Frequent sexual reproduction and high intraspecific variation in Salix arctica: Implications for a terrestrial feedback to climate change in the High Arctic



[1] Genetic variation at molecular loci may underlie important variation in the phenotypes of arctic plants. Such intraspecific variation may be a neglected but important component of biological diversity in the Arctic that could impact how arctic ecosystems respond to climate change. Here, we characterized genetic and phenotypic variation in Salix arctica and evaluated the effect of S. arctica on ecosystem CO2 exchange, a process by which terrestrial ecosystems in the Arctic feedback to the global climate system. We found high genetic variation at microsatellite loci of S. arctica collected from an inland and a coastal site in Greenland that indicates sexual reproduction has occurred frequently as the ice sheet has retreated. Across the North American range of S. arctica, ten chloroplast DNA haplotypes were identified. Haplotype diversity and allelic richness were high overall and similar across regions with different glacial histories. Phenotypic variation in ecologically important traits varied substantially in a High Arctic population of S. arctica. In a widespread High Arctic ecosystem, a net loss of CO2 to the atmosphere was observed except where S. arctica was present. We suggest that high genetic variation in S. arctica is in part a result of frequent sexual reproduction, and that the phenotypic variation we observed is likely to be at least partially genetic-based. This would enable a productive High Arctic species to adapt and potentially prosper as climate changes, and thus affect the terrestrial feedback of the Arctic to the climate system.

1. Introduction

[2] Our understanding of the feedbacks between the Arctic and the global climate system has focused primarily on the dynamics of sea ice, shifts in freshwater export to the ocean, plant species shifts that change ecosystem structure and albedo, and changes in the rate of ecosystem CO2 exchange [Chapin et al., 2005; McGuire et al., 2006; Peterson et al., 2006; Zhuang et al., 2006; Serreze et al., 2007]. These oceanic and terrestrial feedbacks to the climate system, which are an inherent characteristic of biocomplexity in the environment, can be extremely strong and the positive ones may accelerate climate warming in the Arctic and globally [ACIA, 2005; Chapin et al., 2005; McGuire et al., 2006; Serreze et al., 2007]. In Arctic terrestrial systems, the underlying processes that regulate plant responses to climate change are still largely unknown, though we are beginning to develop a conceptual understanding of how fire, depth to permafrost, and nutrient cycling may control some of the observed shifts in species composition [Chapin et al., 1995; Epstein et al., 2000; Lloyd, 2005; Sturm et al., 2005]. Absent, however, has been an appreciation for the evolutionary processes that also may control the response of arctic plants to climate change.

[3] It has been shown that genetic-based variation in ecologically important plant traits allows plants to respond evolutionarily to climate change [Rehfeldt et al., 1999; Davis et al., 2005; Jump and Penuelas, 2005; Franks et al., 2007]. In a region such as the Arctic where plant species diversity is low, variation within species might be particularly important. It could contribute substantially to phenotypic diversity and thus affect processes such as ecosystem CO2 exchange as climate changes [Norberg et al., 2001]. Yet, genetic variation is not well studied in the Arctic compared to temperate and tropical floras in part because it is thought to be relatively low. There are two mechanisms underlying this idea. First, sexual reproduction, an important means of generating genetic variation, is considered infrequent in tundra ecosystems relative to asexual reproduction [Billings and Mooney, 1968; Bliss, 1971; Callaghan and Collins, 1976; Bell and Bliss, 1980]. Many arctic plants produce viable seed [Wager, 1938; Bliss, 1971; Billings, 1987], but germination often requires open space on a soil surface [Gough, 2006]. Frequent disturbance by physical processes creates an abundance of these surfaces in the High Arctic compared to the Low Arctic, but the harsher climate may limit seedling establishment and thus reduce the success of sexual reproduction [Wager, 1938; Bell and Bliss, 1980]. Even infrequent bouts of sexual reproduction, however, could contribute to the creation of new variations.

[4] A second reason that genetic variation is thought to be low in the Arctic is that glaciation during the Pleistocene likely led to the loss of some populations and an initial reduction in the genetic variation of arctic plant species [Abbott et al., 2000; Abbott and Brochmann, 2003; Alsos et al., 2005; Skrede et al., 2006]. However, glaciation can also generate genetic variation through enabling the divergence of populations in isolated glacial refugia. Recent evidence on European temperate tree species suggests that glacial refugia are hot spots of genetic diversity and that where populations converge following migration from isolated refugia genetic diversity can be even greater [Petit et al., 2003]. Patterns of genetic variation for arctic species support a similar glacial history of isolation in refugia, divergence, and postglacial expansion leading to high genetic variation among glacial refugia and within convergence zones [Abbott et al., 2000; Alsos et al., 2005; Skrede et al., 2006]. Many arctic species are widely distributed across glacial refugia and convergence zones [Hulten, 1968; Molau and Mølgaard, 1996; Murray, 1997], and their phenotypes vary within and across these regions [Mooney and Billings, 1961; Teeri, 1973; Chapin and Chapin, 1981; McGraw and Antonovics, 1983; Dawson and Bliss, 1989a; Fetcher and Shaver, 1990; Philipp and Siegismund, 2003]. Thus, genetic variation and the consequent variation in plant phenotypes may not be low for at least some Arctic species, and could affect plant production and ecosystem CO2 exchange as climate changes.

