Fig. 3a). The second axis (with 21.1% of the genetic variation) focused on genetic variability within the two subgroups. With Structure, only one of the four models tested (no admixture with independent allele frequencies) detected any structure, with Kbest = 2. The two subgroups detected by Structure were identical to those resolved by PCO. Genetic divergence between these two subgroups, as measured by both MI and Nei’s genetic distance, shows them to be more closely related to each other than they were to other regional populations (Table 2, Fig. 3b). Based upon the data set with no missing data across eight polymorphic loci, a maximum of seven pairwise differences (pwds) between members of the two subgroups was detected; the maximum number of pwds within each subgroup was four. Extrapolations based on MLHs with missing data at one or more of the nine loci with ≥ 2 alleles suggest a possible maximum of eight pwds between subgroups and five pwds within each subgroup. Based upon the data set with no missing data across eight polymorphic loci, the mean (± SE) number of pwds among MLHs within each subgroup (subgroup A: 3.2 ± 0.1; subgroup B: 2.9 ± 0.3) was roughly half that detected between subgroups (5.6 ± 0.1). A third subgroup (subgroup C) was represented by one MLH which could not be used in the above analyses because of null alleles at six loci. This MLH had alleles at seven polymorphic loci, with a mean of 4.8 ± 0.2 pwds with MLHs of subgroup A and 5.7 ± 0.2 pwds with MLHs of subgroup B. Allele frequencies for the three subgroups are listed in Table S2. All three subgroups were detected on Kohala Mountain (Table 1). Subgroups A and B also occurred at the two non-Kohala sites which were sampled (Table 1), with all of the samples from the Mt Ka’ala Reserve (site HI-8) belonging to subgroup A and all from Wailuku (HI-7) belonging to subgroup B (Table 1).

Figure 3.

Principal coordinates analysis of multilocus haplotypes (MLHs) of Sphagnum palustre. (a) Ordination of Hawaiian MLHs, with the primary axis having 41.5% of the genetic variation and the second axis having 21.1%. (b) Ordination of S. palustre MLHs from four different regions, with the primary axis having 36.4% of the genetic variation and the second axis having 19.2%. Squares, Hawaii subgroup A; circles, Hawaii subgroup B; diamonds, eastern Asia; red triangles, western North America; green triangles, eastern North America.

Table 2.   Pairwise population matrix of mean mutual information (MI) values (below diagonal) and Nei’s genetic distance (above diagonal) over 25 microsatellite loci for regional populations of Sphagnum palustre in Hawaii
PopulationHI-AHI-BHIe. Asiaw. NAe. NA
  1. HI, all Hawaiian samples; HI-A, subgroup A; HI-B, subgroup B; e. Asia, eastern Asia; w. NA, western North America; e. NA, eastern North America. MI was calculated using log base = 2.

e. Asia0.3190.3970.2880.6130.309
w. NA0.5520.6240.4710.4780.488
e. NA0.4510.3750.4790.3130.280

Many of the MLHs in each subgroup differed from other MLHs in that subgroup by just one pwd, and these were probably descended from the same ramet (Ally et al., 2008). We define a MLH cluster (MLHC) as a group of almost identical multilocus haplotypes, with each MLH having no more than one pwd with the MLH that is most similar to all other MLHs in that MLHC. Thus, pwds among MLHs comprising a given MLHC range from 0 to 2. There were nine MLHCs detected in subgroup A (out of 33 samples) and three MLHCs detected in subgroup B (out of seven samples). A total of 14 MLHs could not be placed in a MLHC because of null alleles at two or more polymorphic loci. Several MLHCs were detected at more than one site, with one co-occurring on Kohala Mountain and at the Mt Ka’ala Reserve (Table 1).

