American Journal of Botany 89(1): 29–36. 2002.
STRATIFIED ANALYSIS OF THE SOIL SEED BANK IN THE
CEDAR GLADE ENDEMIC ASTRAGALUS BIBULLATUS:
EVIDENCE FOR HISTORICAL CHANGES IN GENETIC
ASHLEY B. MORRIS,2,4REGINA S. BAUCOM,3,5AND
MITCHELL B. CRUZAN2,3,6
2Department of Botany and3Department of Ecology and Evolutionary Biology, University of Tennessee,
Knoxville, Tennessee 37996 USA
Persistent seed banks may provide information on historical changes in the genetic composition of populations. We used stratified
sampling of the soil seed bank of Astragalus bibullatus (Pyne’s ground plum) to assess levels of temporal variation in population
genetic structure and to infer historical changes in the levels of inbreeding and relative gene flow. This species has an extremely
limited distribution in the Central Basin of Tennessee, where it is found in open areas and along the edges of cedar glades. Protein
electrophoresis was conducted on seedlings grown from seeds that had been recovered from three successive 1 cm thick layers of soil
sampled from six sites. Analyses of seven polymorphic allozyme loci indicated that there were substantial levels of genetic differen-
tiation among soil layers and sites. Higher levels of genetic diversity were found in seed than in vegetative populations that had been
sampled in a previous study. Seed populations from the uppermost soil layer had higher heterozygote deficiencies, displayed higher
levels of differentiation among sites, and had higher private allele frequencies than seed populations from the lower two layers. The
change in heterozygosity and distribution of genetic variation among sites for the youngest soil layer is consistent with a pattern of
increased selfing, sib mating, and decreased gene flow among populations. These changes in inbreeding and relative levels of gene
flow are corroborated by information on historical land use practices in the region and support the hypothesis that loss of appropriate
habitat has led to smaller population sizes and a more fragmented distribution of this cedar glade endemic.
flow; seed bank; temporal variation.
Astragalus bibullatus; fragmentation; inbreeding; landscape genetics; population size; private alleles; relative gene
Theoretical analyses indicate that shifts in the geographic
distribution of organisms can precipitate changes in the level
and distribution of genetic variation. For example, it has been
suggested that increased fragmentation of distributions would
disrupt patterns of gene flow among populations, increase the
severity of inbreeding, reduce levels of local genetic diversity,
and ultimately affect the pattern and mode of evolution (Tem-
pleton et al., 1990; Fahrig and Merriam, 1994; Young, Boyle,
and Brown, 1996). The predicted effects of increased frag-
mentation have been supported by a number of empirical stud-
ies of populations that have recently become isolated from
previously contiguous distributions (Young, Merriam, and
Warwick, 1993; Prober and Brown, 1994; Hall, Walker, and
Bawa, 1996; Young, Boyle, and Brown, 1996; Nason and
Hamrick, 1997; Morden and Loeffler, 1999; Turner et al.,
2000; Cruzan, in press). While these investigations have made
important contributions to our understanding of landscape ge-
netic processes, their conclusions are inferential because com-
parisons of historical and contemporary distributions of ge-
1Manuscript received 16 March 2001; revision accepted 17 July 2001.
The authors thank D. Thonnard, S. Wright, A. Shea, M. Webber, and S.
Case for assistance in the field and the lab; and J. Estill, C. Murren, and H.
Delcourt for improving the manuscript through discussions. Financial support
for this project was provided by the Fish and Wildlife Service and the Natural
Heritage Program of the Tennessee Department of Environment and Conser-
4Current address: Department of Biology, University of Florida, Gaines-
ville, Florida, USA.
5Current address: Department of Genetics, University of Georgia, Athens,
6Author for reprint requests (e-mail: Cruzan@utk.edu).
netic variation for the same populations were not performed.
However, direct comparisons of temporally separated popula-
tions occurring at the same site could provide more accurate
assessments of the consequences of landscape modifications
for the level and distribution of genetic variation and could
measure the potential for fragmented populations to maintain
One approach to the examination of historical patterns of
population genetic structure would be to analyze the genetic
variation present in populations of dormant individuals (e.g.,
Bosbach, Hurka, and Haase, 1982; Vavrek, McGraw, and Ben-
nington, 1991; Tonsor et al., 1993; Cabin, 1996; Cabin, Mitch-
ell, and Marshall, 1998; McCue and Holtsford, 1998; Schnell-
er, 1998). In particular, temporal analyses may be possible in
systems where dormant representatives of past populations are
present in vertical strata that allow their relative age to be
inferred (e.g., van der Valk and Davis, 1979; McGraw, 1993).
Such stratigraphic depositions could provide historical records
of changes in populations and communities. For example,
stratified sampling of seed banks from different soil depths has
been used to infer changes in vegetation composition (Kell-
man, 1970; Leck and Simpson, 1987; Archbold, 1989). While
it would be feasible to use soil seed populations to assess
historical changes in the level and distribution of genetic var-
iation, there have been few attempts to infer temporal variation
in population genetic structure from stratified samples of dor-
mant individuals (e.g., Schneller, 1998).
Here we use stratified sampling of the soil seed bank to
examine historical changes in population genetic structure of
the perennial cedar glade endemic, Astragalus bibullatus Bar-
neby and E. L. Bridges (Fabaceae). This species is ideal for
30[Vol. 89AMERICAN JOURNAL OF BOTANY
an investigation of a temporal analysis of genetic variation
because characteristics of its seeds and habitat favor the de-
velopment of a persistent seed bank that is stratified by age.
First, as with many species in this group, seeds of A. bibullatus
possess hard, impermeable seed coats that impose a strong
physical germination barrier (Rolston, 1978; Baskin and Bas-
kin, 1998). Second, species of legumes that have hard seed
coats are relatively long-lived and are known to remain viable
in the soil longer than seeds of most other species (Toole and
Brown, 1946; Quinlivan, 1968; Baskin and Baskin, 1998).
