A space-for-time substitution reveals the long-term decline in genotypic diversity of a widespread salt marsh plant, Spartina alterniflora, over a span of 1500 years

Authors

  • STEVEN E. TRAVIS,

    Corresponding author
    1. USGS National Wetlands Research Center, 700 Cajundome Boulevard, Lafayette, LA 70506, USA,
      Steven Travis (fax +1 337 266 8592; e-mail steven_travis@usgs.gov).
    Search for more papers by this author
  • MARK W. HESTER

    1. USGS National Wetlands Research Center, 700 Cajundome Boulevard, Lafayette, LA 70506, USA,
    2. Coastal Plant Sciences Laboratory, Department of Biological Sciences and Pontchartrain Institute for Environmental Sciences, University of New Orleans, New Orleans, LA 70148, USA
    Search for more papers by this author

Steven Travis (fax +1 337 266 8592; e-mail steven_travis@usgs.gov).

Summary

  • 1Clonal populations face a trade-off between sexual recruitment and vegetative growth and, once established, may undergo continuous declines in genotypic diversity if their sexual recruits make poor competitors. The geological history of delta formation in the Lower Mississippi River Valley was used to age eight S. alterniflora marshes for use in a space-for-time substitution ranging over 1500 years, in order to determine the long-term effects of clonal growth on genotypic diversity in natural populations.
  • 2We also predicted that highly heterozygous clones are competitively superior, leading to an increase in the overall level of genetic diversity as a marsh ages and/or to an increasingly positive relationship between clone size and individual heterozygosity, and that the clumping of ramets within clones will occur over increasingly large distances as populations age, while the clumping of genetically related clones will become less pronounced as intraclonal competition begins to obscure the initial effects of localized seedling recruitment.
  • 3Using molecular markers to differentiate clones, we documented a decline in clonal richness at the rate of approximately 1% 100 years−1 that was accompanied for the first 300–500 years by an increase in the distance over which clumping of ramets within genets occurred. Older populations, in the 500–1500-year range, showed evidence of clone fragmentation.
  • 4The spatial clustering of kin was observed for only two marshes, and exhibited no clear relationship with marsh age.
  • 5Whereas the overall level of genetic diversity was consistent among marshes and showed no clear relationship with marsh age, the relationship between heterozygosity and individual clone size became increasingly pronounced within older marshes.
  • 6Our results suggest that under natural conditions S. alterniflora marshes will rarely reach ages sufficient for the loss of all clonal diversity, or for the effects of inbreeding and drift to pose a significant threat to population viability.

Introduction

The population dynamics of partially clonal species are principally determined by intraspecific interactions occurring at the level of genets (Harper 1977, 1981; Eriksson 1993; McLellan et al. 1997; Pan & Price 2001). For those species characterized by limited seedling recruitment under conditions of environmental stability (i.e. the so-called initial seedling recruitment or ‘ISR’ species of Eriksson 1989), intergenet competition may be the single most important factor affecting clonal, i.e. genotypic, diversity (Soane & Watkinson 1979; Watkinson & Powell 1993; McLellan et al. 1997). In the complete absence of seedling recruitment, clonal diversity is constrained to the level of the initial cohort and can only decline over time (Eriksson 1993, 1997; Watkinson & Powell 1993). Thus, depending on the life span and rate of outward spread of competitively superior genets, long-lived populations may be characterized by larger and larger differences in clone size over time (e.g. McClintock & Waterway 1993; Jonsson 1995; McLellan et al. 1997), ultimately culminating in the development of genetic monocultures (e.g. Kemperman & Barnes 1976; Worthen & Stiles 1986; Eckert & Barrett 1993). Model simulations suggest this is particularly true for isolated populations that are initially founded by a small number of clones (Watkinson & Powell 1993). From an adaptive perspective, such a population would certainly represent an evolutionary dead end but, in reality, natural levels of disturbance, even in the most stable of environments, are probably sufficient to maintain clonal diversity at a quantifiable level (e.g. Hartnett & Bazzaz 1985; Eriksson & Bremer 1993; Kudoh et al. 1999). Nevertheless, competitive exclusion could reduce effective population size to the extent that genetic drift and inbreeding come to pose a significant threat of extinction (Handel 1985; Williams & Davis 1996; Procaccini & Mazzella 1998; Reusch 2001).

As a first step in characterizing the threat of declining clonal diversity to the viability of ageing populations, it is necessary to adequately quantify the actual rate of decline, as well as the rate of genet turnover due to purely stochastic environmental processes affecting large sectors of populations (Eriksson 1997). It may be that the prevailing tendency in natural populations is for environmental stochasticity to partially or wholly eliminate populations before clonal diversity can decline to the extent that drift and inbreeding are brought into play. In spite of the current availability of various molecular tools for assessing clonal diversity, the accurate estimation of population age remains a significant challenge to resolving this issue, particularly with respect to very old populations. Thus, there have been very few studies aimed at characterizing the relationship between clonal diversity and population age in clonal species, and these have been entirely limited to populations of less than several hundred years in age (Hartnett & Bazzaz 1985; Verburg et al. 2000; Barsoum 2002). Our approach is to focus on a clonal species growing on the Mississippi River deltaic plain, where a well-characterized geological record of land formation provides an ideal natural laboratory.