[5] Deciduous shrubs may be particularly important in the response of vegetation to climate change across the Arctic [Tape et al., 2006]. They have increased in abundance in tundra ecosystems as the climate has warmed [Chapin et al., 1995; Sturm et al., 2001], and respond similarly to experimental warming [Walker et al., 2006]. In the High Arctic, where several experimental studies on plant response to climate change have been done [Jones et al., 1997; Welker et al., 1997; Robinson et al., 1998; Marchand et al., 2004; Welker et al., 2004; Sullivan et al., 2008], the deciduous dwarf-shrub Salix arctica may have the greatest potential to respond to climate change as its abundance and photosynthetic rate have been shown to increase in response to experimental warming [Jones et al., 1997; Robinson et al., 1998; Sullivan et al., 2008]. S. arctica has a circumpolar distribution [Hultén, 1937; Murray, 1995], and like many willows, has a high photosynthetic rate with correspondingly low cost leaves (i.e., low leaf mass per unit area) [Raven, 1992; Dawson and Bliss, 1993; Jones et al., 1999; Sullivan and Welker, 2007]. However, it is the only willow species that extends to the northern limit of land in the Arctic. Given its ability to fix carbon and respond to warming, its importance in the High Arctic, and its circumpolar distribution it may be a particularly important species in determining the terrestrial feedback to climate change in the High Arctic.

[6] The specific aims of our study were to: (1) Evaluate whether S. arctica primarily spreads asexually or sexually in a High Arctic landscape in northwestern Greenland by evaluating similarity among multilocus microsatellite genotypes; (2) Evaluate whether genetic variation at microsatellite loci differs between two sites within Greenland of different age and fertility; (3) Characterize diversity in chloroplast DNA (cpDNA) sequences in northwestern Greenland and determine the relationship of cpDNA sequences to a nearby Canadian location and more distant locations in Alaska and Colorado where arctic populations could have survived during the last glaciation; (4) Characterize the phenotypic variation in ecologically important plant traits in a high arctic population; and (5) Determine whether S. arctica affects ecosystem CO2 exchange.

2. Methods

2.1. Study Species

[7] Salix arctica Pall. is a dioecious, deciduous, prostrate shrub that occurs throughout the Arctic and in alpine tundra in the Northern hemisphere [Hulten, 1968]. Its abundance is especially high in prostrate dwarf-shrub herb tundra, a widespread High Arctic ecosystem [CAVMteam, 2003], but it occurs in diverse tundra ecosystems, including young and old ecosystems that differ in fertility, and dry and wet environments [Dawson and Bliss, 1989b; Hodkinson et al., 2003; Cooper et al., 2004; Sullivan and Welker, 2007]. Individuals are long-lived [Wilson, 1964] and can reproduce sexually and vegetatively through the growth of attached and detached stems. There are three geographical races of S. arctica. Each race is morphologically distinct, but individuals often have blended morphological characteristics [Hulten, 1968; Murray, 1997]. S. arctica is tetraploid in eastern North America and hexaploid in western North America and Eurasia [Suda and Argus, 1969].

2.2. Sampling Sites

[8] Plants were sampled at seven sites in northwestern Greenland and at an additional seven sites across the species range in North America (Table 1 and Figure 1). The seven coastal and inland sites in northwestern Greenland differed in age and fertility as a result of the retreat of the Greenland Ice Sheet and increased nitrogen inputs at coastal sites with bird colonies. The additional seven sites across the species' North American range included regions with different glacial histories during the Pleistocene (Figure 1). Plants from all sites were genotyped using cpDNA. Microsatellite genotypes were characterized at two sites on Kap Atholl. Phenotypic variation was measured at five high arctic sites in northwestern Greenland (Table 1). Ecosystem CO2 exchange was measured on Kap Atholl in prostrate dwarf-shrub herb tundra (Table 1).

Figure 1.

Locations sampled and frequencies of cpDNA haplotypes of S.arctica at each location in (a) northwestern Greenland and (b) at additional sites across its North American range. The area mapped in Figure 1a, is marked ‘Greenland’ in Figure 1b, and the current extent of the Greenland Ice Sheet is in white. In Figure 1b, the area in white is the maximum extent of ice across North America at the last glacial maximum [Frenzel, 1992]. Locations are numbered as in Table 1, the size of each pie is scaled by the number of individuals sampled, and the frequency of each haplotype is represented by a section of the pie. (c) Network showing relationships among haplotypes based on parsimony. Lines represent individual base-pair changes or insertion/deletions sites that differ between the haplotypes. One long insertion/deletion (25bp) is marked with a double hatch. Dashed lines indicate multiple alternate most-parsimonious connections between haplotypes. Haplotypes are lettered and color-coded to match the pie graphs in Figures 1a and 1b.