Sporophytes were not present in any of the Hawaiian herbarium specimens of S. palustre that were examined. No sporophytes were observed by the first two authors at any Big Island site and none have been detected at the Mt Ka’ala Reserve (HI-8) (S. Joe & C. Smith, unpublished data). Degener et al. (1969) also noted that plants collected in the Honaunau Forest Reserve on the Big Island lacked sporophytes. In addition, no Sphagnum spores have been found in peat from Kohala Mountain despite the occasional presence of Sphagnum macrofossils (Hotchkiss, 1998). Because antheridial and archegonial branches were lacking, it was not possible to determine the sex of Hawaiian plants. The absence of sporophytes at HI-8 suggests that the 1.25-ha carpet of S. palustre that formed there resulted solely from vegetative propagation (S. Joe & C. Smith, unpublished data). Microsatellite analysis of 13 samples collected in 2007 along a 1.2-km transect of the Sphagnum carpet at this site showed just three MLHs (MLH1, MLH2 and MLH3) to be present; MLH1 had seven copies, MLH2 had five copies, and MLH3 was a singleton. One additional MLH from this site was obtained from a 1976 specimen and it was a copy of MLH1. Thus, MLH1 has persisted at HI-8 for at least 31 yr. Excluding copies, there were one to four pairwise differences (pwds) among the three MLHs: MLH1 and MLH3 formed an MLHC, having a pwd at the hypervariable locus 14. This MHLC also included an MLH present on Kohala Mountain (Table 1). Both MLH1 and MLH3 had four pwds with MLH2, which represented a second MHLC. MLH2 closely matched an MLH present on Kohala Mountain, but null alleles in the latter precluded its placement in an MHLC. When compared with MLHs from other Hawaiian sites, the allele associated with MLH3 at locus 14 was the only allele private (unique) to HI-8; it is probably a mutation that arose in situ. Thus, the microsatellite data suggest that the founding plants at HI-8 were probably limited to two MLHs (MLH1 and MLH2) representing two MLHCs, and that subsequent population growth of S. palustre there was solely the result of vegetative propagation. This is consistent with the absence of sporophytes at this location.

All six samples from Wailuku (HI-7) belonged to subgroup B, with five MLHs being detected (one sample had an ambiguous MLH). Unlike the samples from HI-8, no MLH copies were detected at HI-7. Two MLHCs with three pwds between them were present at HI-7 and both MLHCs were unique to this site (no members were detected on Kohala Mountain).


As the four loci that were fixed across all samples of S. palustre (loci 1a, 12, 29a and 29b) appear to have alleles that have been stable (retained the same size) for longer than the species has existed, they were excluded from the TRMCA analyses. This left a total of 20 loci that were polymorphic in S. palustre (with samples from all regions sampled treated as one population) that could be used for TRMCA analysis of the Hawaiian plants. As the maximum pwd among Hawaiian MLHs was seven, πmax = 7/20 = 0.35. Upper and lower bound estimates of the microsatellite mutation rate were thus μupper = 7.3 × 10−6 and μlower = 1.5 × 10−6 per locus yr−1, with μmean = 4.4 × 10−6. Assuming only vegetative propagation, the most likely TMRCA for the most divergent pairs of MLHs detected among the Hawaiian plants was 49 kyr (Fig. 4). This represents the most likely time of the founding event for the Hawaiian plants. The most likely TMRCA for the most divergent pairs of MLHs within both subgroups A and B (based on four pwds) was 25 kyr (Fig. 4). This suggests that a genetic bottleneck may have occurred at this point in time, with at least three divergent MLHs surviving to give rise to the current population structure. Assuming that the maximum pwd between and within the two subgroups was eight and five pwd, respectively (based on extrapolations from MLHs with missing data), the corresponding most likely TMRCAs would be 51 and 29 kyr.

Figure 4.

Likelihood profile of estimated time to origin of the clone of Hawaiian Sphagnum palustre (red line) and time to genetic bottleneck (blue line). A mutation rate of 4.4 × 10−6 yr−1 was applied across 20 microsatellite loci that are polymorphic in S. palustre.

Three Hawaiian sites had a sufficient number of samples to allow a comparison of the genetic diversity of S. palustre among sites: HI-2 (Kohala Mountain), HI-7 (Wailuku), and HI-8 (Mt Ka’ala Reserve). Sphagnum palustre was a prominent species at each of these sites; all had also been affected by anthropogenic activity. HI-2 was situated along the Upper Hamakua Ditch as well as being adjacent to the terminus of a jeep road, HI-7 was associated with an intermittent pond and also along an adjacent jeep road, and HI-8 was associated with a trail/boardwalk and subject to many years of attempts to arrest the plant’s spread (see the Introduction section). Genetic diversity was highest across all six measures of diversity at HI-2 (Table 3). The two non-Kohala sites had comparable genetic diversities across all but one of the six measures of diversity. However, even though HI-7 had less than half as many samples as HI-8, a higher number of MLHs were detected at HI-7 (Table 3).