Wetland plants provide a prime example of the trade-offs inherent in a partially clonal lifestyle (Les 1988; Grace 1993). For example, in Spartina alterniflora (smooth cordgrass), a dominant inhabitant of brackish and saline marshes along the Gulf of Mexico and Atlantic coasts of North America, we have determined that high rates of interpopulational gene flow and rapid clonal propagation lead to the extremely rapid colonization of created marshes, even when they exceed 100 ha (Proffitt & Young 1999; Proffitt et al. 2003). However, because seedling recruitment is limited to the first 30 years or so in the life of these marshes, clonal diversity declines in the classic pattern of an ISR species, at least up to the age of the oldest marshes in the area (50–60 years). From an adaptive standpoint, the ability of S. alterniflora to rapidly invade disturbed areas (considered one of the major advantages of clonality; Eriksson & Jerling 1990; Grace 1993; van Groenendael et al. 1996) is partially offset by a geitonogamous selfing rate as high as 30% in young marshes and nearly complete mortality of inbred seedlings. The spatial clumping of genetically related clones, which develops early in the history of these marshes as a result of highly localized seedling recruitment, further increases the likelihood of inbreeding (Travis et al. 2004).

We make use of existing information on the geological history of delta formation in the Lower Mississippi River Valley (Frazier 1967; Penland et al. 1988) to infer the approximate ages of eight S. alterniflora marshes, with the primary goal of using a space-for-time substitution to determine the long-term effects of clonal growth on genotypic diversity. We predict that a continuously decreasing number of clones will monopolize S. alterniflora populations increasingly over time, so that clonal richness will decline. Any variation in disturbance history or patterns of seedling recruitment across marshes would also be expected to exert an effect on richness, but this is not addressed here.

We also predict that genetic diversity at the level of individuals and/or populations will be maintained or even increase as clonal diversity declines, because highly heterozygous clones are likely to dominate competitive interactions (e.g. Lee & Chung 1999; Erickson & Hamrick 2003; Hämmerli & Reusch 2003a). The clumping of ramets within clones (e.g. Chung & Epperson 1999; Hämmerli & Reusch 2003b; Ziegenhagen et al. 2003) is expected to occur over increasingly large distances as populations age, while the clumping of genetically related clones will become less pronounced as intraclonal competition begins to obscure the initial effects of localized seedling recruitment (Hämmerli & Reusch 2003b).

Methods

study sites

The coastal marshes of Louisiana's Mississippi River deltaic plain have been formed as sequential episodes of delta building have been followed by abandonment (Fisk 1944; Frazier 1967; Penland & Boyd 1981). During the past 7000 years, the Mississippi River has built six major delta complexes, within which there may have been more than 17 smaller subdeltas (Frazier 1967). Active delta building is currently restricted to approximately 20% of the delta plain and corresponds to the Balize delta of the Modern (Mississippi River) complex and the Atchafalaya delta complex.

The process of channel abandonment by the Mississippi River is gradual, resulting in considerable age overlap from one delta complex to another. Abandonment is followed by a trend towards conversion of fresh marshes to open water bays and salt marshes in areas proximate to the Gulf of Mexico (Frazier 1967). Currently, salt marshes are associated with only the four youngest delta complexes, as the two oldest existed when sea levels were 3–6 m lower. Radiocarbon dating of delta-plain peat obtained from more than a thousand sediment cores covering all contiguous and overlapping lobes of the Mississippi River deltaic plain (Frazier 1967), has shown that, at 3900–5700 years old, the Teche complex is the oldest delta complex that remains above sea level. This complex is currently represented by only a single remnant marsh, Marsh Island. Marsh Island may historically have been influenced by freshwater inflows from the Vermilion and Atchafalaya Rivers (Orton 1959), and thus probably represents a salt marsh ranging in age from 1000 to 1500 years. The St Bernard delta complex, believed to be 1700–4700 years old, is represented by salt marshes believed to range in age from 500 to 1000 years. Salt marshes in the Lafourche delta complex, believed to be 600–3500 years old, are more difficult to age because of the existence of four recognized distributaries (Penland et al. 1988), and may range in age from as little as 500 to as much as 1500 years.

The youngest salt marshes in the Mississippi River delta plain lie at the seaward extent of the modern Balize complex and do not exceed 200 years of age. One particularly young marsh has been identified in the Red Pass area of the Balize delta, created when an adjacent canal was overfilled during a restoration project. This sediment burial resulted in the death of all existing Spartina alterniflora, followed by extensive seed recruitment the following year (I. Mendelssohn, personal communication). Additional opportunities for sampling young sites existed in south-western Louisiana, where dredging of the Calcasieu River has gradually converted freshwater to saline marshes over the past 50–60 years. These sites have been described previously by Proffitt and coworkers (Proffitt & Young 1999; Travis et al. 2002, 2004; Edwards & Proffitt 2003; Proffitt et al. 2003).