Table 1. Locations of Sampling Sites, Number of Individuals Sequenced (and Genotyped), and Haplotypes Present
SiteLocationCountryLatitude, °Longitude, °nHaplotypesa
  • a

    Number of individuals harboring each haplotype is indicated.

  • b

    Sites where plant traits were sampled.

  • c

    Site where plant cover, LAI and ecosystem CO2 exchange were measured.

  • d

    Sites where microsatellite variation was measured.

1Kap Atholl-coastalb,c,dGreenland76.5568.573 (17)2A, 1G
2Kap Atholl-betweenbGreenland76.4868.4261A, 5B
3Kap Atholl-inlandb,dGreenland76.4368.284 (18)3A, 1G
4Kap Russell-coastalbGreenland78.8868.7541B, 2E, 1J
5Kap Russell-inlandbGreenland78.4769.8362A, 1E, 1G, 1J
6Kap Atholl-bird cliffsGreenland76.2869.0764A, 1B, 1G
7Agpat Agpai - coastalGreenland76.0868.3952A, 1D, 1G, 1J
8Bellot Strait - coastalCanada72.0294.252A, 3J
9Brooks Range - AlaskaU.S.68.42149.2884A, 2B, 1G, 1J
10Wrangell Mtns-AlaskaU.S.61.53142.8611J
11Banff-AlbertaCanada52.19117.1531A, 1H, 1I
12Front Range Mtns - ColoradoU.S.40.42105.7561A, 1E, 4F
13San Juan Mtns 1 - ColoradoU.S.37.88107.8362A, 4C
14San Juan Mtns 2 - ColoradoU.S.37.90107.7341A, 3E

2.3. Plant Collections and DNA Extraction

[9] Ten to twenty individuals were sampled at each site. In the prostrate dwarf-shrub herb tundra common in northwestern Greenland, a male and a female plant were sampled at each of five locations evenly spaced across a 100 m transect in 2003 and 2004. At the other sites, where S. arctica was not as abundant, plants were sampled by walking in a line until ten individuals at least 3 m apart could be located. Either fresh leaf tissue or woody stems were collected from individual plants. Leaf tissue was immediately stored in silica to dry, and stems were shipped to Colorado State University and planted. When stems leafed-out, leaves were collected. Genomic DNA was extracted from 1 to 18 individuals per site (Table 1) using approximately 300 mg of leaf tissue, dried, fresh or frozen at −80°C. DNA was extracted either using the CTAB method with reagents from AutoGen (Plant DNA, Ver. 1.01) or using Dneasy Plant Mini kits from Qiagen®.

2.4. Development of Microsatellite Markers and Genotyping

[10] We focused our microsatellite genotyping on two locations. We genotyped 17 individuals from a coastal site on Kap Atholl approximately 21 km from the ice sheet, and 18 individuals from an inland site that is less than 1 km from the ice sheet (Table 1).

[11] Microsatellite cloning and sequencing was performed at the Evolutionary Genetics Core Facility of Cornell University following the protocol of Hamilton et al. [1999] with modifications following in Marrs et al. [2006]. Of 37 clones with possible microsatellites, four gave robust, relatively easy to score loci (Table 2). PCR amplifications for genotyping was performed in 10μl reaction volumes with 1 μl extracted DNA, 1 μl 10 X buffer PCR, 0.8 μl MgCl2 (25 mM), 0.08 μl dNTPs (100 mM), 0.2μl Hotstart taq (5 U/μl), 0.04 μl forward primer (10 nmole/ml), 0.17 μl reverse primer (10 nmole/ml), 0.17 μl M13 sequence labeled with dye (10 nmole/ml). After a preliminary denaturation step at 95°C for 8 min, PCR amplification was performed for 35 cycles of: 50 s denaturing at 95°C, 1min of annealing at locus specific temperatures (Table 2) and 1min 30 s of extension at 72°C, with a final 10 min extension step at 72°C. PCR products were stored at −20°C until genotyping. Diluted PCR products were mixed with 8.9 μl formamide and 0.1μl GS 650 Liz Ladder, and denatured for 3 min at 90C then separated on an ABI 3100 Genetic Analyzer and analyzed with the program GENEMARKER 1.5 (SoftGenetics LLC ®, 2004).