Table 3.   Mean (± SE) sample size (n), number of multilocus haplotypes (MLHs), number of MLH clusters (MHLCs), haplotype subgroup (HS), mean (± SE) number of alleles per locus (Na), mean (± SE) effective number of alleles per locus (Ne), and per cent polymorphic loci (% poly) across 24 microsatellite loci for Sphagnum palustre at three Hawaiian sites
SitenMLHsMLHCsHSNaNe% poly
  1. The total number of samples is indicated in parentheses, and may differ from ‘n’ because of null alleles. See Table 1 for site locations.

HI-216.5 ± 0.2 (19)13 8A and B1.8 ± 0.31.4 ± 0.329.2
HI-75.4 ± 0.2 (6)5 2B1.3 ± 0.11.2 ± 0.116.7
HI-813.0 ± 0.0 (13)3 2A1.2 ± 0.11.1 ± 0.116.7

The genetic diversity detected in the Hawaiian plants was much higher than that detected in S. palustre in western North America across the four diversity indices used; it was much lower than that detected in eastern North America (Fig. 5). The level of genetic diversity detected in S. palustre in eastern Asia and in S. cristatum on South Island was comparable to that detected in Hawaii (Fig. 5). The geographic range of the population systems studied varied, with the Hawaiian population having the smallest area; the maximum distance among sample sites on Kohala Mountain was 16 km. In contrast, the maximum distance among sites sampled in the other regions ranged from 645 km in north western North America to 2000 km in eastern Asia.

Figure 5.

Mean (+ SE) number of alleles (blue columns), mean (+SE) effective number of alleles (green columns), maximum number of pairwise differences among multilocus haplotypes within a region (solid black line), and number of polymorphic loci (dashed red line) across 25 microsatellite loci in Sphagnum palustre from four regions and S. cristatum from one region (South Island, New Zealand). WNA, western North America; EA, eastern Asia; HI, Hawaii; SI, South Island; ENA, eastern North America. Hawaiian S. palustre was based on 24 microsatellite loci.


All Hawaiian plants of Hawaiian S. palustre share a genetic trait that was either rare or not detected in the other regional populations that were studied. This strongly suggests the occurrence of a single founding event for the Hawaiian population, with the founder, by chance, having gene loss/silencing in one of the homeologous chromosomes associated with marker 30. Evidence of other cases of variation in gene loss/silencing among individuals within S. palustre was detected. For instance, although S. palustre typically has two alleles per plant at marker 18, four MLHs comprising an MHLC at one site in eastern North America each had one allele at this marker.

As S. palustre is dioicous, having one founder would result in the absence of sexual reproduction. The natural distribution of Hawaiian S. palustre is strongly consistent with this possibility. Habitats suitable for S. palustre occur outside of the summit region of Kohala Mountain, both on the Big Island and on other Hawaiian islands (Karlin & Andrus, 1995). However, these sites occur as high-elevation ‘sky islands’ of temperate or boreal climate (Baldwin & Wagner, 2010) separated by large expanses of hot, dry, and barren lava landscapes and/or ocean. Sphagnum spore dispersal is highly effective (Sundberg et al., 2006; Sundberg, 2010; Whitaker & Edwards, 2010; Karlin et al., 2011) and spores could easily reach other Hawaiian sky islands. If locally produced spores were available, one would expect S. palustre to have dispersed to many of the other Hawaiian sky islands, especially given the thousands of years that the plant has been resident in Hawaii. Thus, the apparent natural occurrence of S. palustre at only one Hawaiian site is evidence that sexual reproduction is lacking. However, this distribution pattern could also be caused by the plant’s extirpation from other sites.

One cannot rule out the possibility that there have been two or more long-distance dispersals to Kohala Mountain. If this were the case, however, then one would expect that long-distance dispersal would have also resulted in the presence of S. palustre at more than one sky island. If one assumes that the plant has had a continuous presence on Kohala Mountain for > 23.9 kyr, then the occurrence of multiple long-distance dispersals to this area would have had to have been by closely related spores (particularly with all sharing the same rare genetic trait). This is far less likely than the alternative hypothesis of just one founding event. It is also possible that S. palustre has not been continuously present on Kohala Mountain for over 23.9 kyr, but instead has been represented by different populations over time. If this is the case, then the high genetic diversity detected in the current population developed within a much shorter time period.