Third, soils in the cedar glade habitats consist almost entirely
of coarse sand and rocks (Quarterman, Burbanck, and Shure,
1993), which, in combination with repeated frost heaving and
sedimentation processes, may promote the migration of the
smooth, hard seeds of A. bibullatus down through the soil
column. While it has been suggested that the digging activity
of rodents and invertebrates can disrupt age stratification of
seed banks (Chambers and Macmahon, 1994), evidence of soil
disturbance by animals appears to be minimal in cedar glade
habitats (M. B. Cruzan, personal observation), probably be-
cause the soil is shallow and rocky (Quarterman, Burbanck,
and Shure, 1993). Hence, we expect the development of an
age-stratified seed bank for populations of A. bibullatus, with
the most recently produced seeds near the soil surface and
average seed age increasing with soil depth.
In this study, we assessed levels of genetic diversity and the
genetic structure of past populations of A. bibullatus by sam-
pling seed populations at different depths in the soil column.
Specifically, we used estimates of heterozygote deficiency and
genetic diversity within populations, and differentiation among
populations, to infer historical changes in levels of inbreeding
and relative gene flow. Our temporal analyses of population
genetic processes provide an example of the consequences of
the effects of increased fragmentation and habitat loss on the
level and distribution of genetic variation.
Astragalus bibullatus is an herbaceous perennial endemic to cedar glades
of the Central Basin of Tennessee (Barneby and Bridges, 1987), USA. The
known distribution of A. bibullatus is limited to seven locations within Ruth-
erford County, Tennessee, USA, with two apparently extirpated populations.
The sparse distribution of contemporary populations of this species and its
limited seed dispersal strongly suggest that it must have been more abundant
in the past. The cedar glade habitats of A. bibullatus are open areas with
varying densities of eastern red cedars (Juniperus virginiana) that are domi-
nated by a sparse cover of annual or perennial forbs, annual grasses, and
cryptogams. These areas are characterized by shallow soils, high levels of
irradiance, and temperature extremes (Quarterman, Burbanck, and Shure,
1993; Baskin and Baskin, 1999). During the winter months soils often remain
saturated for long periods and frost heaving is common (A. Shea, Tennessee
Department of Environment and Conservation, personal communication). Pre-
sumably the expansion that occurs during freezing produces fissures in the
soil that promote the burial of seeds. Within glades, A. bibullatus is often
restricted to transition areas along the edges or associated with scattered trees
within glades where they are partially shaded by overstory vegetation. The
flora of cedar glades in the Central Basin is relatively rich with a high inci-
dence of endemism and taxa with broader distributions in the grasslands of
Midwestern North America (Quarterman, Burbanck, and Shure, 1993; Baskin
and Baskin, 1999; Estill and Cruzan, 2001).
Astragalus bibullatus (Pyne’s ground plum) is an herbaceous perennial,
flowering from late April to early May and fruiting in early June. Plants are
acaulescent low-growing rosettes, up to 25 cm in diameter. Their leafy rosettes
arise from fleshy roots that may be branched beneath the soil surface, so larger
plants may consist of several closely spaced rosettes. Inflorescences remain
close to the ground and bear compact racemes of 10–16 pink flowers (?1
cm in length) with darker purple markings. Fruits are inflated pods, 1.5–3 cm
in length and 1–1.5 cm in diameter, which acquire a characteristic reddish
‘‘plum’’ color as they mature. Observations indicate that each inflorescence
typically produces only one or two fruits. Ovaries have up to 40 ovules, but
fruits rarely contain ?30 seeds (M. B. Cruzan, unpublished data). The kidney-
shaped seeds are 2–4 mm in length and are shiny and black, which facilitate
their identification and extraction from soil samples. Primary seed dispersal
is by gravity, but secondary dispersal by water is possible during the winter
months when surface flow is common in cedar glades. Germination in the
field occurs during February and March (M. B. Cruzan, personal observation).
Differences in levels of reproduction among sites have been noted and are
thought to be a result of shading from the encroachment of woody vegetation
(Baskauf and Snapp, 1998). Observations suggest that shaded plants produce
greater vegetative growth and fewer fruits, while plants in full sun produce
large numbers of fruits and go dormant earlier. The pollinators of A. bibullatus
have not been identified, but casual observations indicate that small-bodied
bees and skippers (Hesperiidae) visit flowers (M. B. Cruzan, personal obser-
Sampling methods—We took stratified samples of the soil seed bank in
February 1999. We used six of the known extant sites for this study, all of
which were in Rutherford County and within 6 km of each other. Four of
these sites are the same as were sampled by Baskauf and Snapp in a study
of genetic variation in extant vegetative populations (their WS ? Flat Rock
B, A ? Alexander, D ? Davis, and O ? Overbridge; Baskauf and Snapp,
1998). At the time that we sampled the Overbridge site, it consisted of native
plants plus transplants from Baskauf and Snapp’s C site, which has been
nearly extirpated by the land owner. Our Flat Rock A site was within 100 m
of another one of Baskauf and Snapp’s sites (WO) and our Flat Rock B site.
The Airfield site was discovered after Baskauf and Snapp’s study was com-
pleted and is within 100 m of the Alexander site. A seventh extant site was
not sampled because it was discovered after the completion of our field stud-
At each site, one (five of the sites) or two (the Airfield site only) 5-m
transects were placed in areas where the largest concentrations of plants were
known to occur. The seed bank was sampled by collecting three layers of soil
from five 30 ? 30 cm quadrats at 1-m intervals along each transect. Fruits
from the previous season, along with moss, lichens, and debris, were removed
from the surface of each plot prior to excavation. Three layers of soil (labeled
A, B, and C going from the surface to the deepest layer), ?1 cm in thickness,
were carefully removed from each plot using flat masonry trowels. Seeds may
have been present in deeper soil strata, but they were not sampled in this
study. As lower layers were removed, care was taken to prevent contamination
from the upper soil layers. Soil layers were stored in separate resealable plastic
bags at 4?C to retard the growth of mold during the 1–3 mo period before
they could be processed. Seeds were extracted from soil samples by sifting
the soil with a No. 10 soil sieve (2-mm openings). The seeds collected from
each layer were stored in separate envelopes at room temperature for several
weeks until they were treated for germination trials.