Table 2. Primer Sequences and Characteristics of the Four Microsatellite Loci Isolated From Salix arctica, Including Locus Name and GenBank Accession Number, Primer Sequences With Dye Indicated, PCR Annealing Temperature, Motif in Cloned Allele, and Size of the Sequenced Allelea
Locus(Accession)Primer Sequence (5′ – 3′)Ta, °CRepeat of Cloned AlleleSize, bpSize Range, bpNumber of AllelesHE
  • a

    Summary information on the size range and numbers of alleles found in 35 individuals and expected heterozygosity for each locus.

S8F: (VIC)(M-13)ATAAACAAGTGTGAAGGGGTGTG60(GT)11420428 – 44690.86
S147F: (PET)(M-13)GGATCTACACGGGTACAGCATTAT56(GT)16287301 – 319170.92
S28F: (VIC)(M-13)TCTCTTGTTTGTACTCTTCCATTT50(CT)10395401 – 425130.89
S37F: (6-FAM)(M-13)AAGCCTTGATACGCGGTCGTGATT65(GAA)8271274 – 28640.64

2.5. cpDNA Sequencing

[12] An intergenic cpDNA region was amplified using primers trnQr and trnK2 from Dumolin-Lapegue et al. [1997]. PCR was performed in 10 μl reaction with 1μl extracted DNA, 1 μl 10 X buffer PCR, 0.8 μl MgCl2 (25mM), 0.08 μl dNTPs, 0.2 μl Hotstart taq, 0.2 μl forward and reverse primer. Cycling parameters were one denaturation cycle at 94°C for 8 min, followed by 30 cycles of 40 s at 94°C to denature, 40 s at 50°C to anneal, 2 min at 72°C to extend and a final cycle at 72°C for 10 min to extend. PCR products were sequenced from both ends into the variable intergenic spacers using BigDye Terminator Cycle Sequencing (version 3.1) primed with the PCR primers. Sequencing reactions were cleaned up using ethanol precipitation, and separated on an ABI 3100 capillary sequencer. Sequences were aligned using the DNAstar package. We sequenced a single individual of each haplotype twice to verify our calls.

2.6. Identification of Asexually and Sexually Produced Individuals

[13] To infer whether sampled individuals were members of the same clone or were sexually reproduced, we compared mutlilocus microsatellite genotypes of individuals within sampling locations. Individuals that share common alleles at all four loci could be part of the same clonal individual, and therefore represent asexual reproduction, while individuals that have unique multilocus genotypes are likely to have originated from separate bouts of sexual reproduction. Somatic mutation in microsatellite loci could lead to variation among asexually produced plants which would inflate our estimate of sexual reproduction. Somatic mutation rates can vary widely by species and locus type [Azaiez et al., 2006], but generally rates are low enough that this source of bias should be small.

2.7. Analysis of Genetic Variation

[14] For the microsatellite loci, the number of alleles and the number of private alleles for the two populations were tallied by hand. Expected heterozygosity corrected for sample size was estimated using SPAGeDi 1.2 [Hardy and Vekemans, 2002], a program that can accommodate data from polyploids such as S. arctica. We also used SPAGeDi to perform the equivalent of an AMOVA for polyploid data to determine whether the two populations differ significantly. A permutation test (20,000 permutations) was used to evaluate significance of the test.

[15] We compared cpDNA haplotype diversity and allelic richness controlling for sample size between Greenland, Alaskan, and Coloradan samples using CONTRIB [Petit et al., 1998].

[16] We constructed haplotype networks from the sequence data using the network building software TCS 1.2.1 [Clement et al., 2000], which uses statistical parsimony to construct an unrooted system of relationships between non-recombining sequences based on the genealogical reconstruction algorithm of Templeton et al. [1992]. In these analyses, individual samples can be internal in the network. We treated indels as a 5th state [Giribet and Wheeler, 1999; Simmons and Ochoterena, 2000], and coded each indel as a single mutational event to prevent longer indels from overwhelming other signal in the data set [Simmons et al., 2001]. We evaluated population structure in cpDNA haplotypes by performing an AMOVA in Arlequin (version 2.0).

2.8. Phenotypic Variation in Plant Traits

[17] We measured three key plant traits that affect plant productivity: leaf mass per unit area (LMA), carbon isotope discrimination (Δ13C) and oxygen isotope discrimination (Δ18O). Our sampling scheme included five of the Greenlandic sites that had also been sampled for cpDNA in a crossed design of region by position from the Greenland Ice Sheet (Table 1). In mid-August 2003 at the end of the growing season, ten plants were sampled for LMA and leaf Δ13C at sites 1–5 and for leaf Δ18O at sites 1 and 2, where stem water Δ18O data was available.

[18] To determine LMA, three leaves from one-year old vegetative stems were collected from each plant. Each leaf was scanned at 300 dpi, then oven-dried for 2 d to constant mass (60°C) and weighed to the nearest microgram. Leaf area was calculated from the scans using an unsupervised classification in ERDAS IMAGINE version 9.0 (Leica Geosystems Geospatial Imaging, LLC., Atlanta, GA). We calculated LMA for each leaf individually by dividing leaf mass by leaf area and then averaged the three values for each plant.