Even if there have been multiple long-distance dispersals by S. palustre to Hawaii, unless there were a large number of dispersal events it would be unlikely for both sexes to be dispersed to the same location. Thus, one would expect to find a single sex at each site and, unless members of two populations of different sex were brought together, sexual reproduction would not occur. Subgroups A and B co-occur at two sites and subgroups A and C co-occur at a third (Table 1). Further, three mixed collections of subgroups A and B were detected. Thus, the lack of any evidence of sexual reproduction suggests that just one sex is present among the three subgroups. Another, but less likely, possibility is that both sexes are present but sterile. In contrast, sporophytes are produced by S. palustre in the Azores, where multiple long-distance colonizations have occurred (H. K. Stenøien et al., unpublished data).

Our findings lead us to conclude that sexual reproduction does not occur in Hawaiian S. palustre at the present time and that vegetative propagation is the sole means of population growth and maintenance. Thus, recombination plays no role in haplotype variability, which instead arises from mutations and their random transmission to ramets via vegetative propagation (Karlin et al., 2011). This suggests that some of the microsatellite alleles detected in the polymorphic markers of the Hawaiian plants arose in situ and may not be directly comparable to microsatellites occurring in S. palustre of other regions because of homoplasy. Haplotypic variability was also found to be mostly caused by mutations and rarely, if ever, by recombination in European populations of the dioicous S. angermanicum Melin (Stenøien et al., 2011).

The approximate age of the population at each of the three sites used to compare intersite levels of genetic diversity in Hawaii is known. The Mt Ka’ala Reserve (HI-8) was established in the late 1960s by transplants from Kohala Mountain (Hoe, 1971) and the earliest collection of S. palustre in the vicinity of HI-7 (Wailuku) was made in 1975 (D. R. Herbst 5144 (NY)). Radiometric dating at both HI-2 and HI-7 (determined after the collection of the plants used in the study) showed that the basal peat at HI-2 formed in c. 1930 while the basal peat at HI-7 dated to the mid-1990s (S. C. Hotchkiss et al., unpublished data ). Thus, the extensive accumulation of Sphagnum at all three sites is of recent origin, with the growth at the two non-Kohala sites occurring after 1965. The similar levels of genetic diversity detected at the two non-Kohala sites that were studied strongly suggest that they share a similar history, especially when compared with HI-2. At both sites, a minimum of two founding MLHs were sufficient to account for the current population structure. The lower number of MLHs detected at HI-8 relative to those at HI-7 may be attributable to the intensive ‘weeding’ of S. palustre that has occurred at the former site. As this activity could result in the elimination of newly formed MLHs by chance, it introduced a strong element of genetic drift at HI-8. It could also be that the population at HI-8 is younger than the one at HI-7. However, neither the collection history nor radiometric dating supports the latter hypothesis. Several lines of evidence thus lead us to conclude that the populations at these two non-Kohala sites were recently established, with the founding plants coming from Kohala Mountain.

As there is no direct evidence that S. palustre has had a continuous presence at HI-2 for a length of time sufficient to account for the high level genetic diversity detected there, it is likely that the high genetic diversity resulted from dispersal from other Kohala Mountain sites. The occurrence of four of the eight MLHCs detected at HI-2 at other Kohala Mountain sites supports this hypothesis (Table 1). Given its proximity to both the Upper Hamakua Ditch and a jeep road, dispersal to HI-2 is likely to have been largely, if not solely, the result of anthropogenic activity, particularly the harvesting of S. palustre for horticultural use. Based on the number of MLHCs present, a minimum of eight MLHs would have been required to establish this population. We conclude that some, or perhaps all, of these ‘founders’ were probably introduced to HI-2 from other Kohala Mountain sites within the past 100+ yr.

Dispersal of S. palustre to sites outside of Kohala Mountain would also be possible via the transport of gametophyte fragments by animals, particularly birds. However, aside from anthropogenic related dispersal, there is no evidence that this has occurred.