In total, we sampled from eight natural salt marshes lying along the Gulf Coast of Louisiana (Fig. 1), ranging in age from 6 to approximately 1500 years (Table 1). Six sites represented the Mississippi River delta plain, including one site in the Teche complex (Marsh Island), one site in the St Bernard complex (Breton Sound), one site in the Lafourche complex (Bay Junop), and three sites in the Balize complex (Mississippi River, Nairn and Red Pass). The two remaining marshes represent the Calcasieu River estuary (Sabine National Wildlife Refuge and Hackberry).

Figure 1.

A map of the Louisiana Gulf Coast showing the locations of eight Spartina alterniflora marshes (stars) sampled for genetic diversity, as well as the approximate locations of four major delta complexes of the Mississippi delta plain (separated by hatched lines). Sampling locations are abbreviated as follows: RP = Red Pass, SR = Sabine NWR, HB = Hackberry, MR = Mississippi River, NC = Nairn, BJ = Bay Junop, BS = Breton Sound, MI = Marsh Island.

Table 1.  Site characteristics within Louisiana Spartina alterniflora marshes ranging in age from 6 to 1500 years, and genetic diversity measured as proportion of polymorphic markers, <P>, average heterozygosity, <H>, and average pairwise dissimilarity, D. Site abbreviations are as follows: RP = Red Pass, SR = Sabine NWR, HB = Hackberry, MR = Mississippi River, NC = Nairn, BJ = Bay Junop, BS = Breton Sound, MI = Marsh Island
SiteAge range (yr)ConfigurationSample interval (m)Distance samples (m)Sample size (n)No. of Clones (G)Maximum diameter (m)Clonal richness (G/N)<P><H>D
RP6Interior0.5–1.012517068 19.20.82 ± 0.030.45460.11500.0746
SR50–60Interior0.5–1.010010158 13.90.84 ± 0.040.57580.11650.0756
HB50–60Edge0.514512939 56.60.95 ± 0.050.66670.13910.1054
MR300Interior1.012510156 14.90.82 ± 0.040.75760.15460.1072
NC300–500Interior1.0–2.025015288 28.90.82 ± 0.030.48480.11820.0749
BJ500–1500Interior2.0200 6031NA0.68 ± 0.100.57580.14540.0942
BS500–1000Edge2.5–5.080013161593.40.62 ± 0.050.66670.12160.0789
MI1000–1500Edge5.0465 7538119.00.70 ± 0.060.51520.14100.0862

sampling

All sampling in the Mississippi River deltaic plain was performed from an airboat, while samples in the Calcasieu estuary were collected on foot. Transects were established and leaf material of S. alterniflora collected at regular intervals. As it was hypothesized that the older marshes would be composed of larger diameter clones, transect and sampling intervals were longer than in younger marshes (2–5 m intervals vs. 0.5–1 m intervals, respectively; Table 1). Samples were collected from two to six transects per marsh in the autumn of 1998 and 1999, except Bay Junop, which was sampled in the summer of 2001. Transects extended well into the interior of all S. alterniflora marshes except those that grew only as a narrow fringe bordering open water (Table 1). During collection, the samples were bagged in ziploc bags and placed on ice prior to further processing in the laboratory.

genotyping

We constructed a multilocus genotype for each individual ramet on the basis of amplified fragment length polymorphisms (AFLPs; Vos et al. 1995). We first extracted DNA from approximately 1 g of green leaf tissue using a CTAB (hexadecyltrimethylammonium bromide)-based method (Saghai-Maroof et al. 1984; Rogers & Bendich 1985; Doyle & Doyle 1987) as described in Travis et al. (2002). Fifty nanograms (50 ng) of each DNA sample was restriction digested with EcoRI and MseI, then primer binding sites were established by ligating custom adapters to the resulting restriction fragments, and an initial (pre-amplification) subset of these fragments was PCR-amplified according to the methods of Travis et al. (1996). For our final selective restriction fragment amplification (SRFA) we used a single primer combination known from a previous study (Travis et al. 2004) to be particularly effective at differentiating among S. alterniflora clones. This primer combination consisted of the selective nucleotides ACG and AGT attached to the 3’ end of the EcoRI- and MseI-primer, respectively. Amplified fragments were resolved by electrophoresis on an ABI PRISM® 310 Genetic Analyser (Applied Biosystems Inc., Foster City, California, USA), according to the methods of Travis et al. (2004). Finally, genetic profiles were scored for the presence or absence of a select set of markers using Genographer Version 1.6 (Benham 2001), according to the methods of Travis et al. (2004).

genetic analyses

In addition to characterizing diversity at the population level as both clonal and overall genetic diversity, we characterized diversity at the local level as spatial autocorrelation within and among genotypes. Except as otherwise noted, all our estimates of genetic diversity were based on the total pool of genotypes within each site, rather than the total set of ramets sampled.