[19] An additional ten leaves were collected randomly from vegetative stems on each plant and pooled for the isotopic analyses. Samples were oven-dried for two days to constant mass (60°C), ground to a uniform texture, and analyzed using a Carlo Erba elemental analyzer (Thermo Electron Corp., Milan, Italy) interfaced with a continuous-flow Isochrom isotope ratio mass spectrometer (Micromass UK Ltd., Manchester, UK) for leaf δ13C and on a EuroVector Pyrolysis unit (Euro Vector, Milan, Italy) interfaced with VG-Optima stable isotope ratio mass spectrometer (Micromass UK Ltd., Manchester, UK) for leaf δ18O in the Natural Resource Ecology Laboratory at Colorado State University. Leaf carbon isotope discrimination (Δ13C) was calculated relative to δ13C atmospheric CO2 [O'Leary, 1993], with δ13C atmospheric CO2 held at −8.0‰. A lower leaf Δ13C corresponds with greater plant water use efficiency (WUE) or a higher amount of CO2 fixed per unit water lost. Leaf δ18O is reported as enrichment above source water [Barbour et al., 2000], using stem water δ18O values for S. arctica at sites 1 (−14.4‰) and 2 (−15.1‰) [Sullivan and Welker, 2007]. A lower leaf Δ18O corresponds with greater stomatal conductance or water loss. The data for each plant trait were analyzed using a one-way analysis of variance (ANOVA) to determine the effect of site and Tukey's HSD to determine differences among the means (SAS 9.1, SAS Institute, Cary, NC). A one-way ANOVA was appropriate, because the design was incomplete and without site-level replication.

2.9. Effect of S. arctica on LAI and Ecosystem CO2 Exchange

[20] At one of the coastal sites on Kap Atholl (site 1), plant cover, the LAI and ecosystem CO2 exchange were measured in 18 vegetated areas (sized 0.125 m2), where plant cover was greater than 70% at peak biomass, that varied in S. arctica abundance. Data were collected on eight dates during the growing season in 2004. Since these measurements were confined to vegetated areas that represent 50% or less of the area on Kap Atholl, we also measured plant cover and LAI at peak biomass in 82 randomly selected plots, in which plant cover ranged from 0 to 96%. Ecosystem CO2 exchange was measured in a subset of these plots by selecting plots across the range of variation in the LAI. This sampling design assured that we could characterize spatial variation in plant cover and the LAI and relate this variation to ecosystem CO2 exchange.

[21] Plant cover and the LAI were measured from multispectral images that were taken of each plot using a tripod-mounted, multispectral camera (Tetracam ADC, Chatsworth, CA). The camera is sensitive to green, red, and near-infrared (G, R, NIR) spectral bands at the following wavelength intervals: 520–620 nm (G), 620–750 nm (R), and 750–950 nm (NIR). The images were taken from nadir, under clear skies, near midday, at a set exposure. Images were also taken of a 99% reflectance white standard (Teflon), which was supplied with the camera and calibrated by Tetracam Inc. These images were used to normalize for incoming radiation and calculate reflectance. Visual estimates of plant cover and the proportion of plant cover composed of S. arctica to the nearest 10% were made from these images. The normalized difference vegetation index (NDVI) was calculated as the ratio of the difference between NIR and R reflectance to their sum. LAI was estimated from NDVI using an NDVI-LAI model developed for the ecosystem [Steltzer and Welker, 2006]. The seasonal change in LAI (ΔLAI) was calculated as the difference between the early season minimum LAI and maximum LAI during peak season. The season-long integration of LAI (iLAI) was calculated by multiplying the LAI by the number of days represented by each measurement and summing these products.

[22] Ecosystem CO2 exchange was measured using a clear, Plexiglass chamber connected to a LiCor 6200 portable photosyntheis system (LiCor, Inc. Lincoln, NE) [Vourlitis et al., 1993]. The Plexiglass chamber was placed over a plot and weighted down with steel bars to create a closed system. Measurements of net ecosystem production (NEP) were made over a 30 s period under field conditions. The chamber was then covered to block all light for the measurement of ecosystem respiration (ER). Gross ecosystem production (GEP) was calculated as the difference between NEP and ER. Positive NEP characterizes a flux of CO2 from the atmosphere into the ecosystem. Least linear squares regression was used to analyze the relationship between the proportion of plant cover composed of S. arctica and the LAI, the LAI and GEP, and GEP and NEP (SAS 9.1, SAS Institute, Cary, NC), which are the most relevant relationships for modeling the effect of S. arctica on ecosystem CO2 exchange.