Preliminary tests indicated that seeds of A. bibullatus possessed a physical
germination barrier (i.e., a hard, impermeable seed coat), but did not require
an extended cold treatment. We determined that treatment with concentrated
H2SO4for 15 min was the most effective method for rendering seed coats
permeable to water after several trials with alternative methods of scarification
(Baskin and Baskin, 1998). Treated seeds were rinsed with deionized water
for 15 min and placed in petri dishes on 1% agar containing Hogland’s basal
medium (Sigma H-2395). Petri dishes were first stored for 1 wk at 4?C before
transferring them to a growth chamber with a 12/12 h alternating light/dark
cycle and a corresponding 20?/10?C alternating temperature. Most seeds
quickly imbibed water and nearly doubled their volume within a few days.
Any seeds that remained small after 1 wk were retreated with sulphuric acid
and returned to the growth chamber. Viable seeds generally germinated within
2 wk, and any seeds that remained ungerminated after 4 wk in the growth
chamber were scored as inviable. Viability tests with tetrazolium chloride
January 2002] 31MORRIS ET AL.—STRATIFIED ANALYSIS OF THE SOIL SEED BANK
calculated based on the total number of seeds found in all of the 30-cm2quadrats sampled at each site. Soil layers A–C (uppermost to the
lowest) were pooled to estimate the proportion of polymorphic loci (p), the effective number of alleles per locus (Ae), and the genetic diversity
(He). No viable seeds were recovered from the missing layers at Davis, Flat Rock A, and Overbridge.
The number of quadrats sampled and seeds collected from each population of Astragalus bibullatus. Seeds per square meter were
170 0.72.3 0.12
490.5 1.6 0.15
Davis5 24 0.3 1.5 0.16
Flat Rock A5
Flat Rock B5 28
753 0.63.0 0.10
Overbridge5 62 0.62.10.26
(Baskin and Baskin, 1998) on a subsample of the ungerminated seeds con-
firmed that none of them contained live embryos (unpublished data), so for
the purposes of this study we assumed that germination is equivalent to via-
bility. Upon germination, seedlings where transplanted into soil flats and
moved to a greenhouse. Leaf material was collected from all seedlings for
allozyme analysis 2–3 wk after transplanting.
Electrophoretic methods—We used horizontal starch gel electrophoresis to
estimate the levels and distribution of genetic variation present in different
strata of the soil seed bank. Approximately 0.5 cm2of leaf material was
ground in 300 ?L of extraction buffer (Cruzan, 2001) in 1.5-mL microcen-
trifuge tubes with plastic pestles. The extracted materials were stored in mi-
crocentrifuge tubes at ?70?C. On each day that assays were conducted, frozen
samples were thawed and absorbed onto 3 ? 10 mm wicks cut from Whatman
#3 filter paper. We made initial screens of 20 enzymes on six gel buffers to
identify two buffer systems that clearly and consistently resolved ten loci: (1)
Tris Borate EDTA pH 8.3 for ME (one locus), LAP (one locus), PGI (one
locus), and G3PDH (two loci); (2) L-Histidine pH 5.7 for PGM (two loci),
6PGD (two loci), and ADH (one locus). Tris Borate EDTA gels were run at
55 mA for 5 h. L-Histidine gels were run at 30 mA for 3.5 h. Gels were
documented using a video camera fitted with a video copy printer. Genotypes
were determined from the video images.
Data analysis—We analyzed the seed sampling data to assess differences
in the numbers of seeds recovered and levels of seed viability among sites
and soil strata. These data were not normally distributed, so we used Fried-
man’s two-way analysis of ranks blocked by site (SAS, 1989) to test for
differences in the number and viability of seeds among soil layers and sites.
Genetic data were analyzed to determine whether there were significant
levels of genetic differentiation among soil layers and sites to assess temporal
changes in population genetic parameters. The distribution of genetic variation
within and among sites was analyzed using both hierarchical (soil layers nest-
ed within each site for the A layer and the B and C layers combined) and
stratified (the A layer and the combined B and C layers compared among
sites) designs with the Genetic Data Analysis (GDA) and PopGene (Yeh and
Boyle, 1997) software packages. These programs use Weir and Cockerham’s
(1984) and Nei’s (1973) methods, respectively, to examine genetic structure
(see Weir, 1996 for a comparison of these methods). The 95% confidence
intervals for genetic structure parameters were estimated with GDA by boot-
strapping (1000 replications) across loci. Estimates of gene flow (Nm) among
the sites sampled for different seed bank strata were made using both private
allele analyses (Slatkin, 1985) and FSTmethods (Hedrick, 1983). Note that
gene flow estimates are generally not accurate (Whitlock and McCauley,
1999) and are used here strictly for comparative purposes. Genetic diversity
and differentiation parameters were compared among layers using Friedman’s
two-way analysis of ranks blocked by locus with ranks weighted by each
sample size (SAS, 1989).
The number of seeds recovered from soil samples and the
average viability of seeds varied among sites and soil layers.
A total of 561 seeds were extracted from the 35 quadrats sam-
pled across six sites. Of these, 311 (55%) germinated and were
used in allozyme assays. The number of A. bibullatus seeds
found in soil samples varied substantially among sites, ranging
from a low of 11 (Davis) to a high of 339 (Flat Rock B; F ?
6.01, df ? 5,96, P ? 0.0001; Table 1). Estimates of the total
number of seeds per square meter also varied dramatically
among sites, from a low of 24 to a high of 753 (Table 1). The
viability of soil seeds differed among sites, ranging from a
low of 23% (Airfield) to a high of 72% (Flat Rock A; F ?
8.25, df ? 5,40, P ? 0.0001). Differences among layers for
the number and viability of seeds were less evident (Fig. 1a).
Seed recovery was somewhat greater for the B layer than for
the other two soil strata (Fig. 1a). However, this pattern was
not significant (F ? 2.85, df ? 2,96, P ? 0.630) and was
primarily due to the relatively large number of seeds collected
from the B layer at the Flat Rock B site. Seed viability tended
to decline with soil depth (Fig. 1b), but the difference in ger-
mination among soil layers was not significant (F ? 0.28, df
? 2,40, P ? 0.760).
Analysis of allozyme variation indicated that relatively high
levels of genetic diversity were present in the seed bank pop-
ulations sampled. Out of the ten allozyme loci assayed, only
the two G3PDH loci and the ME locus were monomorphic for
all of the populations sampled, leaving a total of seven poly-
morphic loci. Levels of genetic diversity as measured by the
proportion of polymorphic loci (p), the effective allele number
(Ae), and the expected heterozygosity (He) were relatively high
and varied to some degree among sites (Table 1). Since sample
sizes for the lower soil layers were relatively small (Table 1),
and because allele frequency differentiation between them was
not significantly different from zero (?p? 0.018, and the 95%
confidence interval [CI] of 0.045 to ?0.002 overlaps zero; Fig.