We identified common members of clones as ramets sharing identical multilocus genotypes, as described in Travis et al. (2004). We developed sufficient site-specific markers to ensure that the probability of erroneously assigning two distinct individuals to the same clone was acceptably small, while restricting the number of markers to limit the possibility that non-identical genotypes could represent common clones, as would be expected with somatic mutation. Douhovnikoff & Dodd (2003) developed 177 AFLP markers for distinguishing among clones of Salix exigua, and found that ramets differed by an average of 2.0% of their markers within clones. Assuming similar values for S. alterniflora, the average number of markers by which ramets differ within clones will be held below 1 as long as the total number of markers is kept below 50. In a separate study (S. E. Travis and C. E. Proffitt, unpublished data), we used nine markers to differentiate among 35 replicate plantings of five distinct S. alterniflora clones, and found it to be 100% accurate in correctly assigning ramets to their respective clones.

Clonal diversity was characterized as richness, or the total number of multilocus genotypes, G, divided by the total number of ramets sampled, N (Ellstrand & Roose 1987). In order to standardize the spatial scale over which richness was calculated within sites, samples were drawn at random, without replacement, from 25-m transects. For those sites represented by transects longer than 25 m, each transect was sectioned into consecutive 25-m segments. Richness was calculated from the first set of five samples drawn from a given transect, then independently for the second set of five samples, and so on until fewer than five samples remained. This process generated up to eight independent estimates of richness per transect depending on sampling intensity. Overall richness was calculated for each site by averaging all estimates within and among transects.

We characterized the overall level of genetic diversity within populations using three separate metrics: the proportion of polymorphic markers, average heterozygosity, and the average coefficient of dissimilarity among individuals, as described in Travis et al. (2002). The first two approaches rely on an assumption of Hardy–Weinberg allele frequencies, while the third represents a purely phenetic approach. Average dissimilarities were based on the coefficient of Lynch (1990). We also compared the relationship between individual heterozygosity and clone size among marshes, with the expectation that, due to competition among clones, clone size will be increasingly associated with heterozygosity as marshes age. Specifically, we used linear regression analysis to determine the slope of the relationship between heterozygosity and clone size within each marsh, and then ran a regression of these slopes on age of marsh. For this analysis, clone size was quantified simply as number of ramets, while an index of individual heterozygosity, I, was calculated according to the following formula inline image, where qi is the frequency of the
recessive (null) allele at each locus and xi is 1 for marker presence and 0 for marker absence. This index of heterozygosity thus ranges from 0, when all population-specific markers are absent from an individual, to 1 when all markers are present.

We used spatial autocorrelation analyses based on genetic distances to determine the distribution of local genetic structure within marshes (cf. Degen & Scholz 1998). With this approach, non-random genetic distributions caused by either clones or familial relationships are visualized as distograms. Each distogram plots the average and 95% confidence intervals (CI) of genetic similarity against distance class. We considered a significant level of autocorrelation to occur when the lower CI exceeded the expected (mean) level of similarity for the entire population. It should be noted that for any given site only those distance classes with > 10 observations were included in the analysis. Spatial autocorrelation due to a clumped distribution of ramets within clones was assessed by including all pairwise comparisons in the analyses, while only those comparisons that revealed non-matching samples were used for the purpose of assessing clumped family distributions.

Dispersal rate, as inferred from genetic population structure, represents a critical element in evaluating the potential role of immigration in the establishment and maintenance of genetic diversity within sites. Therefore, we further assessed differentiation among populations using a statistical method specifically designed to overcome the shortcomings inherent in dominant marker data (foremost among them the necessity of assuming Hardy–Weinberg equilibrium frequencies of alleles). This method computes traditional F-statistics, equivalent to Weir & Cockerham's (1984) θ, on a pairwise basis using a recently developed Bayesian method which requires no prior knowledge of the level of inbreeding occurring within populations (Holsinger et al. 2002). Assuming a common level of inbreeding, f, and population structure, θB, for all loci sampled, individual allele frequencies are modelled as a Beta distribution, in keeping with the single locus expectations for an allele subject to drift, migration and mutation under Wright's infinite-island model. Prior distributions for f and θB are approximated through the use of Markov Chain Monte Carlo simulations, after a fixed number of initial samples have been discarded to ensure that the chain has converged to its stationary distribution. This analysis was performed using the computer program Hickory Version 0.7.

As a means of determining whether S. alterniflora populations adhere to a model of isolation-by-distance, we assessed the relationship between genetic and geographical distance among sites. In this case, geographical distance was measured as the minimum distance over open water that a seed or vegetation mat would have to travel between a pair of sites. A Mantel test (Mantel 1967) was used to test for a significant correlation.