3. Results

3.1. Microsatellite Variation and Frequency of Sexual Reproduction

[23] All four microsatellite loci of S. arctica were highly variable (Table 2). The alleles from S8 and S37 conformed well to models of microsatellite mutation given the repeat motifs (e.g., the allele sizes varied by twos for the dimer S8 and by threes for trimer S37). Although variable, the other two microsatellite loci did not conform as well. Most alleles from the dimer S28 followed a pattern consistent with insertions or deletions (indels) of a single repeat motif, with a few additional odd alleles suggesting indels of other sizes. S14, also a dimer, was highly variable, and had many alleles that varied by a single base pair, rather than the expected two. Evaluating all four loci, none of the individuals shared the exact multilocus microsatellite genotype. Even restricting the evaluation of multilocus genotypes to S8 and S37 to be conservative, only two individuals of 17 at the coastal site on Kap Atholl and two of 18 at the inland site nearest the ice sheet shared a multilocus genotype. Thus, it is likely that the individuals sampled were almost entirely produced by sexual reproduction. Average heterozygosity across alleles did not differ between the older, more fertile coastal site (mean ± SD = 0.80 ± 0.16) and younger, less fertile inland site on Kap Atholl (0.84 ± 0.08). Additionally, pairwise FST was not significantly different (FST = 0.0039, P = 0.59).

3.2. Intraspecific Variation in cpDNA Haplotypes

[24] Chloroplast DNA haplotype diversity and allelic richness were high overall and similar among the regions sampled despite different glacial histories (Table 3). In all, we found six chloroplast haplotypes in only 39 samples in northwestern Greenland. Two to four cpDNA haplotypes were identified at each site (Table 1 and Figure 1a). There was one most common haplotype (A) that was found at all locations sampled except for the coastal site on Kap Russell. In 28 samples from across much of the North American range of S. arctica, eight cpDNA haplotypes were identified (Table 1 and Figure 1b). The most common haplotype was the same found to be most common within Greenland (A) and was found in all but one location (Site 10, at which only a single individual was sampled). Four new haplotypes were found in the additional samples across the range of S. arctica in North America (C, F, H, I), while two found in Greenland (D, E) were not present within these samples. The four, unique haplotypes all occurred in the alpine populations sampled in Colorado and in Banff, Canada. All haplotypes from Alaska were also found in Greenland. This geographical arrangement of haplotypes represents significant population structure (Table 4). Most of the variation was found within populations, but there was significant structuring associated with individual sample locations within the geographical groupings. Somewhat surprisingly, there was no significant variation attributable to differences between Alaska, Colorado and Greenland.

Table 3. Haplotype Diversity, Number of Haplotypes, and Allelic Richness of S. arctica
GroupanMean Haplotype Diversity (95% CI)Number of HaplotypesAllelic Richness
  • a

    Site 8 (Bellot Strait, Canada) and site 11 (Banff, Canada) were not included in this analysis.

Alaska90.78 (0.56–1.0)43.0
Colorado160.80 (0.72–0.88)42.9
Greenland390.77 (0.67–0.89)63.3
Table 4. Analysis of Molecular Variance of the Chloroplast DNA Sequences From Alaska, Colorado, and Greenland
Source of Variationd.f.Sum of SquaresVariance ComponentsPercent of Variation
  • a

    P < 0.00001.

Among groups21.58−0.002−0.6
Among populations within groups117.780.08420.5a
Within populations5317.380.32880.0a

3.3. Phenotypic Variation in Plant Traits

[25] Leaf mass area and two integrative measures of plant water relations (leaf Δ13C and Δ18O) varied substantially among the S. arctica plants sampled in northwestern Greenland (Figure 2). Site explained 4.6% to 37.3% of the variance in the plant traits. The remaining variance, 62.7% to 95.4%, occurred within sites (Table 5). Variation in LMA was nearly continuous across its range from 62 g m−2 to 109 g m−2 with one value higher than this range (Figure 2a). Average LMA was lower at the inland sites on Kap Atholl and Kap Russell (Figure 2b). The range of variation was greater for leaf Δ13C (18.6–22.8 ‰) than for leaf Δ18O (27.5–30.6 ‰), but leaf Δ13C was measured at more sites (Figures 2c and 2e). Average leaf Δ13C was lower for plants sampled at the inland site on Kap Russell (Figure 2d). Average leaf Δ18O did not vary between the two sites on Kap Atholl where it was measured (Figure 2f).

Figure 2.

Phenotypic variation in plant traits of S. arctica in northwestern Greenland. (a, b) Leaf mass area (LMA); (c, d) leaf Δ13C; and (e, f) leaf Δ18O. Results of a one-way ANOVA are reported, and significant differences among sites are noted by different letters. Bars are means ± 1 SE (n = 10).