The putative clonal system of Hawaiian S. palustre has apparently persisted in the summit region of Kohala Mountain for over 23.9 kyr, with the most likely time of origin being some 49–51 kyr ago. This has been a sufficient length of time for the development of a comparably high genetic diversity based on microsatellites. Major changes in the climatic environment associated with the summit of Kohala Mountain occurred during this time frame. Although montane rain forest has been associated with the summit of Kohala Mountain during the Holocene, open grassland was present there during the middle- to late-glacial period, when the climate was cooler and drier (Hotchkiss, 1998, 2004). In addition, considerable morphological variation also occurs in Hawaiian S. palustre, with most of the range of morphological variation noted for S. palustre on a global basis being present (Karlin, 2001). The persistence through major changes in climate and the presence of high morphological variation both suggest that Hawaiian S. palustre has a high phenotypic plasticity, a trait that has also been detected in the gametophytically haploid S. subnitens (Karlin et al., 2011). The range of morphological variation in Hawaiian S. palustre probably reflects the differential expression of alleles from the two component genomes present in this allopolyploid more than it does the evolution of new selected alleles since the clone’s inception. And as S. palustre is gametophytically allodiploid, the long-term persistence of the Hawaiian clone may in part be attributable to the presence of two genomes. The notable development of genetic diversity in microsatellites suggests that de novo mutations in selected genes may have also played a role in the clone's long term persistence and morphological variability. However, as microsatellites are likely to mutate faster than selected genes, genetic diversity among the selected genes in Hawaiian S. palustre would be expected to be much lower than that detected for microsatellites. Thus, the contribution of de novo mutations in selected genes to both the long-term survival of this clonal system and the extensive morphological variation present is probably quite small.

Our data suggest that a genetic bottleneck appears to have occurred c. 25–29 kyr ago, during the last glacial maximum (LGM; Clark et al., 2009), a time when temperature and precipitation both reached minima (Hotchkiss et al., 2000; Clark et al., 2009). Pollen records show that four sites on the Kohala summit preserved little or no peat during the LGM (Hotchkiss, 1998 and S. C. Hotchkiss, unpublished data). This strongly implies the occurrence of a climate regime unfavorable to S. palustre, which is an obligate wetland species, at the time of the bottleneck. At least three divergent MLHs survived this bottleneck to give rise to the current population structure.

A widespread population system in northwestern North America based on a single gametophytic founder has been reported (Karlin et al., 2011). In this case, the monoicous S. subnitens founded a system of populations across thousands of kilometers by a combination of vegetative propagation and sexual reproduction (via complete selfing). As no genetic diversity was detected across this extensive regional system of populations, it appears that its origin was so recent that there has not been sufficient time for the development of new alleles among the microsatellite markers utilized (Karlin et al., 2011). By contrast, the high genetic diversity detected among the Hawaiian plants (using the same set of microsatellite markers) underscores the great age of this putative clonal system. As bryophytes are extreme clonal ‘splitters’, reliable age records are difficult, and in many cases impossible, to obtain (Jónsdóttir, 2011). This study presents several lines of evidence that collectively suggest that a bryophyte clone may persist in excess of 23.9 kyr, with the current generation of ramets being robustly fit. Among the ranks of the most long-lived organisms, the putative clonal system of Hawaiian S. palustre is far older than Sphagnum populations occurring in the extensive glaciated landscapes of Eurasia and North America.

Selfing, asexual reproduction, and vegetative reproduction are associated with the development of low genetic diversity and have been considered to be evolutionary ‘dead ends’. Johnson et al. (2011) have recently shown that the absence of recombination and segregation need not be an evolutionary dead end in flowering plants and Karlin et al. (2011) found that a low level of genetic diversity and complete selfing do not necessarily prevent evolutionary success in the short term in Sphagnum. Moreover, in this study we show that an apparent absence of sexual reproduction and immigration has not prevented either the long-term persistence of the Hawaiian population of S. palustre or the development of significant microsatellite genetic diversity (when compared with other regional populations of the same species). Thus, vegetative propagation in a moss does not necessarily preclude evolutionary success in the long term.


Genetic lab work was undertaken at the Duke Bryology Laboratory. We greatly appreciate the collections made for this study by Betsy Gagne at the Mt Ka’ala Reserve. We thank Falon Cartwright for her assistance in lab analyses and Peter Vitousek for his assistance in the field. We are grateful for the helpful comments made by the two anonymous reviewers. Funding was provided by Ramapo College, the Ramapo College Foundation, and NSF grant no. DEB-0515749-002 to A.J.S.