Results

From the combined data of eight collection sites, composed of 1260 S. alterniflora samples, we scored 33 variable DNA fragments. Of these, a subset of 13–33 fragments was used to conduct pairwise comparisons among multilocus genotypes within sites, with an average of 21 ± 7 (SD) markers used per site. Thus, the average number of markers by which ramets would be expected to differ within clones was held at just 0.42. Using these markers, we identified 608 distinct genets, or clones (Table 1). Within any given collection site, the average probability of an erroneous match occurring between the genotypes of distinct clones was 3.44 × 10−3 (or 1 in 291), with the likelihood for the most abundant multilocus genotype averaging 0.0191.

clonal diversity

As predicted, clonal diversity, measured as clonal richness (G/N), declined with age of marsh. Given the uncertainty of the geological record, we ran three separate regressions to test this relationship, one based on the mean estimated age of each marsh, one on the minimum estimated age, and one on the maximum estimated age. Maximum estimated age provided the most reliable predictor of clonal richness, although all three regressions were significant (Table 2, Fig. 2a). Richness ranged from a high of 0.95 ± 0.05 (SE) for the 50–60 year old Hackberry site, to a low of 0.62 ± 0.05 (SE) for the 500–1000 year old Breton Sound site. The youngest site included in the study, Red Pass, estimated at 6 years of age, exhibited a richness of 0.82 ± 0.03 (SE), while richness for the oldest site, Marsh Island, aged at 1000–1500 years, was 0.70 ± 0.06.

Table 2.  Results of regression analyses of maximum clone diameters (log-transformed) and clonal richness on age of marsh for Spartina alterniflora marshes in Louisiana
Dependent variablesd.f.MSF ratioPR2Regression equation
Log(maximum clone diameter)
Mean age1, 50.941 4.050.1000.448y = 0.0009x + 1.314
Minimum age1, 50.716 2.580.1690.340y = 0.0010x + 1.353
Maximum age1, 51.080 5.280.0700.514y = 0.0007x + 1.296
Clonal richness
Mean age1, 60.05211.200.0160.651y = 0.0002x + 0.868
Minimum age1, 60.041 6.420.0440.517y = 0.0002x + 0.859
Maximum age1, 60.05412.870.0120.682y = 0.0001x + 0.867
Figure 2.

The relationship of clonal richness (a) and maximum clone diameter (b) with age of marsh among eight Spartina alterniflora marshes in Louisiana. Because of the uncertainty regarding the precise age of each marsh, each dependent variable was plotted independently against the mean estimated age, minimum estimated age, and maximum estimated age. Bay Junop, for which maximum clone diameter could not be reliably predicted on the basis of regression analysis, is omitted from (b).

In order to determine whether there was a significant effect of marsh age on maximum clone diameter, we first drew predictions from regressions of the proportion of matching pairs of ramets on their physical distance of separation (cf. Chung & Epperson 1999). A negatively exponential regression provided a significant fit to these data for nine of the 10 sites studied (Fig. 3), the sole exception being the Bay Junop site. Thus, we were able to predict maximum clone diameters for seven of the sites from regression equations (Table 1), and to plot these values against age of marsh. As before, we ran three separate regressions of the dependent variable on marsh age using average, minimum and maximum estimates. In each case, a logarithmic function provided the best fit to these data (Fig. 2b depicts the log-transform of the maximum clone diameter at each site). This relationship was non-significant (at an alpha of 0.05) regardless of the method of estimating age, although P-values were in the 0.07–0.10 range for two of the three regressions (Table 2).

Figure 3.

The proportion of ramets sharing identical multilocus genotypes plotted against distance for eight Spartina alterniflora marshes in Louisiana. Trend-lines are shown for all regressions, which were significantly exponential for all marshes except Bay Junop.

spatial autocorrelation

The distance over which spatial autocorrelation occurred as a result of the clumping of ramets within clones was predictable on the basis of age of marsh, although for some of the older marshes, specifically Bay Junop and Breton Sound, spatial autocorrelation did not occur in the conventional sense (Fig. 4). The maximum estimated age of each marsh, chosen for its explanatory power in the two prior analyses, accounted for 67% of the variation in the distance over which spatial autocorrelation occurred for the remaining six sites (F = 8.25; d.f. = 1, 4; P < 0.05; Fig. 5), which ranged from roughly 3.5–17.5 m. This indicates that many clones maintain a clumped distribution of ramets as they increase in size over time, although the sporadic occurrence of autocorrelation over the entire extent of the areas sampled at Bay Junop and Breton Sound indicates that considerable fragmentation of large, old clones must have occurred at these sites.

Figure 4.

Distograms for eight Spartina alterniflora sites in Louisiana showing spatial autocorrelation due to clumping of ramets within clones. Mean genetic similarity values are plotted by distance interval. Smoothed lines represent upper and lower 95% confidence intervals, while the flat line bisecting each plot represents random expectation.

Figure 5.

The approximate distance over which spatial autocorrelation due to clumping of ramets within clones occurs, plotted against the maximum estimated age of marsh for six Spartina alterniflora marshes in Louisiana. Note that sites failing to show a coherent pattern of autocorrelation relative to spatial distance (Bay Junop and Breton Sound) were not included in this analysis.

Spatial autocorrelation due to kinship, or the spatial clumping of genetically related clones, was rarely observed (Fig. 6) and was not associated with age of marsh. This form of spatial autocorrelation was observed when clones were growing within 1–2 m at the Hackberry site, and there was a barely discernible degree of autocorrelation in the shortest distance class (5 m) at the Marsh Island site.