Table 5. One-Way Analysis of Variance of Plant Traits in Greenland
Plant TraitSource of VariationdfSum of SquaresMean SquaresFP ValuePercent of Variation
Error445776131  64.7%
Δ13CError4517.620.39  62.7%
Δ18OError1813.920.77  95.4%

3.4. Effect of S. arctica on LAI and Ecosystem CO2 Exchange

[26] The seasonal change in the leaf area index (ΔLAI) and the season-long integration of LAI (iLAI), which were measured in vegetated areas in the ecosystem, were both positively correlated to the proportion of plant cover composed of S. arctica leaves (Figures 3a and 3b). Since S. arctica is the most abundant deciduous species in this ecosystem, it was not surprising that the proportion of plant cover composed of S. arctica was correlated to ΔLAI. However, this seasonal increase in LAI was large enough and/or was sustained long enough that iLAI was also correlated to the abundance of S. arctica. GEP was correlated to LAI throughout the growing season (Figure 3c), and NEP was correlated to GEP (Figure 3d). These vegetated areas of the ecosystem were generally a source of CO2 to the atmosphere during the growing season except where S. arctica was abundant and thus LAI and GEP were highest. Similarly, positive rates of NEP across bare and vegetated areas of the ecosystem were dependent on high rates of GEP, which only occurred where the LAI was high (Figures 4b and 4c). High values of the LAI were restricted to locations where S. arctica was abundant (Figure 4a). S. arctica was 58% of plant cover on average where it was present, but was not present in over half of the plots. Overall, the proportion of variation in LAI explained by the abundance of S. arctica was low (R2 = 0.21 and 0.23 for seasonal and spatial variation, respectively). However, these relationships are independent of total plant cover, which in large part determines LAI and thus GEP and NEP.

Figure 3.

Relationships between (a, b) the proportion of plant cover composed of S. arctica and the LAI; (c) the LAI and GEP; and (d) GEP and NEP. ΔLAI is the seasonal maximum LAI minus the early season minimum LAI, and iLAI is the sum of LAI across the growing season. Data are for measurements throughout the growing season in vegetated areas where plant cover was greater than 70% at peak biomass. Least linear squares regressions are reported (n = 18 or n = 162).

Figure 4.

Relationships between (a) the proportion of plant cover composed of S. arctica and the LAI; (b) the LAI and GEP; and (c) GEP and NEP. Plant cover and LAI were measured at peak biomass in 82 randomly selected plots that varied in plant cover (0 to 96%). GEP and NEP were measured on a subset of 46 plots selected across the range of variation in LAI. Least linear squares regressions are reported (n = 82 or n = 46).

4. Discussion

4.1. Frequent Sexual Reproduction and High Genetic Variation

[27] In northwestern Greenland, the genetic variation in S. arctica at microsatellite loci indicates that sexual reproduction has been common as the ice sheet has retreated. No more than two individuals of the 17 sampled at the coastal site on Kap Atholl and two of the 18 at the inland site near the ice sheet shared a multilocus genotype and could have been reproduced asexually. The high heterozygosities we observed reflect both sexual reproduction and polyploidy. Only one of the six copies of a microsatellite in a hexaploid needs to differ from the others for that individual to be labeled a heterozygote. Few microsatellite data from natural populations of hexaploids are published, but one recent study of Ranunculus cassubicus shows that a hexaploid had heterozygosities of 1 across six different populations. Thus, by comparison our reported heterozygosities of 0.8 are modest. The lack of significant differentiation between the two sites suggests that similar colonization events have occurred over time, coastal sites were the source of seed for inland sites, and/or genetic exchange has occurred between these two sites.

[28] Rampant sexual reproduction in a High Arctic population of S. arctica challenges the long-held idea that sexual reproduction is rare in such an extreme environment [Wager, 1938; Bliss, 1971; Bell and Bliss, 1980; Billings, 1987]. Yet, the primary observation that led to this idea, high seedling mortality, may not be inconsistent with this result. In cold, dry environments, plant growth and thus clonal reproduction has likely been so slow that few seedlings may have needed to survive for sexually reproduced individuals to be common. Additionally, high seedling mortality is common to many environments and does not preclude sexual reproduction [Forbis, 2003]. Repeated demographic studies would be necessary to identify the frequency of sexual reproduction in a population, but have rarely been practical to do in the Arctic. Studies of seedling demography have rarely if ever exceeded two years [Wager, 1938; Mooney and Billings, 1961; Callaghan and Collins, 1976; Bell and Bliss, 1980; Bliss and Gold, 1999]. Genetic studies do not require multiple years of observation, produce more definitive results, and are becoming a common approach. Our study contributes to a growing body of genetic evidence confirming that frequent sexual reproduction can occur in arctic and alpine plants [Gabrielsen and Brochmann, 1998; Linhart and Gehring, 2003] and has led to the evolution of more species than previously thought [Grundt et al., 2006].