2a), we pooled the B and C soil layers for genetic analyses.
Comparison among soil layers indicated that the effective
number of alleles per locus (Ae) and the level of genetic di-
32[Vol. 89AMERICAN JOURNAL OF BOTANY
layer of individual 30 ? 30 cm quadrats and (b) the proportion of seeds that
germinated across sites of Astragalus bibullatus. Vertical lines represent the
standard error of each mean.
(a)Average number of seeds recovered from each 1 cm thick soil
(b) hierarchical analysis of genetic differentiation among sites and soil layers
(b) in Astragalus bibullatus. The letters A–B represent seed populations re-
covered from successively deeper 1 cm thick soil layers collected at six dif-
ferent sites. Values of ?pare based on allozyme data and were calculated with
the GDA software package (Genetic Data Analysis; Lewis and Zaykin, 2001)
using methods described by Weir (1996). Vertical bars represent 95% confi-
dence intervals that are from 1000 bootstraps of the data across loci. Values
of ?pfor which the confidence intervals do not overlap zero are considered
significantly different from zero.
(a)Levels of differentiation among sites stratified by soil layer and
TABLE 2.Levels of genetic diversity, inbreeding, gene flow, and differentiation among sites for soil seed bank layers collected from populations
of Astragalus bibullatus. Soil layers B and C were combined for analyses. Measures of genetic variation include the proportion of polymorphic
loci across all sites (p), effective allele number (Ae), and expected heterozygosity (He). Inbreeding is indicated by the degree of heterozygote
deficiency, i.e., (He? Ho)/He, or the fixation index, F.
B ? C
Note: Gene flow (Nm) was estimated using the level of genetic differentiation (NmFST) and by private allele analysis (Nmpriv). The latter is
calculated from the average frequency (PI) of alleles found for each soil layer that were unique to one site (NP? the total number of private alleles;
Slatkin 1985). These gene flow estimates were adjusted (Nmadj) to account for differences in the average sample size per population (Nsamp). Levels
of genetic differentiation were estimated using both Weir and Cockerham’s (?p:1984) and Nei’s (FST:1973) methods. Tests for differences between
layers (F values) were made with Friedman’s two-way analysis of ranks (? ? F value too large to be defined, — ? single values per layer so no
test was conducted). Asterisks associated with ?pvalues indicate 95% confidence intervals based on 1000 bootstraps across loci that did not overlap
zero (* ? P ? 0.05; ** ? P ? 0.01; *** ? P ? 0.001).
versity (He) tended to be highest in the A layer and declined
with increasing soil depth (Table 2). Hierarchical analysis of
genetic variation with the GDA program indicated that genetic
differentiation (?; Weir, 1996) was significantly greater than
zero for the among-soil-layers estimate (?S? 0.148, 95% CI
? 0.221–0.049 for differentiation among subpopulations), and
nearly significantly greater than zero for the among-sites es-
timate (?p? 0.082, 95% CI ? 0.159 to ?0.015 for differen-
tiation among populations; Fig. 2b). Confidence intervals are
based on 1000 bootstraps across loci.
The level of genetic differentiation among sites differed de-
pending on the stratum in the seed bank being examined (Table
2). Both the FSTand ?pestimates of population differentiation
from stratified analyses indicated that differences in allele fre-
quencies among populations were much higher for the A than
for the combined B and C layers (Table 2; Fig. 2a). The higher
level of differentiation for the uppermost soil stratum suggests
lower levels of gene flow among sites were prevalent when
these seed populations were formed. The same pattern was
indicated by private allele estimates of gene flow; the two
January 2002] 33MORRIS ET AL.—STRATIFIED ANALYSIS OF THE SOIL SEED BANK
lower soil strata displayed much lower private allele frequen-
cies, and hence higher levels of gene flow, than seed popula-
tions in the A layer (Table 2).
There were differences in the overall level of heterozygosity
and the apparent level of inbreeding among the three soil strata
(Table 2). The expected (He) was higher for the A than for the
combined B and C layers (Table 2). Differences in the level
of heterozygote deficiency were also apparent among layers,
resulting in a fixation index (i.e., the inbreeding coefficient: F
? (He? Ho)/He; Hedrick, 1983) that was more than ten times
greater for the seed populations in the A layer, but the high
variance among loci rendered this difference insignificant (Ta-
Seed populations of Astragalus bibullatus from the three
soil layers displayed differences in their levels of among-pop-
ulation differentiation. Stratified sampling of the soil seed bank
has allowed us to analyze both spatial and temporal variation
in population genetic structure and has provided an historical
perspective on the ecological factors that may have led to the
restricted distribution and reduced abundance of this cedar
glade endemic. The observation of more similar allele fre-
quencies among populations from the older (lower) soil strata
indicates that those seeds were formed under the conditions a
of higher rate of gene flow. A reduction in the level of gene
flow among seed populations in the youngest soil layer may
have been coincident with an increase in the level of local
inbreeding. These most recent seed populations tended to have
higher levels of heterozygote deficiency, indicating there may
have been higher frequencies of selfing and sib mating at the
time that they were produced. These data provide evidence of
historical changes in the level and distribution of genetic var-
iation in A. bibullatus and indicate that these populations may
have been subjected to changes in aspects of their physical
and biological environments that have affected basic popula-
tion genetic processes.
The results of this study have potentially important impli-
cations for our understanding of population genetic processes
and for the management of this endemic taxon. However, it is
important to recognize the assumptions and limitations of the
data presented. First, we are assuming that each soil layer con-
tains seeds of different age and that there have not been sig-
nificant amounts of soil disturbance, which would have ho-
mogenized the seed bank. While it is possible that there has
been some mixing, the high level of genetic differentiation
among soil layers indicates that seed populations in different
layers have remained largely distinct. Ideally, we would like
to obtain the actual ages of seeds from different soil depths.