Figure 6.

Distograms for eight Spartina alterniflora sites in Louisiana showing spatial autocorrelation due to clumping of genetic relatives within populations. Mean genetic similarity values are plotted by distance interval. Smoothed lines represent upper and lower 95% confidence intervals, while the flat line bisecting each plot represents random expectation.

genetic diversity

Using a simple linear regression analysis of genetic diversity on the age of a marsh (with the qualitative variable, marsh configuration, either marsh edge or interior, as a covariate), we found no significant relationship. Overall genetic diversity was relatively consistent from site to site regardless of how it was measured. The proportion of polymorphic loci ranged from 0.48 to 0.76, with a mean of 0.59 ± 0.04 (SE); average heterozygosity ranged from 0.115 to 0.155, with a mean of 0.131 ± 0.005; and the mean coefficient of dissimilarity ranged from 0.075 to 0.107, with a mean of 0.087 ± 0.005 (Table 1).

The relationship between clone size and individual heterozygosity was found to increase with age of marsh. Whereas among young marshes (aged < 100 years) the slope of the regression of number of ramets on heterozygosity was essentially zero, in the oldest marshes (estimated at 1500 years) each unit increase in the number of ramets per clone was accompanied by an increase in heterozygosity of as much as 0.16. The maximum estimated age of each marsh explained 68% of the variation in the regression slopes among marshes (Fig. 7), which was significant (F = 12.55, d.f. = 1, 6, P = 0.0122). It should be noted that we removed one outlier from the determination of slope for the Bay Junop site, specifically, the clone for which 26 ramets were observed. This clone accounted for 43% of the ramets sampled at Bay Junop, whereas no single clone accounted for > 7% of the ramets sampled at any other site.

Figure 7.

The regression slope of individual heterozygosity on clone size, plotted against the maximum estimated age of marsh for eight Spartina alterniflora marshes in Louisiana.

The overall level of differentiation between collection sites was relatively small (θ = 0.0854 ± 0.0113 SD), given that the average span of open water between sites was relatively large (210 km). Assuming an island model of migration, this represents an average migration rate among sites of 2.68 genets per generation. Pairwise comparisons of genetic differentiation between sites ranged over nearly an order-of-magnitude from a low of θ = 0.0295 ± 0.0191 (SD) for Mississippi River vs. Red Pass, which were separated by just 3 km, to a high of θ = 0.2082 ± 0.0529 (SD) for Hackberry vs. Red Pass, which were separated by 355 km (Table 3). Genetic differentiation was significantly correlated with geographical distance (matrix correlation, r = 0.660, Mantel t = 2.94, P = 0.003 based on 1000 permutations; Fig. 8).

Table 3.  Geographic distances in km (above diagonal) and fixation indices based on the Bayesian estimator of Holsinger et al. (2002) (below diagonal), for all pairwise comparisons of Spartina alterniflora marshes surveyed in Louisiana
 RPSRHBNCMRMIBSBJ
Red Pass (RP)341355 42  3217122149
Sabine NWR (SR)0.17 20341338136436192
Hackberry (HB)0.210.14355352150450206
Nairn (NC)0.050.120.20 39217151149
Mississippi River (MR)0.030.100.130.05214123146
Marsh Island (MI)0.150.100.110.140.09313 97
Breton Sound (BS)0.040.100.120.040.040.07244
Bay Junop (BJ)0.100.080.110.080.040.060.04
Figure 8.

A plot of genetic distance, measured as the Bayesian estimator θ, vs. geographical distance for all pairwise comparisons of eight Spartina alterniflora marshes in Louisiana.

Discussion

genet dynamics

Our data provide general support for the hypothesis that, among the estimated 60% of all clonal plant species believed to be characterized by initial seedling recruitment (Eriksson 1989), recruitment does not occur at a sufficient frequency to offset clone mortality. This leads to a steady decline in clonal diversity over time. Our comparison of clonal richness among eight natural populations of S. alterniflora growing in and around the Mississippi River Delta demonstrated a gradual, but significant, decline of approximately 1% per 100 years over a 1500-year period when the maximum estimated age was used for each site (R2 = 0.68; Table 2, Fig. 2). Consistent with our earlier work in which we compared clonal diversity among four marshes aged less than 60 years (Travis et al. 2004), we found that richness increased over a relatively short span of years very early in marsh development, from 0.82 ± 0.03 at 6 years to a peak of 0.95 ± 0.05 (SE) at roughly 50 years, before the onset of the gradual decline (Table 1). Thus, we have provided evidence from two separate studies covering both short and long spans of time demonstrating that S. alterniflora fits the classic pattern of an ISR species. To our knowledge, this is the first study of its kind to use a space-for-time substitution to follow partially clonal populations over a time-scale stretching over more than a millenium, as opposed to several hundred years (see Verburg et al. 2000). Unfortunately, the close dependency of clonal diversity estimates on sampling scale (Reusch et al. 1998; Stehlik & Holderegger 2000) makes it difficult for us to compare the absolute values obtained in our study with those from similar studies.