[29] Genetic variation in cpDNA haplotypes of S. arctica was high in northwestern Greenland and across its range in North America. Haplotype diversity was higher than that found in another Arctic species, Saxifraga oppositifolia [Abbott and Comes, 2003]. Dispersal of sexually reproduced propagules may contribute to this variation. Similarly, such dispersal is likely to account for the widespread distribution of some haplotypes of S. arctica within and across regions of the North American Arctic. In species that mainly reproduce clonally, widespread genotypes are relatively rare [Ellstrand and Roose, 1987]. Genetic variation can occur in predominantly clonal populations, because often they are comprised of multiple clones, a result of being founded through sexually reproduced individuals [Ellstrand and Roose, 1987]. However, such populations tend to have fewer genotypes, and these genotypes tend to be found in only one population [Holsinger, 2000]. Of the ten haplotypes identified across the North American range of S. arctica, four were found at only one site, and all four were at alpine sites in Colorado and southern Canada. This pattern may indicate that sexual reproduction in S. arctica occurs more frequently in the Arctic than in the alpine. However, we think it more likely that the alpine sites, as ecological islands, are simply more geographically isolated than most arctic sites [DeChaine and Martin, 2005].

[30] On Kap Atholl, common haplotypes of S. arctica were identified at younger sites (sites 2 and 3) near the Greenland Ice Sheet and at older, more fertile sites along the coast (sites 1 and 6) (Figure 1a). This suggests that plants at coastal sites on Kap Atholl may be the source of plants that have established at inland sites as the ice sheet has retreated and that these haplotypes can establish in diverse environments. Interestingly, haplotypes E and J, the two haplotypes common to both sites on Kap Russell, were both not found in our Kap Atholl samples.

[31] Overall, the pattern of variation in cpDNA haplotypes across North America suggests that glaciation did not lead to a permanent decrease in the genetic variation of S. arctica. Haplotype diversity and allelic richness were similar among the three regions despite their different glacial histories (Table 3), and many of the haplotypes found in Alaska and northern Canada were also found in northwestern Greenland. The genetic composition of alpine sites in the southern range of S. arctica was most distinct. Shared haplotypes between the alpine and Arctic regions of North America are consistent with an alpine origin for this species. The frequent presence of unique haplotypes at alpine sites suggests that divergence may have contributed to the genetic variation in the southern range of S. arctica, where it now only occurs in alpine environments. The patterns seen in S. arctica in North America are similar to Saxifraga oppositifolia and Dryas integrifolia, two other High Arctic species, in revealing relatively high genetic variation [Tremblay and Schoen, 1999; Abbott et al., 2000; Abbott and Comes, 2003; Philipp and Siegismund, 2003]. Our findings differ from those of Abbott and Comes [2003] for Saxifraga oppositifolia with respect to population structure. We found most cpDNA variation was within and among populations of S. arctica, while they found variation at those levels and between broader geographical groupings. The lack of support for differentiation between Colorado, Alaska, and Greenland could stem from both our small sample sizes and high rates of gene flow between regions.

4.2. Implications for a Terrestrial Feedback to Climate Change in the High Arctic

[32] Gross ecosystem production in prostrate dwarf-shrub herb tundra was affected by variation in the seasonal and spatial abundance of S. arctica. The LAI was highest where S. arctica was abundant and could explain 55% of the temporal variation and 82% of the spatial variation in GEP. Across diverse tundra ecosystems, the LAI explains much of the variation in GEP [Shaver et al., 2007; Street et al., 2007], because GEP is more sensitive to LAI than to foliar N [Williams and Rastetter, 1999]. Thus, the abundance of species such as S. arctica that influence the LAI affect ecosystem CO2 uptake and in part determine whether an ecosystem is a source or sink of CO2 to the atmosphere. We suggest that frequent sexual reproduction and high genetic variation in S. arctica increase the likelihood that this species will be able to survive and potentially thrive as climate changes. Sexual reproduction could lead to the formation of new genotypes that are better adapted to the new climatic conditions [Davis et al., 2005; Jump and Penuelas, 2005; Parmesan, 2006], and high genetic variation increases the likelihood that successful genotypes already exist. Phenotypic variation on which selection can act was also high in S. arctica. Because the abundance of S. arctica was linked to carbon sequestration, adaptation to the changing climate could facilitate a negative feedback that could help counter the many positive feedbacks to climate change in the Arctic.


[33] We thank: W. Morris, P. Ray and K. Westergaard for assisting with the collection of S. arctica; M. Smith, J. Decant, D. Banks and M. Welker for field assistance; and E. Steltzer for the maps. D. Reuss provided assistance with the isotopic analyses. Thule Airbase and VECO Polar Resources provided logistical support in Greenland, and the Mountain Studies Institute and the Center for Snow and Avalanche Studies provided logistical support in the San Juan Mountains. Funding for this research included NSF-OPP grant 0221606 to J.M.W. and funding from the Colorado Agricultural Experiment Station to R.A.H.