The methods to conduct such assays are available (Moriuchi
et al., 2000), but it would be prohibitively expensive to obtain
estimates for reasonable samples of seeds from different lay-
Second, because of the high variance often apparent in soil
seed densities (Leck, Parker, and Simpson, 1989; Baskin and
Baskin, 1998), the sample sizes for some of the populations
studied were minimal. Reanalysis of the data using only the
two populations with the largest sample size yields results that
are qualitatively the same as those presented; there was strong
differentiation among layers (?p? 0.103, 95% CI ? 0.199–
0.007) and gene flow was substantially lower among the youn-
gest (Nm ? 1.08) than among the older (Nm ? 2.60) seed
Third, we are assuming that soil strata at different sites rep-
resent equivalent seed age ranges. For example, it is possible
that differences in soil structure could have resulted in varia-
tion among sites in the rates of seed migration down through
the soil column. Different rates of vertical seed migration
among sites would be expected to produce the largest errors
in the relative age, and hence the highest level of genetic dif-
ferentiation, for the deepest seed populations. However, the
pattern found in the present study was just the opposite, with
the strongest differences in allele frequencies among popula-
tions found in the uppermost soil layer, so it is unlikely that
variation in the vertical migration rates of seeds through the
soil had a substantial impact on the stratigraphic differences
in population genetic structure detected.
While the pattern of genetic differentiation among soil lay-
ers is striking, it is possible that these differences represent
the effects of differential selection rather than a temporal re-
cord of differences in relative levels of gene flow among pop-
ulations. For example, seed germination has been reported to
be nonrandom with respect to genotype (Cabin, Mitchell, and
Marshall, 1998), which could affect allele frequencies in older
seed populations. Hence, the older (deeper) seeds could be-
come genetically differentiated from seeds in the upper layer
because of differential germination. However, note that lower
layers were also more homogeneous across sites. While ge-
notype-dependent germination may be expected to be neutral
or to increase the level of differentiation, it is extremely un-
likely that it would result in a conversion of allele frequencies
among sites. Since the primary homogenizing force that is
generally recognized in population genetics is gene flow, it is
much more reasonable to assume that similarity in allele fre-
quencies among sites in the deeper soil layers is the result of
historically higher rates of dispersal.
Seed bank genetic diversity—It has been suggested that
seed banks may act as reservoirs of genetic variation that
would buffer populations from the loss of genetic diversity
during bottlenecks (Templeton and Levin, 1979). Seed bank
populations of A. bibullatus are apparently consistent with this
expectation and contain higher levels of genetic variation than
vegetative populations. Comparison of seed populations of A.
bibullatus to adult individuals sampled in the same region
(Baskauf and Snapp, 1998) indicates that seed banks contained
higher levels of genetic diversity (He? 0.063 for vegetative
vs. 0.156 for seed populations), a larger proportion of poly-
morphic loci (p ? 25.6% for vegetative vs. 50.4% for seed
populations), and averaged more alleles per locus (An? 1.4
for vegetative vs. 1.97 for seed populations). Note that indi-
vidual plants of A. bibullatus can live for several years (A.
Shea, Tennessee Department of Environment and Conserva-
tion, personal communication), so these studies should be
comparable because sampling for both was done within a sin-
gle generation. Moreover, the difference in genetic diversity
was maintained even when the seed population data were re-
stricted to the same subset of allozyme loci and sites used in
the Baskauf and Snapp study of the vegetative populations
(i.e., He? 0.147, p ? 75.0%, and Ae? 3.9 for the restricted
sample of soil seed populations). However, it is not clear that
this pattern is general because some studies have reported
equal levels of genetic variation in seed bank and vegetative
populations (e.g., Tonsor et al., 1993; Mahy, Vekemans, and
34[Vol. 89AMERICAN JOURNAL OF BOTANY
Jacquemart, 1999). In A. bibullatus, the seed bank appears to
represent a significant genetic reservoir that may help preserve
genetic diversity when vegetative populations are small or ab-
The capacity of seed banks to retain higher levels of genetic
diversity may be dependent on seed dormancy characteristics.
Species with strong germination barriers would be expected to
have seed populations that contain a wider range of seed age
classes and to contain genetic variants from a larger number
of vegetative generations than species with seeds that germi-
nate within a few years (Templeton and Levin, 1979). As the
seeds of A. bibullatus have thick, impermeable seed coats, they
are likely to persist in the soil for a long period of time, pro-
ducing seed banks that contain a wide range of seed ages.
Unfortunately, the dormancy characteristics of the other spe-
cies studied are not readily available, so it is difficult to de-
termine whether the lower levels of seed bank diversity re-
ported in some studies are due to differences in the ages of
The relative size of the seed bank population compared to
the vegetative population may also determine its capacity to
act as a genetic reservoir. Depending on dormancy character-
istics, large seed bank populations may be more likely to se-
quester rare genetic variants that are not present in the respec-
tive vegetative population. However, as pointed out by Cabin
et al. (1998), the large size and aggregated spatial distribution
of seed bank populations render them inherently difficult to
sample. Hence, it is likely that even relatively large sample
sizes will miss much of the variation present, and this may
explain why some studies have not detected higher levels of
seed bank genetic diversity (e.g., Tonsor et al., 1993; Mahy,
Vekemans, and Jacquemart, 1999). It is notable that the studies
that did find seed banks that were genetically diverse com-
pared to the extant vegetative populations (i.e., the present
study and McCue and Holtsford, 1998) were on endemic spe-
cies with relatively small population sizes. In both cases it is
possible that higher seed bank genetic diversities reflect his-
torically greater abundances than are evident from the distri-
butions of contemporary populations.
Variation among soil strata—The A. bibullatus seed pop-
ulations from different soil strata differed with respect to their
seed densities, levels of among-site genetic differentiation, ex-
pected heterozygosity, and heterozygote deficiency. In partic-
ular, the uppermost soil layer contained lower densities of
seeds than expected (see below) and seed populations from
this stratum had the highest levels of among-site differentiation
and the highest heterozygote deficiencies. As seed populations
age, their numbers would be expected to decline as individuals
are lost through germination and mortality (Leck, Parker, and
Simpson, 1989). Hence, with a constant rate of input, we
would expect that the youngest seed banks would be the larg-
est and that seed numbers would decrease with soil depth. In
A. bibullatus, the observation that the youngest seed popula-
tions were smaller than populations from the second layer sug-
gests that contemporary rates of seed input have been lower
than earlier seed input rates.