Given the rate of decline that we observed, environmental stability would need to persist for nearly 10 000 years for the complete loss of clonal diversity from a S. alterniflora population, which is extremely improbable from a geological standpoint. In fact, low intensity disturbances occur naturally in wetland ecosystems with relatively high frequency (Middleton 1999; Mitsch & Gosselink 2000; Zedler 2003; K. R. Edwards, S. E. Travis and C. E. Proffitt, unpublished manuscript), and would be expected to provide recurrent opportunities for seedling establishment. In addition, although clonal plants can grow quite large (Reusch et al. 1999; Ruggiero et al. 2002), they are not known to achieve life spans of more than several thousands of years (Cook 1985; Watkinson & White 1986), although rhizomatous species may be particularly long-lived (Eriksson 1989). Thus, if we could follow populations for more than 2000 years, we would probably begin to see clonal diversity approaching asymptotic values as a dynamic equilibrium is established between vegetative growth during periods of stability and seedling recruitment following disturbance. Competition could further contribute to the maintenance of clonal diversity by selecting for adaptive differentiation among clones (e.g. McLellan et al. 1997; Skalova et al. 1997), particularly if environmental heterogeneity exists on a microenvironmental scale (Vrijenhoek 1979).

Although we found the relationship between maximum clone diameter and age of marsh to be non-significant at P = 0.07 (R2 = 0.51 using the maximum estimated age for each marsh; Table 2, Fig. 2), there was a significant relationship between the distance over which spatial autocorrelation occurred because of clonality and age of marsh (R2 = 0.67; Fig. 5). Thus, we feel confident that clone size is at least partially explained by marsh age, especially given our finding that clonal richness declines with age. Assuming that overall ramet density remains relatively constant over the life span of a marsh, at least beyond the initial colonization phase, a decline in clonal richness must necessarily be accompanied by an increase in the area covered by at least a subset of the remaining clones. Our finding that the size of the largest clones increases as a logarithmic function of marsh age suggests that relative competitive ability is highly skewed towards a relatively small number of clones that increase rapidly in size at the expense of other clones, which may be poorer competitors.

Two of the oldest sites included in our study, Bay Junop and Breton Sound, did not exhibit spatial autocorrelation within clones to the extent that a clear pattern of decline was evident as the sampling scale increased. Rather, these sites showed an erratic pattern of autocorrelation that occurred repeatedly over nearly the entire area sampled (Fig. 4). We take this as an indication that the largest clones representing these two sites have become increasingly fragmented over time. In fact, at the Bay Junop site, one highly fragmented clone had become so large that sampling over a distance of 200 m was insufficient to cover its entire extent. This explains why there was no significant decline at this site in the proportion of ramet pairs representing like genotypes as the distance between them increased (Fig. 3). It is not surprising that we observed clone fragmentation at our oldest sites, given that we have previously observed even very young clones undergoing partial senescence beginning from their initial point of establishment and subsequently expanding to within several metres of their most active zone of growth (Travis et al. 2004; Proffitt et al. 2005). Such fragmentation is commonly observed in clonal plants (see Pitelka & Ashmun 1985; van Groenendael & de Kroon 1990), and may enhance long-term clone survival by preventing mortality due to the spread of disease via rhizomatous connections (McCrea & Abrahamson 1985).

Fragmentation may be of importance to the long-term viability of S. alterniflora marshes by improving opportunities for outcrossing. As clones age, they may increasingly accumulate somatic mutations that significantly reduce their reproductive vigor (Klekowski 1988, 1997; Klekowski & Godfrey 1989; Thompson et al. 1991) and heighten the effects of inbreeding depression. Fragmentation may have the beneficial effect of decreasing the incidence of geitonogamous selfing within large, old clones (Hämmerli & Reusch 2003b), considerably improving their chances of producing viable progeny with the potential for colonizing new substrates made available through disturbance.

Counter to expectations, we did not detect a higher incidence of spatial autocorrelation due to kinship in young S. alterniflora marshes compared with older marshes (Fig. 6). In a previous study we found that local genetic structure resulting from localized seedling recruitment and assessed on the basis of F-statistics, was highly negatively correlated with clonal diversity, such that once diversity had begun to decline beyond a marsh age of approximately 30 years, local genetic structure could account for between 5 and 10% of overall genetic variation within 1–6 m diameter sampling plots (Travis et al. 2004). These estimates were based on 6–12 ramets per plot. On the other hand, structure at the 12-m scale dropped to less than 5%. Thus, our ability to detect spatial autocorrelation is clearly dependent on both the scale and intensity of sampling (see also Hämmerli & Reusch 2003b), and it may be that our current study failed on one or both of these levels. It may also be that the process by which a new marsh is formed will largely determine the dynamics of seedling recruitment, and that the three marshes included in our previous work, which were created through the deposition of dredged sediments, are subject to a much different set of dynamics than are natural marshes associated with the formation of a river delta. A notable characteristic of created marshes is the availability of large areas of bare substrate for localized seedling recruitment in the vicinity of early adult colonizers, whereas far fewer opportunities for seedlings may be available in natural marshes, as suitable substrate would be expected to develop much more gradually over time.