An alternative explanation for the higher number of seeds
in the lower soil layers is variation in vertical migration rates
among soil strata. For example, frost heaving may lead to
more rapid migration rates of seeds through the upper soil and
accumulation in lower layers. However, this is unlikely as it
would lead to homogenization of the soil seed bank, which is
inconsistent with the observed high levels of genetic differ-
entiation among layers. Furthermore, if seeds were accumu-
lating in lower soil layers, then it would be difficult to explain
the observation of high frequencies of unique alleles in the
upper soil layer that were not present in lower layers at the
same sites. While it is not possible to entirely exclude the
possibility that the patterns of seed abundance and genetic var-
iation are due to the accumulation of seeds in deeper soil, the
level of genetic variation and distribution of unique alleles
among layers and sites is more consistent with the hypothesis
that these patterns are due to historical changes in mating pat-
terns and gene flow among populations.
Hierarchical analysis of genetic variation among sites and
soil strata indicated that the level of differentiation among soil
layers was as great or greater than the level of differentiation
among sites sampled. Similar patterns of genetic differentia-
tion between seed bank and seedling populations (Cabin,
1996), seed bank and vegetative populations (McGraw, 1993),
and among seed bank populations of different age (Benning-
ton, McGraw, and Vavrek, 1991) have been observed in other
species. Such variation in the genetic composition of soil seed
banks may be due to fluctuations in allele frequencies in veg-
etative populations (Templeton and Levin, 1979) and nonran-
dom patterns of germination with respect to seed genotype
(Cabin, Mitchell, and Marshall, 1998). In the case of A. bi-
bullatus, the genetic differentiation among seed bank popula-
tions may also have been influenced by historical changes in
mating patterns. The apparent increase in inbreeding in the
uppermost soil seed populations would be expected to de-
crease effective population sizes and increase the probability
of local fixation of alleles. Furthermore, our sampling design
may have been particularly sensitive to the effects of increased
levels of selfing and biparental inbreeding. With very restricted
seed dispersal, our relatively small quadrats (30 ? 30 cm)
would have included seeds from only a few individual plants.
Hence, reduced outcrossing would be expected to lead to in-
creased levels of differentiation among quadrats at a site (i.e.,
a Walhund effect; Hedrick, 1983). Such small-scale differen-
tiation due to rates of inbreeding would also help explain the
higher level of expected heterozygosity (i.e., because of higher
variation in the frequencies of alleles among quadrats within
sites) and the greater heterozygote deficiency observed in the
youngest seed populations. However, note that even if a Wal-
hund effect were responsible for a portion of the observed
heterozygote deficiency, this pattern is still indicative of in-
creased levels of inbreeding for the youngest seed populations.
An increase in the apparent frequency of heterozygous ge-
notypes in older seed populations could also be due to higher
mortality rates for more homozygous seeds (e.g., Del Castillo,
1994), possibly because they were produced through selfing
rather than by outcrossing. However, the alternative hypothesis
that differences in the level of heterozygote deficiency is due
to changes in historical levels of inbreeding is also supported
by the distribution of private alleles among the soil seed pop-
ulations. Both of the older soil layers contained larger numbers
of private alleles that were present at lower frequencies than
in the youngest seed populations. This pattern is consistent
with a recent increase in the level inbreeding since higher fre-
quencies of selfing and sib mating would lead to the random
loss of some rare alleles and the development of local patches
with higher frequencies of other rare alleles. While it is pos-
sible that some of the change in the relative number of het-
erozygous genotypes among soil strata is due to different rates
January 2002] 35MORRIS ET AL.—STRATIFIED ANALYSIS OF THE SOIL SEED BANK
of mortality of inbred and outbred seeds, the observed differ-
ences in unique allele frequencies suggests that at least a por-
tion of the increase in heterozygote deficiency in younger seed
populations is due to an increase in the level of inbreeding
It is notable that some unique alleles in the oldest seed pop-
ulations could be the result of novel somatic mutations, which
are known to occur at relatively high frequencies in aged seeds
(Levin, 1990). For example, 8 out of 24 private alleles in the
two older seed populations were only found once and could
be due to mutations that arose after the seeds were produced.
However, removing these alleles does not produce substantial
changes in our gene flow estimates for the two older soil layers
(Nmadjbecomes 5.14 and 6.02 for the B and C layers, respec-
tively). Hence, the relatively high frequency of rare alleles in
the youngest seed populations is most likely due to a recent
history of increased levels of inbreeding and restricted gene
flow among populations.
Temporal variation—The stratified analysis of seed bank
genetic diversity in A. bibullatus has provided insights into
possible historical changes in processes affecting population
genetic structure. The lack of genetic differentiation among
sites for the oldest soil seed layers indicate that levels of gene
flow were probably higher in the past and that populations
have recently become isolated. Decreased levels of gene flow
among cedar glade populations could be the result of several
factors. For example, it is likely that cedar glades were his-
torically more widespread and had lower densities of trees
(DeSelm, 1994; Heikens and Robertson, 1994). Several lines
of evidence suggest that aboriginal inhabitants of this region
may have used fire to clear these areas of woody vegetation
(Delcourt, 1987; Delcourt et al., 1998). Fire suppression pol-
icies in the last century have apparently led to higher densities
of cedar trees (DeSelm, 1994), and may have increased the
levels of fragmentation of A. bibullatus populations as the hab-
itat quality eroded due to increased shading.
The apparent effects of woody vegetation encroachment on
the viability of A. bibullatus populations can be seen in some
of the extant populations. For example, the Flatrock B site has
one of the lowest census population sizes (A. Shea, Tennessee
Department of Environment and Conservation, personal com-
munication) and the highest soil seed density, suggesting that
plants in this area were much more abundant in the past. The
plants at this site are located along an abandoned road bed that
is surrounded by dense stands of cedars. Numbers of plants at
this site have decreased in recent years, and flowering rates of
these plants are generally very low compared to populations
at more open locations (A. Shea, Tennessee Department of
Environment and Conservation, personal communication).
Such extensive overgrowth by woody species may eventually
lead to extinction of vegetative populations of A. bibullatus,
suggesting that residual seed populations may exist at many
sites in this region where habitat conditions are currently in-
hospitable to their growth and survival.
A possible example of the recovery of such a cryptic pop-
ulation is evident at the Airfield site, where a large population
of A. bibullatus was only recently discovered. Sites in the local
area (e.g., Alexander and the Flat Rock sites) were regularly
counted and surveys made for additional populations in this
area since 1979. However, the large abundance of A. bibullatus
plants at the Airfield site only became apparent in 1996 after
the land owner commenced regular mechanical removal of the
woody vegetation in the area. Whether or not a few vegetative
individuals had persisted at this site and were simply over-
looked, the high density of seeds throughout the soil strata
indicates that a large population of A. bibullatus was present
at this site at some point in the past and that the majority of
the extant population was probably derived from the soil seed
bank within the last few years.
The temporal changes in the population genetic structure of
A. bibullatus observed in this study are consistent with pat-
terns expected under increased fragmentation (Templeton et
al., 1990; Fahrig and Merriam, 1994; Young, Boyle, and
Brown, 1996). While the absolute timescale of these changes
is unknown, based on the studies of seed longevity in the soil
for other species (reviewed in Baskin and Baskin, 1998) and
the thick seed coat of this species, we can surmise that the
oldest seeds in this study could have been produced as much
as a 100 yr ago. In any case, it is probable that intrusion by
woody vegetation and increased urbanization of cedar glades
have contributed to the decreased rates of gene flow and re-
duced population sizes inferred for the uppermost seed layer.
However, the relatively low seed densities and increased in-
breeding apparent in the youngest seed populations suggest
that pollinator availability may also have changed in recent
years. Lack of adequate pollinator service would be expected
to result in lower levels of seed production, higher frequencies
of selfed seeds, and lower rates of gene flow among popula-
tions (Kearns, Inouye, and Waser, 1998), all of which are con-
sistent with the changes observed in the uppermost seed soil
layer compared to the older layers. Additional studies on the
reproductive biology of A. bibullatus may help elucidate the
possible contribution of pollination conditions to the historical
changes in the mating patterns and the prospects for continued
maintenance of genetic diversity in this endemic species.
ARCHBOLD, O. W. 1989. Seed banks and vegetation processes in coniferous
forests. In M. A. Leck, V. T. Parker, and R. L. Simpson [eds.], Ecology
of soil seed banks, 107–122. Academic Press, New York, New York,
BARNEBY, R. C., AND E. L. BRIDGES. 1987. A new species of Astragalus
(Fabaceae) from Tennessee’s Central Basin. Brittonia 39: 358–363.
BASKAUF, C. J., AND S. SNAPP. 1998. Population genetics of the cedar-glade
endimic Astragalus bibullatus (Fabaceae) using isozymes. Annals of the
Missouri Botanical Garden 85: 90–96.
BASKIN, C. C., AND J. M. BASKIN. 1998. Seeds: ecology, biogeography, and
evolution of dormancy and germination. Academic Press, New York,
New York, USA.
BASKIN, J. M., AND C. C. BASKIN. 1999. Cedar glades of the southeastern
United States. In R. C. Anderson, J. S. Fralish, and J. M. Baskin [eds.],
Savannas, barrens, and rock outcrop communities of North America,
206–219. Cambridge University Press, New York, New York, USA.
BENNINGTON, C. C., J. B. MCGRAW, AND M. C. VAVREK. 1991. Ecological
genetic-variation in seed banks. 2. Phenotypic and genetic-differences
between young and old subpopulations of Luzula parviflora. Journal of
Ecology 79: 627–643.
BOSBACH, K., H. HURKA, AND R. HAASE. 1982. The soil seed bank of Cap-
sella bursa-pastoris (Cruciferae): its influence on population variability.
Flora 172: 47–56.
CABIN, R. J. 1996. Genetic comparisons of seed bank and seedling popula-
tions of a perennial desert mustard, Lesquerella fendleri. Evolution 50:
CABIN, R. J., R. J. MITCHELL, AND D. L. MARSHALL. 1998. Do surface plant
and soil seed bank populations differ genetically? A multipopulation
study of the desert mustard Lesquerella fendleri (Brassicaceae). Ameri-
can Journal of Botany 85: 1098–1109.
CHAMBERS, J. C., AND J. A. MACMAHON. 1994. A day in the life of a seed:
36 [Vol. 89AMERICAN JOURNAL OF BOTANY
movements and fates of seeds and their implications for natural and man-
aged systems. Annual Review of Ecology and Systematics 25: 263–292.
CRUZAN, M. B. 2001. Population size and fragmentation thresholds for the
maintenance of genetic diversity in the endemic, Scutellaria montana
(Lamiaceae). Evolution 55: 1569–1580.
DEL CASTILLO, R. F. 1994. Factors influencing the genetic structure of Pha-
celia dubia, a species with a seed bank and large fluctuations in popu-
lation size. Heredity 72: 446–458.
DELCOURT, H. R. 1987. The impact of prehistoric agriculture and land oc-
cupation on natural vegetation. Trends in Ecology and Evolution 2: 39–
DELCOURT, P. A., H. R. DELCOURT, C. R. ISON, W. E. SHARP, AND K. J.
GREMILLION. 1998. Prehistoric human use of fire, the eastern agricultural
complex, and Appalachian oak-chestnut forests: paleoecology of Cliff
Palace Pond, Kentucky. American Antiquity 63: 263–278.
DESELM, H. R. 1994. Tennessee barrens. Castanea 59: 214–225.
ESTILL, J. C., AND M. B. CRUZAN. 2001. Phytogeography of rare plant species
endemic to the southeastern United States. Castanea 66: 3–23.
FAHRIG, L., AND G. MERRIAM. 1994. Conservation of fragmented popula-
tions. Conservation Biology 8: 50–59.
HALL, P., S. WALKER, AND K. BAWA. 1996. Effect of forest fragmentation
on genetic diversity and mating system in a tropical tree, Pithecellobium
elegans. Conservation Biology 10: 757–768.
HEDRICK, P. W. 1983. Genetics of populations. Van Nostrand Reinhold, New
York, New York, USA.
HEIKENS, A. L., AND P. A. ROBERTSON. 1994. Barrens of the Midwest: a
review of the literature. Castanea 59: 184–194.
KEARNS, C. A., D. W. INOUYE, AND N. M. WASER. 1998. Endangered mu-
tualisms: the conservation of plant–pollinator interactions. Annual Re-
view of Ecology and Systematics 29: 83–112.