genetic diversity

Whereas overall levels of genetic diversity were similar for all natural marshes sampled, showing no clear relationship with marsh age, the relationship between heterozygosity and individual clone size became increasingly pronounced within older marshes. Early colonists of a newly formed marsh could comprise an essentially random sample of the local gene pool, with the size of each clone depending largely on stochastic processes related to colonization itself. Thus, in areas of newly available substrate, early colonists would be relatively free of intraspecific competition and the winnowing effects of genetic diversity. This situation might be expected to persist throughout the period of marsh establishment, which in the case of S. alterniflora in Louisiana would primarily depend on the combined effects of subsidence and sea level rise following channel abandonment by the Mississippi River (Frazier 1967). As we have observed clonal diversity to peak at approximately 30–60 years within newly established marshes (Travis et al. 2004; this study), indicating that this is generally the time period required for the winnowing process to begin, we would only expect to see a positive relationship between heterozygosity and clone size among somewhat older marshes. Indeed, we observed a positive relationship beginning with our two marshes estimated at 300–500 years of age, and this relationship only became strongly positive in our very oldest marshes, with maximum ages approaching 1500 years (Fig. 7).

Why did overall levels of genetic diversity not increase in our oldest populations, as might be expected with the preferential spread of highly heterozygous clones? One important consideration is the method by which genetic diversity is calculated and compared at the individual vs. the population scale. At the population scale, genetic diversity differences are based largely or wholly on the incidence of monomorphic loci, whereas monomorphic loci are entirely uninformative in (and therefore omitted from) calculations of heterozygosity at the individual scale. Thus, the most prolific clones may be highly heterozygous in comparison with other individuals within their respective population, but if this same population is characterized by a relatively low incidence of polymorphic markers, its overall genetic diversity may not be particularly high. It is also possible that, over time, overall genetic diversity may be balanced between the loss of individuals carrying genes that are poorly adapted to the local set of environmental conditions, and the competitive gains made by highly heterozygous clones. Finally, it is worth considering that, even among our least clonally diverse sites, the majority of clones in our sample were represented by just a single ramet, suggesting that small-scale disturbances may frequently promote the germination of locally produced seed. We have established previously (Travis et al. 2004) that outcrossing in S. alterniflora may occur at a rate approaching 90% in young marshes with high clonal diversity. While this level almost certainly drops off to some extent in older, less diverse marshes, it may still be maintained at a fairly high level by clone fragmentation, and when selfing does occur, we and others (Somers & Grant 1981; Daehler & Strong 1994; Daehler 1996, 1998, 1999) have established that inbreeding depression is sufficiently intense that few, if any, of the inbred seedlings survive. This may then serve as a mechanism by which relatively high levels of heterozygosity are maintained for all adult S. alterniflora clones regardless of their competitive ability.

In a previous study (Travis et al. 2002), we determined that population-wide genetic diversity develops rapidly within young created marshes, such that it equals or exceeds that of natural marshes in the same area within 2–3 years, and we attributed this finding to high rates of gene flow among marshes. Our current findings indicate that genetic diversity is maintained at similar levels over at least several thousand years, and that the average degree of genetic differentiation among populations is not especially pronounced, with an expectation of 2.68 migrants per generation. In reality, migration rates among marshes are likely to be highly dependent on the actual degree of geospatial separation, given that S. alterniflora adheres to a model of isolation by distance (Fig. 8), and colonization status, as there should be virtually no further recruitment once a population is fully established. A high migration potential is not surprising given the widespread distribution of S. alterniflora and the efficiency of water as a dispersal mechanism for both seeds and vegetative fragments (Hamrick & Godt 1990). In addition, as an ISR species (Travis et al. 2004), S. alterniflora is expected to be well adapted for long-range dispersal of seeds (Eriksson 1993).

Acknowledgements

We thank E. Mouton for providing airboat transportation, and K. Edwards and R. Lowenfeld for assistance with field collections. T. Hibbs, A. Ingalls, J. Kemmerer and S. Reymann provided further field and laboratory assistance; S. Stevens provided additional laboratory assistance. We thank N. Anthony and two anonymous reviewers for helpful comments on a previous version of this manuscript. We thank G. Linscombe and B. Savoie of the Louisiana Department of Wildlife and Fisheries for granting us access to Marsh Island State Wildlife Refuge. The Louisiana Environmental Research Center, McNeese State University, provided the facilities and personnel to conduct portions of this research. This work was supported by a US Environmental Protection Agency (EPA) grant to the Louisiana Environmental Research Center (US EPA Office of Research and Development, National Center for Environmental Research, Science to Achieve Results (STAR) Program, Grant No. R825990). The use of trade names is for descriptive purposes only and does not imply endorsement by the US Government.

Ancillary

Advertisement