KELLMAN, M. 1970. The viable seed content of some forest soil in coastal
British Columbia. Canadian Journal of Botany 48: 1383–1385.
LECK, M. A., V. T. PARKER, AND R. L. SIMPSON. 1989. Ecology of soil seed
banks. Academic Press, San Diego, California, USA.
LECK, M. A., AND R. L. SIMPSON. 1987. Seed bank of a freshwater tidal
wetland: turnover and relationship to vegetation change. American Jour-
nal of Botany 74: 360–370.
LEVIN, D. A. 1990. The seed bank as a source of genetic novelty in plants.
American Naturalist 135: 563–572.
LEWIS, P. O., AND D. ZAYKIN. 2001. Genetic data analysis: computer program
for the analysis of genetic data. http://lewis.eeb.uconn.edu/lewishome/
MAHY, G., X. VEKEMANS, AND A. L. JACQUEMART. 1999. Patterns of allo-
zymic variation within Calluna vulgaris populations at seed bank and
adult stages. Heredity 82: 432–440.
MCCUE, K. A., AND T. P. HOLTSFORD. 1998. Seed bank influences on genetic
diversity in the rare annual Clarkia springvillensis (Onagraceae). Amer-
ican Journal of Botany 85: 30–36.
MCGRAW, J. B. 1993. Ecological genetic-variation in seed banks. 4. Differ-
entiation of extant and seed bank derived populations of Eriophorum
vaginatum. Arctic and Alpine Research 25: 45–49.
MORDEN, C. W., AND W. LOEFFLER. 1999. Fragmentation and genetic dif-
ferentiation among subpopulations of the endangered Hawaiian mint
Haplostachys haplostachya (Lamiaceae). Molecular Ecology 8: 617–625.
MORIUCHI, K. S., D. L. VENABLE, C. E. PAKE, AND T. LANGE. 2000. Direct
measurement of the seed bank age structure of a Sonoran Desert annual
plant. Ecology 81: 1133–1138.
NASON, J. D., AND J. L. HAMRICK. 1997. Reproductive and genetic conse-
quences of forest fragmentation: two case studies of neotropical canopy
trees. Journal of Heredity 88: 264–276.
NEI, M. 1973. Analysis of gene diversity in subdivided populations. Pro-
ceedings of the National Academy of Sciences, USA 70: 3321–3323.
PROBER, S. M., AND A. H. D. BROWN. 1994. Conservation of the grassy
white box woodlands: population genetics and fragmentation of Euca-
lyptus albens. Conservation Biology 8: 1003–1013.
QUARTERMAN, E., M. P. BURBANCK, AND D. J. SHURE. 1993. Rock outcrop
communities: limestone, sandstone, and granite. In W. H. Martin, S. G.
Boyce, and A. C. Echternacht [eds.], Biodiversity of the southeastern
United States: upland terrestrial communities, 35–86. John Wiley &
Sons, New York, New York, USA.
QUINLIVAN, B. J. 1968. Seed coat impermeability in the common annual
legume pasture species of Western Australia. Australian Journal of Ex-
perimental Agriculture and Animal Husbandry 8: 695–701.
ROLSTON, M. P. 1978. Water impermeable seed dormancy. Botanical Review
SAS. 1989. SAS/STAT user’s guide, version 6, 4th ed. SAS Institute, Cary,
North Carolina, USA.
SCHNELLER, J. J. 1998. How much genetic variation in fern populations is
stored in the spore banks? A study of Athyrium filix-femina (L) Roth.
Botanical Journal of the Linnean Society 127: 195–206.
SLATKIN, M. 1985. Rare alleles as indicators of gene flow. Evolution 39: 53–
TEMPLETON, A. R., AND D. A. LEVIN. 1979. Evolutionary consequences of
seed pools. American Naturalist 114: 232–249.
TEMPLETON, A. R., K. SHAW, E. ROUTMAN, AND S. K. DAVIS. 1990. The
genetic consequences of habitat fragmentation. Annals of the Missouri
Botanical Gardens 77: 13–27.
TONSOR, S. J., S. KALISZ, J. FISHER, AND T. P. HOLTSFORD. 1993. A life-
history based study of population genetic structure: seed bank to adults
in Plantago lanceolata. Evolution 47: 833–843.
TOOLE, E. H., AND E. BROWN. 1946. Final results of the Duvel buried seed
experiment. Journal of Agricultural Research 72: 201–210.
TURNER, T. F., J. C. TREXLER, J. L. HARRIS, AND J. L. HAYNES. 2000. Nested
cladistic analysis indicates population fragmentation shapes genetic di-
versity in a freshwater mussel. Genetics 154: 777–785.
VAN DER VALK, A. G., AND C. B. DAVIS. 1979. A reconstruction of the
recent vegetational history of a prairie marsh, Eagle Lake, Iowa, from
its seed bank. Aquatic Botany 6: 29–51.
VAVREK, M. C., J. B. MCGRAW, AND C. C. BENNINGTON. 1991. Ecological
genetic-variation in seed banks. 3. Phenotypic and genetic differences
between young and old seed populations of Carex bigelowii. Journal of
Ecology 79: 645–662.
WEIR, B. S. 1996. Genetic data analysis II. Sinauer, Sunderland, Massachu-
WEIR, B. S., AND C. C. COCKERHAM. 1984. Estimating F-statistics for the
analysis of population structure. Evolution 38: 1358–1370.
WHITLOCK, M. C., AND D. E. MCCAULEY. 1999. Indirect measures of gene
flow and migration: FST1/(4Nm ? 1). Heredity 82: 117–125.
YEH, F. C., AND T. J. B. BOYLE. 1997. Population genetic analysis of co-
dominant and dominant markers and quantitative traits. Belgian Journal
of Botany 129: 157.
YOUNG, A., T. BOYLE, AND T. BROWN. 1996. The population genetic con-
sequences of habitat fragmentation for plants. Trends in Ecology and
Evolution 11: 413–418.
YOUNG, A. G., H. G. MERRIAM, AND S. I. WARWICK. 1993. The effects of
forest fragmentation on genetic variation in Acer saccharum Marsh. (sug-
ar maple) populations. Heredity 71: 277–289.